Collected Articles of Kishori Lal

Catalogue no. 13-598-XIE

Collected Articles

of Kishori Lal

Statistics Statistique Canada Canada

Collected Articles of Kishori Lal

Associate Assistant Chief Statistician National Accounts and Analytical Studies Field Statistics Canada June 2003

PREFACE

How does one summarize a lifetime of professional accomplishment? In some instances, one good way is to compile a book, as we have done here. This volume brings together in one place the substantial number of papers written by Kishori Lal during his lengthy career as a national accounts statistician at Statistics Canada.

Kishori's papers cover a range of subject matter, responding to the twisting current of events through parts of five decades. They have one thing in common: All of the papers focus on one or another aspect of the development of Canada's System of National Accounts. Kishori believes deeply in and is utterly devoted to that system. It grew and evolved quite radically during Kishori’s long career. The changes Canada’s SNA went through followed closely, or in some cases led the development of the international SNA standard. He has left his mark indelibly on both.

The advent of the 1993 SNA gave the impetus for several papers. These examined the implications of the new standard for Canada’s national accounts and explored issues associated with its practical implementation in the 1997 historical revision. ‘Production’ was always a central focus of his work and many of the papers in this volume consider aspects of Canada’s input-output accounts. Over the years he also turned his attention to several specific production measurement issues, such as the treatment of ‘financial intermediation services indirectly measured’ (FISIM) and inventory change. International comparisons were a special interest. Indeed one of his best papers, written in the year before he retired, contrasted the United States national accounts with the Canadian accounts. This detailed and authoritative paper was widely acclaimed south as well as north of the border. Subsequently the Australian Bureau of Statistics indicated they intended to prepare a similar paper, extending the comparison to include the Australian national accounts, and the Organisation for Economic Co-operation and Development in Paris asked if they could publish Kishori’s work to give the study even wider exposure.

On a personal note, I reported to Kishori for eight full years, when I was Director of the National Accounts and Environment Division and he was the Director-General of the System of National Accounts Branch. I could not ask for a better boss. The man knows, understands and loves the national accounts more profoundly than anyone else in Canada and a better mentor-teacher on the mysteries of this subject one could not hope to find. He is a true professional, yet also a kind and supportive person. On behalf of all of us at Statistics Canada, I wish him well in his retirement. On behalf of all my fellow citizens, I thank him for his contribution to Canada’s statistical system.

Philip Smith Assistant Chief Statistician National Accounts and Analytical Studies Field June 2003

TABLE OF CONTENTS

1. Preliminary Papers on the 1993 International System of National Accounts, September 1994

1.1 Preliminary Comments on the 1993 System of National Accounts vis-à-vis the Canadian System of National Accounts (1992) ...... 3

1.2 The 1993 International System of National Accounts: Its Implementation for Developing Countries...... 43

1.3 The 1993 International System of National Accounts: Preliminary Views on its Implementation in Canada ...... 63

2. Papers on Measurement of Production, March 1998

2.1 Papers on Measurement of Production: Gross Domestic Product ...... 75

2.2 Services Industries in the Business Sector of the Canadian Economy...... 85

2.3 Financial Intermediation Services Indirectly Measured (FISIM) and the Canadian System of National Accounts (CNSA) ...... 97

2.4 FISIM Revisited: Calculating and Allocating FISIM: Canadian Approach...... 113

2.5 Recording of Change in Inventories in the SNA and in the Business Accounts: A Case Study of Canadian Practices...... 119

2.6 A Note on the Measurement of Industrial Production ...... 133

3. Remaining Differences Between the 1997 Canadian System of National Accounts and the 1993 International System of National Accounts, September 1998...... 141

4. The 1997 Historical Revision of the Canadian System of National Accounts: Record of Changes in Classification of Sectors and Transactions, Concepts and Methodology, October 1998...... 165

5. Value Added by Industry: A Problem of International Comparison, June 1999 ... 247

6. Papers on Canadian Input-Output Tables, August 2000

6.1 Input-Output Accounts: The Rectangular Format...... 257

6.2 Compilation of Input-Output Tables: Canada...... 275

6.3 Canadian Input-Output Tables and Their Integration With Other Sub-Systems of the National Accounts...... 295

6.4 Canadian System of National Accounts: An Integrated Framework ...... 307

6.5 The 1993 International System of National Accounts and The Canadian Input-Output Tables...... 327

6.6 Evolution of the Canadian Input-Output Tables 1961 to date ...... 345

7. The 2001 Revision of the Canadian System of National Accounts: Record of Revisited Issues, September 2001 ...... 351

8. Implementation of the 1993 SNA in Canada Backcasting Issues, June 2002 ...... 383

9. Aggregating Sub-Annual Current Price Value of Changes in Inventories to Annual Totals Under Condition of Inflation: An Inherent Dilemma, July 2002...... 389

10. Measurement of Output, Value added, GDP in Canada and the United States: Similarities and Differences, February 2003 ...... 399

Preliminary Comments on The 1993 System of National Accounts vis-à-vis The Canadian System of National Accounts

by Kishori Lal Director General System of National Accounts Statistics Canada

The International System of National Accounts 1993 (1993 SNA) was published in 1993. This voluminous document, spread over 760 pages, comprising 21 chapters and 6 annexes, was prepared under the auspices of the Inter-Secretariat Working Group on National Accounts, consisting of the Statistical Office of the European Communities (Eurostat), International Monetary Fund (IMF), the Organization of Economic Cooperation and Development (OECD), the Statistical Division (UNSTAT) and regional commissions of the United Nations Secretariat, and the World Bank. Adoption of the 1993 SNA was unanimously recommended to the United Nations Economic and Social Council by its Statistical Commission at its twenty-seventh session, held in New York from February 22 to March 3, 1993. This paper examines the 1993 SNA through the experience of the Canadian System of National Accounts (CSNA). Our comments are organized to highlight certain important areas where the CSNA will need to revise its present practices to conform to the 1993 SNA. The impact of these revisions on the CSNA is noted and an implementation strategy is summarized. There are some areas where we may not be able to implement the recommendations of the 1993 SNA; in such cases, our reasons are stated and alternatives are suggested. Our occasional departures from the 1993 SNA guidelines are primarily prompted by pragmatic considerations. We fully recognize the importance of promoting international comparability, but it should also be recognized that the specific circumstances existing at a given time in different countries can vary, often substantially.

General comments

The last UN SNA manual was published in 1968 and the balance sheet guidelines were published in 1977. An Expert Group Meeting on the Review and Development of the SNA, held in New York in 1982, recommended that a long term review of the SNA be undertaken to produce a revised SNA. The Group recommended that the main objectives of the SNA Review be clarification and simplification of concepts, further harmonization with other international guidelines, such as the IMF Balance of Payments Manual and A Manual on Government Financial Statistics, and updating of the system to fit new circumstances. The 1993 SNA has met these objectives.

The framework of the CSNA bears a close relationship to the one described both in the last UN SNA manual and the 1993 SNA manual. The CSNA, like the 1993 SNA, is both a comprehensive and an integrated system, in the sense that all aspects of the economy are included and all of its components are interrelated. The CSNA comprises production accounts, income and outlay accounts, capital finance accounts, and financial flow and balance sheet accounts. The balancing item in the

Statistics Canada - 3 - Collected Articles of Kishori Lal production accounts, value added, is the starting point of the income and outlay accounts; that of the income and outlay accounts, saving, is the take-off point for the capital finance accounts; and the balancing item of the capital finance accounts, net lending, is the point of departure for the financial flow accounts. Through these linkages, the CSNA is integrated both conceptually and statistically.

The 1993 SNA and the CSNA both feature gross domestic product (GDP) rather than gross national product (GNP) as the primary focus of the production accounts.

Introductory chapters

1. A Thoughtful Introduction

The first three chapters of the 1993 SNA - Chapter 1, Introduction; Chapter II, Overview; and Chapter III, Flows, stocks and accounting rules - are introductory chapters. Chapter I, Introduction, is excellent. It explains the integrating and comprehensive nature of the SNA and provides a general and thoughtful rationale of what is and is not included in the SNA production, consumption and assets boundary. It also links the SNA with economic theory. It is a stand-alone chapter which provides sufficient information to readers to enable them to proceed further or, alternatively, stop there and still have a good appreciation of the SNA. All students of macroeconomics and the SNA will appreciate this chapter.

2. Overview

One of the major virtues of the national accounting system is that it provides a relatively simple macro-economic overview of the economic system. It is important that the overview chapter of the 1993 SNA bring into sharp focus the integrating and interlocking nature of the SNA framework in an easy-to-understand and non-intimidating way. Instead, the present overview chapter, is very long, comprising 54 pages, and the amount of detail is intimidating. In spite of the chapter's length, the summary explanations of concepts, valuation principles, and balancing items are out of context and do not adequately introduce the discussion in the later chapters.

Statistics Canada - 4 - Collected Articles of Kishori Lal 3. Flows, stocks and accounting rules

Chapter III is another introductory chapter which deals with the basic nature of the entries in the accounts and tables of the system. The discussion in this chapter is a general one; definitions of specific flows and stocks and specific applications of the accounting rules appear in later chapters. Here, too, everything discussed in this chapter summarizes what is much better explained in later chapters, where the items are discussed in their proper context.

Chapter IV. Institutional units and sectors

4. Mutually exclusive sectors

The 1993 SNA defines an institutional unit "as an economic entity that is capable, in its own right, of owning assets, incurring liabilities and engaging in economic activities and in transactions with other entities" (paragraph 4.2). The resident institutional units that make up the total economy are grouped into the following five mutually exclusive sectors in the 1993 SNA (paragraph 4.6):

i) the non-financial corporations sector; ii) the financial corporations sector; iii) the general government sector; iv) the non-profit institutions serving households(NPISHs) sector; v) the household sector.

In the CSNA, the financial corporations sector is combined with the non-financial corporations sector, and the two together are called the corporate and government business enterprise or, in short, the corporate sector. The government sector is very similar to the one in the 1993 SNA. In the CSNA, the household sector is combined with the NPISHs sector and the unincorporated business enterprises to form the persons and unincorporated business sector. These three sectors are used for the income and outlay accounts but not for the production accounts. (See also item 11 below, production account for institutional sectors). In the accumulation accounts, such as financial flows and balance sheets, these main sectors are disaggregated into many sub-sectors.

5. Residence

"The total economy is defined as the entire set of resident institutional units. It is divided into sectors that consist of groups of resident institutional units. An institutional unit is resident in a country when it has a centre of economic interest in the economic territory of that country" (paragraph 4.15). However, "Owners of land and building, in the economic territory of a country are deemed always to have a centre of economic interest in that country, even if they do not engage in other economic activities or transactions in their country. All land and buildings are therefore owned by residents" (paragraph 4.16d).

Statistics Canada - 5 - Collected Articles of Kishori Lal 6. Conglomerates

Conglomerates (groups of corporations) are not treated as "...single institutional units... It may be difficult to obtain data for groups whose activities are not closely integrated. Moreover, many conglomerates are much too large and heterogenous for them to be treated as single units, and their size and composition may be continually changing over time as a result of mergers and take-overs" (paragraph 4.39).

7. Quasi-Corporations

The 1993 SNA, like the 1968 UN SNA, recommends that quasi-corporations be treated as if they were corporations. Quasi-corporations include, among others, unincorporated enterprises, ". owned by households which are operated as if they were privately owned corporations" (paragraph 4.50). The CSNA does not distinguish quasi-corporations from other unincorporated enterprises owned by households. In the production account, all unincorporated enterprises form part of the business sector.

8. Sector classification of government establishments

The 1993 SNA recommends that "if a government establishment or group of establishments ... a) charges prices for its outputs that are economically significant, b) is operated and managed in a similar way to a corporation, and c) has a complete set of accounts that enable its operating surpluses, savings, assets and liabilities to be identified and measured, it should be treated as a quasi corporation ... classified, sectored and sub-sectored in the same way as public corporations" (paragraph 4.107).

"...The producer units that remain integrated with the government units that own them are those that cannot be treated as quasi-corporations. They are likely to consist largely, or entirely, of non-market producers: that is, producers most or all of whose output is supplied to other units free or at prices that are not economically significant" (paragraph 4.109).

The operating phrase in the above-quoted paragraphs "charges prices for its output that are economically significant" appears again and again in the 1993 SNA. It is crucial to have an unambiguous understanding of this term. Government grants to government establishments, which form part of the general government sector, have a different impact, per present conventions, on the economy-wide GDP at market prices than they would if provided to government establishments forming part of public corporations. What determines the sector classification of a government establishment fundamentally depends upon the operational meaning of "charging prices for its outputs that are economically significant". The issue of subsidies is further discussed at the end of this paper under the heading "Outstanding Issues".

The 1993 SNA states that prices are not economically significant when they "...do not have a significant influence on the amounts the producers are willing to supply or on the amounts purchasers wish to buy" (paragraph 4.60). This definition suffices to enable national accountants to demarcate the boundary between government establishments and public corporations in most cases.

Statistics Canada - 6 - Collected Articles of Kishori Lal Chapter V. Establishments and industries

There are two areas in which present Statistics Canada practice is not the same as that suggested in the 1993 SNA.

9. Central ancillary activities.

The 1993 SNA recommends: "For example, the purchasing, sales, accounts, computing, maintenance or other departments of an enterprise may all be the responsibility of a head office which is located separately from the establishments in which the principal or secondary activities of the enterprise are carried out. In such a case, the costs of the central ancillary activities must be distributed to the establishments which they serve, for example in proportion to the latter's outputs or costs, and added to the latter's own costs" (paragraph 5.29).

This may be conceptually desirable, but if it is to be to be done properly, the head office has first to be identified as a unit and given a geographical location to which employment and capital expenditures can be assigned. To retain the identity of the unit, its output can be given an imputed value which can be assigned to the establishments that it serves as if they bought its services. This method of handling head offices would lead to the correct industrial distribution of value added, as the costs of the central ancillary activities would be allocated to the establishments served and would be added to their costs. However, total gross output would also have to be increased by the imputed value of head office output. It is not legitimate only to divide up and allocate the costs of the head office to the establishments that it serves because the head office has a geographical location where employment occurs and capital expenditure is made. If head office costs are divided up among establishments the employment and capital expenditures would also have to be divided up, and the head office would have to be made to disappear; otherwise productivity measures would be distorted.

The alternative way of handling head offices is the way it is presently done in Statistics Canada: they are identified as separate units with a geographical location to which employment and capital expenditures can be assigned. Then, for purposes of industrial classification, the whole unit is assigned to a single industry, the one in which the bulk of the value added of the establishments it serves is generated.

10. Units of Homogeneous Production.

The 1993 SNA states that: "For purposes of input-output analysis, the optimal situation would be one in which each producer unit were engaged in only a single productive activity so that an industry could be formed by grouping together all the units engaged in a particular type of productive activity without the intrusion of any secondary activities" (paragraph 5.46). It further states: "Although the unit of homogeneous production may be the optimal unit, ... it may not always be feasible to partition establishments... into a series of mutually exclusive units of homogeneous production. In situations of this kind, it will not be possible to collect directly from the enterprise or establishment the accounting data corresponding to units of homogeneous production. Such data may have to be estimated subsequently by transforming the data supplied by enterprises on the basis of various assumptions or hypotheses" (paragraph 5.47). In the Canadian input-output tables, we do not subdivide establishments to create units of homogeneous production except in the case of

Statistics Canada - 7 - Collected Articles of Kishori Lal construction. The proponents of pure commodity technology perhaps forget that such conceptual perfection would require a separate vector of inputs for each of the twenty thousand or so commodities identified in the market. It is completely unrealistic to seek to achieve such a target. Aggregating twenty thousand commodities into a manageable set of 500-1000 commodity groups can hardly be called generating homogeneous production units.

Chapter VI. The production account

11. Production account for institutional sectors

The 1993 SNA differs from the 1968 UN SNA regarding a proposal to compile production accounts for establishments and industries as well as for institutional units and sectors. According to the 1968 UN SNA, separate production accounts are required for establishments and industries but not for individual institutional sectors: only one consolidated production account is recommended for the economy as a whole. The 1993 SNA further recommends the following accounts for each of the sectors of the economy:

a) The primary distribution of income account; b) The secondary distribution of income account; c) The use of income account. (In the CSNA, the above three accounts are combined into a single income and outlay account). d) The capital account (corresponding to the CSNA's capital finance account). e) The financial account (corresponding to the CSNA's financial flow account). f) Other changes in assets account. g) The balance sheet. (The CSNA's balance sheet account reflects items f) and g), although f) is not fully specified).

The CSNA has implemented these recommendations, although with modifications. The CSNA has a consolidated production account, but the income and outlay and capital finance accounts are published for three resident sectors rather than five. These are:

a) Corporations, including both financial and non-financial enterprises; b) General government; c) Persons, including non-profit institutions serving persons, and unincorporated enterprises;

To complete the system, a non-resident or rest of the world sector is added as the fourth sector.

Financial flow (including most items in the capital finance account) and balance sheet accounts are prepared for many additional sub-sectors of the four main sectors.

Statistics Canada - 8 - Collected Articles of Kishori Lal The 1993 SNA recommends, as noted above, that there should be full production accounts, for institutional units and sectors, full in the sense of gross output, intermediate expenses, and value added (paragraphs 6:1-4). Full production accounts for an institutional sector are not necessary to analyse its income and outlay account or its capital finance account. Their additional utility is quite marginal, especially when the very significant resources that would be required to develop such estimates for the current period are considered.

It is worth noting that the CSNA does produce full production accounts for all years for which input- output tables are compiled but the sector classification is different from that of the 1993 SNA. The CSNA produces annual input-output tables with a lag of 22 years after the reference year. A business sector is created, which comprises all corporate and unincorporated enterprises producing goods and services for sale in the market. Two additional sectors, the general government sector and the household sector (including non-profit institutions serving households), produce goods and services primarily not for sale in the market but for their own consumption. The sector classification in the CSNA input-output accounts is not radically different from the 1993 SNA sector classification, if all unincorporated enterprises are assumed to be quasi-corporations. However, the issue of producing full production accounts for all institutional sectors for the current period remains. Existing data sources do not allow the calculation of such accounts; only value added and its components can be and are compiled, again only for the consolidated account, not for the separate institutional sectors.

A far more useful alternative would be to develop a comprehensive production account for all the producing enterprises in three sectors: business (both corporate and unincorporated enterprises), general government, and non-profits institutions serving households. The only additional production generated in the strictly household sector is that of domestic and personal services produced by employing paid domestic staff. The own-account production of housing services by owner-occupiers is considered to belong to the unincorporated enterprise part of the business sector. Furthermore, all three producing sectors should be subdivided by economic activity. The above alternative (embodied in the CSNA production account) is similar to the comprehensive enterprise sector account recommended by Professor Richard Ruggles in his recent paper, "The United Nations System of National Accounts (SNA): Its Implementation for Developing Countries" (Yale University, May, 1992). He does not support the utilisation of a production account for institutional sectors. His preference, like ours, is to have a comprehensive production account. He writes: "A simpler solution to the institutional sectoring problem would be to have a comprehensive enterprise sector composed of corporations (both financial and non-financial), unincorporated enterprises, government enterprises and non-profit institutions. Such consolidation would reduce the articulation of transactions required between the major sectors of the economy. The total of this enterprise sector would be coincident with the total of the industrial sectors used for measuring production and input- output" (p. 5). There is no point in having delineated production sectors if their inter-sectoral transactions cannot be articulated. It is our judgement that the institutional sectors for production accounts recommended in the 1993 SNA are not very useful for inter-sectoral articulation and if such inter-sectoral articulation is attempted, it will be very costly.

Statistics Canada - 9 - Collected Articles of Kishori Lal 12. The Production boundary in the System

A vitally important issue is the demarcation of the boundary for valuing production for SNA purposes. The 1993 SNA lists the following activities that fall within the production boundary of the System (paragraph 6.18):

a) the production of all individual or collective goods or services that are supplied to units other than their producers, or intended to be so supplied, including the production of goods or services used up in the process of producing such goods or services; b) the own-account production of all goods that are retained by their producers for their own final consumption or gross capital formation; c) the own-account production of housing services by owner-occupiers and of domestic and personal services produced by employing paid domestic staff.

Note that the own-account production of domestic and personal services by members of the household for their own final consumption is excluded from the production boundary (paragraphs 6.19-22). Included in the production boundary, however, is the so-called illegal production. Illegal production was always implicitly contained within the boundary, and in the 1993 SNA this has been properly clarified. Illegal production (paragraph 6.30) comprises. "a) the production of goods and services whose sale, distribution or possession, is forbidden by law; b) production activities which are usually legal but which become illegal when carried out by unauthorized producers; e.g. unlicensed medical practitioners". Further: "Examples of activities which may be illegal but productive in an economic sense include the manufacture and distribution of narcotics, illegal transportation in the form of smuggling ... and services such as prostitution" (paragraph 6.32).

It has generally been assumed that concealed production and the underground economy form part of the production boundary. This, too, has been clarified (paragraphs 6.34-36) in the 1993 SNA.

13. Financial intermediation services indirectly measured (FISIM)

Banks and other financial institutions provide a variety of services. Those that are specifically charged for include currency exchange, handling of cheques etc; and the corresponding revenues form part of the institutions' output. An additional, and very significant, part of their income comes from charging higher interest rates to borrowers and paying lower rates to depositors than they would need to if they charged explicitly for all their services. This "hidden" charge (known as imputed banking service in the 1968 UN SNA) is called financial intermediation services indirectly measured (FISIM) in the 1993 SNA. "The total value of FISIM is measured in the System as the total property income receivable by financial intermediaries minus their total interest payable, excluding the value of any property income receivable from the investment of their own funds..." (paragraph 6.125). The CSNA has not been able to exclude property income from investment of own funds since it is almost impossible to segregate various risk bearing income generating assets, related to own funds, from borrowed funds.

Statistics Canada - 10 - Collected Articles of Kishori Lal the business sector rather than to all borrowers and lenders, as is now recommended in the 1993 SNA. Within the business sector, the 1968 UN SNA does not allocate these services to each industry; instead, it uses a dummy industry which buys the entire service as an intermediate expense and generates an equivalent negative value added.

The 1993 SNA recognizes that certain countries may not be able to allocate FISIM among the various users. Thus it states: "In principle, the total output should, therefore, be allocated among the various recipients or users of the services for which no explicit charges are made. In practice, however, it may be difficult to find a method of allocating the total output among different users in a way which is conceptually satisfactory from an economic viewpoint and for which the requisite data are also available. Some flexibility has therefore to be accepted in the way in which the total output is allocated" (paragraph 6.126). Various ways are suggested by the 1993 SNA to allocate FISIM among users. The following provides a summary of the estimation and allocation procedures of financial intermediation services used in the CSNA. A fuller discussion is presented in my paper entitled "Financial Intermediation Services Indirectly Measured (FISIM) and the Canadian System of National Accounts (CSNA)", May 1994.

In the CSNA, the financial sector is disaggregated into a) banks and other deposit-accepting institutions, b) trust and deposit-accepting mortgage companies, c) credit unions, and d) consumer finance companies. For the banks and other deposit-accepting institutions, estimates are made for intermediation services, as in the 1993 SNA, but with modifications. The intermediation services provided to non-residents are removed from output. This reduced output is allocated to all sectors (business, persons and government) using such services. The value of intermediation services is calculated separately for borrowers and depositors. (Within the CSNA, part of the financial intermediation services relating to borrowers in the personal sector is also called the administrative portion of interest). The allocation of FISIM to the above sectors is based on their share of deposits with the banks and on their borrowings (except government borrowings) from them. At the time of the implementation of the 1993 SNA, the CSNA will need to change its present practice of not imputing and allocating services to non-residents and to government borrowing activity.

In the case of trust and deposit-accepting mortgage companies, intermediation services are estimated only for the personal sector, based on its share of deposits and borrowings. The intermediation services provided to the business and government sectors are removed from output. The business sector, particularly the rental industry (both with respect to rented dwellings and to owner- occupied dwellings) uses these services extensively but no allocation is made to it. Consequently, GDP for the trust and deposit-accepting companies is understated, and there is a corresponding overstatement in GDP for the rest of the business sector, particularly the rental industry. The implementation of the 1993 SNA will lead to a reallocation of the business sector's GDP between trust and mortgage companies on the one hand and the rest of the business sector on the other hand. The GDP of trust and deposit-accepting mortgage companies will further increase by an amount equal to the intermediation services allocated to the government sector.

Statistics Canada - 11 - Collected Articles of Kishori Lal service charge is lowered by the amount of the operating profit. As a result, total output equals the operating expenses including consumption of fixed capital. The present practice of allocating the entire output to the personal sector will need to be revised as the credit unions do provide services to the business sector, particularly the rental industry. In the CSNA, a reduction in the allocation to the personal sector counterbalanced by an allocation to the business sector will reduce GDP for the total economy and, particularly, for the rental industry.

The consumer finance companies are non-deposit-accepting enterprises, and it is assumed that all their lending is to the personal sector. The value of their intermediation services is equated to their operating expenses (intermediate consumption, labour costs, and consumption of fixed capital), and the total is allocated entirely to the personal sector.

The allocation of financial intermediation services does not affect the level of saving in the sector accounts. The additional expenditure on FISIM for deposits by the personal and government sectors is balanced on the income side by an imputed income equal to such expenditures. The additional expenditure on FISIM for borrowers in the personal sector is counterbalanced by an equal reduction in loan interest expenditure (within the CSNA, this reduced loan interest, equivalent to pure interest, is also called the transfer portion of loan interest).

To summarize, the present CSNA practice with respect to the estimation and allocation of financial intermediation services to all users will need to be changed to align it fully with the recommendations of the 1993 SNA. The required changes will affect the valuation of:

(i) services provided to non-residents (ii) services provided to governments on their borrowing activities (iii) services provided to the rental industry by the trust and deposit accepting mortgage companies (iv) services provided to the rental industry by the credit unions.

To complete the accounts, services received by the residents from the rest of the world will have to be included in imports and as inputs for all the users.

14. Insurance

In the 1993 SNA, the calculation of the total output of both life and non-life insurance services is described in paragraph 6.135-140 and further elaborated in Annex IV, The treatment of insurance, social insurance and pensions, (pages 569-583).

Output equals total actual premiums earned plus income from investment of the insurance technical reserves (also called total premium supplements) less total claims due, and less increases (plus decreases) in actuarial reserves and reserves for with-profit insurance. Note that actuarial reserves only apply to life insurance services, not to non-life insurance. Note also that the output definition in paragraph 6.140 is incorrectly worded.

Statistics Canada - 12 - Collected Articles of Kishori Lal The value of output of insurance services in the 1993 SNA is different from that in the CSNA. In the CSNA, income from investment of insurance technical reserves is ignored in the calculation of output. This will need to be corrected at the time of the implementation of the 1993 SNA. Also, at present, output of life insurance is calculated as the administrative costs of insurance corporations, and the entire output is allocated to final consumption of the household sector. The method recommended in the 1993 SNA will generate a similar value of output (apart from the distributed profits to policy holders) and will similarly be allocated entirely to final consumption of the household sector. The output of non-life insurance in the CSNA will increase by the full amount of income from investment of technical reserves.

15. The boundary between intermediate consumption and gross fixed capital formation

"Intermediate consumption measures the value of goods and services that are transformed or entirely used up in the course of production during the accounting period. It does not cover cost of using fixed assets owned by the enterprise nor expenditures on the acquisition of fixed assets" (paragraph 6.157). Whether certain outlays are treated as intermediate consumption or capital expenditures is an important consideration in setting prices. Market prices could be different, and most probably lower, for products whose intermediate expenses are capitalized or recouped over a period longer than one year. Unless there are compelling reasons to deviate from them, SNA conventions should reflect the actual practices of businesses in the market economy. We now examine some important 1993 SNA recommendations: a) Mineral exploration

In another departure from the 1968 UN SNA, the 1993 SNA recommends: "Expenditures on mineral exploration are not treated as intermediate consumption. Whether successful or not, they are needed to acquire new reserves and are, therefore, all classified as gross fixed capital formation" (paragraph 6.166). The CSNA practice in this respect is already largely in line with this recommendation. In the CSNA, mineral prospecting (called exploration and development ) outlays are largely capitalized. More specifically, outlays for exploration and development drilling in petroleum and natural gas industries and on-property exploration and development in other mining industries are capitalized. Off-property mineral prospecting, such as geological and geophysical outlays in petroleum and natural gas industries and general exploration in other mining industries, are, however, treated as intermediate consumption. Capitalized expenditures make up about 80% of total Canadian outlays on mineral prospecting. The present CSNA practice of not capitalizing all expenditures on mineral exploration will need to be changed at the time of the implementation of the 1993 SNA. b) Military equipment

The 1993 SNA states that: "The construction of buildings for use by military personnel, including hospitals and schools, and also of roads, bridges, airfields, docks, etc., for use by military establishments should be treated as fixed capital formation. In addition, machinery and equipment of the same type as used by civil establishments for non-military purposes should also be treated as fixed capital formation" (paragraph 6.171). Conversely, military hardware such as rockets, missiles and their warheads, warships, submarines, fighter aircraft and bombers, should continue to be treated as intermediate consumption.

Statistics Canada - 13 - Collected Articles of Kishori Lal In the 1968 UN SNA, all military expenditures on goods are treated as intermediate consumption. Although the CSNA follows the 1968 UN SNA, the recommendation in the 1993 SNA is reasonable and we support its implementation. When it is implemented, GDP will increase by the amount of consumption of fixed capital. It is worth noting that one would need details of military expenditures classified by product in order to separate the goods to be added to capital formation.

16. Consumption of fixed capital

There are two issues worth noting under the topic of consumption of fixed capital; one is its valuation and the other its coverage. a) Valuation

We support this recommendation. In the CSNA, consumption of fixed capital for the government sector, housing and agriculture is calculated using current costs while for other industries we use what firms report in their financial statements. The new treatment will not affect GDP, but it will change the value of consumption of fixed capital which will be counterbalanced by an equal change in the value of net operating surplus. b) Coverage

The 1993 SNA states that: "Capital consumption is calculated for all fixed assets.... Fixed assets must themselves have been produced as outputs from processes of production as defined in the System. Consumption of fixed capital does not, therefore, cover the depletion or degradation of non- produced assets such as land, mineral or other deposits, or coal, oil, or natural gas" (paragraph 6- 185).

The 1993 SNA further states that: "Losses of fixed assets due to normal accidental damage are also included under consumption of fixed capital, that is, damage caused to assets used in production resulting from their exposure to the risk of fires, storms, accidents due to human errors, etc. When these kinds of accidents occur with predictable regularity they are taken into account in calculating the average service lives of the goods in question.... On the other hand, losses due to war, or to major natural disasters which occur very infrequently - major earthquakes, volcanic eruptions, tidal waves, exceptionally severe hurricanes, etc. - are not included under consumption of fixed capital. There is no reason for such losses to be charged in the production accounts as cost of production. The values of the assets lost in these ways are recorded in the other changes in volume of assets accounts." (paragraph 6.187). We completely agree that the destruction to property due to the exceptionally severe hurricanes or major earthquakes should have no impact on consumption of fixed capital.

Statistics Canada - 14 - Collected Articles of Kishori Lal In the 1993 SNA, the value of output of insurance companies is primarily equal to total premiums earned plus net income from investment of insurance technical reserves less claims paid to compensate for losses of fixed assets due to both normal accidental damage and major natural disasters. Such claims lead to an increase in the surplus of the insured enterprises and to an identical decrease in the surplus of the insurance companies, leaving the total surplus in the economy unchanged. In the CSNA income and expenditure accounts, such payments received by the insured enterprises form part of miscellaneous valuation adjustments to consumption of fixed capital in the calculation of their value added accounts. (By implication such claims are deducted from their reported intermediate consumption of total premiums paid, as without this adjustment the output of the insurance companies, which is net of claims, will not balance with its use). This procedure will need to be changed to be in conformity with the 1993 SNA. The miscellaneous valuation adjustment will need to be added to profits or surplus rather than to consumption of fixed capital. This change will affect only net domestic product, leaving unchanged both GDP at market prices and GDP at factor cost; additions to profits will be counterbalanced by a decrease in consumption of fixed capital.

17. General valuation principles

In the 1993 SNA, the preferred method of valuation of output of goods and services produced for the market is at basic prices, especially when a system of VAT, or similar deductible tax, is in operation (paragraph 6.218). Note that the concept of basic prices in the 1993 SNA is the same as that of approximate basic prices in the 1968 UN SNA. In the Input-Output Accounts, the CSNA already uses a basis of valuation that is close to this one. a) Taxes and subsidies on products

The 1993 SNA states that: "The basic price is the amount receivable by the producer from the purchaser for a unit of a good or service produced as output minus any tax payable and plus any subsidy receivable, on that unit as a consequence of its production or sale. It excludes any transport charges invoiced separately by the producer" (paragraph 6.205a). Further: "When output is recorded at basic prices, any tax on the product actually payable is treated as if it were paid by the purchaser directly to the government instead of being an integral part of the price paid to the producer. Conversely, any subsidy on the product is treated as if it were received directly by the purchaser and not the producer" (paragraph 6.206).

In its input-output tables, the CSNA reports taxes on production separately, as recommended; however, subsidies on products are not excluded from the basic prices, contrary to the recommendation. We agree that subsidies should be treated as negative margins on products, just as taxes on products are treated as positive margins for balancing commodity flows in the input-output tables. This change will not affect GDP at market prices but will change all calculations at basic prices.

Statistics Canada - 15 - Collected Articles of Kishori Lal b) Valuation of intermediate consumption

The 1993 SNA further recommends that: "Expenditures on goods and services intended to be used for intermediate consumption should be valued at purchasers' prices" (paragraph 6.220). We have no fundamental problem with this recommendation; this is how the basic records are kept in the industrial surveys. However, for deriving gross value added by industry at market prices (which in total is equal to total expenditure on GDP at market prices), one needs to value both outputs and intermediate consumption at the same basic prices. Otherwise, incomes and expenditures in the economy will not balance. Thus, one would need to add an aggregate of taxes less subsidies on products to the value added (derived through output at basic prices minus intermediate consumption at purchasers' prices) to balance it with GDP at market prices.

Chapter VII. The primary distribution of income account

"The general purpose of the primary distribution of income account is to show how primary incomes (such as compensation of employees, taxes on production and imports, subsidies and operating surplus) are distributed among institutional units and sectors. Primary incomes are incomes that accrue to institutional units as a consequence of their involvement in processes of production or ownership of assets that may be needed for purposes of production. They are payable out of the value added created by production" (paragraph 7.2). The primary distribution of income account "... consists of two consecutive accounts: the generation of income account and the allocation of primary income account" (paragraph 7.1). The generation of income account "...represents a further extension or elaboration of the production account in which the primary incomes accruing to governments units and to the units participating directly in production are recorded. Like the production account, it may be compiled for establishments and industries as well as for institutional units and sectors" (paragraph 7.3).

As noted earlier in our comments on Chapter VI, The production account, the CSNA produces production accounts (in the sense of output, intermediate consumption and value added) for establishments and industries but not for institutional units or sectors. We do not find very useful or cost effective to produce a generation of income account by institutional units or sectors.

The generation of income account shows the sectors in which primary incomes originate as distinct from the sectors destined to receive such incomes. The allocation of primary income account focuses on resident institutional units or sectors in their capacity as recipients of primary incomes rather than as producers whose activities generate primary incomes. The allocation of primary income account is really a part of the CSNA income and outlay accounts. For example, compensation of employees is attributed entirely to households in the allocation of primary income account.

The chapter continues by defining and clarifying the various components of primary incomes. The following changes will need to be made in the CSNA to bring it into conformity with the 1993 SNA.

Statistics Canada - 16 - Collected Articles of Kishori Lal 18. Payroll taxes

Payroll taxes are taxes payable by the employer on the wage or salary bill. Such taxes are quite commonly applied by the various governments in Canada. The CSNA treats such taxes as part of compensation of employees, whereas the 1993 SNA recommends that such taxes be "treated as taxes on production in the same way as taxes on buildings, land or other assets used in production" (paragraph 7.21). Implementation of this recommendation will not change the GDP at market prices, but it will reduce GDP at factor cost which will be counterbalanced by an increase of taxes on production, and it will also reduce personal income.

19. Mixed income

Mixed income in the 1993 SNA corresponds to net income of unincorporated business in the CSNA. The 1993 SNA states that: "Mixed income is the term reserved for the balancing item of the generation of income account of unincorporated enterprises owned by members of households ..." (paragraph 7.81). In the CSNA production accounts, mixed income is not a balancing item but, rather, is explicitly calculated for most industries using information from tax records, surveys etc. As the CSNA does not have a separate production account for unincorporated enterprises, but only a consolidated production account of the business sector as a whole, the balancing item is the operating surplus of the entire sector.

20. Property income

Property income is defined in the 1993 SNA as: "... the income receivable by the owner of a financial asset or a tangible non-produced asset in return for providing funds to, or putting the tangible non-produced asset at the disposal of another institutional unit" (paragraph 7.88). Property incomes are paid out of the operating surplus of the institutional units using such assets. The payment, for example, for the use of land and sub-soil assets (typically called rent for land and royalties for sub-soil assets) is, in the 1993 SNA, part of the value added of the enterprise using such assets. This means that land rent is not part of intermediate consumption of, say, agriculture. Similarly, royalty payments are not part of intermediate consumption of the mining or forestry industries. Such payments are part of their value added.

This procedure is not followed in the CSNA. Instead, these expenditures are treated as intermediate consumption of the user industries. The implementation of this recommendation will not affect GDP of the economy or even that of the business sector, but it will lead to an increase in the GDP of agriculture, mining and forestry industries counterbalanced by a decrease of value added in the real estate and royalty industries.

21. Reinvested earnings on direct foreign investment

Statistics Canada - 17 - Collected Articles of Kishori Lal while the other is the imputed reinvestment of those earnings" (paragraph 7.120. The CSNA does not follow this procedure.

Starting with the first quarter of 1994, reinvested earnings will be incorporated in the Canadian balance of international payments in both the current and capital accounts as per the recommendations in both the 1993 SNA and the 1993 IMF Balance of Payments Manual. The Balance of Payments will specifically show the data on reinvested earnings in both current and capital accounts. The rest of the CSNA, particularly national income and expenditure accounts and financial flows accounts, will not incorporate the new treatment but, instead, will provide reconciliation accounts to allow users to go from one approach to the other. This arrangement will remain in place until the full implementation of both the 1993 SNA and IMF Balance of Payments Manual, at the time of the forthcoming CSNA historical revision.

The implementation of reinvested earnings on foreign direct investment, in the Canadian balance of international payments, will lead to an increase in its current account payments, counterbalanced by an increase of capital inflow in the capital account.

Chapter VIII. The secondary distribution of income account

Apart from the balance of primary incomes, which is the balancing item carried forward from the primary distribution of income accounts, all the entries in the secondary distribution of income account consist of current transfers such as: (i) current taxes on income, wealth, etc., (ii) social contributions and benefits, or (iii) other current transfers. As noted earlier, the CSNA income and outlay accounts are a consolidation of three accounts in the 1993 SNA , namely the primary distribution of income account, the secondary distribution of income account and the use of income account (Chapter IX).

22. Social transfers in kind

This chapter defines, clarifies and classifies various entries forming part of the secondary distribution of income account. An important proposed change, both for the CSNA and the 1968 UN SNA, is the inclusion in this account of social transfers in kind. As the 1993 SNA states: "In the System, final consumption expenditure is incurred only by general government, NPISHs and households. All of households' consumption expenditure is incurred on their own behalf. Consumption expenditure by general government, on the other hand, is either for the benefit of the community at large (collective consumption) or for the benefit of individual households. By convention, all consumption expenditure by NPISHs is treated as being for the benefit of individual households... Consumption expenditures by general government and NPISHs on behalf of households (their individual consumption expenditures) are undertaken for the purpose of making social transfers in kind. They cover the non-market output of both general government and NPISHs delivered to households free or at prices that are not economically significant, as well as goods and services bought from market producers and provided to households free or at prices that are not economically significant" (paragraph 8.38).

Statistics Canada - 18 - Collected Articles of Kishori Lal We welcome this recommendation as it would help international comparability of household income and household consumption. The inclusion of social transfers in kind (important examples are education and health) will add an identical value to the income and consumption of households, without affecting their savings. Implementing this change will require a functional distribution of government expenditures. There has been a long-standing need to develop a functional distribution of government expenditures on an SNA basis. This new requirement should give a further impetus to the development of such estimates.

23. Income, capital gains, taxes on capital gains and savings

The 1993 SNA incorporates capital gains ( also called holding gains) in an accumulation account but taxes thereon are recorded in an income account. This seems inconsistent. The issues are fairly complex and it is important to clarify them. "From a theoretical point of view, income is often defined as the maximum amount that a household, or other unit, can consume without reducing its real net worth. However, the real net worth of a unit may be changed as a result of the receipt or payment of capital transfers and as a result of real holding gains or losses that accrue on its assets or liabilities. It may also be changed by events such as natural disasters that change the volume of assets. Capital transfers, real holding gains or losses and other changes in the volume of assets due to the effect of events such as natural disasters are specifically excluded from disposable income as measured here. Capital transfers are recorded in the capital account of the System, while other changes in the volume of assets and real holding gains or losses are recorded in the other changes in assets account" (paragraph 8.15).

The 1993 SNA concept of income is equivalent to the above concept only when net worth at the beginning of the period is not changed by capital transfers, real holding gains or losses or events such as those that destroy assets. Thus the 1993 SNA defines its concept of income as "... the maximum amount that a household or other unit can afford to spend on consumption goods or services during the accounting period without having to finance its expenditures by reducing its cash, by disposing of other financial or non-financial assets or by increasing its liabilities" (paragraph 8.15).

At current prices, capital gains, both realised and unrealised, are recorded in the other changes in assets account. This is necessary in order to reconcile the market values of assets at the end of the accounting period with those at the beginning. The question remains, however, as to where to record taxes on realised capital gains. It is inappropriate to record these taxes as a deduction in the balance sheet or in the other changes in assets account, as that would violate the principle of valuing assets at market value.

One possibility would be to record taxes on capital gains in the same place as capital taxes namely, the capital account. This approach, however, is open to the same objection as the above approach, because the capital gains and the taxes thereon are still not recorded in the same account. Furthermore, unlike the taxes that are included in the capital account, such as capital levies and taxes on capital transfers (see Chapter X, The Capital Account, paragraph 10.136), taxes on capital gains are essentially income taxes rather than taxes on capital.

Statistics Canada - 19 - Collected Articles of Kishori Lal According to the 1993 SNA, these taxes should be recorded as a deduction from current income, as are all income taxes (see paragraph 8.52). At first glance, deducting taxes on capital gains from current income, which itself does not include capital gains, seems anomalous. It also seems to distort saving. One must, however, consider the measurement of saving -- an increase in the net worth of assets -- in the context of the entire sequence of SNA accounts. This increase in the net worth of assets appears in successive entries in the production, distribution and use of income, and accumulation accounts. At the end of the process, the increased net worth is the same whether the capital gains tax has been recorded in the income account or in some other account.

Since there are no compelling conceptual reasons to record taxes in any of the accumulation accounts, and it is difficult to separate income taxes from taxes on capital gains, we fully support the convention adopted in the 1993 SNA to record such taxes in the income account. In the CSNA, taxes on capital gains are recorded as a deduction in the income account, as recommended. There is, however, one small item, succession duties and estate taxes, which appears in the CSNA income and outlay account which will need to be transferred to the CSNA capital account to be consistent with the 1993 SNA.

Chapter IX. The use of income account

The purpose of the use of income account is to show how households, government units and non-- profit institutions serving households allocate their disposable income between final consumption and saving. This chapter clarifies various issues in the final consumption area, such as the distinction between goods and services, durable and non-durable goods, new and used goods, individual and collective goods, individual and collective services, consumption expenditures, and actual consumption, etc.

24. Final consumption expenditure and actual final consumption

An important development, both in relation to the CSNA and the 1968 UN SNA, is the specification in the 1993 SNA of the interrelationship between final consumption expenditure and actual final consumption for the three sectors (general government, NPISHs and households) in which final consumption takes place (paragraphs 9.93-99). This treatment is a counterpart of that of social transfers in kind in the secondary distribution of income account. On a practical level, it may be noted that each of the aggregates, whether referring to consumption expenditure or actual final consumption, has to be derived from data on expenditures.

Thus, final consumption expenditure consists of the following components:

Statistics Canada - 20 - Collected Articles of Kishori Lal Actual final consumption of households is measured by the value of all the individual consumption goods or services acquired by resident households. There are three sets of goods and services entering into household actual final consumption:

a) those acquired through expenditure by households themselves: their value is given by item (1) above; b) those acquired as social transfers in kind from NPISHs: their value is given by item (2) above; c) those acquired as social transfers from general government: their value is given by item (3.1) above.

The value of actual final consumption of general government is given by item (3.2) above. NPISHs have no actual final consumption.

It should be repeated that actual final consumption for the whole economy is exactly equal to final consumption expenditures.

We welcome this recommendation as it would help international comparability of household consumption.

Chapter X. The capital account

The capital account is the first of four accounts dealing with changes in the values of assets and liabilities held by institutional units. The purpose of the capital account is to record the values of non-financial assets acquired or disposed of by resident institutional units, the change in net worth due to saving and capital transfers and, as a balancing item, net lending or borrowing . The capital account in the 1993 SNA is the same as the capital finance account in the CSNA.

Five categories of changes in assets are distinguished (paragraph 10.25):

a) Gross fixed capital formation; b) Consumption of fixed capital; c) Changes in inventories; d) Acquisitions less disposals of valuables; e) Acquisitions less disposals of non-produced non-financial assets.

25. Gross capital formation

The 1993 SNA states: "Gross capital formation is measured by the total value of the gross fixed capital formation, changes in inventories and acquisitions less disposals of valuables" (paragraph 10.32). Acquisitions less disposals of valuables is a brand new item in the definition of gross capital formation, for both the CSNA and the 1968 UN SNA. As is later explained: "Valuables consist of: a) precious stones and metals such as diamonds, non-monetary gold, platinum ... held by any units including enterprises provided that they are not intended to be used as intermediate inputs into processes of production; b) paintings, sculptures, etc. recognized as works of art and antiques; c)

Statistics Canada - 21 - Collected Articles of Kishori Lal other valuables, such as jewellery fashioned out of precious stones, metals and collections" (paragraph 10.116).

This is an important additional element in the capital account which also applies to the expenditure side of GDP. Although we are receptive to revising our definition of gross capital formation, we recognize that it will be difficult to estimate expenditures on valuables.

26. Gross fixed capital formation

In most cases, the CSNA and the 1993 SNA agree as to what constitutes fixed capital formation. However, there are certain cases where the CSNA will need to revise its conventions to conform to those of the 1993 SNA. Military equipment is one such case and it has already been discussed in item 15.b above. Others cases follow: a) Inventories of structures

According to the 1993 SNA: "When there is no contract of sale agreed in advance, the output produced by the construction enterprise must be recorded as work-in-progress or as additions to the producers' inventories of finished goods, depending upon whether the construction is completed. For example, finished dwellings built speculatively remain as additions to producers' inventories of finished goods until they are sold or otherwise acquired by users" (paragraph 10.75). In the CSNA, all structures, completed or unfinished, with or without contract of sale, are classified as fixed capital formation. Implementation of the approach recommended in 1993 SNA will have an effect on gross fixed capital formation counterbalanced by an identical change in inventories. This will require additional information which is not collected at present. b) Machinery and equipment and progress payments

"Gross fixed capital formation is not recorded until the ownership of the fixed assets is transferred to the unit that intends to use them in production. Thus, new machinery and equipment that has not yet been sold forms part of additions to inventories of finished goods held by the producers of the assets. Similarly, imported machinery and equipment is not recorded as gross fixed capital formation until it is acquired by the unit that intends to use it" (paragraph 10.81).

When progress payments are made for equipment such as ships, aircraft and rolling stock which take a long time to complete, the ownership is not assumed to be transferred in stages as the payments are made, even if there is a contract of sale agreed in advance. Progress payments made under a contract of sale or otherwise are recorded as a financial transaction in the financial account, and not as fixed capital formation. The same rule applies to progress payments made on imported machinery and equipment. This will require changes to the estimation of gross fixed capital formation in the CSNA. c) Livestock

The 1993 SNA recommends (actually it repeats the recommendation of the 1968 UN SNA) that livestock used in production year after year, such as breeding stock, dairy cattle, sheep reared for

Statistics Canada - 22 - Collected Articles of Kishori Lal wool and draught animals, be treated as fixed assets. On the other hand, animals raised for slaughter, including poultry, are not fixed assets but are included in inventories (see paragraph 10.83). The CSNA has not followed this recommendation; instead the value of acquisitions less disposals of livestock during an accounting period is allocated to inventories.

We support the recommendation as it would better reflect economic reality and also promote better international comparability. When it is implemented, the increase in fixed capital formation will be counterbalanced by a decrease in inventories. GDP will also increase by the amount of consumption of fixed capital. d) Plantations, orchards, etc.

The 1993 SNA recommends (again, it repeats the recommendation of the 1968 UN SNA) that trees cultivated in plantations for the products they yield year after year - such as fruit trees, vines, rubber trees, palm trees, etc. - be treated as fixed assets (see paragraph 10.83). The value of such fixed assets may be approximated, if necessary, by the costs incurred in their production during the period (paragraph 10.88).

The CSNA has not followed this recommendation. Instead, it treats these costs as intermediate consumption. We support the recommendation as it, too, would better reflect economic reality and promote better international comparability. When it is implemented, GDP will increase by the amount of this capital formation which will, of course, be subject to capital consumption allowance. e) Entertainment, literary or artistic originals

The 1993 SNA recommends treating original creations - such as original films, sound recordings, manuscripts, etc. - as capital formation (paragraph 10.94). The CSNA does not capitalize the value of such originals. In some cases, it is added to inventories and in others, it is treated as intermediate consumption. Our data sources are quite weak in this area and we may not be able to implement the recommendation, although, in principle, we support it.

27. Acquisitions less disposals of valuables

As indicated above, the 1993 SNA recommends that the acquisition less disposals of valuables be reflected in the capital account. "Valuables are assets that are not used primarily for production or consumption, that do not deteriorate over time under normal conditions and that are acquired and held primarily as stores of value" (paragraph 10.116). We support this recommendation in principle. Due to data problems, however, the CSNA has not been able to record explicitly such expenditures in the capital finance account; instead, they form part of the balancing item, net lending or borrowing of the sectors.

28. Acquisitions less disposals of non-produced non-financial assets

The 1993 SNA recommends that acquisitions less disposals of non-produced non-financial assets be reported in the capital account. These assets consist of land, sub-soil assets that may be used in the

Statistics Canada - 23 - Collected Articles of Kishori Lal production of goods and services and intangible assets such as patented entities, leases, other transferable contracts, etc. (paragraphs 10.120-130). We also support this recommendation in principle. Here, too, due to data problems, the CSNA has not been able to record explicitly such expenditures in the capital finance account; instead they form part of the balancing item, net lending or borrowing of the sectors.

29. Net lending or borrowing

As noted, in the CSNA, the balancing item of the capital finance account, net lending or borrowing, at present includes net purchase of existing non-financial assets. This is equivalent to the 1993 SNA definition of acquisitions less disposals of both valuables and non-produced, non-financial assets, by every institutional sector. At the level of the total economy, acquisitions less disposals of these assets would cancel out but they could be of significant value for different institutional sectors. We recognize that this is a major weakness in the capital finance accounts of the CSNA.

Chapter XI. The financial account

The financial account records transactions involving financial assets and liabilities between institutional units or between institutional units and the rest of the world. "The financial account is ...the final account, in the full sequence of accounts, that records transactions between institutional units. The financial account does not have a balancing item that is carried forward to another account... Rather, net balance of the financial account is equal in magnitude ...to the balancing item of the capital account" (paragraph 11.3).

Net lending or net borrowing, the balancing item of the capital finance account, is carried forward into the financial account. The financial account indicates how the net borrowing sectors obtain the necessary financial resources, by incurring liabilities or reducing assets, and how the net lending sectors allocate their surpluses by acquiring assets or reducing liabilities.

30. Presentational or definitional differences

As financial innovation has led to the development and proliferation of new, and often complex, financial instruments to meet the needs of the investors with respect to maturity, yield, avoidance of risk, and other factors, the 1993 SNA rightly suggests that: "A substantial degree of flexibility in presentation is therefore appropriate to meet national needs and to reflect national practices" (paragraph 11.15). This is a sensible guideline. a) Banker's acceptance (BA)

The 1993 SNA states that: "A banker's acceptance involves the acceptance by financial institutions of drafts or bills of exchange and the unconditional promise to pay a specific amount at a specified date. The banker's acceptance represents an unconditional claim on the part of the holder and an unconditional liability on the part of the accepting bank; the bank's counterpart asset is a claim on its customer. For this reason, the SNA recommends that the banker's acceptance be treated as an actual financial asset even though no funds may have been exchanged" (paragraph 11.27). The CSNA does

Statistics Canada - 24 - Collected Articles of Kishori Lal treat the banker's acceptance as an asset of the holder and a liability of the issuer, but does not find it useful to further gross-up the transaction by showing the banker's acceptance as an asset and a liability of the accepting bank. b) Gold

The 1993 SNA has demystified the treatment of gold in the various accounts. For example, when the monetary authorities acquire gold, this transaction "... is recorded in the capital account as a positive entry under acquisition less disposals of valuables or change in inventories and the counterpart entries are recorded in the accounts of the institutional units or the rest of the world supplying the gold. When non-monetary gold is acquired from abroad, the entry is recorded under imports. Monetization or demonetization itself does not give rise to entries in the financial accounts; instead, the change in balance sheet positions is accounted for by entries in the other changes in volume of assets account as a reclassification, i.e., the reclassification of gold in inventories or gold as valuables to monetary gold" (paragraph 11.65). The transactions are portrayed exactly as they occur between institutional units, whereas in the present treatment, any export or import of gold sold or purchased by domestic monetary authorities is not shown at all in the merchandise trade statistics. Monetization or demonetization of gold, in the 1993 SNA, is purely a classification issue; there is no transaction involved. See also item 40, Further on gold. c) Certificates of deposit

The 1993 SNA recommends that certificates of deposit be treated as short-term paper, just like other securities such as treasury bills, bonds, etc. (paragraph 11.74). The CSNA treats such certificates as part of currency and deposits. There is merit in the proposed classification of the 1993 SNA.

Chapter XII. Other changes in assets account

31. The Reconciliation statement in the balance sheets

There are two kinds of changes in assets: "The first kind consists of changes that are due to factors such as discoveries or depletion of sub-soil resources, destruction by war or other political events, or destruction by natural catastrophes, all of which actually change the volume of assets. The second kind consists of changes in the values of assets, liabilities, and net worth due to changes in the level and structure of prices, which are reflected in holding gains and losses. Thus, the other changes in assets accounts are subdivided into the other changes in the volume of assets account and the revaluation account" (paragraph 12.2).

Statistics Canada - 25 - Collected Articles of Kishori Lal There is no separate other changes in assets account in the CSNA. Many of the items discussed, in the 1993 SNA, in the context of the other changes in assets account form part, in principle, of the reconciliation statement (not yet fully developed) in the CSNA balance sheets. The development and compilation of this account in the CSNA will help to specify several items which still remain consolidated or unexplained, thus permitting better analysis of balance sheets.

32. Link to environmental satellite accounts

Several of the items reported in the other changes in the volume of assets account, such as the economic appearance and disappearance of non-produced tangible assets and the environmental degradation of fixed assets, provide a link to the emerging environment satellite accounts (paragraph 12.9). The CSNA is in the process of developing such estimates in its environment satellite accounts. Thus, the recommendation of the 1993 SNA is very timely and welcome.

33. Decomposition of holding gains and losses

The revaluation account records the nominal holding gains or losses accruing during the accounting period to the owners of financial and non financial assets and liabilities. Nominal holding gains are further decomposed in the 1993 SNA into neutral holding gains and real holding gains. A neutral holding gain is defined as the value of the holding gain that would accrue if the price of the asset changed in the same proportion as the general price level. A real holding gain is defined as the value of the additional command over real resources accruing to the holding of an asset as a result of a change in its price relative to the prices of goods and services in general in the economy (paragraphs 12.63-64).

Estimation of changes in the nominal holding gains or losses is very important for a realistic interpretation of a balance sheet and income statement. However, given the resource constraints in many national statistical organizations, we question the need to give a high priority to decomposing the nominal holding gains into neutral and real.

34. Holding gains for financial assets and land

The definition and measurement of holding gains for financial assets and liabilities are both problematic and controversial, as illustrated below: a) Bonds

"A bond is a security that gives the holder the unconditional right to a fixed money income or contractually determined variable money income over a specified period of time and also the right to a fixed sum as repayment of principal on a specified date or dates..." (paragraph 12.109). "The prices of marketable bonds change, however, when the market rates of interest change, the prices varying inversely with the interest rate movements. The impact of a given interest rate change on the price of an individual bond is less, the closer that bond is to maturity. Changes in bond prices that are attributable to changes in market rates of interest constitute price, and not quantum changes. They therefore generate nominal holding gains or losses for both the issuers and the holders of the bonds" (paragraph 12.111). The Government of Canada, for example, is a major issuer of bonds. As

Statistics Canada - 26 - Collected Articles of Kishori Lal interest rates decline, the market prices of bonds increase. Should higher values of assets be reported for the holders of these bonds and higher values of liabilities and a higher debt position for the government sector when the face value of the debt position has not changed? Although we are rather uncomfortable with the 1993 SNA recommendation, we agree that there must be full symmetry in the valuation of assets and liabilities of the same instrument. See also item 37. b) Shares

There is no discussion in this chapter of holding gains on tradable shares of corporations. Shareholders receive substantial holding gains or suffer substantial holding losses, as evidenced by the recent turbulence in the world stock markets. Again, the issue of symmetry in the valuation by holders and issuers of corporate shares cannot be ignored. c) Land

There is no discussion of holding gains on land and other non-depreciating fixed assets. There are, in fact, substantial holding gains and losses accruing to such assets.

Chapter XIII. The balance sheet

The 1968 UN SNA did not develop guidelines for the national and sectoral balance sheet and reconciliation accounts. Such guidelines were published later as Provisional International Guidelines on the National and Sectoral Balance Sheet and Reconciliation Accounts of the System of National Accounts, Series M, No. 60, United Nations, New York, 1977. The CSNA publishes national and sectoral balance sheet accounts annually, starting with the reference year 1961.

"For an institutional unit or sector, the balance sheet provides an indication of economic status - that is, the financial and non-financial resources at its disposal that are summarized in the balancing item, net worth. For the economy as a whole, the balance sheet shows what is often referred to as national wealth -the sum of non-financial assets and net claims on the rest of the world" (paragraph 13.2). The 1993 SNA integrates the balance sheets in the overall System. As it explains: "The balance sheet completes the sequence of accounts, showing the ultimate result of the entries in the production, distribution and use of income, and accumulation accounts" (paragraph 13.3). A basic accounting identity links the opening balance sheet and the closing balance sheet for a given asset. Table 13.2 (pages 300-301) of the 1993 SNA lists the items which link the opening and closing balance sheet. This table is quite untidy and at places incorrectly specified.

This chapter provides a very useful list and definitions of entries in the balance sheet. There are, however, three points which, from the CSNA's perspective, are especially important, namely, coverage and classification, valuation, and memorandum items.

Statistics Canada - 27 - Collected Articles of Kishori Lal 35. Coverage and classification

The comments on the preceding three chapters - The capital account, The financial account, and Other changes in assets account apply equally well to The balance sheet chapter. It is useful here to list items already commented upon under the following headings: a) Coverage

The CSNA will need to include the following items in the balance sheet to conform to the 1993 SNA.

i) Part of military equipment and structures - see also item 15.b ii) Plantations, orchards, etc. - see also item 26.d iii) Entertainment, literary or artistic originals - see also item 26.e b) Classification

The CSNA includes the following items in the balance sheet but its classification is different from the one proposed in the 1993 SNA.

i) Inventories of unsold structures - see also item 26.a ii) Livestock - see also item 26.c iii) Banker's acceptances - see also item 30.a iv) Certificates of deposits - see also item 30.c c) Recording of certain items

Due to data problems, the CSNA has not been able to record several items in the balance sheet; they either implicitly form part of other financial assets or are ignored.

i) Acquisitions less disposals of valuables. see also item 27 ii) Acquisitions less disposals of non-produced, non-financial assets - see also items 28 and 32

Only agricultural land and land underneath buildings and structures are recorded at present in the CSNA balance sheet. Other non-produced non-financial assets, such as subsoil assets, forests, mineral resources are either implicitly reflected in the value of other financial assets or ignored. The valuation of some of these assets is presently underway and they will form part of the CSNA Environment Satellite Account.

36. Valuation

The 1993 SNA states that: "For the balance sheets to be consistent with the accumulation accounts of the System, a particular item in the balance sheet should be valued as if it were being acquired on the date to which the balance sheet relates, including any associated costs of ownership transfer in the case of non-financial assets. This implies that assets and liabilities (and thus net worth) are to be

Statistics Canada - 28 - Collected Articles of Kishori Lal valued using a set of prices that are current on the date to which the balance sheet relates and that refer to specific assets". (Paragraph 13.25).

The above valuation principle is not consistently applied in the CSNA balance sheet account. All non-financial assets covered in the balance sheet are valued at current prices but the valuation of financial assets and liabilities is mixed; some are at current prices while others, such as bonds and corporate shares, are at book value. See also items 34a and 34b above for further details.

37. Memorandum items

The 1993 SNA recommends that consumer durables, direct foreign investment, net equity of households in unfunded pension schemes and face values of long term bonds be included as memorandum items to the balance sheet (paragraph 13.84-89). We regard this as a useful recommendation. a) Consumer durables

In the CSNA, consumer durables already form part of the balance sheet and can easily be shown as a memorandum item, if so desired. b) Direct foreign investment

In the CSNA balance sheet, we record holdings of financial assets and liabilities for the rest of the world sector but do not separate foreign direct investment from foreign portfolio investment. The inclusion of foreign direct investment as a memorandum item seems sensible. c) Net equity of households in unfunded pension schemes

Unfunded pension schemes, which are common in the public sector, are, by definition, defined benefit schemes. However, there is no accumulated pool of assets from which to pay benefits. The 1993 SNA recommends that the present value to households of promises made by these schemes to pay future pension benefits be shown as a memorandum item in the balance sheet as assets of households.

The CSNA does not include such assets in the balance sheet. This will need to be revised at the time of the implementation of the 1993 SNA. d) Bonds and corporate shares

The 1993 SNA recommends that, for long-term bonds and corporate equity securities, alternatives to current market value such as nominal (face) value may be shown as memorandum items in the balance sheet.

The CSNA values such bonds and shares only at face value. Providing both face value and market value is a very sensible recommendation. See also items 34a and 34b above.

Statistics Canada - 29 - Collected Articles of Kishori Lal Chapter XIV. The rest of the world account

"It should be noted that the rest of the world account is presented from the point of view of the rest of the world so that a resource for the rest of the world is a use for total economy or vice versa. If a balancing item is positive, it signifies a surplus of the rest of the world and a deficit for the total economy..." (paragraph 14.4).

38. Progress payments and exports and imports

"In line with the general principles adopted in the System, the exports and imports of goods should be recorded at the time at which ownership of the goods in question passes from a resident to a non- resident and vice versa. This principle is sufficient to determine the coverage of international trade in goods but in practice certain exceptions, which are specified below, are made to the principle" (paragraph 14.55). Exceptions to the change of ownership principle are described in paragraphs 14.57-14.64

i) The system imputes a change of ownership from lessor to lessee when a financial lease is arranged even though legally the leased good remains the property of the lessor, at least until the termination of the lease when legal ownership of the good is usually transferred to the lessee (paragraph 14.58). ii) The second exception to the change of ownership principle concerns goods shipped by an enterprise to a branch or subsidiary in a foreign country. Legally, the ownership of the goods may remain unchanged in such circumstances, but a de facto change of ownership is imputed (paragraph 14.59). iii) The third exception is one in which a change of ownership may occur but is ignored in the accounts. The exception relates to merchants or commodity dealers who buy commodities from non-residents and then sell them again to non-residents, within the same accounting period, without the commodities actually entering the country in which the merchants are resident (paragraph 14.60). iv) The fourth and final exception to the change of ownership principle relates to goods which are sent for processing abroad. It is recommended that such goods be treated on a gross basis in exports and imports even if no change of ownership has occurred (paragraph 14.61- 64).

These are the only exceptions to the rules on ownership accepted in the System and the IMF Balance of Payments Manual. Progress payments are treated as financial transactions even if there is a contract of sale. Such goods will be valued in exports and imports only when they are "physically" in the hands of owners. At present, progress payments for machinery and equipment are treated as fixed capital formation in the CSNA and as exports or imports of goods in the Canadian balance of international payments. This will need to be changed to conform to the new guidelines.

Processing is another case where the present Canadian BOP statistics will need to be changed to conform to the new guidelines.

Statistics Canada - 30 - Collected Articles of Kishori Lal 39. Harmonization

An important feature of the 1993 SNA is the harmonization of its concepts and valuation rules with other international guidelines such as the IMF Balance of Payments Manual. This chapter has fully succeeded in doing so, as is demonstrated in section F, Relationships between the current external transaction and accumulation accounts and the balance of payments accounts (paragraphs 14.147- 155). One important item which is not harmonized is the financial intermediation services. In the 1993 SNA, "...financial intermediation services indirectly measured (FISIM) is included in the external account of goods and services, reflecting services that are not explicitly charged, but is not shown under imports and exports of services in the balance of payments accounts. However, it is included indistinguishably under "investment income, interest" in those accounts..." (paragraph 14.149). As noted earlier, this lack of harmonization does not affect the current account balance in the balance of payments.

The issue of f.o.b. valuation of exports and imports and the c.i.f. valuation of detailed imports is dealt with in my paper, "The 1993 International System of National Accounts and the Canadian Input-Output Tables", May 1994.

40. Further on gold

The issue of the treatment of gold has been dealt with in our comments on Chapter XI The financial account (see item 30b). One point which needs to be further clarified refers to transactions in monetary gold between the authorities of different countries. Such transactions are recognized as financial items and are reflected in the financial account; they are not included in imports or exports of goods (paragraph 14.93). The 1993 SNA distinguishes three types of gold: monetary gold owned by monetary authorities as a component of international reserves; gold held as a store of value; and other gold used for industrial purposes. A transaction in the second or third category is always treated as a transaction in a commodity, no matter who the transactor is. A transaction in monetary gold can only be conducted between the monetary authorities of one country and the monetary authorities of another country or with international monetary authorities; such a transaction is always a financial transaction. All other transactions in gold are recognized as transactions in gold as a commodity.

Chapter XV. Supply and use tables and input-output

"The System includes an integrated set of supply and use tables, or matrices, as well as symmetric input-output tables, or matrices. They provide a detailed analysis of the process of production and the use of goods and services (products) and the income generated in that production. The concepts and definitions in the input-output tables of the SNA are the same as in the rest of the system" (paragraph 15.1). As in the 1968 UN SNA, "the integration of the input-output in the overall system of national accounts is an important feature of the SNA" (paragraph 15.2). We fully agree with the importance attached to the input-output tables in the SNA.

Statistics Canada - 31 - Collected Articles of Kishori Lal This chapter deals with several important issues:

i) definition of basic prices; ii) valuation of product flows, particularly of exports and imports; iii) gross value added at factor cost; iv) presentation of non-market producers; v) gross fixed capital formation, stocks of fixed assets and labour inputs; vi) conversion of supply and use tables into symmetric tables; vii) cross-classification of value added by sectors and industries; and viii) double deflation.

The valuation of exports and imports has been fully harmonized with that of the revised IMF Balance of Payments Manual, 1993. The presentation of non-market producers is quite useful. Adding three rows, as supplementary information, to the value added quadrant of the use tables, one each for gross fixed capital formation, stocks of fixed assets, and labour inputs, will facilitate productivity analysis and provide useful additional information to a broad range of users. The 1993 SNA recommendation that the supply and use tables be converted into symmetric tables using commodity technology assumption is, however, problematic.

All these items have been commented upon in detail in my paper, "The 1993 International System of National Accounts and the Canadian Input- Output Tables", May 1994. However, the comments upon the last two items -cross classification of value added by sectors and industries and double deflation- are repeated below, to provide a link to our comments upon items in other chapters of the 1993 SNA.

41. Cross-classification of value added by sectors and industries

The 1993 SNA recommends that value added by industry in the use tables be cross-classified by institutional sectors so that production can be related to income and consumption (see paragraphs 15.106-110). This is a laudable recommendation but it cannot be implemented by most statistical organizations. The approach requires a well functioning business register linking establishments to enterprises and sectors. Such information needs to be current, so that one can relate the current profile of establishment-based input-output industry statistics to the institutional-based structure of the economy. It also demands a highly articulated micro database which can be readily added to provide aggregates for the SNA. These conditions are rigorous. Given that the approach is costly to implement and of limited usefulness to relate production with income and consumption, an alternative presentation is to by-pass the production account and the generation of income account by institutional sector and to go direct to primary allocation of income account by institutional sector.

42. Double deflation

Statistics Canada - 32 - Collected Articles of Kishori Lal framework by using the double deflation method, as the difference between: a) The value of output deflated by a price index of output; b) The value of intermediate consumption deflated by a price index for these inputs" (paragraph 15.162).

The double deflation method in the context of inputs and outputs at basic prices is a very efficient technique. A similar recommendation was made earlier in the UN Manual on National Accounts at Constant Prices, Series M, No. 64, New York, 1979 (in short, UN Constant Price Manual). The Canadian practice conforms to the recommendation in the UN Constant Price Manual , which states: "In an ideal world real product by kind of activity would always be derived from an input-output table by double deflation" (p. 55). In Canada, the input-output tables form the core of the production accounts. Input-output tables in full detail are produced annually both in current and constant prices. Real output by industry is produced using the preferred double deflation approach.

In the double deflation approach, one deflates commodity outputs at basic prices of an industry, its intermediate uses of commodity inputs at basic prices as well as taxes on products and other taxes on production, net of subsidies. The difference between the deflated value of outputs and the total deflated value of commodity inputs and net taxes on products and production equals GDP at factor cost at constant prices. The double deflation approach satisfies the requirement of an identity between GDP income and expenditure based estimates in constant prices. However, there are certain hazards in using double deflation for deriving the GDP of an industry whose value added represents a small proportion of total gross output.

For such an industry, GDP estimated by double deflation might be erratic, because small shifts in the relative prices of intermediate inputs and gross output could translate into big shifts in the resultant value added at constant prices. Here, the UN Constant Price Manual guidelines are not entirely satisfactory. They state: "The solution to this problem, however, may simply be to consolidate industries with very small ratios of value added to gross output with related industries at earlier or later stages of production. In other words, the problem of instability may be solved by aggregation into larger units whose values added are large enough in relation to gross output not to be too sensitive to the effects of changes in prices or technology" (p. 53). In one year, value added in one industry may be erratic but in the next, a different industry might suffer. Ad hoc aggregation into large units would disturb the continuity of time series. Thus, one needs additional guidelines. In the CSNA, we have solved this problem as follows: Values added are combined as suggested by the UN Constant Price Manual. The combined value added is then redistributed using gross output or any other indicator as a proxy, but the combined value added for a given sub-aggregate remains unchanged. Without this constraint, the above noted GDP identity requirement will not be satisfied. These comments on the UN Constant Price Manual apply equally to the 1993 SNA recommendation on double deflation.

Chapter XVI . Price and volume measures

Statistics Canada - 33 - Collected Articles of Kishori Lal mentioned above. The 1993 SNA differs from both the 1968 UN SNA and the 1979 UN Constant Price Manual in two important respects: it recommends the use of chain indices, and it incorporates as an integral part of the system the calculation of trading gains and losses from changes in the terms of trade. The issue of double deflation, also considered in this chapter, is not commented upon here as it has just been addressed.

43. Chain indices

It is self- evident that the more remote the base year becomes, the less relevant are its prices for purposes of deflating the value of current flows of goods and services. When the base year is changed, there are two ways in which the series on either side of the new base year may be linked. The first method is to revalue not only the series for all years subsequent to the new base year at new prices, but also for all the years preceding it in order to have an unbroken series expanding on either side of the new base year. The second method is to leave unchanged the constant price growth rates for all years up to and including the new base year, and simply use the new base year prices for valuing all flows of goods and services from the new base year onwards. In order to develop an unbroken series expanding on either side of the new base, one would simply link the series through a chain index. The CSNA rejects the first method in favour of the second on conceptual grounds. A new base year is required because the old one is not relevant anymore: thus one cannot effectively argue for the first method. Both the 1968 UN SNA and the 1979 UN Constant Price Manual opt for the first method whereas the 1993 SNA recommends the use of chain indices for linking the series.

The 1993 SNA recommends: "In general, constant price series should not be allowed to run for more than five, at the most, ten years, without rebasing" (paragraph 16.76). It further recommends: "The preferred measure of year to year movements of GDP at constant prices is a Fisher volume index, changes over longer periods being obtained by chaining: i.e., by cumulating the year to year movements" (paragraph 16.73a). In the same paragraph, it also notes: "Chain indices should only be used to measure year to year movements and not quarter to quarter movements" (paragraph 16.73e).

While we agree that the quarterly chaining of unadjusted seasonal data is not desirable, we believe it can sometimes be useful to chain seasonally adjusted data on a quarterly basis. A recent example occurred in Canada when the government introduced the Goods and Services Tax (a VAT), which caused a major shift in relative prices. Users wanted quick information about the impact of the new relative price regime on the indices. The quarterly chain series, derived using seasonally adjusted data, met the requirement. The CSNA provides two sets of quarterly chain indices, one rebased and linked annually and the other quarterly.

44. Additivity problem

When the series preceding the new base year are chained to the series succeeding the new base year, "the problem that emerges ... is that the constant price values for the components do not add up to the constant price values of the aggregates after the series have been linked" (paragraph 16.37).

Statistics Canada - 34 - Collected Articles of Kishori Lal In the CSNA's fixed weight indices for final expenditure (rebased and linked every five years) additional "adjusting entry" series are shown for each published aggregate, calculated as the difference between the linked aggregate and the sum of its published, linked components. This approach has two advantages: it alerts users to the problem by displaying explicitly the extent of non-additivity, and it reassures them that no elements are missing from the accounting system. The CSNA's monthly fixed weight estimates of real value added by industry, on the other hand, while they are linked in a similar manner, contain no such "adjusting entry" series. The reason is a practical one. There are a great many industry aggregates in the system and including all the "adjusting entry" series explicitly would probably do more to confuse than to clarify the picture.

45. Terms of trade

The 1993 SNA recommends the estimating of trading gains and losses due to changes in the terms of trade for calculating real gross domestic income (GDI). The 1993 SNA states: "... an improvement in the terms of trade makes it possible for an increased volume of goods and services to be purchased by residents out of the incomes generated by a given level of domestic production. Real GDI measures the purchasing power of the total incomes generated by domestic production so that when the terms of trade change there may be a significant divergence between the movements of GDP at constant prices and real GDI. The difference between the change in GDP at constant prices and real GDI is generally described as the trading gain (or loss). The differences between movements in GDP at constant prices and real GDI are not always small. If imports and exports are large relative to GDP, and if the commodity composition of the goods and services which make up imports and exports are very different, the scope for potential trading gains and losses may be large" (paragraph 16.152). The recommendation to calculate real GDI is not currently implemented in many countries, including Canada. Its implementation will be very useful for all countries, particularly developing countries, where the differences in the commodity composition of exports and imports are typically large. It will provide very useful additional analytical constructs.

Chapter XVII. Population and labour inputs

46. Total hours worked

The 1993 SNA requires that the measurement of labour input variables (both salaried employment and self employment) be consistent with the production of goods and services within its production boundary, in order to examine productivity. It states that "Output per job would be an excessively crude measure of productivity and total hours worked is the preferred measure of labour inputs for the System" (paragraph 17.11). It also specifies in the same paragraph which items should be included in, and excluded from, statistics of hours worked. Statistics of hours worked should exclude: a) hours paid for but not worked, such as paid annual leave, paid public holidays, paid sick leave; b) meal breaks and c) time spent on travel from home to work and vice versa. It is laudable that the 1993 SNA has harmonized its definition of hours worked with the one adopted by the International Labour Organization. This is one of the many examples where the various international statistical organizations have, in this round of revisions, combined their efforts, and sometimes made compromises, in order to develop identical definitions for identical transactions, thus enhancing the usefulness of the statistics.

Statistics Canada - 35 - Collected Articles of Kishori Lal We fully support the view that total hours worked is a preferred measure of labour inputs for productivity, but we also need to report total hours paid in order to trace average labour compensation.

Chapter XVIII. Functional classifications

This chapter lists four functional classifications in order to report certain transactions of producers and of three institutional sectors: households, general government and non-profit institutions serving households. They are called functional classifications because they identify the functions - in the sense of the purposes or objectives - for which these transactors engage in certain transactions. These are (paragraph 18.1):

Classification of individual consumption by purpose (COICOP) Classification of the functions of the government (COFOG) Classification of the purposes of non-profit institutions serving households (COPNI) Classification of outlays of producers by purpose (COPP)

47. Present status

The 1993 SNA states: "the classifications of individual consumption by purpose and the classification of outlays of producers by purpose are provisional and classification of outlays of producers by purpose in particular may be substantially revised before it can be published. The classification of the purposes of non-profit institutions serving households is unchanged from the 1968 version of the SNA. ...The classification of the functions of government needs modifying both to identify more precisely social transfers in kind and to identify more fully functions that are of increasing policy concern such as relief of poverty, preventive health care and repair and prevention of environment damage" (paragraph 18.5). It is not clear when will we receive a final version of the above classifications.

Applications and extensions of the 1993 SNA

The last three chapters of the 1993 SNA - Chapter XIX, Application of the integrated framework to various circumstances and needs; Chapter XX, Social accounting matrices; and Chapter XXI, Satellite analysis and accounts - demonstrate by way of examples how a country may adapt the central framework, by using it in a flexible way, to meet special types of analysis.

48. Chapter XIX. Application of the integrated framework to various circumstances and needs

The 1993 SNA states: "The purpose of this chapter is not to elaborate in detail how a country may design its system following the SNA, but to show how, by using it in a flexible way, a country may adapt the central framework to special circumstances or types of analysis. The chapter provides a number of examples of such adaptations. In order to avoid misunderstanding, it must be clear that these adaptations are not outside the central framework but rather they are ways and means of building up a central framework with specific features according to national circumstances and

Statistics Canada - 36 - Collected Articles of Kishori Lal needs" (paragraph 19.6). We applaud the inclusion of this chapter in the 1993 SNA. It provides reassurance to national accountants at various statistical organizations, who may need more sub- sectoring in certain areas and less detail in others, given their unique policy and analytical needs. The various applications are accompanied by sound practical advice, and provided in a flexible mode. Among the issues discussed are the sub-sectoring of the household sector, expenditures by enterprises in transition economies, high inflation, quarterly and regional accounts, etc.

49. Chapter XX. Social accounting matrices

This chapter provides another example of the flexibility of the central framework to meet special needs or to allow special types of analysis. Flexibility is provided by presenting the SNA accounts in a matrix format called "social accounting matrix" (SAM) rather than in the traditional T format. "A SAM is defined as the presentation of SNA accounts in a matrix which elaborates the linkages between a supply and use table and institutional sector accounts. In many instances, SAMs have been applied to an analysis of interrelationships between structural features of an economy and the distribution of income and expenditure among household groups" (paragraph 20.4). SAMs have so far been compiled and applied in a few developing countries. As the 1993 SNA explains: "A SAM applies the properties of a matrix format to incorporate specific details on various economic flows. Traditionally, it has been applied to specific types of analysis, focussing on causes and consequences of various aspects of inequality among household groups" (paragraph 20.26). It is further suggested that: "... casting accounts into a SAM framework implies that matrix algebra can be applied to balance them" (paragraph 20.121). The CSNA has no experience in compiling SAMs, hence we have no particular comments on their presentation. A matrix presentation, with the help of matrix algebra, permits the calculation of the inverse of the matrix for modelling and policy analysis. However, a mere matrix presentation is no substitute for an integrated and a coherent data base for policy analysis.

50. Chapter XXI. Satellite analysis and accounts

Applications of the integrated framework (as noted in Chapter XIX) and SAMs (as noted in Chapter XX) are within the constraints of the central framework of the SNA. The central framework presents a number of characteristics which give it the advantages of an integrated accounting structure. It is exhaustive and consistent within the boundary of the economic activities it covers. On the other hand, there are certain limitations as to what may be accommodated directly in the central framework. "The SNA does not claim that its categories and concepts are in all cases the only right ones. Additional or different requirements necessitate the development of complementary or alternative categories and concepts" (paragraph 21.3). Such additional or different requirements are now accommodated in satellite accounts.

This chapter provides examples of satellite accounts, some which fall within the boundary of economic production in the SNA and others which extend it. The tourism satellite account is an example of the first kind and the valuation of household work and environment accounts are examples of the second kind. These examples are for illustrative purposes only. Countries are encouraged to adapt or modify them to serve their own specific needs for data development or analysis. We applaud the inclusion of this additional flexibility in the 1993 SNA.

Statistics Canada - 37 - Collected Articles of Kishori Lal Outstanding issues

51. Subsidies

The treatment of subsidies has not been given an adequate airing in this revision process. The Core Expert Group of the 1993 SNA met several times during the revision process to resolve the troublesome treatment of subsidies in the SNA but could not reach a satisfactory agreement. Consequently, the 1968 UN SNA treatment is retained. Given that the 1993 SNA is applicable to every country (unlike the 1968 UN SNA which applied to market oriented economies), and given the overwhelming importance of subsidies in developing countries and, even more so, in the former centrally planned economies and the highly visible and significant subsidies on products such as agriculture in several developed countries, the retention of the 1968 UN SNA treatment is highly regrettable. Subsidies were not a major issue when the 1968 UN SNA was drafted but, today, their present treatment distorts our understanding of market valuation of agricultural products, rental on housing, food, transportation charges and many other consumer products in many countries.

When a government subsidizes a product, its market price is reduced by the subsidy. The government, in principle, could provide the same support to consumers of the product by giving them "coupons" exactly equal in value to the subsidy, but this fiscal rearrangement increases GDP at market prices though the cost to the government has not changed. Let us take another example where a government establishes a unit, say, a broadcasting corporation, in which only 30% of total costs are covered through sale of advertisements and the other 70% are covered by the government from general revenues. Let us say that the total cost is $100. This unit is part of general government and its total cost of $100 forms part of final demand and, hence, of GDP at market prices. If this unit is reclassified as a government business enterprise and the government still covers 70% of its cost through subsidies, the valuation of GDP at market prices drops from $100 to $30. The net cost to the government has not changed.

The underlying economic activity in the above examples has remained the same, and the total cost of providing service has not changed, yet the valuation of GDP at market prices has changed. This is a troublesome situation. National accountants of both developing countries and transition economies have also expressed reservations about the present treatment of subsidies.

A pragmatic solution would be to treat some of the highly subsidised products as non-market products, thus mitigating the impact on GDP at market prices. This can be accommodated with some minor modifications to the present guidelines of the 1993 SNA. Countries in which subsidies cover a significant part of the overall cost of production could treat such subsidies as a purchase of goods and services, rather than as transfers from government to business toward current costs of production. Then, GDP at market prices would not change following a fiscal rearrangement or a reclassification of units from the government sector to the business sector or vice versa.

52. Valuation of own-account production of building services by government units.

It is regrettable that the own-account production of building services by government units has remained outside the production boundary of the 1993 SNA. There is hardly any conceptual difference in this instance with the own- account production of housing services by owner occupiers,

Statistics Canada - 38 - Collected Articles of Kishori Lal yet the latter is included in the 1993 SNA production boundary, while the former is not. The exclusion does not seem logical.

Government owned and occupied buildings are a far more common occurrence in developing and former centrally planned economies than in industrialised countries. Given the increasing acceptance and implementation of liberal economic reforms and privatization, in both developing and former centrally planned economies, the number of such buildings is expected to decline. Without the inclusion of services they provide in production, both inter-temporal and international comparisons would be less accurate.

In the spirit of flexibility in the implementation of the 1993 SNA, both developing and ex-centrally planned economies could apply the following approach in estimating the value of the own-account production of building services by government units and include it in their production accounts.

Government units may own or rent the buildings they occupy. When the buildings are rented, the rentals paid are one of the components of the intermediate consumption whose costs enter into the calculation of the value of the non-market output produced by government establishments. When the buildings are owned by government, the building services produced for own consumption by government should be similarly valued at the rentals that would be paid if they were rented in order to obtain estimates of the value of intermediate consumption of government that are invariant to whether its buildings happen to be owned or rented.

Several components of the costs of producing rental services are already included in government expenditures, namely wages and other costs incurred in maintaining, repairing and cleaning buildings, capital consumption on buildings, and property or other taxes paid etc. A most important item, not included at present, is notional interest charges on the total value of a building. We believe that it should be dealt with along the following lines. Rather than creating a separate vector of actual and notional expenditures that are equivalent to rental value, our proposal is to simply add to the government expenditures two notional items: capital consumption on buildings, and interest charges on the total value of buildings. (These interest charges are for productive purposes and are analogous to those in the rental industry). These items will form part of the gross operating surplus of the government sector. Capital consumption on government buildings is already imputed in the 1993 SNA. Adding these items to actual maintenance and repair costs and property taxes would cover most of the cost elements of rentals in the market. This alternative pragmatic proposal does not require a direct estimation of imputed rental, yet the result comes close to it. The only additional statistical requirement is information on long term bond rates or mortgage rates, which is readily available.

Statistics Canada - 39 - Collected Articles of Kishori Lal Concluding remarks

The 1993 SNA is a vast improvement over the 1968 UN SNA in terms of clarification and simplification of concepts and harmonization with other international guidelines. Some recommendations affect the level of GDP, some affect the balance of international payments, while others simply provide a thoughtful articulation of certain important measures. The new measure of actual final consumption, which has already been used in the international purchasing power parities project and in the Material Product System (MPS) applied in the former centrally planned economies, will help international comparability of household consumption. The 1993 SNA has demystified the treatment of gold. It has incorporated a new measure of acquisition less disposals of valuables and of non-produced non-financial assets in the accumulation accounts, thus articulating the net lending or borrowing position of the institutional sectors as well as providing a link to the emerging environmental satellite accounts. It has recommended that financial intermediation services be allocated to all users of such services, and that military structures and equipment of the same type as those used by civil establishments be treated as gross fixed capital formation.

From the perspective of the CSNA, only the treatment of subsidies and building services produced on own account by government units, however, remains unsatisfactory. By and large, however, the recommendations are implementable and will provide a coherent, integrated analytical framework and credible international comparability.

Statistics Canada - 40 - Collected Articles of Kishori Lal Note:

I acknowledge with gratitude the help that was provided in the preparation of this paper by several of my colleagues, particularly Abe Tarasofsky, Philip Smith, John Joisce and Gylliane Gervais; their comments were most useful.

An earlier draft of this paper was presented at the 22nd General Conference of International Association for Research in Income and Wealth (IARIW), Flims, Switzerland, August 30 - September 5, 1992. At that time, all the chapters of the revised SNA had not been received. A revised draft including comments on all the chapters was presented at the UN sponsored Interregional Seminar on the Revised System of National Accounts, Aguascalientes, Mexico, October 5 - 9, 1992.

Statistics Canada - 41 - Collected Articles of Kishori Lal

The 1993 International System of National Accounts: Its Implementation for Developing Countries

by Kishori Lal Director General System of National Accounts Statistics Canada

The International System of National Accounts 1993 (1993 SNA) was published in 1993. This voluminous document, spread over 760 pages, comprising 21 chapters and 6 annexes, was prepared under the auspices of the Inter-Secretariat Working Group on National Accounts (ISWGNA). ISWGNA consisted of the Statistical Office of the European Communities (Eurostat), International Monetary Fund (IMF), the Organization of Economic Cooperation and Development (OECD), the Statistical Division (UNSTAT) and regional commissions of the United Nations Secretariat, and the World Bank. Adoption of the 1993 SNA was unanimously recommended to the United Nations Economic and Social Council by its Statistical Commission at its twenty-seventh session, held in New York from February 22 to March 3, 1993. This paper describes the revision process leading to its publication and the participation of the experts from developing countries in this process. It further examines some of the major changes proposed in the 1993 SNA vis-à-vis the 1968 edition of the U.N. SNA. Its implementation is examined, largely, though not exclusively, from the perspective of developing countries. There are some areas where we do not support the recommendations of the 1993 SNA; in such cases, our reasons for differing are stated and alternatives are suggested.

Revision process

The U.N. Statistical Commission at its Twenty-first Session, in 1981, considered the question of a revision to the 1968 edition of the UN SNA. An Expert Group meeting on the Review and Development of the SNA was convened in New York, in March 1982. The Expert Group recommended that a long term review of the SNA be undertaken to produce a revised SNA. The Group further recommended that the main objectives of the SNA Review be clarification and simplification of concept, further harmonization with other international guidelines, such as the IMF Balance of Payments Manual and A Manual on Government Financial Statistics, and the updating of the system to fit new circumstances. The Expert Group felt that future work could be accomplished more efficiently if a continuous working group were established by the secretariats of the international agencies generally concerned with the SNA. The Inter- Secretariat Working Group on National Accounts originally consisted of the UN Statistical Office, the Organization for Economic Cooperation and Development (OECD), the Statistical Office of the European Communities (Eurostat) but was later expanded to include the International Monetary Fund (IMF) and the World Bank.

The Inter-Secretariat Working Group decided that in order to cover all areas of the SNA, a series of subject matter Expert Group meetings would be desirable. It further decided that the Expert Group would consist of twelve people. Six of these would represent core experts and be present

Statistics Canada - 43 - Collected Articles of Kishori Lal at all meetings; the other six would be subject specialists drawn in for each meeting in turn. Of the core experts, three would represent developed countries, and three developing countries. In view of the limited number of people who could participate in the Expert Group meetings, it was decided that the process should start by having an inter-regional meeting with representation exclusively from developing countries so that the core experts from these countries would be able to represent a wide range of views, it being made clear to all of the experts involved that they would attend the meetings in a personal capacity rather than as representatives of their government or institution. This inter-regional meeting, called the Inter-regional Seminar for Developing Countries on the SNA Review, was held in Geneva, in June 1986. At this meeting, the first major discussion point was whether the SNA could be equally applicable for both industrialized and developing countries. All participants very strongly endorsed the view that there should be a single system that would embrace the needs of the whole range of countries. This is in sharp contrast to the 1968 UN SNA, which carries a "full" version for developed countries and a "cut-down" version for developing countries. The cut-down version is elaborated in the 1968 UN SNA in chapter IX, Adaptation of the Full System to the Developing Countries.

Immediately following the Inter-Regional Seminar for Developing Countries, the first Expert Group meeting on the SNA Structure was held in Geneva, in June 1986. The next Expert Group meeting on Prices and Quantity Comparisons was held in Luxembourg, in November 1986. The meeting on External Sector was held in Washington, in March 1987. The Household Sector meeting was held in Florence, Italy, in August 1987. The Public Sector meeting was held in Washington, in January 1988. The Expert Group meeting on Production Accounts and Input- Output tables was held in Vienna, in March 1988. I attended this meeting as a subject matter specialist. The last Expert Group meeting on Financial Flows and Balances was held in Washington, in September 1988. Since then, several Coordinating Group meetings have been held to pull together the discussions of the previous meetings. The last meeting, the Inter- regional Seminar on the SNA Revisions, was held in Aguascalientes, Mexico, in October 1992. This was the largest meeting, with representation from all regions of the world. Again, I attended this meeting as a subject matter specialist. Readers interested in having a more detailed report on the SNA revision process may refer to "The SNA; 1968 to 1993 and Beyond", a paper submitted by Anne Harrison (OECD), at the 22nd General Conference of the International Association for Research in Income and Wealth, Flims, Switzerland, 1992.

The 1993 SNA was supposed to establish links and reconciliation with the Material Product System (MPS) used by the centrally planned economies of Central and Eastern Europe and elsewhere. Indeed, a meeting on Reconciliation of SNA and MPS was held in Moscow, in December 1989. With the collapse of central planning in the economies of Central and Eastern Europe, the need for reconciliation of the two Systems, SNA and MPS, disappeared as MPS became redundant. The ex-centrally planned economies have now accepted to use SNA. With this, the 1993 SNA has become the universal system of national accounts, applicable in full to all countries.

Major changes

The 1993 SNA has proposed several major changes vis-à-vis the 1968 UN SNA. These are detailed, chapter by chapter, in my two papers, "The 1993 International System of National Accounts vis-à-vis The Canadian System of National Accounts" and "The 1993 International System of National Accounts and the Canadian Input-Output Tables". These papers (available

Statistics Canada - 44 - Collected Articles of Kishori Lal upon request) examine the 1993 SNA through the experience of the Canadian System of National Accounts (CSNA). Our comments are organized therein to highlight certain important areas where the CSNA will need to revise its present practices to conform to the 1993 SNA. The impact on the CSNA is noted and an implementation strategy is summarized. The present paper selects only those issues which, in my judgement, are significant as to their usefulness, impact and implementability in developing countries, in the near future. Although the explicit focus of this paper is on developing countries, its contents, I believe, are also applicable to countries whose economies are in transition from central planning to a market system.

1. Production account for institutional sectors

The 1993 SNA differs from the 1968 UN SNA regarding a proposal to compile production accounts for the various institutional sectors, such as: the non-financial corporate sector; the financial corporate sector; the general government sector, the non-profit institutions serving households sector (NPISHs); and the household sector. According to the 1968 UN SNA, separate production accounts for the individual sectors are not required; only a consolidated production account for the nation as a whole is recommended. The 1993 SNA recommends that there should also be full production accounts for institutional sectors, full in the sense of gross output, intermediate expenses and value added (paragraphs 6.1-4).

We do not support this recommendation. Full production accounts are not necessary to relate with and analyse the sector income and outlay and finance accounts. Their additional utility is quite marginal, especially when the very significant resources that would be required to develop such estimates are considered, particularly in many developing countries. Even in a developed country such as Canada, support is lacking, and Canada is not alone in its opposition to this recommendation.

A far more useful alternative would be to develop a comprehensive production account for all the producing enterprises in three sectors: business (both corporate and unincorporated enterprises), general government, and non-profits institutions serving households. The only additional production generated in the strictly household sector is that of domestic and personal services produced by employing paid domestic staff. The own-account production of housing services by owner-occupiers is considered to belong to the unincorporated enterprise part of the business sector. Furthermore, all three producing sectors should be disaggregated, and classified by economic activity. The above alternative (embodied in the CSNA production account) is similar to the comprehensive enterprise sector account recommended by Professor Richard Ruggles in his recent paper, "The United Nations System of National Accounts (SNA): Its Implementation for Developing Countries" (Yale University, May 1992). He does not support the utilisation of a production account for institutional sectors. His preference, like ours, is to have a comprehensive production account. He writes: "A simpler solution to the institutional sectoring problem would be to have a comprehensive enterprise sector composed of corporations (both financial and non-financial), unincorporated enterprises, government enterprises and non- profit institutions. Such consolidation would reduce the articulation of transactions required between the major sectors of the economy. The total of this enterprise sector would be coincident with the total of the industrial sectors used for measuring production and input- output" (p. 5). There is no point of having delineated production sectors if their inter-sectoral transactions cannot be articulated. It is our judgement that the institutional sectors for production

Statistics Canada - 45 - Collected Articles of Kishori Lal accounts recommended in the 1993 SNA are not very useful for inter-sectoral articulation and, if such inter-sectoral articulation is attempted, it will be very costly.

2. The production boundary in the System

A vitally important issue is the demarcation of the boundary for valuing production for SNA purposes. The 1993 SNA lists the following activities (paragraph 6.18) that fall within the production boundary of the System:

Note that the own-account production of domestic and personal services by members of the household for their own final consumption is excluded from the production boundary (paragraphs 6.19-22). The production boundary in the 1993 SNA is slightly enlarged from the one in the 1968 SNA. It includes the production of all goods within the production boundary, and services in so far as they are exchanged in the market and/or generate incomes for other economic units. With regard to own-account production of goods by households, the 1993 SNA removes the 1968 SNA limitations which exclude the production of goods not made from primary products, the processing of primary products by those who do not produce them and the production of the goods by households who do not sell any part of them on the market. This extension of the boundary will, of course, increase the value of production for SNA purposes, and this extension will have a bigger impact in developing countries than in the industrialized economies, as in the latter, the role of own-account household production of goods is insignificant. Its implementation in developing countries will add a significant part to the GDP, a part which will, however, require strenuous efforts to calculate.

Included in the production boundary, however, is so-called illegal production. Illegal production was always implicitly contained within the boundary, and in the 1993 SNA this has been properly clarified. Illegal production (paragraph 6.30) comprises: "a) the production of goods and services whose sale, distribution or possession, is forbidden by law; b) production activities which are usually legal but which become illegal when carried out by unauthorized producers; e.g. unlicensed medical practitioners". Further: "Examples of activities which may be illegal but productive in an economic sense include the manufacture and distribution of narcotics, illegal transportation in the form of smuggling ... and services such as prostitution" (paragraph 6.32). Manufacture and distribution of illicit liquor in the Indian countryside may not be insignificant; its inclusion will raise the GDP and improve estimates of savings of the household sector. Estimating the value of this activity, however, will be very difficult. It has generally been assumed that concealed production and the underground (black market) economy form part of the production boundary. Again, this has been clarified (paragraphs 6.34-36) in the 1993 SNA. The underground economy forms a significant part of the overall GDP in many countries, and perhaps a dominant part in some developing countries. It needs to be estimated and included in

Statistics Canada - 46 - Collected Articles of Kishori Lal the country's GDP as, otherwise, the two sides of the accounts, the income side and the expenditure side, will not balance.

3. Financial intermediation services indirectly measured (FISIM)

Banks and other financial institutions provide a variety of services. Those that are specifically charged for include commissions, charges on cheques etc; and the corresponding revenues form part of the institutions' output. An additional, and very significant part of their income comes from charging higher interest rates to borrowers and paying lower rates to depositors than they would need to if they charged explicitly for all their services. This "hidden" charge (known as imputed banking services in the 1968 UN SNA) is called financial intermediation services indirectly measured (FISIM) in the 1993 SNA. Its value is measured as "...the total property income receivable by financial intermediaries minus their total interest payable, excluding the value of any property income receivable from the investment of their own funds..." (paragraph 6.125). Several countries, including Canada, have not been able to exclude property income from investment of own funds.

The recommendation for allocating intermediation services to all users in the 1993 SNA is quite different from the approach specified in the 1968 UN SNA. In the 1968 UN SNA, imputations for banking intermediation services are made; however, such services are allocated exclusively to the business sector rather than to all borrowers and lenders, as is now recommended in the 1993 SNA. Within the business sector, the 1968 UN SNA does not allocate these services to each industry; instead, it uses a dummy industry which buys the entire service as an intermediate expense and generates an equivalent negative value added. Because it did nothing to resolve the issue, this procedure was perhaps the single most important weakness of the 1968 UN SNA.

We support the recommendation of the 1993 SNA. Households, the general government and the rest of the world sector use the financial intermediation services just as market enterprises do. The statistical difficulties involved in estimating sectoral allocation should not discourage national accountants from following the right path. Implementation of this proposal will raise overall GDP. However, it is worth noting that the implementation of the 1993 SNA recommendation will not affect the level of savings in general government, households or the rest of the world sectors.

4. The boundary between intermediate consumption and gross fixed capital formation

Intermediate consumption measures the value of goods and services that are transformed or entirely used up in the course of production during the accounting period. It does not cover the use of fixed assets owned by the enterprise or expenditures on gross fixed capital formation. Whether certain outlays are treated as intermediate consumption or capital expenditures is an important consideration in setting prices. Market prices could be different, and most probably lower, for products whose intermediate expenses are capitalized or recouped over a longer period than one year. Unless there are compelling reasons to deviate from them, SNA conventions should follow the actual practices of businesses in the market economy. We now examine some important 1993 SNA recommendations:

Statistics Canada - 47 - Collected Articles of Kishori Lal a) Mineral exploration

In contrast to the 1968 UN SNA, the 1993 SNA recommends: "Expenditures on mineral exploration are not treated as intermediate consumption. Expenditures on mineral exploration, whether successful or not, are needed to acquire new reserves and are therefore all classified as gross fixed capital formation" (paragraph 6.166). We support this recommendation. Expenditures on mineral exploration are very significant in many developing countries. Implementation of this recommendation with raise GDP. b) Military equipment

The 1993 SNA states that: "The construction of buildings for use by military personnel, including hospitals and schools, and also of roads, bridges, airfields, docks, etc., for use by military establishments should be treated as fixed capital formation. In addition, machinery and equipment of the same type as used by civil establishments for non-military purposes should also be treated as fixed capital formation" (paragraph 6.171). Conversely, military hardware such as rockets, missiles and their warheads, warships, submarines, fighter aircraft and bombers should continue to be treated as intermediate consumption.

In the 1968 UN SNA, all military expenditures on goods were treated as intermediate consumption. The recommendation in the 1993 SNA is reasonable and we support its implementation. When it is implemented, GDP will increase only by the amount of consumption of fixed capital. It is worth noting that one would need details of military expenditures classified by products in order to separate the goods to be added to capital formation. The level of military expenditures is quite significant in many developing countries.

5. Valuation at basic prices

The 1993 SNA recommends that the preferred method of valuation of output of goods and services produced for the market be at basic prices (paragraph 6.218). Note that the concept of basic prices in the 1993 SNA is the same as that of approximate basic prices in the 1968 UN SNA. Valuation at basic prices is very useful as it provides the most homogenous valuation for all users. We strongly support its implementation. The 1993 SNA states that: "The basic price is the amount receivable by the producer from the purchaser for a unit of a good or service produced as output minus any tax payable and plus any subsidy receivable on that unit as a consequence of its production or sale. It excludes any transport charges invoiced separately by the producer" (paragraph 6.205a). Further: "When output is recorded at basic prices, any tax actually payable on the products is treated as if it were paid by the purchaser directly to the government instead of being an integral part of the price paid to the producer. Conversely, any subsidy on the product is treated as if it were received by the purchaser and not the producer" (paragraph 6.206). The above definition of the basic price can be further clarified with the help of an arithmetical example. Let us assume that a purchaser pays $103 for a product and that this value includes a tax of $5 and a subsidy of $2. The basic price is $100 in this case. Suppose that there is no tax payable on product; the purchasers' price is $98 but the basic price is still $100. In the developing countries and in the transition economies, where subsidies form a significant part of the value of a product, the most sensible valuation from the perspective of cost to society is at basic prices. See further discussion of subsidies, item 19, under Outstanding issues.

Statistics Canada - 48 - Collected Articles of Kishori Lal 6. Reinvested earnings on direct foreign investment

According to the 1993 SNA, both systems (SNA and balance of payments) "... require the saving or retained earnings of a foreign direct investment enterprise to be treated as if they were distributed and remitted to foreign direct investors in proportion to their ownership of the equity of the enterprise and then reinvested by them. In other words, two additional entries are required in the accounts of the enterprises and their foreign owners, one of which is the imputed remittance of retained earnings, while the other is the imputed reinvestment of those earnings" (paragraph 7.120). This is a departure from the 1968 UN SNA. There has been some controversy around this item. On the whole, it would be reasonable to implement the 1993 SNA recommendation, as this will harmonize the SNA with the IMF Balance of Payments Manual and permit international comparability.

In the case of retained earnings of foreign direct investment in any country, the implementation of this recommendation will increase current account payments of the balance of payments, counterbalanced by an increase of capital inflow in the capital account.

7. Social transfers in kind

An important proposed change from the 1968 UN SNA is the inclusion of social transfers in kind in the distribution of income. As the 1993 SNA states: "In the System, final consumption expenditure is incurred only by general government, NPISHs and households. All of households' consumption expenditure is incurred on their own behalf. Consumption expenditure by general government, on the other hand, is either for the benefit of the community at large (a collective consumption) or for the benefit of individual households. By convention, all consumption expenditure by NPISHs is treated as being for the benefit of individual households... Consumption expenditures by general government and NPISHs on behalf of households (their individual consumption expenditures) are undertaken for the purpose of making social transfers in kind. They cover the non-market output of both general government and NPISHs delivered to households free, or at prices that are not economically significant, as well as goods and services bought from market producers and provided to households free or at prices that are not economically significant" (paragraph 8.38). The role of social transfers in kind is far more important in economies in transition and in developing countries than it is in the developed market economies. The inclusion of social transfers in kind (important examples are education and health) will add identical value to income and consumption of households, without affecting their savings. We welcome this recommendation as its inclusion would help international comparability of household income and household consumption. Implementing this change will require however, a functional distribution of government expenditures.

8. Final consumption expenditure and actual final consumption

An important change from the 1968 UN SNA is the specification in the 1993 SNA of the interrelationship between final consumption expenditure and actual final consumption for the three sectors, general government, non-profit institutions serving households (NPISHs) and households in which final consumption takes place (paragraphs 9.93-99). This treatment is a counterpart of that of social transfers in kind in the distribution of income just noted above. On a practical level, it may be noted that each of the aggregates, whether referring to consumption

Statistics Canada - 49 - Collected Articles of Kishori Lal expenditure or actual final consumption, has to be derived from data on expenditures. Thus, final consumption expenditure consists of the following components:

1) Household final consumption expenditure; 2) Final consumption expenditure of NPISHs; 3) Government final consumption expenditure: 3.1 Government expenditures on individual consumption goods and services; 3.2 Government expenditures on collective consumption services. Actual final consumption of households is measured by the value of all the individual consumption goods or services acquired by resident households. There are three sets of goods and services entering into household actual final consumption:

The value of actual final consumption of general government is given by item (3.2) above. NPISHs have no actual final consumption.

It should be repeated that actual final consumption for the whole economy is exactly equal to final consumption expenditures.

This is an important recommendation from the point of view of international comparability of household consumption. Inclusion of social transfers in kind in consumption would help comparability between developed and developing countries. In centrally planned countries following the MPS System, such social transfers were always allocated to final consumption of households.

9. Gross capital formation

The 1993 SNA states: "Gross capital formation is measured by the total value of the gross fixed capital formation, changes in inventories and acquisitions less disposals of valuables" (paragraph 10.32). In the definition of gross capital formation, acquisitions less disposals of valuables is a brand new item compared with the 1968 UN SNA. As is later explained, "valuables consist of: a) precious stones and metals such as diamonds, non-monetary gold, platinum ... held by any units including enterprises provided that they are not intended to be used as intermediate inputs into processes of production; b) paintings, sculptures, etc. recognized as works of art and antiques; c) other valuables, such as jewellery fashioned out of precious stones, and metals and collections" (paragraph 10.116). This is an important additional specification and its impact will differ widely, from country to country. In the case of India, where gold and gold jewellery have captivated its people throughout the ages, this item assumes a very large proportion, difficult to estimate but unwise to ignore.

Statistics Canada - 50 - Collected Articles of Kishori Lal 10. Gross fixed capital formation

In most cases, the 1993 SNA definition of gross fixed capital formation corresponds to the common sense view. However, there are certain cases where the 1993 SNA is different from the 1968 UN SNA. Military equipment is one such case and it has already been discussed in item 4.b above. Others cases follow: a) Inventories of structures

According to the 1993 SNA: "When there is no contract of sale agreed in advance, the output produced by the construction enterprise must be recorded as work-in-progress or as additions to the producers' inventories of finished goods, depending upon whether the construction is completed. For example, finished dwellings built speculatively remain as additions to producers' inventories of finished goods until they are sold or otherwise acquired by users" (paragraph 10.75). Implementation of this approach will affect gross fixed capital formation counterbalanced by an identical change in inventories. This will require additional information which is not readily collected in most countries, developed or developing. b) Entertainment, literary or artistic originals

The 1993 SNA recommends treating original creations - such as original films, sound recordings, manuscripts, tapes, models, etc. - as capital formation (paragraph 10.94). Our data sources in Canada are quite weak in this area and we may not be able to implement the recommendation, although, in principle, we support it. We suspect that its implementation in most countries, developed or developing, will be problematic.

11. Harmonization with the balance of payments

An important feature of the 1993 SNA is the harmonization of its concepts and valuation rules with other international guidelines such as the IMF Balance of Payments Manual. The rest of the world account has fully succeeded in doing so, as is demonstrated in section F, Relationships between the current external transaction and accumulation accounts and the balance of payments accounts (paragraphs 14.147-155). One item which is not harmonized is financial intermediation services. In the 1993 SNA, "... financial intermediation services indirectly measured (FISIM) is included in the external account of goods and services, reflecting services that are not explicitly charged, but is not shown under imports and exports of services in the balance of payments accounts. However, it is included indistinguishably under "investment income, interest" in those accounts" (paragraph 14.149). As noted earlier, this lack of harmonization does not affect the current account balance in the balance of payments. It is very useful to harmonize the concepts in the two systems, SNA and balance of payments, as it would help economic analysis.

12. Gold

The 1993 SNA (paragraph 11.65 and paragraph 14.93) distinguishes three types of gold: monetary gold owned by monetary authorities as a component of international reserves; gold held as a store of value; and other gold used for industrial purposes. A transaction in the second or third category of gold is always treated as a transaction in a commodity, no matter who the transactor is. A transaction in monetary gold can only be conducted between the monetary

Statistics Canada - 51 - Collected Articles of Kishori Lal authorities of one country and the monetary authorities of another country or with international monetary authorities; such a transaction is always a financial transaction. All other transactions in gold are recognized as transactions in gold as a commodity. The impact of this approach will differ from country to country, depending upon the role of the banking sector (other than monetary authorities) in the bullion market. Following the 1993 SNA, such banking transactions will be classified in merchandise trade in the balance of payments, rather than in its capital account as financial transactions, thus affecting the current account balance.

13. Valuation of product flows

We fully agree with the 1993 SNA that "the preferred method of valuation is at basic prices" (paragraph 15.33) for product flows in the input-output tables. The preference for basic prices is further justified in the same paragraph as follows:

i) Basic prices provide the most homogenous valuation for all the users. ii) Basic prices are found most useful when a system of Value Added Tax (VAT) or similar deductible tax is in operation. iii) Basic prices record the amounts available to the producer.

"For exports and imports, the System adopts analogous price concepts: the free on board (f.o.b.) for exports and total imports and the cost insurance and freight (c.i.f.) price for detailed imports" (paragraph 15.35). Note that the f.o.b. price is at the customs frontier of the exporting country both for exports and imports. Further, the 1993 SNA states: "The f.o.b. price is considered to be a special purchasers' price applied to flows of exports. The c.i.f. price is considered to be a basic price applied to flows of imports, equivalent to the basic price of a domestically produced good or service" (paragraph 15.36). This is a change from the 1968 UN SNA where "the protective duties form part of the basic value of imports" (paragraph 2.16 of the 1968 UN SNA). Its impact will differ from country to country, depending upon the role of the rest of the world sector in its economy.

To remain consistent with the f.o.b. valuation of exports and imports of goods and services in the balance of payments statistics, several adjustments are needed to present these series at basic prices in, say, the input-output tables of the SNA. a) Exports

The f.o.b. price for exports, as noted above, is considered to be a special purchasers' price and includes basic price plus any net taxes (taxes less subsidies on products) and trade and transport margins from the producing establishment to the customs frontier of the exporting country. The trade and transport margins can be supplied both by resident and non-resident producers of such services. All components of f.o.b. price have to be fully delineated in product detail to balance the commodity flows in basic prices. The delineation of the value of exports at f.o.b. price in the balance of payments statistics with their value in the input-output tables at basic prices will require close attention, if non-resident transporters play a role in delivering exports from the producing establishment to the customs frontier of the exporting country. Such charges will have to be noted simultaneously as exports and imports of transport services in the input-output tables valued at basic prices, unless this has already been done in the export flows in the balance of payments statistics. Valuation at basic prices brings into sharp focus the need to impute an

Statistics Canada - 52 - Collected Articles of Kishori Lal equal amount for exports and imports of transport services provided by non-resident transporters of exports. b) Imports

The c.i.f. price for detailed imports includes f.o.b. price for imports (valued at the customs frontier of the exporting country) plus the cost of insurance and freight between the frontier of the exporting country and the frontier of the importing country. Again, these services can be provided by both resident and non-resident producers of such services. In the balance of payments statistics, the System adopts f.o.b. price for valuation. One will need to add insurance and freight charges provided by both non-resident producers and resident producers to bring this valuation at c.i.f. price into the input-output tables at basic prices. The part charged by the non- resident producers should be removed from the total for services, as it is part of the c.i.f. price of imports of goods, such that the two together (goods and services) remain the same. But the insurance and freight charged by resident producers of the importing country to bring imports of goods from the frontier of the exporting country to the frontier of the importing country need to be properly delineated. The implementation of the newly recommended c.i.f. valuation will raise the total level of imports of goods by the amount of insurance and freight supplied by the carriers of the importing country and, at the same time, it will require an imputation to raise the level of exports of services by an equal amount to balance the System. There will be no change in the net external balance in the exports and imports of goods and services, but the gross valuation will be higher at basic prices compared with the f.o.b. valuation in the balance of payments.

The 1993 SNA provides an alternative presentation which obviates the need to increase the gross flows of goods and services in the input-output tables at basic prices. Rather than imputing an addition to exports of insurance and freight services provided by resident carriers, the 1993 SNA recommends creating a new c.i.f./f.o.b. global adjustment on imports. The recording in the supply table of imports and the c.i.f./f.o.b. adjustment item are explained in the 1993 SNA (paragraphs 15.69) as follows:

i) Imports of goods detailed by products are valued c.i.f. ii) All transport and insurance services on imports, provided by both resident and non- resident producers, which are included in the c.i.f. value of imports by products, are globally deducted, thus the total of imports of goods in the System is always recorded f.o.b. in the table. iii) Those transport and insurance services on imports, provided by non-resident producers, are recorded under imports of services.

The above methods of recording plus a certain presentational device (see Table 15.1, Chapter XV) permit showing total imports at f.o.b., the same valuation as the one recorded in the balance of payments statistics.

Whether the provision of transport and insurance services to bring imports from the frontier of the exporting country to the frontier of the importing country by the resident producers is shown as an addition to exports of services or as a c.i.f./f.o.b. adjustment to imports, it requires the same detailed calculation. Problems associated with f.o.b. valuation of exports and imports will differ from country to country depending upon their trade patterns.

Statistics Canada - 53 - Collected Articles of Kishori Lal 14. Gross value added at factor cost

In the 1993 SNA, gross value added at factor cost "... is not recommended as a measure of value added in the System, since there are no observable prices such that output minus intermediate consumption equals gross value added directly ..." (paragraph 15.39). The measure recommended throughout the System is gross value added at basic prices, which is defined as "... output valued at basic prices less intermediate consumption valued at purchasers' prices" (paragraph 15.37). It is noted, however, that gross value added at factor cost "... could be derived from gross value added at basic prices by subtracting other taxes less subsidies on production" (paragraph 15.39). It is further asserted: "Other taxes less subsidies on production, by definition, are in fact taxes or subsidies that cannot be eliminated from the prices of outputs and inputs. Therefore, gross value added at factor cost is essentially a measure of income and not output" (paragraph 15.39).

We do not agree with this position. Information on other taxes less subsidies is no more difficult to obtain from industry records than information on intermediate consumption. Other taxes less subsidies cannot be eliminated from the prices of outputs and inputs but neither can compensation of employees, profits or capital consumption. Gross value added at factor cost has been and remains a standard measure of output for analysis of productivity by industry. We do not see any logic in calling gross value added at basic prices an output measure, and gross value added at factor cost an income measure. Both are output measures, as GDP at purchasers' prices is an output measure in the System. We at Statistics Canada have been producing and publishing GDP by industry at factor cost for decades at both annual and monthly frequencies. Our alternative proposal is to fully delineate the contribution of each industry to the aggregate gross domestic product at purchasers' prices in terms of three prices: factor cost, basic prices and purchasers' prices. The flexibility of presentation in the CSNA may be emulated by other national statistical organizations.

15. Cross-classification of value added by sectors and industries

The 1993 SNA recommends that value added by industry in the use tables be cross-classified by institutional sectors, such that income and consumption can readily be related (see paragraphs 15.106-110). This is a laudable recommendation but it is not implementable at most statistical organizations. It requires a well functioning business register which can readily relate establishments to enterprises and sectors. Such information needs to be current, so that one can relate the current profile of establishment-based input-output industry statistics to the institutional-based structure of the economy. It also demands a highly articulated micro database which can be readily added to provide aggregates for the SNA. These conditions are rigorous and very costly to implement. Given the recommendation's demand for resources and its limited usefulness to relate production with income and consumption, an alternative presentation is to by-pass the production account and the generation of income account by institutional sectors and to go direct to the primary allocation of income account by institutional sector. See also item 1, Production account for institutional sectors.

Statistics Canada - 54 - Collected Articles of Kishori Lal 16. Conversion of supply and use tables into symmetric input-output tables

Supply and tables are typically rectangular in the sense that the number of products (commodities) is greater than the number of industries. Symmetric tables are tables which have an identical number of rows and columns for either industries or products. There are important advantages to the rectangular format of the input and output tables over the symmetric tables. As noted by Anne Carter and Wassily Leontief in their article "Goals for the Input-Output Data System in the Seventies", published in the US Department of Commerce Survey of Current Business, July 1971, page 31: rectangular tables permit as much detail as is available in the basic census or survey records, and the meaning of each entry is straightforward because observed transactions are not combined with fictitious transfers (a feature of symmetric tables).

In the 1993 SNA, several procedures are listed (see paragraphs 15.137-143) to convert the rectangular supply and use tables into symmetric input-output tables. One procedure suggested for producing symmetric tables is transferring secondary products from the activities in which they are produced to the activities for which they are principal products, such that all "off- diagonal" entries in the supply table are eliminated. Transferring the corresponding inputs within the industries in the use table is properly noted as being more complicated, as the basic data on inputs relate to industries and not to particular products produced in those industries. To overcome this difficulty, it is suggested that supplementary information be used to make such a transfer. Failing the above approach of using supplementary information to generate symmetric tables, the 1993 SNA recommends using a mechanical approach by adding certain analytical assumptions (see paragraphs 15.144-149). It notes that the mathematical methods used when transferring outputs and associated inputs hinge on two types of technology assumptions:

i) industry (producer) technology, assuming that all products produced by an industry are produced with the same input structure; and ii) product (commodity) technology, assuming that a product has the same input structure in whichever industry it is produced.

The 1993 SNA opts for the product technology assumption. We have serious reservations about this recommendation. The product technology assumption may be valid if one can develop a vector of inputs for each of the twenty thousand or so products identified in the market. It is, however, completely unrealistic to seek to achieve such a data base. Aggregating twenty thousand products into a manageable set of 200-300, or even 500-1000, product groups can hardly be called a replication of the commodity technology. At this level of aggregation of products or industries, there is hardly any difference between the two sets. Furthermore, it is evident that at least some aspects of industry technology are more plausible than commodity technology. Such things as R&D, management skills and management style, which are industry specific or even firm specific, affect similarly all the products produced within an industry. The commodity technology assumption is partly a hangover from a perception that one needs the same physical things to produce a given product, no matter who produces it.

Another reason why we do not support the commodity technology assumption is that for model building purposes, the algebraic manipulation is easier using the industry technology assumption. The industry technology assumption does not require artificially transforming the basic rectangular supply and use tables into square tables. On the other hand, the commodity technology assumption can only work with a square table (see also the 1968 UN SNA paragraph

Statistics Canada - 55 - Collected Articles of Kishori Lal 3.84 where it is stated: "... The assumption of a commodity technology can only be used if the number of industries is equal to the number of commodities".

17. Double deflation

We agree with the 1993 SNA that "... the supply and use tables are the most complete consistent framework for constant price estimation and provide: a) interdependent measures of prices and volumes; b) an important check on the numerical consistency and reliability of the entire set of such measures, interlinking values at constant and current prices, value and volume indexes and deflators" (paragraph 15.161). "Constant price measures for gross value added are possible in the input-output framework by using the double deflation method, as the difference between: a) the value of output deflated by a price index of outputs; 2) the value of intermediate consumption deflated by a price index for these inputs" (paragraph 15.162).

The double deflation method in the context of inputs and outputs at basic prices is a very efficient technique. A similar recommendation was made earlier in the UN Manual on National Accounts at Constant Prices, Series M, No. 64, New York, 1979 (in short, UN Constant Price Manual). Let me further discuss this issue, using the Canadian practice. The Canadian practice conforms to the recommendation in the UN Constant Price Manual. The UN Constant Price Manual states: "In an ideal world real product by kind of activity would always be derived from an input-output table by double deflation" (p. 55). In Canada, the input-output tables form the core of the production accounts. Input-output tables in full detail are produced annually both in current and constant prices. Real output by industry is produced using the preferred double deflation approach.

In the double deflation approach, one deflates commodity outputs of an industry at basic prices, its intermediate uses of commodity inputs at basic prices as well as taxes on products and other taxes on production, net of subsidies. The difference between the deflated values of outputs and the total of commodity inputs and net taxes on products and production equals GDP at factor cost in constant prices. The double deflation approach satisfies the requirement of an identity between GDP income and expenditure based estimates in constant prices. However, there are certain important hazards in using double deflation for deriving the GDP of an industry whose value added represents a small proportion of total gross output.

For such an industry, GDP estimated by double deflation might be erratic, because small shifts in the relative prices of intermediate inputs and gross output could translate into big shifts in the resultant value added at constant prices. Here, the UN Constant Price Manual guidelines are not entirely satisfactory. They state: "The solution to this problem, however, may simply be to consolidate industries with very small ratios of value added to gross output with related industries at earlier or later stages of production. In other words, the problem of instability may be solved by aggregation into larger units whose values added are large enough in relation to gross output not to be too sensitive to the effects of changes in prices or technology" (p. 53). In one year, value added in one industry may be erratic but in the next, a different industry might suffer. Ad hoc aggregation into large units would disturb the continuity of time series. Thus, one needs additional guidelines. In the CSNA, we have solved this problem as follows: Values added are combined as suggested by the UN Constant Price Manual. The combined value added is redistributed using gross output or any other indicator as a proxy, but the combined value for a given sub-aggregate remains unchanged. Without this restriction, the above noted GDP identity

Statistics Canada - 56 - Collected Articles of Kishori Lal requirement will not be satisfied. These comments on the UN Constant Price Manual apply equally to the 1993 SNA recommendation on double deflation.

18. Price and volume measures

The 1993 SNA provides a framework within which an integrated set of price and volume measures can be compiled which are conceptually consistent and analytically useful. The 1968 UN SNA provides a summary discussion of price and volume comparisons, but it is limited to the production accounts. More detailed guidelines were issued in 1979 by the United Nations in the Manual on National Accounts at Constant Prices, Series M, No. 64, United Nations, New York, 1979. The 1993 SNA differs from both the 1968 UN SNA and the 1979 UN Constant Price Manual in its recommendation to use chain indices. a) Chain indices

It is self evident that the more remote the base year becomes, the less relevant are its prices for purposes of deflating the value of current flows of goods and services. When the base year is changed, there are two ways in which the series on either side of the new base year may be linked. The first method is to revalue not only the series for all years subsequent to the new base year at new prices, but also the series for all the years preceding the new base year in order to have an unbroken series expanding on either side of the new base year. The second method is to leave the data for all years up to and including the new base year unchanged and simply use the new base year prices for valuing all flows of goods and services from the new base year onwards. In order to develop an unbroken series expanding on either side of the new base, one would simply link the series through a chain index. We reject the first method in favour of the second on conceptual grounds. A new base year is required because the old one is not relevant anymore: thus one can not effectively argue for the first method. Both the 1968 UN SNA and the 1979 UN Constant Price Manual opt for the first method whereas the 1993 SNA recommends the use of chain indices for linking the series.

The 1993 SNA recommends: "In general, the constant price series should not be allowed to run for more than five, at the most, ten years without rebasing" (paragraph 16.76). It further recommends: "The preferred measure of year to year movements of GDP volume is a Fisher volume index, changes over longer periods being obtained by chaining: that is, by cumulating the year to year movements" (paragraph 16.73a). This is a very significant recommendation. Its implications for all countries, particularly developing countries, are very significant. There is a lot of debate in the media, in the legislatures, and in the scholarly journals about growth rates. A fair amount of this debate is meaningless because the growth rates measured with reference to fixed weights established a decade or so earlier are misleading. As the relative prices have become very volatile, fixed weight growth rates are not reliable. A five percent growth rate could easily be three or four percent when the base year is changed. Recently, Allan H. Young, Deputy Director of the Bureau of Economic Analysis, US Department of Commerce has published a paper, "Alternative Measures of Change in Real Output and Prices", Survey of Current Business, April 1992 in which he compares (constant price) average annual rate of change in the manufacturing sector of the United States, using three different fixed weight periods. The average annual rate of change for total manufacturing for the full decade 1977-87 was 4.7% using 1977 weights, 2.6% using 1982 weights, and a mere 1.6% using 1987 weights. These are horrendous differences for any analysis. As both the overall rate of inflation and the

Statistics Canada - 57 - Collected Articles of Kishori Lal relative price changes are much higher in most developing countries and economies in transition, than they are in the United States, the growth rate changes due to a shift in base years will be even more dramatic. To get a sensible reading of a growth rate, one must change the fixed weight to as current a period as possible, or preferably use chain indices. b) Additivity problem

When the series preceding the new base year is chained to the series succeeding the new base year, "the problem that emerges ... is that the constant price values for the components do not add up to the constant price values of the aggregates after the series have been linked" (paragraph 16.37).

Addressing the additivity problem, the 1993 SNA argues in favour of publishing linked, rebased data "... without adjustment leaving it to users to decide whether, or how, to deal with the resulting discrepancies" (paragraph 16.59). This is the approach followed in the Canadian system but with a modification. In the CSNA's fixed weight indices for final expenditure (rebased and linked every five years) additional "adjusting entry" series are shown for each published aggregate, calculated as the difference between the linked aggregate and the sum of its published, linked components. Our approach has two advantages: it alerts users to the problem by displaying explicitly the extent of non-additivity, and it reassures them that no elements are missing from the accounting system. c) Terms of trade

The 1993 SNA recommends the estimating of trading gains and losses due to changes in the terms of trade for calculating real gross domestic income (GDI). The 1993 SNA states: "... an improvement in the terms of trade makes it possible for an increased volume of goods and services to be purchased by residents out of the income generated by a given level of domestic production. Real GDI measures the purchasing power of the total incomes generated by domestic production so that when the terms of trade change there may be significant divergence between the movements of GDP at constant prices and real GDI. The difference between the change in GDP at constant prices and GDI is generally described as the trading gain (or loss). The differences between movements in GDP at constant prices and real GDI are not always small. If imports and exports are large relative to GDP, and if the commodity composition of the goods and services which make up imports and exports are very different, the scope for potential trading gains and losses may be large" (paragraph 16.152). The recommendation to calculate real GDI is not currently implemented in many countries, including Canada. Its implementation will be very useful for all countries, particularly developing countries where the differences in the commodity composition of exports and imports are typically large. It will provide very useful additional analytical constructs.

Outstanding issues

19. Subsidies

Statistics Canada - 58 - Collected Articles of Kishori Lal Consequently, the 1968 UN SNA treatment is retained. Given that the 1993 SNA is applicable to every country in the world (unlike the 1968 UN SNA which is applicable to market oriented economies), and given the overwhelming importance of subsidies in developing countries and even more so in Eastern European countries and ex-Soviet Republics, and the highly visible and significant subsidies on products such as agriculture in several developed countries, the retention of the 1968 UN SNA treatment is highly regrettable. Subsidies were not a major issue when the 1968 UN SNA was drafted but, today, its present treatment distorts our understanding of market valuation of agricultural products, rental on housing, food, transportation charges and many other consumer products in many countries.

When a government subsidizes a product, its market price is reduced by the subsidy. The government, in principle, could provide the same support to consumers of the product by giving them "coupons" exactly equal in value to the subsidy, but this fiscal rearrangement increases GDP at market prices though the cost to the government has not changed. Let us take another example where a government establishes a unit, say, a broadcasting corporation, which covers only 30% of its total cost through sale of advertisements and the other 70% is covered by the government from general revenues. Let us say that the total cost is $100. This unit is part of general government and its total cost of $100 forms part of final demand and, hence, of GDP at market prices. If this unit is reclassified as a government business enterprise and the government still covers 70% of its cost through subsidies, the valuation of GDP at market prices drops from $100 to $30; again in this case; the net cost to the government has not changed.

The underlying economic activity in the above examples has remained the same, and the total cost of providing service has not changed, yet the valuation of GDP at market prices has changed. This is a troublesome situation. A similar discomfort with the present treatment of subsidies has been expressed by several national accountants of both developing countries and transition economies.

In the absence of a full reopening of the issue of subsidies, a pragmatic solution would be to treat some of the highly subsidised products as non-market products, thus mitigating the impact on GDP at market prices. This can be accommodated with some minor modifications to the present guidelines of the 1993 SNA. In the spirit of flexibility in its implementation, the 1993 SNA might even encourage those countries, where subsidies cover a significant part of the overall cost of production, to treat such subsidies as a purchase of goods and services rather than as transfers from government to business toward current cost of production. Then, GDP at market prices would not change due to either a reclassification of units from the government sector to the business sector or vice versa, or a fiscal rearrangement to support a program.

20. Valuation of consumption of building services produced on own account by government units

It is regrettable that the output of building services produced on own account by government units has remained outside the production boundary of the 1993 SNA. There is hardly any conceptual difference between the building services produced on own account by government units and the own account production of housing services by owner occupiers. The latter is included in the 1993 SNA production boundary while the former is excluded. Its exclusion does not seem logical.

Statistics Canada - 59 - Collected Articles of Kishori Lal The phenomenon of government owned and occupied buildings is far more significant in the developing and ex-centrally planned economies than in the developed industrialised world. Given the increasing acceptance and implementation of liberal economic reforms, in both the developing and ex-centrally planned economies, the role of government owned and occupied buildings is expected to decline. In the absence of the inclusion of building services produced on own account by government units, both inter-temporal and international comparison of the performance of the economies will become suspect.

In the spirit of flexibility in the implementation of the 1993 SNA, both developing and ex- centrally planned economies may utilize the following approach in estimating the value of output of building services produced on own account by government units and include it in their production accounts.

Several components of the costs of producing rental services are already included in government expenditures. These consist of wages and other costs incurred in maintaining, repairing and cleaning buildings, capital consumption on buildings, and property or other taxes paid etc. A most important item, not included at present, is notional interest charges on the total value of a building. We believe that it should be dealt with along the following lines. Rather than creating a separate vector of actual and notional expenditures that are equivalent to rental value, our proposal is to simply add to the government expenditures two notional items: capital consumption on buildings, and interest charges on the total value of buildings. (These interest charges are for productive purposes and are analogues to those in the rental industry). These items will form part of the gross operating surplus of the government sector. Capital consumption on government buildings is already imputed in the 1993 SNA. Adding these items to the actual costs incurred in maintaining, repairing, cleaning, etc. and property taxes would cover most of the cost elements of rentals in the market. This alterative pragmatic proposal does not require a direct estimation of imputed rental, yet the result comes close to it. The only additional statistical requirement of this proposal is information on long term bond rates or mortgage rates, which is readily available.

Statistics Canada - 60 - Collected Articles of Kishori Lal Concluding remarks

The 1968 UN SNA did not develop guidelines for the national and sectoral Balance Sheet and Reconciliation Accounts. Such guidelines were developed later and published as Provisional International Guidelines on the National and Sectoral Balance Sheets and Reconciliation Account of the System of National Accounts, Series M, No. 60, United Nations, New York, 1977. The 1993 SNA incorporates and integrates the balance sheets in the SNA. There are four chapters (chapter X, The Capital account; chapter XI, The Financial account; chapter XII, Other changes in assets account; and chapter XIII, The Balance sheet) devoted to the accumulation accounts in the 1993 SNA. The last three chapters of the 1993 SNA - Chapter XIX, Application of the integrated framework to various circumstances and needs; Chapter XX, Social accounting matrices; and Chapter XXI, Satellite analysis and accounts - demonstrate by way of examples how a country may adapt the central framework, by using it in a flexible way, to meet special types of analysis. Countries are encouraged to develop satellite accounts which would incorporate, for example, the valuation of household work in the economic estimates calculated strictly in the market economy. This enlarged measure would permit a better analysis of the real growth of the economy. Another example would be to develop a satellite account to measure the role of R&D (Research and Development) in any sector of the economy. An environment satellite account would be another important development to understand the underlying sustainable economic growth of an economy. It is my judgment that the issues discussed in this paper will assume a higher priority in most developing countries in their implementation strategy.

The 1993 SNA is a vast improvement over the 1968 UN SNA in terms of clarification, simplification of concepts and harmonization with other international guidelines. Some recommendations affect the level of GDP, some affect the balance of international payments, while others simply provide a thoughtful articulation of certain important measures. The new measure of actual final consumption, a measure which has already been used in the international purchasing power parities project and in the Material Product System (MPS) used by the centrally planned economies will help international comparability of household consumption. The 1993 SNA has harmonized its treatment of reinvested earnings of direct foreign investment with that in the IMF Balance of Payments Manual. It has recommended that financial intermediation services be allocated to all users of such services, and that military structures and equipment of the same type as those used by civil establishments be treated as gross fixed capital formation. Its recommendation to produce production accounts for institutional sectors may not prove very useful, while its preference for chain indices for measuring year to year constant price growth rates is a great step forward. The treatment of subsidies and the valuation of consumption of building services produced on own account by government units remains troubling, particularly from the perspective of developing countries and economies in transition.

Statistics Canada - 61 - Collected Articles of Kishori Lal Note:

I am grateful to Abe Tarasofsky for his very useful comments and suggestions in the preparation of this paper.

An earlier draft of this paper was published in Policy Perspectives in Indian Economic Development (essays in honour of Professor G.S. Bhalla), edited by G.S. Cathay, Had-Anand publications, New Delhi, 1994. In that version, all references were made to the revised SNA draft, distributed under the auspices of the United Nations Secretariat (Provisional ST/ESA/STAT/Ser. F/2/Rev 4, various dates from March to September 1992). The present version includes references to the published 1993 SNA. The earlier version of my paper included an item 3, valuation of the consumption of building services produced on own account by government units. The published 1993 SNA has dropped this recommendation. Hence, this item has been rewritten under Outstanding issues in the present version of my paper. Apart from this and minor editorial changes, the two versions of my paper remain substantially the same.

Statistics Canada - 62 - Collected Articles of Kishori Lal The 1993 International System of National Accounts: Preliminary Views on Its Implementation in Canada

by Kishori Lal Director General System of National Accounts Statistics Canada

The International System of National Accounts 1993 (1993 SNA) was prepared and published under the auspices of the Inter-Secretariat Working Group on National Accounts, consisting of the Statistical Office of the European Communities (Eurostat), International Monetary Fund (IMF), the Organization of Economic Cooperation and Development (OECD), the Statistical Division (UNSTAT) and regional commissions of the United Nations Secretariat, and the World Bank. Its claim as a document for universal implementation derives from the fact that its adoption was unanimously recommended to the United Nations Economic and Social Council by its Statistical Commission at its twenty-seventh session, held in New York from February 22 to March 3, 1993. The plan for its implementation, however, does not seem to be as well organised as was its production.

We have made very detailed comments on this document in our two papers, The 1993 International System of National Accounts vis-à-vis The Canadian System of National Accounts and The 1993 International System of National Accounts and the Canadian Input-Output Tables. The present paper highlights, in a summary fashion, certain important areas where the Canadian System of National Accounts (CSNA) will need to revise its practices to conform to the 1993 SNA; the reader is encouraged to refer to these two papers for further details. There are some areas where we may not be able to implement the recommendations of the 1993 SNA; in such cases, our reasons are stated and alternatives are suggested. Our occasional departures from the recommended guidelines are primarily prompted by pragmatic considerations. We fully recognize the importance of promoting international comparability, but it should also be recognized that the specific circumstances existing at a given time in different countries can vary, often substantially. We believe that it would prove useful, from the standpoint of encouraging wide implementation of the 1993 SNA, if countries share their own adaptations of the recommended guidelines with the larger international community. Our adaptations are presented in this spirit.

We plan to implement the 1993 SNA at the time of the forthcoming rebasing of the CSNA to the year 1992 from the present base of 1986. The results of this exercise will be released in the summer of 1997. Our comments, below, follow the sequence of chapters in the 1993 SNA.

1. Production Account for Institutional Sectors

The 1993 SNA differs from the 1968 UN SNA regarding a proposal to compile production accounts for establishments and industries as well as for institutional units and sectors. In the 1968 UN SNA, separate production accounts for establishments and industries are required but not for individual institutional sectors. The 1993 SNA recommends that there should be full

Statistics Canada - 63 - Collected Articles of Kishori Lal production accounts, full in the sense of intermediate expenses, value added and gross output (paragraphs 6.1-4) for individual institutional sectors. Considering the very significant resources that would be required to develop such estimates for the current period and their limited usefulness, we do not plan to implement the recommendation.

Given our data sources, our alternative implementation plan has the following elements:

i) Develop a comprehensive production account for all the producing units in three sectors- business (both corporate and unincorporated enterprises), general government, and non-profit institutions serving households (NPISHs). ii) Disaggregate all three producing sectors, classified by economic activity in terms of establishment-based classification. iii) Produce full production accounts for the period for which input-output tables are prepared and construct production accounts limited to value-added for the current period.

Our current practices will require some minor changes in order to implement the above alternative in 1997.

2. Financial Intermediation Services Indirectly Measured (FISIM)

The recommendation for allocating FISIM to users in the 1993 SNA is quite different from the approach specified in the 1968 UN SNA. In the 1968 UN SNA, imputations for banking intermediation services are made; however, such services are allocated exclusively to the business sector rather than to all borrowers and lenders, as is now recommended in the 1993 SNA ( paragraphs 6.124-126). Within the business sector, the 1968 UN SNA does not allocate these services to each industry; instead, it uses a dummy industry which buys the entire service as an intermediate expense and generates an equivalent negative value added.

The CSNA present practice on the estimation and allocation of financial intermediation services to all sectors is similar to the 1993 SNA. Some minor changes will be required in the CSNA to fully implement the 1993 SNA. The reader is encouraged to refer to our paper, Financial Intermediation Services Indirectly Measured (FISIM) and the Canadian System of National Accounts, for further details.

3. Insurance

The 1993 SNA recommends to include income from investment of insurance technical reserves in the valuation of output of insurance services (paragraphs 6.135-140). In the CSNA, such income is ignored in the calculation of output. We support the 1993 recommendation and plan to implement it.

4. Operating Leasing versus Financial Leasing

The 1993 SNA states that operating leasing and financial leasing are totally different kinds of activity, the first being a process of production while the second is a method by which funds are channelled from a lender to a borrower ( paragraph 6.119 ). It further states in the same paragraph that "It is therefore essential to distinguish between the two types of leasing, even

Statistics Canada - 64 - Collected Articles of Kishori Lal though financial arrangements may be devised which are hybrids of the two and which are consequently difficult to classify". This recommendation is a departure from the 1968 UN SNA which does not recognize financial leases and, as a consequence, treats them in the same manner as operating leases.

We support the position taken by the 1993 SNA and will try to implement it, though fully recognizing that in practice it may be very difficult to distinguish the two leasing arrangements. In the CSNA, the present practice, with rare exceptions, is to treat all leases as operating leases. Our plan is to give higher priority to develop estimates for those leases which affect the rest of the world account than the ones which involve the transactors within the domestic economy of Canada. An example of such an arrangement with the rest of the world would be the leasing of aircraft by some Canadian airline company from abroad. Leasing arrangements within the domestic economy are more difficult to distinguish partly because of the evolving hybrid nature of leasing and the very large number of lessees, many having contracts of small values. Typically, professionals such as doctors and lawyers would lease cars rather purchase them.

5. Mineral Exploration

6. Military Equipment

The 1993 SNA states that: "The construction of buildings for use by military personnel, including hospitals and schools, and also of roads, bridges, airfields, docks, etc., for use by military establishments should be treated as fixed capital formation. In addition, machinery and equipment of the same type as used by civil establishments for non-military purposes should also be treated as fixed capital formation" (paragraph 6.171). Conversely, military hardware such as rockets, missiles and their warheads, warships, submarines, fighter aircraft and bombers should continue to be treated as intermediate expenditures.

In the 1968 UN SNA and the CSNA, all military expenditures on goods are treated as intermediate expenditures. We support the recommendation in the 1993 SNA and plan to implement it in the CSNA.

7. Consumption of Fixed Capital, Major Earthquakes, Hurricanes, etc.

The 1993 SNA states that: ".... losses due to war, or to major natural disasters which occur very infrequently - major earthquakes, volcanic eruptions, tidal waves, exceptionally severe hurricanes etc. - are not included under consumption of fixed capital. There is no reason for such losses to be charged in the production account as cost of production. The values of the assets lost in these ways are recorded in the other changes in volume of assets accounts." (paragraph 6.187). We completely agree with the 1993 SNA that the destruction of property due to hurricanes or major earthquakes should have no impact on consumption of fixed capital. We plan to implement this recommendation.

Statistics Canada - 65 - Collected Articles of Kishori Lal

8. General Valuation Principles

The 1993 SNA states that the preferred method of valuation of output of goods and services produced for the market is at basic prices, especially when a system of VAT, or similar deductible tax, is in operation (paragraph 6.218). The concept of basic prices in the 1993 SNA is the same as that of approximate basic prices in the 1968 UN SNA. The CSNA Input-Output Accounts already use a basis of valuation that is close to this one. The required change with respect to the treatment of subsidies will be made to fully implement the 1993 SNA recommendation in 1997.

9. Payroll Taxes

Payroll taxes are taxes payable by the employer on the wage or salary bill. Such taxes are quite commonly applied by the various governments in Canada. The CSNA mostly treats such taxes as part of compensation of employees whereas the 1993 SNA recommends such taxes to be "treated as taxes on production in the same way as taxes on buildings, land or other assets used in production" (paragraph 7.21). The implementation of this recommendation will not change GDP at market prices, but it will reduce GDP at factor cost counterbalanced by an increase of taxes on production, and it will also reduce personal income. We plan to implement it in 1997.

10. Reinvested Earnings on Direct Foreign Investment

According to the 1993 SNA, both systems (SNA and Balance of Payments) "... require the saving or retained earnings of a foreign direct investment enterprise to be treated as if they were distributed and remitted to foreign direct investors in proportion to the ownership of the equity of the enterprise and then reinvested by them. In other words, two additional entries are required in the accounts of the enterprises and their foreign owners, one of which is the imputed remittances of retained earnings while the second is the imputed reinvestment of those earnings" (paragraph 7.120).

Starting with the first quarter of 1994, reinvested earnings have been incorporated in the Canadian Balance of Payments in both the current and capital accounts. The rest of the CSNA, particularly national income and expenditure accounts and financial flows accounts, has not incorporated the new treatment but, instead, provides reconciliation accounts to allow users to go from one approach to the other.

11. Final Consumption Expenditure and Actual Final Consumption

An important departure, both from the CSNA and the 1968 UN SNA, is the specification in the 1993 SNA of the interrelationship between final consumption expenditure and actual final consumption for the three sectors (general government, NPISHs and households) in which final consumption takes place (paragraphs 9.93-99). On a practical level, it may be noted that each of the aggregates, whether referring to consumption expenditure or actual final consumption, has to be derived from data on expenditures. It should be emphasized that actual final consumption for the whole economy is exactly equal to final consumption expenditures.

Statistics Canada - 66 - Collected Articles of Kishori Lal We welcome this recommendation and plan to implement it, as it would help international comparability of household consumption.

12. Inventories of Structures

The 1993 SNA states: " When there is no contract of sale agreed in advance, the output produced by the construction enterprise must be recorded as work-in-progress or as additions to the producers' inventories of finished goods, depending upon whether the construction is completed" ( paragraph 10.75). In the CSNA, all structures, completed or unfinished, with or without contact of sale, are classified as fixed capital formation. The implementation of the 1993 SNA recommendation will affect gross fixed capital formation counterbalanced by an identical change in inventories. This will require additional information which is not collected at present. The only effect on GDP will be in terms of timing in the allocation of consumption of fixed capital. Given that there is minimal effect on measured GDP, and that there would be additional cost to collect information, we do not give high priority to the implementation of this recommendation.

13. Plantations, Orchards, etc.

The 1993 SNA recommends (repeating the recommendation of the 1968 UN SNA) that trees cultivated in plantations for the products they yield year after year - such as fruit trees, vines, rubber trees, palm trees, etc. - be treated as fixed assets. The value of such fixed assets may be approximated, if necessary, by the costs incurred in their production during the period (paragraph 10.88). The CSNA has not followed this procedure. Instead, it treats these costs as intermediate expenditure.

We support the 1993 SNA recommendation, in principle, as it would better reflect economic reality and promote better international comparability. However, given its minimal effect on measured GDP and the additional cost to collect information, we do not give high priority to its implementation.

14. Entertainment, Literary or Artistic Originals

The 1993 SNA recommends treating original creations - such as films, sound recordings, manuscripts, renderings, literary and artistic works, etc. - as capital formation (paragraph 10.94). The CSNA does not capitalize the value of such originals. In some cases they are added to inventories, and in others they are treated as intermediate consumption. Our data sources are quite weak in this area. Given its minimal effect on measured GDP and the additional cost to collect information, we do not give high priority to implementing the recommendation, although, in principle, we support it.

15. Acquisitions Less Disposals of Valuables

The 1993 SNA recommends that acquisitions less disposals of valuables be reflected in the Capital Account. "Valuables are assets that are not used primarily for production or consumption, that do not deteriorate over time under normal conditions and that are acquired and held primarily as stores of value" (paragraph 10.116). We support this recommendation in principle and will try to implement it. Due to data problems, however, the CSNA may not be

Statistics Canada - 67 - Collected Articles of Kishori Lal able to specify such expenditures in the Capital Finance Account; instead, they may continue to form part of the residual balancing item, net lending or borrowing of the sectors.

16. Acquisitions Less Disposals of Non-produced Non-financial Assets

The 1993 SNA recommends that acquisitions less disposals of non-produced non-financial assets be reported in the Capital Account. These assets consist of land, sub-soil assets that may be used in the production of goods and services and intangible assets such as patented entities, leases, other transferable contracts, etc. (paragraphs 10.120-130). We also support this recommendation in principle and will try to implement it. Here, too, due to data problems, the CSNA may not be able to specify such expenditures in the Capital Finance Account; instead they may continue to form part of the residual balancing item, net lending or borrowing of the sectors. It is worth reporting that at present, the CSNA is developing estimates of some sub-soil and forestry assets as part of its natural resource accounting work and its environment satellite account.

17. The Reconciliation Statement in the Balance Sheets

The 1993 SNA states that: "This chapter is concerned with the recording of changes in assets, liabilities, and net worth between opening and closing balance sheets that result from other flows -that is, flows that are not transactions, the transactions being recorded in the capital account and financial account of the System" (paragraph 12.1). There are two kinds of changes: "The first kind consists of changes that are due to factors such as discoveries or depletion of sub-soil resources, destruction by war or other political events, or destruction by natural catastrophes, all of which actually change the volume of assets. The second kind consists of changes in assets, liabilities, and net worth due to changes in the level and structure of prices, which are reflected in holding gains and losses" (paragraph 12.2). The changes of the first kind are further elaborated in the Other Changes in the Volume of Assets Account and those of the second kind, in the Revaluation Account.

Many of the items discussed in the 1993 SNA, in the context of the Other Changes in Assets Account, form, in principle, part of the Reconciliation Statement (not yet fully developed) in the CSNA Balance Sheets. The development and compilation of this account is of high priority in the CSNA as this will help specify several items which still remain consolidated or unexplained, thus permitting better analysis of balance sheets.

18. Progress Payments, and exports and imports

Progress payments are treated as financial transactions even if there is a contract of sale. Such goods will be valued in exports and imports only when they are "physically" in the hands of the owners ( paragraphs 14.57-64). At present, progress payments for certain machinery and equipment purchased by the government are treated as fixed capital formation in the CSNA and as exports and imports in the Canadian Balance of Payments. We plan to implement the 1993 SNA recommendation at the time of the rebasing of the CSNA.

Processing is another case where the present Canadian Balance of Payments statistics will need to be changed. The 1993 SNA recommends that the goods which are sent for processing abroad should be treated on a gross basis in exports and imports (paragraphs 14.61-64).To the extent that it can be identified, processing, at present, is treated as a service in the Canadian

Statistics Canada - 68 - Collected Articles of Kishori Lal Balance of Payments. This treatment will be changed to bring it in line with the 1993 SNA guidelines.

19. Gold

The 1993 SNA , in the Rest of the World Account , differentiates the transactions in monetary gold conducted between the monetary authorities of the different countries from other transactions in gold as a commodity. It recommends that the transactions in monetary gold be recognized as financial items and not included in the imports or exports of goods (paragraph 14.93). The 1993 SNA distinguishes three types of gold: (1) monetary gold owned by monetary authorities as a component of international reserves; (2) gold held as a store of value; and (3) other gold used for industrial purposes. A transaction in the second and third category of gold is always treated as a transaction in a commodity, no matter who the transactor is. A transaction in monetary gold can only be conducted between the monetary authorities of one country and the monetary authorities of another country or international monetary authorities; such a transaction is always a financial transaction. All other transactions in gold are recognized as transactions in gold as a commodity.

We support this recommendation and plan its implementation in the CSNA.

20. Cross-Classification of Value Added by Sectors and Industries

21. Conversion of the Supply and Use Tables into Symmetric Tables

The 1993 SNA opts for the product (commodity) technology assumption to convert the rectangular supply and use tables into symmetric tables (see paragraphs 15.137-143). The product technology assumption may be valid if one can develop a vector of inputs for each of the twenty thousand or so products identified in the market. It is, however, completely unrealistic to seek to achieve such a database. Aggregating twenty thousand products into a manageable set of 200-300, or even 500-1000, product groups can hardly be called a replication of the commodity technology. At this level of aggregation of products or industries, there is hardly any difference between the two sets. Furthermore, it is evident that at least some important expenditures such as R&D, as well as management skills and management style, which are industry specific or even firm specific, affect importantly and similarly all the products produced within an industry.

Statistics Canada - 69 - Collected Articles of Kishori Lal Our alternative presentation for symmetric tables is to use an industry technology assumption. For model building purposes, the algebraic manipulation is easier using this assumption, which does not require artificially transforming the basic rectangular Supply and Use tables into square tables. On the other hand, the commodity technology assumption can only work with a square table. (See also the 1968 UN SNA paragraph 3.84 : "... The assumption of a commodity technology can only be used if the number of industries is equal to the number of commodities").

22. Chain Indices

The 1993 SNA recommends: "In general, the constant price series should not be allowed to run for more than five, at most ten years, without rebasing" (paragraph 16.76). It further recommends: "The preferred measure of year to year movements of GDP at constant prices is a Fisher Volume index, changes over longer periods being obtained by chaining: i.e., by cumulating the year to year movements" (paragraph 16.73a). The CSNA already provides chain indices, one chain is typically linked every five years ( the present weights refer to 1986) and two sets of quarterly chain indices are published, one rebased and linked annually and the other linked quarterly.

23. Terms of Trade

The 1993 SNA recommends estimating trading gains and losses from changes in the terms of trade in calculating real Gross Domestic Income (GDI). The 1993 SNA states: "... an improvement in the terms of trade makes it possible for an increased volume of goods and services to be purchased by residents out of the income generated by a given level of domestic production. Real GDI measures the purchasing power of the total incomes generated by domestic production, so that when the terms of trade change there may be a significant divergence between the movements of GDP at constant prices and real GDI. The difference between the change in GDP at constant prices and real GDI is generally described as the trading gain (or loss). The differences between movements in GDP at constant prices and real GDI are not always small. If imports and exports are large relative to GDP, and if the commodity composition of the goods and services which make up imports and exports are very different, the scope for potential trading gains and losses may be large" (paragraph 16.152). The recommendation to calculate real GDI is not currently implemented in many countries, including Canada. We plan its implementation, as it will provide a very useful additional analytical construct.

24. Satellite Accounts

The 1993 SNA encourages countries to develop satellite accounts in a more flexible manner than the mainstream SNA accounts. The 1993 SNA provides examples of satellite accounts, some which are within the economic boundary of the SNA and others which extend the boundary. The tourism satellite account is an example of the first kind and the valuation of household work and environment accounts are examples of the second kind. At Statistics Canada, both kinds are under development.

Statistics Canada - 70 - Collected Articles of Kishori Lal Concluding Remarks

Our several adaptations of the recommended guidelines and our proposed alternatives are primarily prompted by pragmatic considerations. Apart from satellite accounts, our implementation approach hardly requires significant development of new data sources based on economic surveys or censuses. Most of the work is of a developmental/ analytical nature. By and large, the 1993 SNA recommendations are implementable in the CSNA and, once implemented, will provide a more coherent, integrated analytical framework and, and also greater international comparability.

Note

Excepting minor editing, this is a reproduction of the paper published in Statistics Canada=s National Income and Expenditure Accounts, Quarterly Estimates, Third Quarter, 1994 (Catalogue No. 13-001).

Statistics Canada - 71 - Collected Articles of Kishori Lal

Gross Domestic Product a

by Kishori Lal Statistics Canada

Gross Domestic Product (GDP) is a concept which occupies a central place in Macro-Economics in general and in the System of National Accounts in particular. Just about every country publishes its GDP estimates. These estimates are regularly reported, even highlighted by the media. The United Nations puts together such estimates from all (almost all) countries and publishes them annually. Contributions to and foreign aid from International Organizations are partly determined with reference to an individual country's GDP. If a country's GDP falls, the pride of its citizens is hurt, its government may fall, its international credit rating may suffer. A general reader would expect a concept so central in Economics and so widely used nationally and internationally to be a precise measure. But it is not.

GDP can be defined as a measure of an unduplicated value of current production of an economy, with all duplications such as intermediate expenses, as well as inter-sectoral and intra-sectoral transfers eliminated. An example of such a transfer could be a payment of social security benefits to persons by the Government but this does not add anything to current production. And there is no unique statistical definition of duplication except with reference to time, institutions, political set-up and (in practice) statistical development. In addition to these constraints, it is impossible to get a direct observation of each and every one of the myriad transactions made daily in any economy. The feasible numerical description resulting from collection, estimation, and imputation of all the transactions in an economy must remain somewhat abstract.

As noted, it is difficult to record all the myriad details of daily economic transactions. But even if one would be able to record them, the record will be bewildering. It is imperative and fundamental that the myriad transactions and the transactors be grouped into some "significant" categories so that one could see a "meaningful picture". Note that significant categories for one scholar or for one time or for one economy could be different from another scholar or another time or a different economy. Only the naive would accept at face value the regularly published international and intertemporal comparisons of GDP. This note gives only a flavour of some of the issues relevant to the subject of gross domestic product, particularly how to interpret it. Let me briefly go over the history of the recent methodological developments so that the present day measurement problems are put in a larger perspective.

Recent Methodological Developments

The Statistical Office of the United nations has been very active in developing and presenting a set of guidelines for compiling economic accounts. In 1953, the United Nations originally published A System of National Accounts and Supporting Tables(or SNA for short). This was the first time that an international set of guidelines was published on the System of National Accounts. But we must recognize that the UN 1953 document was not produced in a vacuum. There was a long intellectual

Statistics Canada - 75 - Collected Articles of Kishori Lal history behind it. Imagine that the first set of national income estimate was prepared for England by Sir William Petty in 1665. Many other scholars in other countries followed.1

The UN 1953 system was applicable (with modifications by individual countries) to market oriented economies such as USA, Britain, Canada, India, etc. This system was not used by the centrally planned economies such as the Soviet Union, China, etc. During the next fifteen years, the original 1953 system was extensively revised and a new version of the SNA was published in 1968.2 This version was completed under the able guidance of Professor Richard Stone of Cambridge University, England. Professor Stone has been very active in the theoretical and statistical developments in the general area of economic accounts, right from 1939 in England. In that year, the Chancellor of the Exchequer in Great Britain, under the influence of Keynes, authorized the preparation of national income estimates. Richard Stone and James Meade, with encouragement and detailed advice from Keynes, completed a set of national income and expenditure estimates for Great Britain which were published with the 1941 Budget.

Methodologically, the British estimates were significant in that national product was grossed to include depreciation allowances and net indirect business taxes, thus representing final output at market prices. Of more importance, the British White Paper of 1941 stressed the "national income and outlay account" idea. That is, income and expenditure were viewed as the two sides of a double entry production account for the entire national economy. National expenditure was thus finally established as a coordinate variable, which together with income was necessary as background for budget formulation as well as projections and broader policy formulation.

Comparable to the SNA, the United Nations launched a major effort to produce and publish similar guidelines, applicable to centrally planned economies such as the Soviet Union, China, etc. The United Nations published these guidelines under the heading .Basic Principles of the System of Balances of the National Economy,3 (also known as Material Product System or MPS for short). Thus, two systems, one SNA and the other MPS, are prevalent now to compile economic accounts. Another effort is going on now to provide guidelines for comparisons between the two systems. Conceptual relationships between the two systems have been delineated.4 As well, statistical reconciliation is progressing between some European countries following SNA and MPS systems.

Another major effort is underway to produce international comparisons of gross product and purchasing power. As a result of work over the past two decades by national statistical offices, the United Nations and other international organizations, data on national income and expenditure are becoming more and more comparable from the standpoint of statistical methodology. However, even where standard methodology has been adopted to produce national estimates of these aggregates, a major limitation to comparability has been the inadequacy of official exchange rates for the purposes of converting estimates in national currencies to a common basis of valuation. The United Nations jointly with the World Bank initiated a project in 1968 to fill this important gap in international statistics by developing detailed inter-country comparisons for gross domestic product and the purchasing power of currencies. Only a handful of countries participated in the first round of this exercise. These results were published in 1975.5 The results were dramatic for a country like India. For example, using the official exchange rate, India's per capita income was about one fiftieth of the USA but using the purchasing power parity methodology of the UN International Comparisons Project, it was only one fourteenth of the USA for the year 1970. The official exchange rate in that

Statistics Canada - 76 - Collected Articles of Kishori Lal year was Rs 7.5 to a dollar but the purchasing power rate was calculated to be Rs 2.15 per dollar. The Statistical Office of the United Nations has given a high priority to the project. A second report was issued in 1978. It is quite possible that in the foreseeable future this new set of statistics would become a regular feature of the United Nations statistics. Note that Canada is now participating in this important new endeavour of the United Nations.

Three Approaches to the Measurement of GDP

There are three basic approaches to the measurement of economic production each of which yields an identical total. These are usually denoted as the income approach, the expenditure approach, and the net value added approach. A simple set of figures will illustrate the matter. Assume an economy in which all output is produced by three firms. Each firm sells its production to another firm for further processing or fabrication, or disposes of it in final markets as sales to final users (Table 1). Thus Firm 1, a grain producer, imports seeds and produces grains for sale to Firm 2, a Flour Mill. The Flour Mill sells a part of his output of flour in final markets (to consumers, governments, etc., who are the ultimate users of the product) and the balance to Firm 3, a bread manufacturer. The bread manufacturer in turn sells all of this production of bread to final users. In this example, each firm generates income only in the form of wages and salaries, and profits. For simplicity, it is assumed that there are no taxes, no depreciation, and no inventories.

From the basic data presented in Table 1, it is possible to measure the unduplicated value of production produced by these three firms in three separate ways. The first is simply to sum all of the factor incomes (wages and salaries, and profits) generated by this productive activity - incomes representing the returns to the labour and capital employed. The results of this summation are shown in Table 2.

A second approach is to sum all sales which these firms have made to final users - to consumers, to governments, to business on capital account, or in export markets. This approach also provides an unduplicated measure of the total value of production produced by these three firms. Imports, of course, have to be deducted from this summation (as in Table 3) since imports are implicitly included in these final sales and should not be counted as a part of domestic production - they represent part of the production of a foreign country. Sales from one firm to another (intermediate production) are not counted since to do so would involve double counting, all intermediate production being embodied in final output sold to final users. (In this particular example, all output is assumed to be sold and none accumulates in the form of unsold inventories).

It can be seen that this "sales to final users" approach (or "sum of expenditures" approach) yields the same value of production as the "sum of incomes" approach.

Finally, the unduplicated value of production can be measured by taking the gross value of production of each firm and subtracting from this each firm's costs of production in the form of its purchases from other firms (including imports), to yield the "net value added" to production by the firm. Again, it can be seen from Table 4 that this procedure yields the same value of "unduplicated production" as is achieved under the first two approaches.

Statistics Canada - 77 - Collected Articles of Kishori Lal

Table 1. Hypothetical Example of a Three-firm economy ______Cost of Production $ Gross value of production $

FIRM 1 (Grain Producer) Imports (seeds) 15 Sales to Firm (2 Grains) 100 Wages and Salaries 60 Profits 25 TOTAL 100 TOTAL 100

FIRM 2 (Flour Mill) Purchases from Firm 1 Sales to final users (flour to (Grains) 100 consumers, governments, etc.) 30 Wages and Salaries 40 Sales to Firm 3 (Flour to Bread Profits 10 manufacturer) 120 TOTAL 150 TOTAL 150

FIRM 3 (Bakery) Purchases from Firm 2 Sales to final users (Flour) 120 (Bread to consumers, Wages and Salaries 900 governments, etc. 1,100 Profits 80 TOTAL 1,100 TOTAL 1,100 ______

Statistics Canada - 78 - Collected Articles of Kishori Lal

Table 2. "Sum of Incomes" Approach ______$ Firm 1 60 + 25 = 85 Firm 2 40 + 10 = 50 Firm 3 900 + 80 = 980 Total incomes earned in current production (total value of production) 1,115 ______

Table 3. "Sum of Sales to Final Users" Approach or "Sum of Expenditures by Final Users" ______$ Firm 1 ...... - Firm 2 ...... 30 Firm 3 ...... 1,100

Sub-total ...... 1,130

Less imports ...... - 15

Total sales of current Canadian production to final users ...... 1,115 ______

Table 4. "Sum of Net Values Added" Approach ______Gross value Purchases from Net value of other firms, added in production or imports production (1) (2) (3)= (1-2)

Firm 1 ...... 10015 85

Firm 2 ...... 150100 50

Firm 3 ...... 1,100 120 980

Total ...... 1,350 235 1,115 ______

Statistics Canada - 79 - Collected Articles of Kishori Lal

It is this approach which is primarily used in estimating Gross Domestic Product by Industry both by the Industry Product and Input-Output Divisions in Statistics Canada. In some European countries, the national accounts are produced using the Input-Output approach. Input-Output Accounts not only produce value-added but as well produce final expenditures from the same database. Total production can be used for intermediate use and/or for final uses and cost of production includes intermediate expenses and primary inputs (including profits). In this equation, intermediate expenses are common, hence final uses equal primary inputs. Primary inputs are nothing but the elements of the income approach above and the final uses depict elements of "Sum of Sales to Final Uses". This approach forces a joint and statistically equal calculation of the two sides of the unduplicated value of current production of an economy.

Gross Domestic Product (GDP) versus Gross National Product (GNP):

Some countries call their unduplicated production Gross Domestic Product while others call it Gross National Product. These are not identical terms. Here is the difference.

Residents of Canada may contribute to production in other countries by investing capital in those countries, which earns them a return. Similarly, non-residents may contribute to production arising within the geographic boundaries of Canada by investing capital in Canada which earns them a return. In the first instance, if the boundary of production is drawn to include all of the production of Canadian factors of production irrespective of what country the production originates in, the measurement is conventionally denoted as the "national product". In the second instance, if the boundary of production is confined to the geographic territory of Canada irrespective of whether the factors of production are resident (of Canada) or non-resident, the measurement is conventionally denoted as the "domestic product". For most practical purposes, the reconciliation between GDP and GNP involves the recording of transfer of interest and dividends across national borders. Normally, Canada's GDP will be higher than GNP because more foreign capital is invested here than Canada's capital abroad.

The revised UN System of National Accounts is built around the domestic concept. It is easier to reconcile with and aggregate the regional and provincial accounts to the accounts for the entire country on domestic basis than on national basis, precisely because there are no records kept for movement of interest and dividends across provincial boundaries within a given country. More and more countries are producing regional accounts. Another consideration is the prevalence of input-output tables in most countries. Input-Output tables depict production or industrial relations, hence are built around production statistics based on "establishment" records rather than "company" records. Establishment-based statistics record production data, inputs and outputs of commodities whereas company-based statistics record financial data such as profits (or losses), service expenditure, etc. Given these considerations, it is preferable to work with the domestic concept of production for the main accounts of an economy.

The growth rate of an economy could be different depending upon whether one measures it on domestic basis or national basis. This is particularly true in times of legal and politic changes of administration or tax laws. When an economy's growth rate is published, one normally does not

Statistics Canada - 80 - Collected Articles of Kishori Lal specify the conceptual intricacies (such as GDP versus GNP) to understand the reasons for growth or lack of it. Only a catch-all phrase, "growth of the economy" is bandied about which could be misleading.

GDP in Constant Prices

A time series of GDP expressed in current prices, that is the prices which obtain in the actual periods being measured, reflect two elements of change - changes in the quantities of commodities produced and changes in the prices attached to those commodities. The process of removing price change from the nominal values of transactions is referred to as deflation.

Certain characteristics of the deflation process should be noted at the outset. Deflation is essentially the construction of a model, or logical abstraction of actual transactions, and as such it has certain interesting limitations. The seemingly ideal approach to deflation would be direct measures of quantity. However such measures are not additive. There is no meaningful way to combine tons of steel and dozens of oranges, for example. The only workable solution to this dilemma is to fix values for these quantities in terms of their prices at some point in time, the base year. Thus the dollar value of a quantity of oranges is additive to the dollar value of a quantity of steel. If all quantities are expressed in terms of their base-year prices, the constant price values for each commodity will be proportional to its quantities in different years, yet commodities of diverse characteristics will remain additive, and the construction of a set of accounts summarizing manifold transactions is possible.

The conventions just described leave one problem unsolved. It is easily observed that relative price changes for different commodities are not identical, and thus the relationship of the prices of diverse commodities will be different at different points in time. The ratio of the price of a litre of oil to the price of a shirt is quite different in the later 1970's than it was in the previous decade. If quantities of oil and shirt were to be combined, the importance of oil, or its "weight", if the quantities are valued in prices of (say) 1976, will be much greater than if valued in 1961 prices. Thus it is important to note that trends in constant prices series which are summations of two or more commodities can be strongly influenced by the choice of the base year.

Given relevant price indices, one can convert current price series of final expenditures into constant price series. But on the Income side of the accounts, it is difficult to even visualize how to factor some of the primary inputs into quantum and price. Imagine the construction of price indexes for profits and investment income!

In the absence of a direct measure of factoring primary inputs into quantum and price, a statistical technique, known as double deflation is applied. Double deflation is a technique whereby the constant price value-added of (say) an industry is inferred as the difference between the quantum of gross output and the quantum of commodity input of that industry, rather than being measured direct. Valued-added thus calculated of all industries is combined to arrive at GDP for the whole economy.

In the Input-Output Accounts, there is an inherent equality between value-added and final expenditure in current prices and the same equality prevails in constant prices. Value-added in

Statistics Canada - 81 - Collected Articles of Kishori Lal constant prices must equal final expenditures in constant prices because the constant price intermediate use is common to both sides of the accounts. Double deflation technique must by definition result in this equality when applied to completely balanced Input-Output tables in current prices.

Interpretation of the GDP Numbers

We have gone over some of the issues of interpretation of numbers on GDP. In this section let us summarize in one place the various signals that an economist should watch before making judgements on GDP numbers.

(i) GDP is not an observable phenomenon. It is a concept, a model. It is not a unique reality because it is not independent of the assumptions and conventions made and used by the investigator. That it is primarily an economic model should be kept in mind before making comments on GDP numbers.

(ii) It is true that most of the countries now use internationally published (both for SNA and MPS) conventions, definitions etc, but institutional rearrangements could lead to differences in growth rate though "production activity" has remained constant. For example it makes a difference in the level of GDP whether activities are carried out by government or by business. Thus for instance in a country where research and development activities are carried out largely by government institutions and the results of these activities are made available free of charge to enterprises, gross domestic product as defined in the SNA will be higher than in a country where these activities are carried out by private institutions and paid for by enterprises as a part of their costs.

(iii) We noted above that growth rates could be different when production is measured on GDP concept than GNP concept though both concepts measure an unduplicated value of production of an economy.

(iv) Constant price GDP or real growth of an economy is not independent of the base year that one chooses. Real growth between 1979 and 1978 could be different depending upon whether one is measuring it in 1971 prices or 1975 prices. Indeed constant price GDP is even more complex to interpret than current price GDP as it is a model twice removed from reality - first for current price GDP and then for the choice of a year for comparison. (v) As noted above the international comparisons of per capita GDP or total GDP can't be adequately made using official exchange rate. One needs to develop purchasing power parity for countries to have a clearer idea of a growth of an economy. Presently published figures are inadequate.

(vi) International comparisons between countries using different accounting systems (SNA versus MPS) are simply not believable without a major research effort to convert one system to the other. For example, Soviet Union's reported economic growth rate may be 5 percent and United States' 4 percent. It is impossible to say which country actually has a higher growth

Statistics Canada - 82 - Collected Articles of Kishori Lal

rate without putting both on the same system. Soviet Unions' material production might grow by 5 percent while the growth for non material production such as services - hospitals, education, government administration, professional services, etc. - might be zero. These statistics give Soviet Union's growth rate 5 percent on MPS but measured by SNA conventions, it might be only 3 percent compared to United States' of 4 percent on SNA. A great caution is needed to interpret results.

Concluding Remarks

Three major issues, in addition to the many nuances of GDP, have been briefly touched in this note: international comparability of the two main systems of accounts, SNA and MPS; international comparisons of levels of output using purchasing power parity values rather than the official exchange rates; and removing price changes from the nominal values of producing a "quantum" measure of GDP. Underlying all these is an assumption of some kind of classification scheme for transactions and transactors. Classification is fundamental to any statistical compilation. Concordance of various (and sometimes conflicting) classifications in use for myriad transactions is a formidable challenge. This topic must be studied in depth by every economist interested in working with empirical results, otherwise one's perception remains vague and confusing.

Statistics Canada - 83 - Collected Articles of Kishori Lal

References

a Excepting minor editing, this is a reproduction from the Canadian Statistical Review, Statistics Canada, Ottawa, April, 1982. An earlier version of this paper was published in the PSE (Punjab School of Economics) Economic Analyst, Amritsar, Vol. II, No. 2, December 1981.

1. John W. Kendrick in his book Economic Accounts and Their Uses (New York, 1972, pp. 10-19) gives an excellent summary of the history of Economic Accounts.

2. United Nations, A System of National Accounts, Studies in Methods, Series F, No. 2, Rev. 3, New York, 1968.

3. United Nations, Basic Principles of the System of Balances of the National Economy, Studies in Methods, Series F, No. 17, New York, 1971.

4. United Nations, Comparisons of the System of National Accounts and the System of Balances of the National Economy, Part 1, Conceptual Relationships, Studies in Methods, Series F, No. 20, New York, 1977.

5. Irving B. Kravis and others, A System of International Comparisons of Gross Product and Purchasing Power, United Nations International Comparison Project Phase 1, Baltimore, 1975.

Statistics Canada - 84 - Collected Articles of Kishori Lal Services Industries in the Business Sector of the Canadian Economy a

by Kishori Lal Statistics Canada

Abstract: The focus of this paper is an examination of concepts and statistics necessary to support the measurement of real output of services industries. Although applied to the business sector of the Canadian economy, the issues raised may in fact be relevant for many other countries. The Canadian practice is judged with reference to the international guidelines and recommendations. General methodological issues of deflation are examined, particularly double deflation, extrapolation and rebasing. The perennial problems such as the deflation of imputed banking services, insurance, trade margins, etc. are analysed with a view to provide some tentative solutions thereof. Quality assessment of statistics and the criteria used for such an assessment are indicated to share our concern with the international colleagues facing similar problems.

The focus of this paper is an examination of concepts and statistics necessary to support the measurement of real output of services industries in the business sector of the Canadian economy. Wherever appropriate, the Canadian practice is judged with reference to the guidelines and recommendations released about ten years ago in the Manual on National Accounts at Constant Prices (hereafter referred in this paper as UN Manual). There are 34 specific business services industries delineated in the Canadian System of National Accounts (CSNA). Appendix A on Constant Price GDP Annual Benchmarks, attached to this note, provides a brief description of the type of deflators used and a judgmental indicator on the overall quality rating of GDP by Industry. For readers who wish to pursue the subject further, some of the important documents are listed at the end of this paper.

General Methodological Issues of Deflation

At the outset it is important to note that measurement of real output by industry and for the total economy is essentially the construction of a model or logical abstraction of actual transactions. "Volume" or "quantum" is always estimated with reference to time and weights. As we are all familiar with the index number problems associated with changing weights and aggregation, the estimated quantum measure at any level of aggregation is not an unambiguous phenomenon even if one has the most reliable value, price and quantity statistics in detail. This nature of deflation applies to goods, services, industries, whatever methods are used. However, the alternative methods e.g. double or single deflation can be examined from the point of view of their effect on the measures of real output.

(i) Double deflation

The Canadian practice conforms to the UN recommendation. The UN Manual states: "In an ideal world real product by kind of activity would always be derived from an input-output table by double

Statistics Canada - 85 - Collected Articles of Kishori Lal deflation (p. 55)." In Canada, the Input-Output Tables form the core of the production accounts. Input-output tables in their full details are produced annually both in current and constant prices. Real output by industry is produced using the preferred double deflation approach.

In the double deflation approach, one deflates commodity outputs of an industry, its intermediate use of commodity inputs and net indirect taxes. The difference between the deflated values of outputs and the total of commodity inputs and net indirect taxes equals GDP at factor cost in constant prices. The double deflation approach satisfies the requirement of an identity between GDP income and expenditure based estimates in constant prices. However, there are certain important hazards using double deflation for deriving GDP of an industry whose value added makes a small proportion of total gross output.

For such an industry, GDP estimated by double deflation might be erratic because small shifts in the relative prices for intermediate inputs and gross output could translate into big shifts in the resultant value added at constant prices. Here the UN Manual guidelines are not entirely satisfactory. They state: "The solution to this problem, however, may simply be to consolidate industries with very small ratios of value added to gross output with related industries at earlier or later stages of production. In other words, the problem of instability may be solved by aggregation into larger units whose values added are large enough in relation to gross output not to be too sensitive to the effects of changes in prices or technology (p. 53)." In one year, value added in one industry may be erratic but in the next, a different industry might suffer. Ad hoc aggregation into large units would disturb the continuity of time series. Thus one needs additional guidelines. In the CSNA, we have resolved this problem as follows: Values added are combined as suggested by the UN. The combined value added is redistributed using gross output or any other indicator as a proxy but the combined value as such for a sub-aggregate remains unchanged. Without this restriction, the above noted GDP identity requirement will not be satisfied.

(ii) Extrapolation of the base year value added

The UN Manual deals essentially with the problem of "extrapolation" of the base year value added by industry. One can extrapolate value added using constant price gross output, physical quantities of gross output, constant price intermediate inputs, employment etc. Should one prefer deflated output over physical quantities of gross output as a projector of base year value added? The UN Manual recommendation is: "The value at constant prices of the goods available in both years should be obtained by deflating the current year value by a price index, rather than by extrapolating the base year value by a volume index .... The justification for this recommendation is that price relatives generally display less variation than quantity relatives. The range of variation of quantity relatives may be anywhere from zero to infinity whereas that for price relatives is much narrower (pp. 46-47)."

Statistics Canada - 86 - Collected Articles of Kishori Lal price and quantity information. Statisticians may still prefer quantities over prices if in their judgement the quantity information is less incomplete than price information.

(iii) Rebasing

It is self evident that the more remote the base year becomes, the less relevant are its prices for purposes of deflating the value of current flows of goods and services. (In the Canadian System of National Accounts, the base year is changed every ten years. However, we are analysing the need to shift to a five year frequency. Our latest base year is 1981).

When the base year is changed, there are two ways in which it may be done in practice. The first method is to revalue not only all years subsequent to the new base year at the new prices, but also all the years preceding the new base year in order to have an unbroken series extending on either side of the new base year. The second method is to leave the data for all years up to and including the new base year unchanged and simply to use the new base year prices for valuing all flows of goods and services from the new base year onwards.

CSNA rejects the first method in favour of the second on conceptual grounds. One requires a new base year because the old one is not relevant anymore. Thus one can't honestly argue for the first method. But the UN Manual recommends the first method: "It is also recommended that the second method of rebasing should be replaced by the first method whenever resources make it possible (p. 18)." CSNA is fully aware that the conceptual argument becomes rather weak at the boundary. For example, the 1981 base year may be more relevant for 1980 than the 1971 base year. However, for operational reasons and convenience, the boundary period is not handled differently.

CSNA, for example, preserved the original growth rates of both total GDP and all the published sub-aggregates for the periods 1961-71 and 1971-81 when the base was shifted to 1981 during the last historical revisions completed in 1986. All the details and totals were rescaled to 1981 base without disturbing the original growth rates. Components do not add to totals but history is not rewritten every time the base year is changed. This issue needs to be re-addressed.

Conceptually Satisfactory

Our understanding and appreciation of problems surrounding the measurement of real output of services industries would be enhanced if country practices in this area were internationally known. This and the following sections detail our practices and our judgements in some of the most important areas of the economy.

Four industries (i, Monetary and Financial Institution; ii, Insurance; iii, Data Processing Services; and iv, Lotteries and Race Tracks) are considered conceptually unsatisfactory and reasons for their being judged so are detailed in a later section called Conceptually Unsatisfactory. All other industries are considered conceptually satisfactory. Though the two trade industries (Wholesale Trade and Retail Trade) whose principal output is trade margins are considered conceptually satisfactory, a nagging feeling of uneasiness remains. Deflation of trade margins has not been

Statistics Canada - 87 - Collected Articles of Kishori Lal resolved internationally or in the CSNA to our entire satisfaction. However, the CSNA approach is worth sharing with others.

Trade Margins

In the CSNA, both the Make Matrix and the Use Matrix of the Annual Input-output accounts are in the preferred approximate basic values as defined in the UN SNA (1968). Approximate basic values are usually recorded for Make Matrix, but rarely for Use Matrix. The Use Matrix records values at purchaser prices which are equal to approximate basic values plus, where applicable, trade, transport and tax margins. Each recorded purchaser price value has to be disaggregated into its component of approximate basic value and the margins, all delineated separately. The matrix for each margin has the same dimension as the Use Matrix. The trade and transport margins remain part of the intermediate matrix, but the tax margins get shifted to the primary inputs matrix to get GDP at market prices. The estimation of margins is made for the current price annual Input-output program of CSNA. Indirect taxes are universally deflated using the base year's rates and the transport margins are usually deflated using base year's rate per ton kilometre. However, there are no internationally accepted guidelines for deflating trade margins. The Canadian experience is interesting in the light of the UN Manual statement: "There seems to be no practical or operational way in which the amount of services per unit of goods can be quantified and measured, and it is not possible to make specific recommendations on this point (p. 88)."

Ideally, constant price trade margins should be derived from the difference between the constant price sales of commodities and the constant price cost of these commodities for the trader. However, the array of price indexes required include buying and selling price indices by both wholesale and retail trade, by commodity, and these are not compiled.

Calculating margins in constant prices as the product of the margin rate in the base year and the volume of the commodity being traded in subsequent years was rejected on conceptual grounds. Constant margin rates imply constant proportionality between the volume of a commodity being sold and the quantity of trade margin or service being attached to that good. Observation of the market place in action suggests that this is not a good assumption. Marketing techniques change and the quantity of distribution service appears to decrease. Service stations require drivers to pump their own gasoline; free delivery is being discontinued; packaging and display are "economized" to reduce costs.

Another approach tried was to use the margin rates for each year from the current price tables. This approach was used for the 1960s. During this period, the range of variation of price relatives was quite narrow. This approach worked quite well and the results were judged quite satisfactory. The above technique was originally carried into the 1970s with less satisfactory results. The use of current year trade margin rates exacerbated the effects of rapidly changing prices. The next step was to emulate Goldilocks and the Three Bears in search of some formula that seemed "just right." The closest to this proved to be using a margin rate calculated as the average of the given year's rates and the base year's rates from the current price tables. The two principal criteria by which different

Statistics Canada - 88 - Collected Articles of Kishori Lal approaches were evaluated were reasonable output measures for the trade industries and an implicit price at purchasers' values that most closely approximated observed purchasers' prices.

The implicit price at purchasers' value is calculated in the CSNA Input-Output Tables as follows: (i) approximate basic values are deflated using product price index; (ii) constant price transport and tax margins are calculated using base year's rates; (iii) trade margins are deflated using the average of the current year's and base year's rates; (iv) sum the above three components to arrive at deflated purchaser price value; (v) divide current price counterpart of the above by (iv), which gives implicit purchaser price; (vi) compare this with recorded purchaser price index such as Consumer Price Index. If the implicit purchaser price approximates the recorded purchaser price, we feel just as happy as Goldilocks was.

The above method of using the average of the two rates has been used for most of the commodities in the Input-Output Tables since 1971. The results are embedded in our published series. Still there is a nagging feeling of uneasiness. This is one area which needs to be further investigated both nationally and internationally, all the more so because trade industries contribute a rather significant part of total GDP. One fourth of GDP within the business services sector of the Canadian economy is contributed by the trade industries and their share may be similar in other internationally advanced economies.

Conceptually unsatisfactory

In the case of an individual good or service, value is expressed as the product of price multiplied by quantity. Quantities are measured in physical units. The quantities may be expressed simply in numbers of items or in graduated units of length, volume or weight. "It is imperative that the physical unit should be identifiable. Otherwise, the concept of price is meaningless (UN Manual, p. 12)." However, there are several rather important industries in the business services sector which do not possess a clearly identifiable physical unit. These include Finance, Insurance industries, rental of data processing and activities connected with gambling (such as lotteries and race tracks).

(i) Monetary and Financial Institutions

Monetary and financial institutions provide a variety of services some of which are specifically charged. They include commissions, charges on cheques, etc. There is a fee and the current revenues are estimated based on transactions which involve a number of physical units. The concept of output is clear and in principle output is deflatable. At times these services are bundled into a single product which has no identifiable measure of physical units. Moreover, the gross output of banks as defined in the SNA also includes imputed services which are not directly charged. These equal revenues from interest received from loans less interest paid on deposits. Accordingly, the SNA defined current price GDP of banks (with the above imputation for gross output) remains the same as recorded in the bank books. However, the issue is how to deflate such services when there is no identifiable unit of measurement.

Statistics Canada - 89 - Collected Articles of Kishori Lal In view of the indirect charging for (imputed) services and the bundling of services, it is sometimes suggested that the best way to obtain information on the content of bundles and to delineate additional product detail is to examine the cost structure of the industry. This should help to compute relative prices and detailed weights for some such services. The forthcoming UN commodity classification, called Central Product Classification (CPC) has listed several financial intermediation services. If current revenues and expenses broken down by each of the CPC categories could be collected and the relative prices computed, it might be possible to approximate a measure of real output. However, the success of this approach depends on solving the conceptual problem of how to allocate profits to each of the many services provided by banks. Without allocation of profit, the components of gross output are not exhausted.

The UN Manual (p. 92) recommended another approach to improve upon the input measures: "Although it may not be possible to obtain a sufficiently comprehensive range of indicators to give an adequate coverage of the entire field of banking, selective and detailed case studies of labour productivity in certain areas of banking seem a promising line of investigation and to offer the prospect of improving on input measures which make no allowance whatsoever for increased productivity." The UN suggestion has not been followed in the CSNA work program primarily because such detailed and selective studies of labour productivity are time consuming, costly and very problematic. The UN Manual makes an identical suggestion for insurance output and it has not been followed for the same reasons as those for banking.

It is quite possible and appropriate that we collectively (i.e. all participating organizations/countries of the Voorburg Group) formulate terms of reference of selective studies for both banking and insurance industries and then distribute them by organization/country. We are aware that another country's productivity study may be inappropriate for incorporation into the national program. However, out of this might emerge a common approach which is cost effective and which adds to our knowledge of this important, growing and increasingly internationalized sector of our economies.

The Canadian SNA practice to deflate imputed bank services for the annual program is to obtain details on various kinds of loans and on their respective interest rates as well as details on the various kinds of deposits and on their interest rates. This is followed by an attempt to match as closely as possible each category of loan with a "corresponding" deposit category. The difference between the deposit rates and the loan rate is estimated to be the margin that the financial institutions charge and in principle these should be equal to the measurement of imputed services of banks. For current years the base year margin rates are applied to the corresponding volume of loans by category to measure the deflated output of imputed services. The uneasiness with this practice derives from the fact that the calculated labour productivity measures are volatile. Sometimes they do not correspond to common sense observations in the market.

(ii) Insurance Industries

Insurance industries share some of the same problems as financial institutions. The current SNA convention is to set the value of the service produced for property and casualty insurance as the difference between claims and premiums. However, the level of premiums is probably established

Statistics Canada - 90 - Collected Articles of Kishori Lal so as to take into account the investment income earned on the policy reserves of the insurer. Moreover, the present SNA defined value of insurance output does not have an identifiable unit of measurement.

The CSNA practice is to deflate both the premium income of life insurance and the premium income of property and automobile insurance by the Consumer Price Index (CPI) and to extrapolate the base premium income of accident and sickness insurance by the number of persons covered. The labour productivity measures resulting from the above calculations are as volatile as those calculated for the finance industries.

(iii) Data Processing Services

Data processing services, similar to banking services are sold as "bundles" or "packages" without identifiable physical units of measurement. The underlying services are changing rapidly and are not amenable to an easy solution. In areas of data processing services, we have simply calculated implicit index of inputs with all the weaknesses embedded into them.

(iv) Lotteries and Race Tracks

The output of lottery business is measured as the difference between sales of lottery tickets and the prizes. It has no physical unit of measurement. In the CSNA, the output is deflated by the overall CPI for goods and services.

Industries which are deflated by conceptually unsatisfactory methods make up about 15 percent of GDP of business services sector of the Canadian economy, or about 6 percent of total GDP in Canada. They are primarily in the finance and insurance sector of the economy.

Overall Quality Assessment

The conceptual foundation for output measures of individual industries may be sound or weak, the statistical foundation of such measures may be good or weak, yet the national accounts must be comprehensive. An estimate must be made for each item that appears in an accounting system. The system of national accounts could not be established in practice if it were restricted to those components which were soundly based on accurately recorded facts. Inevitably, published statistics will include estimates for which evidence is relatively weak. However, even a relatively weak series is strengthened after it is analysed, adjusted and generally put through the checks and balances of the national accounts. It is, nonetheless, the duty of statisticians to warn users of the weaknesses of the underlying statistics. After all, the users must not (except at their own risk) draw substantial conclusions about the economy from differences between components, or differences in the size of a given component between one period and another if such differences may be due wholly or mainly to errors of measurement.

SNA estimates are synthetic. They use series based on random samples, universe collection, projections based on circumstantial evidence plus adjustment for undercoverage, all constrained by

Statistics Canada - 91 - Collected Articles of Kishori Lal the checks and balances of the system. To date, we have not found a way of summarizing the quality attributes of the basic data and the quality of the resulting SNA estimates in an objective manner. It is, however, possible from knowledge of the data and concepts to form very rough and partly subjective judgements about the range of reasonable doubt that should be attached to such estimates. Such quality ratings are included in the appendix.

On a scale of 1 to 3, rating of 1 is most reliable. All service industries whose output measures are conceptually unsatisfactory carry a rating of 3. In addition, all service industries for which there is no substantial annual data carry the rating 3. This sub-set includes industries which are not surveyed on a regular basis even quinquennially. A rating of 1 is attached to those industries where the undercoverage is minimal; very good commodity output detail exists; and the intermediate inputs are well specified. Furthermore, to stay within the rating of 1, the deflators of the main outputs or inputs of an industry are based on price indices of commodities directly covered in a price survey or the deflators are constructed from data on quantities and values from production surveys or other records in which the information is adequate in coverage and detail. In between the two ratings are all other industries. These carry the rating 2.

Concluding Remarks

We have listed four industries whose output measures are conceptually deficient. An additional four industries are listed which have a poor statistical base for the annual measures. These eight industries are: SNA Industry Code 137 Banks, Credit Unions and Other Deposit Institutions; 139 Insurance Industries; 142 Other Business Services Industries (Including Data Processing); 150 Other Amusement and Recreational Services (Including Lotteries and Race Tracks); 124 Taxicab Industry; 126 Highway and Bridge Maintenance; 141 Owner Occupied Dwellings; and 145 Educational Service Industries. The first four are conceptually deficient and the last four are statistically deficient. Other countries with similar structure to Canada may have an almost identical list.

Trade industries are also deficient in many countries, but the Canadian practice may contribute to an improvement on the current condition. Issues relating to rebasing and double deflation need to be readdressed and again here the Canadian practice might be worth considering. The Canadian practice for deflating banking imputed services, though deficient, might be an improvement over the recommendations included in the 1979 UN Manual.

Statistics Canada - 92 - Collected Articles of Kishori Lal APPENDIX A

Constant Price GDP for Annual Benchmarks, Business Services Industries, Canada (Double deflation is the method used throughout)

CSNA Overall Industry Brief Description of Quality Code CSNA Industry Title Principal Deflators Assessment

118 Air Transport & Services Incidental Passenger Kilometres by class 1 119 Railway Transport & Rel. Services 2 Passengers Consumer Price Index Freight Rate per ton kilometre 120 Water Transport & Rel. Services Freight Freight rate indexes 2 Incidental Services Base Year rates 121 Truck Transport Industries 2 Freight Rate per ton kilometre 122 Urban Transit System Industry Consumer Price Index 1 123 Interurban & Rural Transit Systems Consumer Price Index 2 124 Taxicab Industry Consumer Price Index 3 125 Other Transport & Serv. to Transp. 2 Other Transport School Bus Consumer Price Index Chartered Buses Consumer Price Index Services Incidental Parking Consumer Price Index Travel Agencies Consumer Price Index Freight Forwarders Truck Transportation Deflator 126 Highway & Bridge Maintenance Ind. Truck Transport Price Index 3 127 Pipeline Transport Industries Volume Indexes of Goods 2 Transported & Base Year Rate 128 Storage & Warehousing Industries 2 Storage Volume Indexes by Type of Grain and Type of Elevator & Base Year Rates Warehousing Rent Deflator 129 Telecommunication Broadcasting Ind. Advertising Rates for Prime Time Advertising 2 Cable Consumer Price Index 130 Telecommunication Carriers & Other Volume Indicators 2 131 Postal Service Industry Volume Indicators 2 135 Wholesale Trade Industries 2 Margins Average of Base and Current Year 136 Retail Trade Industries 2 Margins Average of Base and Current Year Rates Automobile Repairs Consumer Price Index 137 Banks, Credit Union & Oth. Dep. Inst. 3 Banks Deflated Bank Loans & Base Year Gross Margin Rate by Category Credit Unions, etc. Bank Deflator 138 Trust, Other Finance & Real Estate 2 Imputed Interest - Non-Banks Bank Deflator Stock and Bond Commissions Base Year Commission Rates & Deflated Value of Stocks Traded Real Estate Commissions Base Year Commission Rate & Deflated Sales Rents - Residential and Other Base Year Rent & Index of Stock of 139 Insurance Industries Dwellings 3 Life Non-Life Deflated Premium Income - CPI Property and Automobile Accident and Sickness Deflated Premium Income - CPI Number of Persons Covered

Statistics Canada - 93 - Collected Articles of Kishori Lal APPENDIX A - continued

CSNA Overall Brief Description of Industry CSNA Industry Title Quality Principal Deflators Code Assessment 140 Govt. Royalties on Nat. Resources Base Year Rate & Deflated Gross 2 Output 141 Owner Occupied Dwellings Index of Stock of Dwellings and 3 Base Year Imputed Rent 142 Other Business Service Industries 3 Rental of Data Processing Equip. Implicit Price Index of Inputs Miscellaneous Business Services Index of Average Weekly Earnings 2 143 Professional Business Services Legal Output Activity & Base Year Fees Accountants Index of Average Weekly Earnings Architects Index of Average Weekly Earnings Engineers Index of Average Weekly Earnings 144 Advertising Services 2 Print Media Advertising Rates Per Line Radio and Television Prime Time Rates 145 Educational Service Industries 3 Private Schools Implicit Price Index of Operating Expenditures of Universities Other Education and Cultural Consumer Price Index Services 146 Hospitals Index of Employment 2 147 Other Health Services 2 Doctors Index of Provincial Government Remuneration Rates Dentists Consumer Price Index Special Care Facilities Index of Employment Applied to Hospital Labour Income Other Health Care Consumer Price Index 148 Accommodation & Food Service Ind. 2 Accommodation Consumer Price Index Meals Consumer Price Index Service Margin on Alcoholic Index of Deflated Sales to Res Beverages taurants and Hotels & Base Year Margin 149 Motion Picture & Video Industries 2 Exhibition Consumer Price Index Production and Distribution Consumer Price Index 3 150 Other Amusement & Recreational Serv. Theatre Consumer Price Index Sports and Recreation Consumer Price Index Race Track and Gambling Attendance Lotteries Consumer Price Index 151 Laundries & Cleaners Consumer Price Index 2 152 Other Personal Services Barbers and Beauty Salons Consumer Price Index 2 Funeral Services Implicit Price Index for Inputs Day Care Centres Consumer price Index 153 Photographers Index of Average Weekly Earnings 2 154 Misc. Service Industries 2 Repair Services Consumer Price Indexes and Indexes of Average Weekly Earnings Rental of Office Equipment Price Index for Machinery and Equipment for Commercial Sector Services to Building and Index of Average Weekly Earnings Dwellings of Janitors Rental of Automobiles and Trucks Consumer Price Index Rental of Other Machinery and Machinery and Equipment Price Equipment Index

Statistics Canada - 94 - Collected Articles of Kishori Lal REFERENCES a Excepting minor editing, this is a reproduction from the Review of Income and Wealth, March 1990. An earlier draft of this paper was presented at the third meeting of the Voorburg Group on Services Statistics, held at Wiesbaden, West Germany, October 4-6, 1988. The Voorburg Group is a body of volunteer countries in association with international organizations to promote methodology and comparability in the field of service statistics.

Additional Readings

Statistics Canada, The Input-Output Structure of the Canadian Economy 1961-81, Catalogue 15-510, Statistics Canada, Ottawa, 1987.

The Input-Output Structure of the Canadian Economy in Constant Prices, 1961-81, Catalogue 15-511, Statistics Canada, Ottawa, 1987.

The Input-Output Structure of the Canadian Economy 1981-84, Catalogue 15-201, Statistics Canada, Ottawa, 1988.

The Input-Output Structure of the Canadian Economy in Constant Prices, 1981-84, Catalogue 15-202, Statistics Canada, Ottawa, 1988.

United Nations, Manual on National Accounts at Constant Prices, Series M, No. 64, United Nations, New York, 1979.

A System of National Accounts, Series F, No. 2, Rev. 3, United Nations, New York, 1968.

Statistics Canada - 95 - Collected Articles of Kishori Lal

Financial Intermediation Services Indirectly Measured (FISIM) and the Canadian System of National Accounts (CSNA) a

by Kishori Lal Director General, System of National Accounts Statistics Canada

Banks and other financial institutions provide a variety of services. Those that are specifically charged for include currency exchange, handling of cheques etc; and the corresponding revenues form part of the institutions' output. An additional, and very significant, part of their income comes from charging higher interest rates to borrowers and paying lower rates to depositors than they would need to if they charged explicitly for all their services. This "hidden" charge (known as imputed banking service in the 1968 UN SNA) is called financial intermediation services indirectly measured (FISIM) in the International System of National Accounts 1993 (1993 SNA). "The total value of FISIM is measured in the System as the total property income receivable by financial intermediaries minus their total interest payable, excluding the value of any property income receivable from the investment of their own funds..." (paragraph 6.125, 1993 SNA). The CSNA has not been able to exclude property income from investment of own funds since it is almost impossible to segregate various risk bearing income generating assets, related to own funds, from borrowed funds.

The recommendation for allocating financial intermediation services to users in the 1993 SNA is quite different from the approach specified in the 1968 UN SNA. In the 1968 UN SNA, imputations for financial intermediation services are made; however, such services are allocated exclusively to the business sector rather than to all borrowers and lenders, as is now recommended in the 1993 SNA. Within the business sector, the 1968 UN SNA does not allocate these services to each industry; instead, it uses a dummy industry which buys the entire service as an intermediate expense and generates an equivalent negative value added.

The 1993 SNA recognizes that certain countries may not be able to allocate FISIM among the various users. Thus it states: "In principle, the total output should, therefore, be allocated among the various recipients or users of the services for which no explicit charges are made. In practice, however, it may be difficult to find a method of allocating the total output among different users in a way which is conceptually satisfactory from an economic viewpoint and for which the requisite data are also available. Some flexibility has therefore to be accepted in the way in which the total output is allocated" (paragraph 6.126). Various ways are suggested by the 1993 SNA to allocate FISIM among users. This paper provides a summary of the estimation and allocation procedures of financial intermediation services used in the CSNA and, then, a detailed calculation using actual data for the year 1989.

Statistics Canada - 97 - Collected Articles of Kishori Lal

Summary of procedures

In the CSNA, the financial sector is disaggregated into a) banks and other deposit-accepting institutions, b) trust and deposit-accepting mortgage companies, c) credit unions, and d) consumer finance companies. For the banks and other deposit-accepting institutions, estimates are made for intermediation services, as in the 1993 SNA, but with modifications. The intermediation services provided to non-residents are removed from output. This reduced output is allocated to all sectors (business, persons and government) using such services. The value of intermediation services is calculated separately for borrowers and depositors. (Within the CSNA, part of the financial intermediation services relating to borrowers in the personal sector is also called the administrative portion of interest). The allocation of FISIM to the above sectors is based on their share of deposits with the banks and on their borrowing (except government borrowing) from them. At the time of the implementation of the 1993 SNA, the CSNA will need to change its present practice of not imputing and allocating services to non-residents and to government borrowing activity.

In the case of trust and deposit-accepting mortgage companies, intermediation services are estimated only for the personal sector, based on its share of deposits and borrowing. The intermediation services provided to the business and government sectors are removed from output. The business sector, particularly the rental industry (both with respect to rented dwellings and to owner- occupied dwellings) uses these services extensively but no allocation is made to it. Consequently, GDP for the trust and deposit-accepting companies is understated, and there is a corresponding overstatement in GDP for the rest of the business sector, particularly the rental industry. The implementation of the 1993 SNA will lead to a reallocation of the business sector's GDP between trust and mortgage companies on the one hand and the rest of the business sector on the other hand. The GDP of trust and deposit-accepting mortgage companies will further increase by an amount equal to the intermediation services allocated to the government sector.

The output of credit unions is equal to FISIM, plus explicit service charges, and other operating revenue. Credit unions are treated differently from chartered banks in the calculation of intermediation services. Since credit unions are non-profit organizations owned by its members, their operating profits are treated as interest paid to depositors. Therefore, the total intermediation service charge is lowered by the amount of the operating profit. As a result, total output equals the operating expenses including consumption of fixed capital. The present practice of allocating the entire output to the personal sector will need to be revised as the credit unions do provide services to the business sector, particularly the rental industry. In the CSNA, a reduction in the allocation to the personal sector counterbalanced by an allocation to the business sector will reduce GDP for the total economy and, particularly, for the rental industry.

Statistics Canada - 98 - Collected Articles of Kishori Lal

To complete the accounts, services received by the residents from the rest of the world will have to be included in imports and as inputs for all the users.

Detailed procedures

1. Banks and other Deposit-Accepting Institutions

This industry consists of the Bank of Canada, the chartered banks and other banking-type institutions. Chartered banks account for approximately ninety percent of the output of this industry. i) Chartered banks:

The Office of the Superintendent of Financial Institutions publishes the Consolidated Return of Revenue, Expenses and Changes in Capital and Reserve for both Schedule A and Schedule B banks. Schedule A banks are Canadian-owned and Schedule B banks are foreign-owned. The value of FISIM is property income received less interest paid on a deconsolidated basis by Canadian residents, but the data received are consolidated to include bank subsidiaries. These subsidiaries primarily include mortgage loan companies, but they also include other financial institutions. Therefore, in order to avoid duplication and to calculate the value of FISIM of the industry, chartered banks, one must deconsolidate the data. Several chartered banks perform real estate activities and payroll operations to businesses, which are not specifically classified as such. The Office of the Superintendent of Financial Institutions provides a measure of "other income" which may include these activities, and these are included among the explicit charges. The FISIM output calculation of chartered banks is provided in Table 1.

Statistics Canada - 99 - Collected Articles of Kishori Lal

Table 1 - FISIM Output, Chartered Banks, 1989

A - Banks B - Banks Total

$ millions $ millions $ millions

1. Total property income 35,586 4,578 40,164 2. Less property income by subsidiaries (11,304) (232) (11,536) 3. Equals property income by banks 24,282 4,346 28,628

4. Total deposit interest paid 23,171 3,730 26,901 5. Less interest paid by subsidiaries (8,563) (148) (8,711) 6. Equals interest paid by banks 14,608 3,582 18,190

7. FISIM by banks (3-6) 9,674 764 10,438 7a. Resident portion (CSNA output) 8,337 764 9,101 7b. Non-resident portion 1,337 - 1,337 8. FISIM less CSNA Output 1,337

A resident ratio is calculated as the ratio of resident holdings of "booked in Canada" deposit liabilities of the banks to total "booked in Canada" liabilities for Schedule A banks. In 1989, this ratio was 0.8618; it changes from year to year. The non-resident portion of FISIM is not included in output but it should be included in it, as non-residents also benefit from bank services. This will need to be revised at the time of the implementation of the 1993 SNA in the Canadian SNA.

Note that the property income of chartered banks includes income from all assets, both loans and securities. ii) Bank of Canada

The Bank of Canada is the central bank of Canada and is responsible for monetary policy. The general public does not make deposits to and get loans from the Bank of Canada as it does with respect to the chartered banks. Thus the CSNA treats the Bank of Canada differently from the chartered banks, and its output is equated to its operating expense. Its operating expense equals investment income less net revenue, and net revenue is paid to the Receiver General for (Government of) Canada. Investment revenue derives primarily from holdings of Government of Canada treasury bills and bonds. iii) Other government business enterprises

Alberta Treasury Branches and Ontario Savings Offices are two small government-owned financial institutions covered under banks and other deposit-accepting institutions in Canada. Their financial

Statistics Canada - 100 - Collected Articles of Kishori Lal statements provide details on their net property income and other revenues. Their net property income in 1989 amounted to $183 millions. iv) Allocation of CSNA Output by Sector

The total resident component of intermediation service output of chartered banks, in 1989, amounted to $9,101 millions, and it was allocated among depositors and borrowers using a ratio based on the value of deposits and loans outstanding. This is based on the assumption that a dollar of deposit or loan has an identical service component. This assumption needs to be reexamined as we know that margins earned vary a great deal amongst various assets and liabilities (see margins earned, Tables 6 and 7).

Intermediation services allocated to depositors are then allocated among persons, government and businesses on the basis of deposit holdings in each group. Note that the personal sector does not include unincorporated business enterprises or the rental industry (both owner occupied and tenant occupied). Unincorporated business enterprises and the rental industry form part of the business sector in Canada.

Intermediation services allocated to borrowers are distributed among the sectors on the basis of outstanding loans attributed to them. These data are provided by the Bank of Canada and Office of the Superintendent of Financial Institutions. By applying these sectoral splits to the depositors' and borrowers' portion of output, the CSNA is able to compute financial intermediation services for each sector. Table 2 shows this allocation.

The sectoral split of output of the Bank of Canada is based on a fixed ratio of distribution: one third is allocated to the personal sector and two thirds to the government sector. The allocation to the personal share is derived from the relative cost of administering Canada Savings Bonds by the Bank of Canada. In most countries, it is quite reasonable to allocate the entire output of the central bank to the government sector.

Statistics Canada - 101 - Collected Articles of Kishori Lal

Table 2 - FISIM Allocation by Sector, 1989, $ Millions

INDUSTRY OUTPUT BORROWER'S DEPOSITOR'S PORTION PORTION

CHARTERED BANKS: 9,101 4,464 4,637 Personal share 4,619 1,382 3,237 Business share 4,275 3,082 1,193 Government share 207 0 207

BANK OF CANADA: 186 0 186 Personal share 62 0 62 Business share 0 0 0 Government share 124 0 124

GOVERNMENT BUSINESS ENTERPRISES: 183 72 111 Personal share 104 27 77 Business share 72 45 27 Government share 7 0 7

TOTAL 9,470 4,536 4,934

Total output of the business sector is $4,347 millions ($3,082 millions plus $45 millions for the borrower's portion and $1,193 millions plus $27 millions for the depositor's portion). This total is distributed by industry in the input-output tables (see Table 8).

2. Trust and deposit-accepting mortgage companies

For trust and deposit-accepting mortgage companies, there are several sources of information, of which the most important is provided by the Office of the Superintendent of Financial Institutions. Their output is calculated as shown in Tables 3 and 4.

Statistics Canada - 102 - Collected Articles of Kishori Lal

The industry is estimated separately for its two components; trust companies and deposit-accepting mortgage companies. Trust companies were still largely independent in 1989, while deposit- accepting mortgage companies have been dominated by the chartered banks for a considerable length of time. Each component has two commodity outputs; FISIM and specific service charges. The FISIM calculation for this industry is analogous to the chartered bank calculations described above. However, only the personal portion of intermediation services is counted as CSNA output. FISIM output of $510 millions of the trust companies (allocatable to the business sector) and of $938 millions of deposit-accepting mortgage companies (allocatable to the business sector, of $897 millions, and to the government sector, of $41 millions) is simply not counted at present. As a result, GDP for this industry is underestimated by $1,448 millions; GDP of $1,407 millions of the business sector, particularly the rental industry, is overestimated. However, GDP for both the business sector and the economy is understated by only $41 millions, with a corresponding non-allocation of $41 millions to the government sector. An allocation to both the business and government sectors will need to be made at the time of the implementation of the 1993 SNA.

Table 3 - FISIM Output of Trust Companies, 1989, $ Millions

1. Total interest received 12,315 1a. Interest on consumer debt 1,156 1b. Interest on other debt, mostly mortgages 11,159

2. Total interest paid 10,388

3. FISIM output (net interest), (1-2) 1,927

4. Net interest allocated to depositors 1,252 4a. Persons 877 4b. Business 375

5. Net interest allocated to borrowers 675 5a. Persons (consumer loans) 540 5b. Business (mostly mortgages) 135

6. CSNA output 4a+5a 1,417

7. FISIM less CSNA output 510

Statistics Canada - 103 - Collected Articles of Kishori Lal

Table 4 - FISIM Output of Deposit-Accepting Mortgage Companies, 1989, $ Millions

1. Total interest income 11,915 1a. Mortgage interest 9,580 1b. Ex-mortgage interest 2,335

2. Total deposit interest paid 10,106

3. FISIM (net interest), (1-2) 1,809

4. Total deposits outstanding 367,879

5. Total loans outstanding 358,690

6. Net interest allocated to depositors 916 6a. Persons 640 6b. Business 235 6c. Government 41

7. Net Interest allocated to borrowers 893 7a. Persons (consumers loans) 231 7b. Business (mortgage and ex-mortgage) 662

8. CSNA output (6a+7a) 871

9. FISIM less CSNA output 938

3. Credit Unions

Credit unions are membership organizations; thus, any operating profits are deemed to be distributed to the members of the organization in the form of rebates on loans, bonus interest on deposits and dividends. The output of credit unions is equal to intermediation services charges plus explicit service charges and other operating revenue, as noted in Table 5. Credit unions are treated differently from chartered banks in the calculation of intermediation services. Since credit unions are non-profit organizations owned by their members, operating profits are treated as interest paid to depositors. Therefore, the total intermediation services charge is lowered by the amount of the operating profit. As a result, total output equals operating expenses and is allocated entirely to the personal sector. At the time of the implementation of the 1993 SNA, FISIM will also be allocated to business, thus reducing the allocation to the personal sector. This will reduce the GDP for the entire economy, particularly the rental industry.

Statistics Canada - 104 - Collected Articles of Kishori Lal

Table 5 - FISIM Output of Credit Unions, 1989, $ Millions

1. Interest income 9,434 Bonds 689 Mortgages 3,725 Consumer Loans 2,599 Other 2,421 2. Interest expense 6,694 2a. Depositors 6,158 2b. Borrowings 536 3. Net interest 1-2 2,740 4. Less operating profits 1,265 5. FISIM output 1,475 5a. Allocated to borrowers 1,177 5b. Allocated to depositors 298 6. Other operating revenue 321 7. CSNA output 5+6 1,796

4. Consumer Finance Companies

Consumer finance companies are non-deposit-accepting enterprises. The value of their output is equated to their operating expenses (intermediate consumption, labour costs and consumption of fixed capital). The entire output is allocated to the personal sector. The total output of consumer finance companies amounted to $2,070 millions in 1989 ($1,585 for consumer loans and $485 millions for administering outstanding receivables from the household sector by retail department stores). We may have overestimated the value of output of these enterprises, as operating expenses seem to include all expenses including interest payments. This will need to be corrected at the time of the next CSNA historical revisions. FISIM output of financial intermediaries cannot be higher than net interest received (interest income less interest payment). Such a valuation will typically be equal to operating expenses (excluding interest payments) plus operating profits.

Suppose that certain enterprises entirely use their own funds for loans. Should their activity be classified as FISIM? Applying the strict definition of the 1993 SNA, such enterprises do not provide financial intermediation services. However, they hire employees, have intermediate consumption for their activities, recognized as such by normal accounting standards, and submit their income and outlay statements and balance sheets to the tax authorities as businesses. Their activity should be recognized, if not strictly as FISIM, at least as FISIM-like. The calculation of their FISIM output may be made on the following lines.

(a) Separate notionally the lending activity of these enterprises from the owners. (b) Notionally, pay to the owners of funds, preferably, "pure interest" or in the absence of this, pay interest which is equal to the difference between interest income received less operating expenses (intermediate consumption, labour cost plus consumption of

Statistics Canada - 105 - Collected Articles of Kishori Lal

fixed capital). Pure interest represents the pure cost of borrowing funds without any intermediation services and from which the risk premium has been eliminated to the greatest extent possible. (c) FISIM output is equal to (a)-(b). FISIM output is understated by the amount of operating profit, if it is equated to operating expenses.

5. Imports

Data on imports of FISIM are very difficult to obtain. One has to obtain them from countries which provide loans to residents of Canada. Such estimates are not made at present but will need to be made to complete the accounts. Any allocation of imports of FISIM to the business sectors will reduce GDP; business sector intermediate consumption of FISIM will increase and its interest payments will decrease exactly by the amount of imports of FISIM. Imports of FISIM by the government and personal sectors are at present not embedded in the estimates of consumption of goods and services. Hence, any allocation of imports of FISIM to the government and personal sectors will add to their expenditures, counterbalanced by an increase in imports, leaving GDP unchanged.

To summarize, with the implementation of the 1993 SNA, GDP will increase by exports of $1,337 millions, now missing from the output of banks (see Table 1), and by the allocation of output of trust and mortgage companies to the government sector of $41 millions (see Table 4). Within the business sector, output and GDP of the financial intermediaries will increase by $2,785 millions, balanced by an increase of $1,378 millions ($1,337 millions of exports and $41 millions of government expenditure) in the final demand and by a reduction of GDP of $1,407 millions in the business sector, particularly in the rental industry (with a corresponding increase of $1,407 millions of intermediation consumption). GDP will also decrease by the allocation of output of credit unions to be shifted from the personal sector to the business sector, and by the amount of allocation of imports of FISIM to the business sector. Output and GDP of consumer loan companies will decrease by the amount of interest payments included in operating expenses.

6. Constant Prices

Detailed procedures for calculating constant price FISIM output are illustrated using Schedule A chartered banks. In 1989, Schedule A chartered banks had a FISIM output of $9,674 millions at current prices (see Table 1, line 7). In 1986, equivalent output of Schedule A chartered banks, amounted to $7,478 millions. Constant (1986) price output of 1989 is calculated using margin rates earned in 1986 multiplied by constant price (using consumer price index) assets and liabilities in 1989. We need to have a detailed breakdown of assets and liabilities in 1986 and 1989 and their corresponding margin rates in 1986. The consumer price index (CPI) is readily available and also uses 1986 as its base year. The CPI for 1989 was 114 with 100 for the base year 1986.

Statistics Canada - 106 - Collected Articles of Kishori Lal i) Assets

FISIM output at constant prices on account of assets, for 1989, for Schedule A chartered banks is separately calculated and shown in Table 6 where C$ means current prices and K$ means constant prices.

Margins earned for assets are the difference between the rates earned and the cost of borrowed funds. The cost of funds represents the "pure interest" cost of borrowed funds in the base year for the appropriate term.

Table 6 - 1989 FISIM Output, Constant (1986) Prices, Schedule A Chartered Banks, Assets

Assets Pure Average Annual Assets Balance FISIM Rate rate of Margin $ Millions Output Earned interest Earned 1986 1989 1989 $ Millions % % % C $ C $ K $ 1986 1989 C $ K $

MORTGAGES: - NHA Insured Residential 11.5 9.7 1.8 3,733 1,167 1,024 67 18 - Conventional Res. Mortgages 11.4 9.7 1.7 2,383 3,390 2,974 41 51 - Non-residential Mortgages 10.8 9.7 1.1 777 0 0 9 0

TO INDIVIDUALS FOR NON- BUSINESS PURPOSES: - To Purchase Canada Savings Bonds 10.4 8.5 1.9 663 54 47 13 1 - Personal Loan Plans 13.4 8.5 4.9 20,565 24,538 21,525 1,008 1,055 - Credit Cards 14.4 8.5 5.9 5,123 7,668 6,726 302 397 - Other Personal 11.2 8.5 2.7 6,425 17,587 15,427 173 417

OTHER LOANS: - Day Loans, Dealers 9.1 8.5 0.6 1,614 1,750 1,535 10 9 - Provincial Governments 8.7 8.5 0.2 468 124 109 1 0 - Municipal or School Corp. in Canada 10.7 8.5 2.2 1,108 990 868 24 19 - Associated Corporations 10.7 8.5 2.2 78 24 21 2 0 - Other Loans for Business Purposes 10.4 8.9 1.5 107,309 99,178 86,998 1,610 1,305

Total 3,259 3,272

The corresponding pure rate of interest for any particular asset category was chosen from the range of interest rates published by the banks. In 1986, this pure rate of interest ranged from 8.5% for non- mortgage loans to 9.7% for mortgage loans. This range is properly narrow, slightly higher for fixed term loans. However, the range of rates, earned by the banks was quite a bit wider, and should be so. The lowest rate earned was 8.7% from loans to provincial governments and the highest was 14.4% for credit cards. The margin earned thus ranged from 0.2 percent to 5.9% - the lowest from provincial governments to the highest on personal credit cards. It is essential to choose margin rates such that when multiplied by assets, the result approximates (does not need to be identical to) the actual (calculated as in Tables 1 and 2) output of FISIM on account of borrowers from banks in the

Statistics Canada - 107 - Collected Articles of Kishori Lal

base year. The choice of margin rates is based on professional judgement, helped by discussion with financial institutions.

ii) Liabilities

FISIM output at constant prices on account of liabilities for Schedule A chartered banks is calculated and shown in Table 7. On the liability side, the rate paid on various deposits is published by the banks. One must relate these rates with the cost of funds or pure interest estimated on the assets of the banks. The range of rates paid was quite wide, all the way from zero % to 9.3%. The interest rate paid on some demand deposits was zero, in 1989, whereas the rate paid on fixed term deposits was 9.3%. The margin earned on some demand deposits was highest at 8.5%, compared with the lowest of 0.4% for fixed term deposits.

Table 7 - 1989 FISIM Output, Constant (1986) Prices, Schedule A Chartered Banks, Liabilities

Liabilities Average Annual FISIM Pure Liabilities Balance Output Rate of Rate Margin $ Millions $ Millions Interest Paid Earned 1986 1989 1989 1986 1989 % % % C $ C $ K $ C $ K $

PERSONAL DEPOSITS: - Checking- Daily Interest 8.5 6.7 1.8 24,505 27,236 23,891 441 430 - Checking- Other Personal 8.5 2.6 5.9 5,214 4,427 3,883 308 229 - Non-Checking- Daily Interest 8.5 5.9 2.6 12,244 44,750 39,254 318 1,021 - Non-Checking- Other Personal 8.5 6.1 2.4 29,509 18,827 16,515 708 396 - Fixed Term 9.7 9.3 0.4 45,898 44,463 39,003 184 156

DEPOSITS BY BANKS: - Checking 8.5 2.0 6.5 60 126 111 4 7 - Non-Checking 8.5 7.6 0.9 424 129 113 4 1 - Fixed Term 9.7 9.3 0.4 19,701 9,957 8,734 79 35

OTHER NON-PERSONAL: - Checking- Provincial Governments 8.5 6.0 2.5 143 1,139 999 4 25 - Checking- Others 8.5 6.7 1.8 8,162 13,647 11,971 147 215 - Non-Checking- Provincial Governments 8.5 6.6 1.9 18 133 117 0 2 - Non-Checking - Others 8.5 5.9 2.6 2,749 3,337 2,927 71 76 - Other Fixed Term - Government of Canada 9.7 8.2 1.5 366 823 722 5 11 - Other Fixed Term - Others 9.7 8.4 1.3 31,691 20,348 17,849 412 232

INTEREST PAID ON DEMAND DEP. - Banks 8.5 3.7 4.8 1,969 1,335 1,171 95 56 - Government of Canada 8.5 6.6 1.9 2,500 642 563 48 11 - Provincial Governments 8.5 5.2 3.3 206 111 97 7 3 - Individual Personal Checking Accounts 8.5 1.6 6.9 15,114 18,713 16,415 1,043 1,133 - Others 8.5 0.0 8.5 2,955 4,083 3,582 251 304

Total 4,128 4,344

Statistics Canada - 108 - Collected Articles of Kishori Lal

Here, again, it is essential to choose (choice has to be judgmental) margin rates such that when multiplied by the liabilities of the banks, the result approximates (does not need to be identical to) the actual (as calculated in Tables 1 and 2) output of FISIM on account of deposits in the base year. Output calculations using margin rates with assets and liabilities are as follows:

1986 1989 Current $ Millions Constant $ Millions

On account of assets 3,259 3,272 On account of liabilities 4,128 4,344 ______Total 7,387 7,616 Growth 100.0 103.1

However, actual output for schedule A chartered banks in 1986, at current prices, amounted to $7,478 millions. Thus, output calculated on the basis of margin, $7,387 millions is 98.8% of the actual output. It is reassuring that the margin rates chosen in 1986 approximated the actual output in 1986, hence are useable for constant price calculations for other years (in our example, for 1989). Constant price output using margins grew by 3.1% in 1989, compared with 1986, and it is this movement which is used to project the actual CSNA output of banks.

Note that the actual CSNA output of schedule A chartered banks did not include the non-resident portion. Out of $9,674 millions, $1,337 was removed for non-resident portion, (see Table 1, line 7b) leaving an output of $8,337 millions in 1989. Add to this the outputs of $186 millions (of Bank of Canada) and of $183 millions (of the two government-owned financial institutions), total output amounted to $8,706 millions at current prices in 1989; its counterpart output for 1986 amounted to $6,211 millions. Multiplying $6,211 millions by the relative growth of margin-calculated output of 103.1%, we reach a constant price CSNA output of $6,403 in 1989.

Parallel calculations are made for schedule B banks. Output of schedule B banks in 1989 was $764 millions (see Table 1) at current price and it equalled $524 millions at constant prices. The two together are added to get constant price output of banks and other deposit-accepting institutions. Against the current price output of $9,470 millions in 1989 (see Table 2), the constant price output amounted to $6,927 (6,403+524) millions. iii) FISIM allocation to users, current and constant prices

1989 FISIM output of $9,470 millions at current prices and of $6,927 millions at constant prices was allocated to industries, persons and governments, and is shown in Table 8. The sector detail at current prices for business, persons, and government is the same as in Table 2. As the constant price output is derived from projecting each of the three separate aggregates for the business, persons and

Statistics Canada - 109 - Collected Articles of Kishori Lal government sectors, the implicit deflators come out differently for each sector, and are listed below, with 1986 = 100

Business sector 155 Personal sector (consumer expenditures) 122 Government sector (government expenditures) 157 Overall (Grand total) 137

The wide divergence amongst the implicit deflators for the three sectors in 1989 suggests that there is a wide divergence of current price growth rates of the margins in the sectors.

Table 8 - 1989 FISIM Allocation to Users, Current and Constant Prices, $ Millions

Current Prices Constant Prices

1. Agriculture 227 145 2. Fishing, Hunting & Trapping 9 5 3. Forestry 39 25 4. Mines, Quarries & Oils Wells 185 118 5. Manufacturing 768 489 6. Construction 483 307 7. Transportation & Storage 157 100 8. Communications 29 19 9. Elec Power, Gas, Other Utilities 39 25 10. Wholesale Trade 528 336 11. Retail Trade 293 186 12. Finance, Insurance, Real Estate 983 655 13. Community Bus, Personal Service 607 386 Sub-Total Business Sector 4,347 2,796

14. Consumer Expend., Services 4,785 3,916 15. Govt. Gross Current Expenditures 338 215 Sub-Total Persons and Governments 5,123 4,131

Grand-Total 9,470 6,927

Statistics Canada - 110 - Collected Articles of Kishori Lal

Concluding remarks

In this paper, we share our procedures with our colleagues at both national and international statistical organizations, for estimating and allocating FISIM to all users, both at current and constant prices, in the CSNA. We are fully aware of some of the weaknesses in our procedures and would welcome suggestions for improvements. In spite of the weaknesses, we believe that it is necessary to allocate FISIM to all users - business, persons and governments - as this portrays economic reality better than no allocation at all. Zero allocation of FISIM to users is a methodology which flies against reality. This is not the only area of SNA where procedures have not been internationally agreed upon. SNA practitioners must make compromises to portray economic reality as best they can, reflecting international guidelines, national institutions and national statistical developments. An estimate must be made for each item that appears in an accounting system. The System of National Accounts could not be established in practice if it were restricted to those components which were soundly based on accurately recorded or measured facts. Inevitably, published statistics will include estimates for which evidence is relatively weak. Sharing this information with others should improve the overall accuracy and relevance of SNA estimates.

Reference a. An earlier draft of this paper was presented at the Joint ECE-OECD Meeting on National Accounts, Geneva, 27-29 April 1994. This note was completed in May 1994. It was further elaborated in May 1996 and is issued as FISIM Revisited.

Statistics Canada - 111 - Collected Articles of Kishori Lal

FISIM Revisited - Calculating and Allocating FISIM: Canadian Approach a

by Kishori Lal, Director General System of National Accounts Statistics Canada

The following notes outline Statistics Canada's current practice as well as our plans for the future.

Current Practice at Statistics Canada

In the 1993 SNA, FISIM is calculated as the total property income receivable by financial intermediaries minus their interest payable, excluding the value of any property income receivable from the investment of own funds. Current practice in the Canadian System of National Accounts , for chartered banks, already follows the recommendations of 1993 SNA but income receivable from own funds is not excluded. The value of FISIM is calculated as the value of property income receivable by chartered banks less the interest payable by them. The total FISIM is divided between borrowers and depositors on the basis of the ratio between deposits and loans (including securities other than shares). For instance if deposits as a proportion of loans plus deposits is 52% then, 52% of FISIM is service to depositors and 48% to borrowers. Both deposit and loan FISIM are allocated by sector in proportion to stocks of deposits and loans.

The fundamental assumption in the split of FISIM between borrower and depositor is that each dollar of loan has the same FISIM as a dollar of deposit. In terms of the pure rate of interest, the assumption is that the pure rate lies midway between the average rate of return on loans and the average rate of return on deposits.

Adaptation of current methodology to the 1993 SNA recommendation

The main issue for us is the removal of the return on own funds from the FISIM calculation. As near as possible we have divided the balance sheet of chartered banks into the categories as outlined in Section 5 of the Eurostat paper and calculated some results over a 30 year period.

We have used the National Balance Sheet Accounts for chartered banks as our data source. At present these data are available annually and are year-end values. In the near future quarterly estimates will be available and will be used. Though we will continue to refine the estimates we feel confident that the process is viable and that the current results are reasonable.

In the Canadian case own funds exceed fixed assets and investments (case 2 in the Eurostat paper) in 30 of the 33 years calculated. From 1978 to 1980 own funds are less than fixed assets and investments (case 3).

Statistics Canada - 113 - Collected Articles of Kishori Lal

The following is a case 2 example of the actual situation in Canada in 1993.

Case 2 Own Funds > Fixed Assets and Investments

A: Fixed assets and investments C: Own Funds

$ 21 Billions $ 41 Billions

B: Securities other than shares, D: Intermediated funds and loans

$ 527 Billions $ 507 Billions

(Average rate of return RA = 8.0%) (Average rate of return RL = 5.0%)

Total assets $548 Billions Total liabilities $548 Billions

For ease of reference each box has been assigned a letter. The values correspond to the actual asset or liability for the relevant box. The average rates of return on the instruments in box B and D are shown.

Eurostat Approach

The following formula is suggested, for case 2, in the Eurostat paper:

FISIM = property income receivable other than interest + interest receivable - interest payable - property income receivable other than interest - that part of interest receivable which comes from the investment of own funds

It is suggested that the last term be calculated as:

(own funds-fixed assets and investments) x r where r is the reference rate.

Statistics Canada Approach

Statistics Canada proposes to use a different approach. Specifically, we will calculate the deduction for own funds differently. The difference lies in the choice of rate applied to own funds. Our choice

Statistics Canada - 114 - Collected Articles of Kishori Lal is to apply a rate calculated from the financial statements of the banks. We will apply the following formula for calculating own funds income. ( C - A ) x ( RAB + RLD ) ) ( B + D ) where RA is rate of return on securities other than shares and loans (assets in box B) and RL is the rate of return on deposits (liabilities in box D). As long as B and D are close in value the rate calculated for own funds will be close to the simple average of the two rates shown in the diagram. The procedure appeals to us because it implies FISIM to both borrowers and depositors and all calculations are easily derived from the financial statements.

Using the figures in the diagram we calculate below FISIM, both following current practice and after adopting the new guidelines.

Current Formulation

FISIM = ( RAB - RLD )

= ( .08 x 527 ) - ( .05 x 507 ) = 16.81 billions

Proposed formulation

FISIM = RAB - RLD - ( C - A ) x ( RAB + RLD ) ) ( B + D )

= ( .08 x 527 ) - ( .05 x 507 ) - ( 20 ) x ( .08 x 527 + .05 x 507) ) (527 + 507)

= 16.81 - 1.31 = 15.50

Implications of the Canadian proposed formulation on the 1993 SNA

The 1993 SNA definition, as noted above, will have to be modified as follows: FISIM is calculated as the total property income receivable by financial intermediaries minus their interest payable excluding the value of any pure interest receivable from the investment of own funds. This change is required following all the three documents noted in Reference at the end. It will be a good idea to rename such financial services as Financial Services Indirectly Measured or FSIM and not FISIM. Intermediation is not a necessary attribute for such services.

Once the 1993 SNA recognises that the economic activity associated with lending own funds is not be excluded, then the issue of the boundary of money lending needs to be settled. In the Canadian practice, we include only those non-deposit-accepting entities which have employees, incur intermediate expenditures but lend mostly or entirely out of their own funds. These are typically consumer finance or household finance companies. We do not include those money lenders for whom the money lending is only a side business and not their principal activity.

Statistics Canada - 115 - Collected Articles of Kishori Lal

Another issue raised by Peter Hill is that property income from securities should be excluded from FISIM. Suppose a financial institution accepts deposits and invests these deposits into securities rather than loans. It has employees and incurs intermediate expenditures to handle its securities portfolio. This activity needs to be valued and allocated. It may not be FISIM but it is FSIM.

Distribution of FISIM

As already mentioned, FISIM will be allocated to borrowers as 527/(527+507) or 51% and 49% to depositors. The necessary data exist for the export calculation and we are still examining data availability for imports. The FISIM will be allocated to sector and industry as per current practice.

Constant Price Estimates

Statistics Canada has used the method, similar to trade margins, for constant price estimates, for some years now. This is similar to the Australian proposal.

The Central Bank

In Canada the central bank called the Bank of Canada serves four broad functions:

C formulate and implement monetary policy C issue and replace bank notes C central banking services C managing the public debt

It is only those activities related to item 3 central banking services that could generate FISIM. The bank does hold deposits and make advances as part of its banking activities but this activity and any associated FISIM would be small. On an operating cost basis, it is estimated that central bank services accounted for 9% of expenses while monetary policy accounted for 16% of costs, bank notes for 48%, and debt management for 27%.

It seems to us more appropriate to measure the output of the central bank as the sum of expenses. The central bank will continue to be classified to the business sector with the Financial Institutions. The entire output of the central bank will be purchased by the central government. The January 1996 SNA News and Notes issued by the ISWGNA clarifies this issue.

Imports

From the Canadian Balance of Payments capital account, we have information on Canadian deposit assets abroad (for chartered bank and non-bank institutions) and loan liabilities of Canadians, outside of Canada. Likewise, from the current account, information is available regarding loan interest payments to foreign institutions and deposit interest receipts from Canadian deposits abroad. We are considering using the same rate for import estimates, as for domestic calculations. For example, the average domestic FISIM rate for 1993 is 15.50 ) (527 + 507) = 1.5%. We will apply this rate to the

Statistics Canada - 116 - Collected Articles of Kishori Lal total of deposit assets abroad plus our loan liabilities. This calculation will provide us an estimate of FISIM import and will be allocated to users in the same fashion as domestic FISIM.

Concluding Remarks

The Canadian approach, mostly Canadian practice, is very much in line with the thrust of the approach and recommendations of both Peter Hill and the Australian paper. It is imperative , as also urged by Australia, that allocation of Financial Services Indirectly Measured be adopted. It will help international comparisons of output, particularly the Financial sector in all economies.

Calculating FISIM by reference to detail is all right, yet we have to measure the total output of a financial institution including: FISIM, directly charged services, non-FISIM services indirectly measured such as handling of loans out of own funds and handling of securities portfolio etc.

Unlike in the Eurostat methodology, FISIM rate is never negative in our calculation.

FISIM in our calculation is unambiguous and not subject to manipulation, an important characteristic for an international standard.

Reference a This note ,finalised late in May 1996 takes into account: the Eurostat document, Calculating and Allocating FISIM: Presentation of a Method, presented at the Joint ECE/Eurostat/OECD Meeting on National Accounts, Geneva, April 30 - May 3, 1996; Peter Hill's paper, The Services of Financial Intermediaries or FISIM Revisited, presented at the same meeting and the most recent Australian paper, FISIM : An Elaboration of Earlier Australian Proposals, revised since the above meeting.

Statistics Canada - 117 - Collected Articles of Kishori Lal

Recording of Change in Inventories in the SNA and in the Business accounts- A case study of Canadian practices a

by Kishori Lal System of National Accounts Statistics Canada

Introduction

In the System of National Accounts (SNA), intermediate inputs are recorded and valued at the time they enter the production process, while outputs are recorded and valued as they emerge from the process. The SNA's rule that inputs must be valued at the prices current at the time they are consumed , and outputs at the prices when they are produced , is equivalent to valuing the goods in question as if they never spent any time in inventories, thereby ensuring that the values of intermediate consumption and output do not include any holding gains. Typically in the business accounts, intermediate inputs are recorded and valued at the time of their purchase, while outputs are recorded and valued as they are sold. Thus the values of inputs used in production and of outputs produced in the SNA have to be derived by adjusting the corresponding purchases and sales for changes in inventories as reported in the business accounts.

The purpose of the adjustment is not simply to derive the correct quantities but, also, to ensure that the inputs and outputs are correctly valued. The SNA's valuation rules require both entries to , and withdrawals from, inventories to be valued at the prices prevailing at times when they take place. When a good is put into inventory at one price and subsequently withdrawn at a higher price , the values of the acquisition and the disposal do not cancel out, so that the monetary value of the change in inventories is not zero, even though the quantity change may be zero. When a good is withdrawn from inventory at a higher price than at which it entered, a nominal holding gain accrues to the owner of the enterprise equal to the difference between the two prices. This holding gain cannot be part of value added as it does not result from productive activity.

Thus the first task of the national accountant is to examine the business account of an enterprise with respect to: the reported value of inventories in the opening and closing balance sheets, the basis of valuation, the turnover period of inventories ( the duration between the entry and the withdrawal points), and of course the price statistics at the entry and withdrawal points. In the business accounts, the reported book value of change of inventories equals the value of inventories in the closing balance sheet less the value of inventories in the opening balance sheet of an enterprise. In the SNA, this reported book value change needs to be split into a value of (volume/quantity) change in inventories and the holding gains. The current price value of change in inventories and the nominal holding gains at period t with given volume/quantities (q) and prices (p) at period t and t-1 are formulated in the 1993 SNA as follows:

Statistics Canada - 119 - Collected Articles of Kishori Lal

Change in inventories = ( qt - qt-1) pt Nominal holding gains = ( pt- pt-1) qt-1

In the business accounts of an enterprise, only the aggregate value of inventories in the opening and closing balance sheet is usually recorded rather than values or quantities of individual commodities. In such a situation, volume must first be estimated using price indices; only then a calculation of change in inventories and nominal holding gains can be made. In the Canadian SNA, change in inventories is called Value of Physical Change in Inventories (VPC) and nominal holding gains are called Inventory Valuation Adjustment (IVA). Canadian Practices

In the Canadian System of National Accounts, inventory data are estimated for the government sector and for the farm and non-farm business sectors. Inventories of natural resources (e.g. ores in the ground, uncut trees) are not included. Only inventories owned and located within Canada are included.

Government inventories are a very small part of total inventory holdings. They include inventories held by government marketing agencies such as the Canadian Dairy Commission. All inventory changes recorded in the Accounts relate only to the federal government. Estimates are based on the accounting records of the agencies.

Farm inventories are calculated separately from non-farm business inventories. Farm-held inventories and grain in commercial channels are estimated by crop and by livestock using quantities obtained from monthly surveys of farms. These are valued using the market prices prevailing during the reference period.

Non-farm business inventories represent by far the largest part of total inventory holdings in the economy. They include all inventories of raw materials, goods-in-process, and finished products, including goods purchased for resale, held by corporations, non-farm unincorporated businesses, and government business enterprises. By industry, the major part of non-farm business inventory stocks is held in manufacturing and in wholesale and retail trade. Non-farm business inventories are estimated mainly from book value data reported by businesses in various Statistics Canada surveys.

Inventories include:

1- Finished goods available for sale by Manufacturers, Wholesalers, Retailers or other industries. Such goods may have been produced by the industry or purchased for resale.

2- Raw materials, i.e. goods intended to be used for intermediate consumption.

3- Goods-in-process, i.e. partially finished goods which must be processed further before they can be used.

Statistics Canada - 120 - Collected Articles of Kishori Lal

4- Construction >work put in place= is classed as capital investment. Consideration is being given to putting >speculative= construction (i.e. construction done before a purchaser has been found) into inventories in order to conform with the 1993 SNA.

5- All gold bullion, including that owned by persons, is supposed to be included in inventories since such gold is re-sellable for use in production and estimated using supply disposition method. Compared to other inventories, the data base for inventory of gold bullion is weaker and hardly involves linking business accounts to national accounts.

Farm inventories

The value of physical change for farm inventories is estimated in three parts: farm-held grains, other farm-held inventories (livestock, tobacco and potatoes) and grain in commercial channels.

The farm-held grains component is estimated by calculating additions to and withdrawals from inventory separately for each of eight crops: wheat, oats, barley, rye, flaxseed, canola, corn and soybeans. The estimates are derived using raw data on physical quantities, which are valued in both current period and base period prices. Additions to grain inventories represent the value of the harvest. They are zero in the first and second quarters and positive in the third and fourth quarters. Withdrawals represent sales from inventory plus amounts accounted for as feed, seed and wastage and occur in a relatively steady stream through the course of the year.

The other component of farm-held inventories comprises potatoes and special crops (lentils, mustard, sunflower, etc.) and livestock and tobacco. Potatoes and special crops are treated in a similar manner to farm-held grains. But for livestock and tobacco, the VPC is derived by valuing changes in stocks directly; additions and withdrawals are not treated separately.

Finally, the grain in commercial channels component is calculated by valuing the change in the physical grain holdings of the Canadian Wheat Board and the grain that is held privately by commercial dealers.

Additions to farm-held grain inventory (the harvest) are seasonally adjusted. Before 1989, the quarterly grain production was seasonally adjusted by dividing the annual harvest value by four and spreading it equally over the four quarters of the year. The main virtue of this procedure was its simplicity. Its main deficiency was that it concentrated the entire transition from one year's crop to the next between the fourth and the first quarters of the two years. For periods with large swings in the size of the crop, the January and the first quarter real GDP growths were seriously under- or overstated. At the same time, variations in the crop size had no influence on the other eleven months' or three quarters' real GDP growth rates. A new approach to the seasonal adjustment problem was implemented with the first quarter of 1989. In this new approach, the transition between annual crops is smoothed over a number of quarters instead of concentrating the entire transition in the first quarter. The seasonally adjusted monthly and quarterly values for grain production are now calculated using a quadratic minimization technique. A more smoothly evolving series of quarterly production values is thereby derived which sums to the given annual crop while minimizing squared

Statistics Canada - 121 - Collected Articles of Kishori Lal quarter to quarter changes. For more details, refer to Technical Paper on the Treatment of Grain Production in the Quarterly Income and Expenditure Accounts, Technical series#2 , published by National Accounts and Environment Division , Statistics Canada, 1989.

The sales portion of withdrawals from farm-held grain inventory is seasonally adjusted using Statistics Canada=s standard X11-ARIMA method. The remainder (feed, seed and wastage) is distributed over the four quarters using a quadratic minimization technique. Seasonally adjusted estimates of the VPC for other farm-held inventory and grain in commercial channels are derived using the standard X11-ARIMA method. X-11-ARIMA

X-11-ARIMA method of seasonal adjustment was developed at Statistics Canada by E.B.Dagum and published as The X-11-ARIMA Seasonal Adjustment Method, Ottawa, Statistics Canada, 1980, Catalogue NO. 12-564. Typically, the seasonally adjusted series are obtained by using some type of moving averages. A limitation inherent in any seasonal adjustment procedure based on moving averages is that the last few observations cannot be smoothed with the same set of symmetric weights (moving averages) applied to central observations . Because of this , the estimates for current observations must be revised as more data are added to the original series. It is desirable to have a method which yields estimates with maximum reliability which is equivalent to estimates with revisions at a minimum to avoid confusing the users of the data. It is the criterion of minimum revisions which led to the introduction of ARIMA extrapolation when seasonally adjusting series. By adding a year of extrapolated data obtained from ARIMA ( Auto-Regressive Integrated Moving Average) models fitted to the series, the current observation becomes more central and thus the seasonally adjusted estimate undergoes smaller revisions.

Note that sales/production, operating surplus and change in inventories for the farming sector of the Canadian economy are calculated using physical volume of crops and livestock: thus no IVA needs to be calculated. The only IVA calculated is for private grain dealers whose profits are reported rather than derived.

Inventories for Manufacturing, Wholesale and Retail industries

Inventory book value data and shipment and sales data for the manufacturing, wholesale and retail industries, are collected by monthly sample surveys. For the retail industry, inventory data are collected only for large or multi-location establishments and are expanded to the complete industry based on sales data. The data from the monthly surveys are bench-marked to the annual surveys which are more extensive and more detailed.

The surveys are at the >establishment= level. An establishment is defined by Statistics Canada as the smallest operating entity which produces as homogeneous a set of goods or services as possible and from which data can be provided on the value of output together with the cost of materials used and quantity of labour resources employed to produce the output.

Statistics Canada - 122 - Collected Articles of Kishori Lal

Financial data, including inventory book values, are reported to Statistics Canada as they would appear on the financial statements of the establishments. The common methods of reporting inventories are:

1. Specific item cost - the actual cost of each item is ascertained separately.

2. FIFO (first in- first out) - the cost of items sold or consumed during a period is computed as though the items were sold or consumed in the order of their acquisition.

3. Average cost - the cost of an item is determined by applying a weighted average of the cost of all similar items available for sale over a period of time.

4. LIFO (last in- first out) - the cost of items sold or consumed during a period is deemed to be the cost of the most recent acquisitions or production.

The method of accounting is not requested on survey forms. An extensive survey of manufacturers in 1975 showed 35% using the FIFO method and 31% using >average cost method. A small survey in 1990 of wholesalers and retailers showed 68% using the >specific item cost= method which was consistent with increased use of computerized inventory control. An annual survey (Financial Reporting in Canada) by the Canadian Institute of Chartered Accountants showed in 1994 that 44% of companies used FIFO, 36% average cost and less than 4% LIFO.

However, for calculating VPC and IVA, the FIFO method is presumed at Statistics Canada.

For some industries (other than manufacturing, wholesale and retail) , sub-annual survey data are not available. In some cases (e.g. logging, gold), changes in inventories are estimated as the difference between supply (production plus imports) and disposition (domestic use plus exports). In other cases (e.g. construction; financial, insurance and real estate), inventory levels are estimated from indicator series such as employment data or sales data.

Prices

The deflation and revaluation of inventory data are done using a variety of prices. For example, industrial selling price indexes are used for manufacturing, industrial selling prices and import prices are used for wholesaling, consumer price indexes are used for retailing, raw material price indexes for logging and mining and other specialized indexes for other industries. These price index series are all monthly and on 1986 base year at present. The base year for constant price calculation will be changed to 1992 with the release of the CSNA historical series in December 1997.

Most of these index series are created with weights based on sales or shipments; however, what is ideally required are indices with weights based on inventory levels. Efforts are made to reweight combinations of the price indexes using weights based on inventory levels wherever possible.

Statistics Canada - 123 - Collected Articles of Kishori Lal

The price series are adjusted when necessary for changes in tax rates or tax structures. The CPI series are adjusted for changes in the various federal and provincial sales taxes. A large change affecting most levels of inventories occurred at the beginning of 1991 when the Federal Sales Tax, a manufacturing sales tax which was included in the cost of inventory, was replaced by the Goods and Services Tax, an end-user value added tax which does not appear in the cost of inventory. This change created a >spike= in the IVA series as the effective prices fell for many series.

The calculation at Statistics Canada

The VPC and IVA are calculated using an eight-line calculation:

1) B book values as reported 2) D deflator Indices to value at base year prices 3) K = B / D constant dollar values at base year prices 4) DK$ = Kt - Kt-1 constant dollar Value of Physical Change 5) R revaluer indices to value at current period prices 6) VPC = R * DK$ current dollar Value of Physical Change 7) CB = Bt - Bt-1 change in reported book values 8) IVA = CB-VPC Inventory Valuation Adjustment

Although the Accounts data are published quarterly, these calculations are performed monthly for the manufacturing, wholesale and retail industries and accumulated to quarters. The working level for the calculations includes about 130 manufacturing industries by stage of fabrication, ten trade groups for wholesale, eighteen trade groups for retail, six >other= industries and gold.

Deflators

Inventory book values are presumed, as noted above, to be reported on a FIFO (First In- First Out) basis. Therefore, each inventory commodity has a >turnover period= which is the average length of time that a good spends in inventory (or, equivalently, the number of days or weeks of inventory that an establishment keeps on hand). The inventory book value reported by an establishment for a given reference period, is the sum of the costs of goods entering inventory during the turnover period of the reference period. To deflate the book value to base year constant dollars, the deflator price index must reflect price movements during the turnover period. Therefore, the deflator is calculated as the average of a price index series during the turnover period at the end of the reference period.

For a commodity with a 3 month turnover period, the 4th quarter deflator would be the sum of the price index for December, November and October, divided by 3. For turnover periods of less than one month, the deflator series equals the price series.

The turnover periods used are constant with the exception of those for retail inventories which are varying, based on stock to sales ratios.

Statistics Canada - 124 - Collected Articles of Kishori Lal

Revaluers

The price index series used to revalue the VPC from constant dollar values to current dollar values reflects the average costs of the inventories during a period. For monthly calculations, the revaluers are the reference price series themselves. For quarterly calculations, a simple average over the quarter is used.

At aggregate levels, the implicit revaluer (the ratio of the current and constant dollar VPC) may bear little resemblance to a proper price series. In an extreme case, two series may have equal but opposite constant dollar movements and the implicit revaluer of their sum would not be defined.

In particular, this is true when aggregating monthly data to quarterly levels and quarterly data to annual levels. The implication is that quarterly VPC, as the sum of months, is different from VPC calculated quarterly and, similarly, for annual VPC. Since the lower level data better reflect the variances of price over time, published annual data are the sum of quarterly data and, where applicable, quarterly data are the sum of monthly data. However, the annual data in the input-output tables do not follow this method , hence a discrepancy exists between the VPC in the input-output tables and that in the accounts published at quarterly frequency.

It should be noted that many monthly indexes are not averages over the month but, rather, spot prices. Usually this is not relevant, but cases do occur in which extraordinary price movements occur within one month. In these cases, extra care is required for deflating and revaluing. Where possible, the reported book value, the deflators and the revaluers are seasonally adjusted using Statistics Canada's X-11 ARIMA seasonal adjustment program The seasonally adjusted VPC and IVA series are derived from these.

Statistics Canada calculation versus the 1993 SNA calculation

For a single commodity , the eight step calculation of Statistics Canada gives an identical value of change in inventories and holding gains compared with the 1993 SNA calculation. We will show this later in the Table . Quantities are not usually available and their volume needs to be calculated using steps 2, 3, and 4. Then one needs revaluer, step 5, which has the same value as deflator for a single commodity. Other steps follow the 1993 SNA .

Rebasing

Constant dollar data are now published using 1986 as the base year. However, constant dollar data on a 1992 base year will be available in December 1997. Current policy is to rebase the Accounts about every five years.

When converting to a new base year, data points prior to the new base year are linked so as to preserve growth rates. When the base year was changed to 1986 from 1981, the constraints for inventory data prior to 1986 were that:

Statistics Canada - 125 - Collected Articles of Kishori Lal

1) current dollar values remain unchanged.

2) the relative movements of the constant dollar VPC series remain unchanged.

Since this calculation was done at the detailed level, the components no longer summed to the aggregates, the traditional non-additivity problem in chain linked constant price series. The differences were entered in the Accounts as adjusting entries, a treatment used for most series in the Accounts.

In the new base year, the annual sums of the constant dollar VPC series were forced to equal the annual sum of the corresponding current dollar VPC.

Annual Input- Output Tables

In the Canadian annual input-output tables, VPC is calculated for approximately 200 industries and 464 commodities. An assumption is made that inventories are turned over every 3 months. Reported inventories thus are assumed to reflect the price level of the last 3 months of the previous year and the reported closing inventories reflect that of the last 3 months of the current year. Reported values of inventories are distinguished by finished goods, goods- in- process and raw materials but no commodity detail is readily available. It is then assumed that the commodity content of each type of inventories of an establishment bears a close relationship to its corresponding shipments or purchase of raw materials as the case may be. Reported opening inventories are revalued using weighted price indices of the reported closing inventories. The same weighted price indices are used to revalue the closing inventories. The difference between the revalued reported opening inventories and the revalued closing inventories is the measure of VPC. The difference between the nominal change in inventories and the VPC is the calculated IVA. VPC and IVA calculated using annual series are not exactly the same as the cumulative annual VPC and IVA using quarterly accounts. The differences are not big because the inflation rate has been quite low in Canada. We do not force the two series to be identical and both are published.

Data detail

VPC, at both current and constant prices, and IVA series are prepared, quarterly, seasonally adjusted and unadjusted, and of course annually, for:

Manufacturing: 23 industries detail

Wholesale: 5 way detail -non-durables; machinery & equipment; building materials; motor vehicles; and other durables

Retail: 3 way detail- non-durables; motor vehicles; and other durables

Other industries: logging; mining; utilities; transportation and communications; construction; financial, insurance, real estate; and gold

Statistics Canada - 126 - Collected Articles of Kishori Lal

Government: As noted above, only inventories calculated are for the Federal Government.

In Canada, an aggregate of holding gains ( IVA) is published in the income based Gross Domestic Product as an adjustment to profits. Typically in an economy with a rise in prices, holding gains are positive, hence a negative adjustment is made to remove the effect of such gains from the reported profits in the business accounts to bring them in line with the SNA concept. An identical adjustment is made to the reported book value change in inventories in the business account to value the change in inventories in the expenditure based GDP.

VPC is also published by industry at the following detail: agriculture; forestry; mines, quarries, and oil wells; manufacturing; construction; transportation, storage and communications; electric power, gas and water utilities; grain in commercial channels; wholesale trade; retail trade; finance, community, business and personal services; and government.

In the annual input-output Tables, changes in inventories are calculated by industry- more than 200 industries- and are published by commodity- about 500 commodities-under two headings- finished goods inventories and raw materials inventories.

The following table, Change in Inventories in Canada 1989-1992 $ millions, provides book value inventories as reported in the business accounts of enterprises. The second column provides the SNA calculation of holding gains. The third column, the SNA concept of value of ( physical) change in inventories is calculated by deducting the calculated holding gains from the reported balance sheet book value change in inventories. All this detail is provided at quarterly frequency in the Canadian System of National accounts. The input-output tables are only produced annually. There is a difference in value between the input-output calculation and the other accounts. However, this difference is not significant given the value of gross flows of inventories.

Statistics Canada - 127 - Collected Articles of Kishori Lal

Change in Inventories in Canada, 1989-1992, $ Millions

Quarterly Accounts Input- at Annual Rates Output Tables

Inventory Valuation Value of Value of Adjustment/ Physical Physical GDP Book Value Holding Change in Change in Implicit Change Gains Inventories Inventories Price Index (1) (2) (3) = ( 1 - 2) (4) 1986 = 100 (5)

1989 - 1st Quarter 4744 4940 -196 112.7

1989 - 2nd Quarter 8984 3108 5876 114.6

1989 - 3rd Quarter 7192 -584 7776 115.7

1989 - 4th Quarter -172 -1144 972 116.5

1989 - Annual 5187 1580 3607 3806 114.9

1990 - 1st Quarter 1768 3944 -2176 117.1

1990 - 2nd Quarter 316 140 176 118.1

1990 - 3rd Quarter -1472 2916 -4388 118.8

1990 - 4th Quarter -10424 -5472 -4952 119.8

1990 - Annual -2453 382 -2835 - 4079 118.5

1991 - 1st Quarter -744 -1340 596 121.3

1991 - 2nd Quarter -9660 -2692 -6968 122.0

1991 - 3rd Quarter -7612 -3400 -4212 122.1

1991 - 4th Quarter -2744 -388 -2356 122.1

1991 - Annual -5190 -1955 -3235 -4415 121.9

1992 - 1st Quarter -2208 2480 -4688 122.5

1992 - 2nd Quarter 1164 3048 -1884 122.9

1992 - 3rd Quarter -1180 1812 -2992 123.9

1992 - 4th Quarter -2356 2884 -5240 124.3

1992 - Annual -1145 2556 -3701 -5583 123.4 Government industries, contributing a very small share of total inventories, are included.

Statistics Canada - 128 - Collected Articles of Kishori Lal

Reliability

The inventory data are considered one of the less steady series in the Accounts. By their very nature, inventory data are volatile. They represent values calculated at certain fixed times and do not enjoy the smoothing acquired by other series which are summed over time.

Although inventory surveys are usually twinned to sales or shipments surveys so that the same level of coverage is attained, the fact remains that inventory levels are not tracked as carefully as sales by many companies, particularly small ones. The response rate for reporting inventories is much lower than for sales, resulting in larger blow-up factors and sampling errors and more heroic imputations for missing data. For small retail stores, only sales data are available.

Nevertheless, data based on surveys are undoubtedly more reliable and more detailed than indirectly derived values. However, indirectly derived values using supply and disposition models fill an important gap when directly related survey data are not available. Such calculations also provide an additional information to assess the reliability of inventory estimates. The accumulated errors from adding and subtracting many large series, together with the many assumptions required to fill gaps in the data availability, leave a high level of uncertainty in the residual representing the change in inventory level.

Change in inventories under conditions of high inflation

Canada is not a high inflation economy but many countries in the world are. Thus it may be useful to examine how the value of change in inventories should be calculated under conditions of high inflation. Calculation of value of change in inventories and nominal holding gains under conditions of high inflation can be best understood by referring to an example from part D of Table 6.1 ( page 79) of a recent OECD manual, Inflation Accounting A Manual on National Accounting under conditions of high inflation (OECD 1996), called Inflation Manual in this paper. We have expanded this table to demonstrate that the Canadian 8 step calculation also generates the same results. Canadian VPC in Step 6 is identical to the Change in inventories in row 6 of the SNA method. Canadian IVA in step 8 is identical to row 7 of the SNA method.

A few comments on the information in this table may be in order to better understand the calculations. The balance sheet values are at market prices. Quantities at the 0 quarter ( last year's 4th quarter) were 100 and after going down for 2 quarters, were up in quarters 3 and 4, resulting again in 100 units at the end of the 4th quarter. This is an example of an economy under severe inflation. Prices are rising from 0.4 per unit to 2.5 per unit in a year or from 100 to 625. Changes in value in the closing inventories minus opening inventories for each quarter are calculated. The total for the 4 quarters of the year is 210, exactly equal to 250 minus 40, as reported in the Book Values for the year. Quantity declines by 40 in each of the first 2 quarters and then increases by 40 in each of the last 2 quarters, resulting in a zero change for the whole year. The balance sheet change of 210 is made up of 98 of change in inventories (of course, in current prices) and 112 of nominal holding gains.

Statistics Canada - 129 - Collected Articles of Kishori Lal

The Canadian method gives the same results, as in the SNA method, for VPC and IVA. The Canadian method , however, does not require data on quantity as such information is usually not available from the surveys. Volume at base year prices is calculated. Change in volume thus calculated multiplied by the revaluer , which is the same as deflator, provides us the values for VPC ( Change in Inventories). Similarly, IVA follows the same principles as the SNA method, and the results are identical.

Comment on results

The results look plausible from the point of view of a business account. The nominal holding gains of 112 are consistent with the cash flow of the enterprise: using the SNA method data, it spent 40 dollars to acquire 100 units in period 0, sold 40 units in Q1 for 20, sold another 40 units in Q2 for 32, purchased 40 units in Q3 for 50, and purchased 40 units in Q4 for 100, with a total net expenditure of 138. The market value of 100 units in Q4 is 250, hence a nominal holding gain of 112 ( 250-138).

Change in balance sheet value minus holding gains equals change in inventories of 98. However, change in inventories at constant prices or change in quantity of inventories for the whole year is zero. Implicit deflator- current price change in inventories divided by constant price change in inventories- is not interpretable for the whole year. However, the implicit deflator for each quarter behaves as expected. Aggregate data for the year seem to provide a different interpretation than the data from the 4 quarters. Where is the problem?

The problem arises because the underlying principles of aggregation are not satisfied in an inflationary situation. The OECD Inflation Manual very wisely notes: "The economic theory underlying the aggregation of the values of different kinds of goods and services is that relative prices should reflect both relative costs of production and relative utilities to users, whether producers or consumers. Market forces may be expected to ensure that relative prices do not diverge very much from these underlying ratios at any given point of time. When there is high inflation, however, the ratio of a given good or service at a later point of time to its price earlier in the same accounting period may simply reflect the general rate of inflation and have nothing to do with relative prices" ( pp 31-32). The Inflation Manual recommends recasting the sub-period ( say quarterly) values of a given accounting period ( say annual), such that the relative weights of the sub-periods are not distorted by inflation: such recasting is called Constant Price Level (CPL) accounts. If a CPL type account is done for change in inventories, the holding gains within the year will disappear. However, the holding gains need to be reflected in the balance sheet as they have occurred in the market. The OECD Inflation Manual does not seem to provide any explicit guideline for handling this problem. It could be a solution to add a reconciliation item in the other changes in assets account, this item being the difference between the valuation of changes in inventories as calculated by CPL accounts and as calculated in the above Table.

Statistics Canada - 130 - Collected Articles of Kishori Lal

Table-Recording Change in inventories and holding gains

Quarters 0 1 2 3 4 Year

1993 SNA Method

Prices 0.4 0.5 0.8 1.25 2.5 Quantities 100 60 20 60 100 Book Values 40 30 16 75 250 Change in Balance Sheet -10 -14 59 175 210 Change in quantity -40 -40 40 40 0 Change in inventory = ( qt - qt-1) pt -20 -32 50 100 98 Nominal holding gain =( pt- pt-1)qt-1 10 18 9 75 112

Canadian Method

Book value 40 30 16 75 250 Deflators-Indices 100 125 200 312.5 625 Constant Dollar Value 40 24 8 24 40 Change in Constant $ Value -16 -16 16 16 0 Revaluers- Indices 100 125 200 312.5 625 Current Price VPC -20 -32 50 100 98 Change in Balance Sheet -10 -14 59 175 210 IVA 10 18 9 75 112

Concluding remarks

This paper describes in a summary fashion the Canadian practice for calculating change in inventories and holding gains and its implementation in the Canadian System of National Accounts; additional detail is always kept in our worksheets. As noted , the reliability of the data base of reported book values and our estimates of inventory changes and holding gains is not of the same high standard as are many other components of the Canadian accounts. It is a learning process for all of us. We strongly urge, nevertheless, that every country remove holding gains from both reported profits and book value inventories: otherwise the calculated GDP will remain deficient both conceptually and statistically.

Statistics Canada - 131 - Collected Articles of Kishori Lal

Reference b. This paper, finalised in September 1997, has extensively incorporated the details on Canadian practices in inventories from the excellent notes prepared by John Butterill of the National Accounts and Environment Division of Statistics Canada and very useful and thoughtful comments by Abe Tarasofsky of the System of National Accounts Branch and to both I am very grateful.

This paper is to form part of the UN Handbook on National Accounting Series F No 76: Links between Business Accounting and National Accounting

Statistics Canada - 132 - Collected Articles of Kishori Lal

A Note on the Measurement of Industrial Production a

by Kishori Lal System of National Accounts Statistics Canada

Measurement of Industrial Production has been there for a long time, much longer than the more widely known and reported measure of Gross Domestic Product (GDP) of the System of National Accounts. Industrial Production is regularly published by the Organisation for Economic Co- operation and Development (OECD) as one of the main economic indicators. The world famous weekly British magazine, The Economist publishes, in every issue, industrial production of not only the advanced industrial countries but of all the 25 emerging economies of the world. The emerging countries list includes India, China , Russia , Taiwan, Turkey and others. There is no doubt about the importance attached to this concept and its measurement. Given its age and its importance, one would expect that this concept is well defined and that it follows internationally accepted guidelines. This note would try to test this hypothesis.

Boundary of Industrial Production

Industrial Production includes the following three divisions of Industrial Classification in Canada: Mining, Manufacturing and Utilities. Most ( but not all) member countries of the OECD include the same set of industrial divisions. Utilities such as Electric Power and Gas Distribution are typically the ones ignored in some countries. For any inter-country on even inter-provincial comparison, the first check needs to be made regarding the units included underlying the concept at hand.

After one has assured oneself about the units, the next big question relates to the meaning of production. Does production mean gross output like cars in a car factory or value added in the production of cars ? There is no international agreement on this point, not even among the OECD member countries. Note that where the value added is an unduplicated concept, the gross output includes duplication. As well, the amount of duplication depends upon how the industrial structure in a given jurisdiction is integrated. The more the integrated , the less the gross output, the limit being full integration or full consolidation which would yield gross output equal to value added only. This is a bit complicated and would require further elaboration later.

Industrial Production is a volume concept, not a current price concept . There is even less agreement at the international level regarding the precise formulation of constant price Industrial Production. Industrial Production at constant prices and the Index of Industrial Production will be different ( and the differences would not always be small) depending upon the base year one chooses, how often one changes the base year, and the base weights one chooses. For example, one might calculate industrial production in 1998 over 1997 using 1997 weights or 1992 weights or even 1981 weights. Each of these formulations would provide different growth rates for 1998 over 1997. Are the weights derived from gross output or value added and this again would result in differences in the 1998

Statistics Canada - 133 - Collected Articles of Kishori Lal

growth rate. Is the statistical agency using a single deflator regime or the preferred double deflation technique?

Units included in Industrial Production

As we noted above, the three divisions of Standard Industrial Classification-Mining, Manufacturing and Utilities- are the norm for the industrial production statistics. The underlying detail at three or four digit industries forming these divisions will always be different amongst the various countries. It is the industrial detail which is really the more interesting aspect of this phenomenon. In Canada, we estimate for about 160 separate industries for industrial production and aggregate them into 100 industries for public release, all at monthly frequency. For inter-temporal comparisons, one must have a consistent definition for each and all aggregations of industries for the entire period under consideration. We have just released the above detail for the period 1961 to date and of course, at the monthly, quarterly and annual detail. It was quite a time consuming exercise to diligently classify the detailed industries from the various vintages of Standard Industrial Classifications used over the last four decades in Canada.

Definition of Production

Production for any aggregation is not a well defined term as it could include any level of duplication. Let us assume that there are two enterprises, one in Delhi and the other in Calcutta, and both produce cars. The two enterprises are identical in technology and scale. But the enterprise in Calcutta is organised in 2 establishments - one producing parts and the other producing cars using parts from unit 1. The enterprise in Delhi has both these units combined into one; other than that , there is no difference. The establishment producing parts in Calcutta buys Rs 300 millions of raw materials to produce parts which it sells to the car producing establishment for Rs 700 millions, thus adding value of Rs 400 millions. The car producing establishment in Calcutta uses those parts and produces cars worth Rs 1500 millions, thus adding value of Rs 800 millions. The value added at the two establishments in Calcutta equals Rs 1200 millions ( 400 in the parts factory and 800 in the car factory). The Delhi enterprise is an integrated unit; it buys raw materials for Rs 300 millions , makes parts and uses them for cars and produces cars worth Rs 1500 millions , an amount identical to the Calcutta establishment. The value added in Delhi is Rs 1200 ( 1500 less 300 ). Both places produce an identical value added but the gross production is different. Value of production in Calcutta is 2200 (1500 for cars and 700 for parts). Value of production in Delhi is 1500, all cars because parts establishment is not separated. Value or weight of Industrial production is 3700 in the above example if one picks up the reports from the enterprises; only 3000 if ignores the administrative difference at the two locations and only concentrates on the finished product ; only 2400 if one counts the value added from these two locations. The internal shuffling does not change the value added but does change the so-called value of production.

Value added weights

A very important industrial country in Europe used to publish value of production different for every different aggregation of statistics. It was quite confusing. Let me rephrase that the only concept which would produce the same value of output, irrespective of the level of aggregation, is value

Statistics Canada - 134 - Collected Articles of Kishori Lal

added. There is a big advantage to use value added as a weight in aggregating the detail for industrial production. At Statistics Canada, we use the value added weights for aggregating each of the 150 separate industries which form part of Canadian industrial production. Another advantage of using value added weights for industrial production is that its importance can then easily be compared with other sectors of the economy such as the services sector or government sector or agriculture. An even more important benefit is that the industrial production program becomes a part of the GDP by Industry program. It is then automatically produced in an integrated set of statistics, giving this program a higher quality rating and placing it in a better analytical framework.

Deflation of Industrial Production

At the outset it is important to note that measurement of industrial production at constant prices or for that matter real output of any sector or for the total economy is essentially the construction of a model or logical abstraction of actual transactions. "Volume" or "quantum" is always estimated with reference to time and weights. As we are all familiar with the index number problems associated with changing weights and aggregation, the estimated quantum measure at any level of aggregation is not an unambiguous phenomenon even if one has the most reliable value, price and quantity statistics in detail. This nature of deflation applies to goods, services, industries, whatever methods are used. Value added of industrial production at constant price can be derived by the preferred double deflation technique. Such constant price value added may be extrapolated using constant price gross output, physical quantities of gross output, constant price intermediate inputs, employment etc. A connected issue is how often one should rebase the constant price series and then how to link the earlier period growth rates when the base year changes.

Double deflation

In principle, any commodity forming part of industrial production can be deflated using price indices or quantities produced as there are units of measurement available. But there is no unit of measurement of value added for industrial production or value added for other sectors of the economy or even the entire economy. Hence it can not be directly deflated. One needs some other technique to convert current price GDP into constant price GDP. The internationally recommended preferred technique is double deflation. In the double deflation approach, one deflates commodity outputs of an industry at basic prices, its intermediate uses of commodity inputs at basic prices as well as taxes on products and other taxes on production, net of subsidies. The difference between the deflated values of outputs and the total of commodity inputs and net taxes on products and production equals GDP at factor cost in constant prices. This approach, called double deflation approach, satisfies the requirement of an identity between GDP income and expenditure based estimates in constant prices. However, there are certain important hazards in using double deflation for deriving the GDP of an industry whose value added represents a small proportion of total gross output. We will come back to this issue later.

Note that the double deflation approach assumes the availability of Input and Output tables or Make and Use tables or in the 1993 SNA terminology the Supply and Use tables. We strongly support the 1993 SNA recommendation to use the double deflation approach to calculate constant price GDP or GDP of any sector such as industrial production. The 1993 SNA states: "... the supply and use tables

Statistics Canada - 135 - Collected Articles of Kishori Lal

are the most complete consistent framework for constant price estimation and provide: a) interdependent measures of prices and volumes; b) an important check on the numerical consistency and reliability of the entire set of such measures, interlinking values at constant and current prices, value and volume indexes and deflators" (paragraph 15.161). "Constant price measures for gross value added are possible in the input-output framework by using the double deflation method, as the difference between: a) the value of output deflated by a price index of outputs; 2) the value of intermediate consumption deflated by a price index for these inputs" (paragraph 15.162).

The double deflation method in the context of inputs and outputs at basic prices is a very efficient technique. A similar recommendation was made earlier in the UN Manual on National Accounts at Constant Prices, Series M, No. 64, New York, 1979 (in short, UN Constant Price Manual). Let me further discuss this issue, using the Canadian practice. The Canadian practice conforms to the recommendation in the UN Constant Price Manual. The UN Constant Price Manual states: "In an ideal world real product by kind of activity would always be derived from an input-output table by double deflation" (p. 55). In Canada, the input-output tables form the core of the production accounts. Input-output tables in full detail are produced annually both in current and constant prices. Real output by industry and constant price Industrial Production are produced using the preferred double deflation approach.

As noted above, there are certain hazards using double deflation for those industries whose value added component makes a very small share of total output. For such an industry, GDP estimated by double deflation might be erratic, because small shifts in the relative prices of intermediate inputs and gross output could translate into big shifts in the resultant value added at constant prices. Here, the UN Constant Price Manual guidelines are not entirely satisfactory. They state: "The solution to this problem, however, may simply be to consolidate industries with very small ratios of value added to gross output with related industries at earlier or later stages of production. In other words, the problem of instability may be solved by aggregation into larger units whose values added are large enough in relation to gross output not to be too sensitive to the effects of changes in prices or technology" (p. 53). In one year, value added in one industry may be erratic but in the next, a different industry might suffer. Ad hoc aggregation into large units would disturb the continuity of time series. Thus, one needs additional guidelines. In the Canadian System of National Accounts (CSNA), we have solved this problem as follows: Values added are combined as suggested by the UN Constant Price Manual. The combined value added is redistributed using gross output or any other indicator as a proxy, but the combined value for a given sub-aggregate remains unchanged. Without this restriction, the above noted GDP identity requirement will not be satisfied. These comments on the UN Constant Price Manual apply equally to the 1993 SNA recommendation on double deflation.

Chain Indices

The 1993 SNA provides a framework within which an integrated set of price and volume measures can be compiled which are conceptually consistent and analytically useful. The 1968 UN SNA provided a summary discussion of price and volume comparisons, but it was limited to the production accounts. More detailed guidelines were issued in 1979 by the United Nations in the Manual on National Accounts at Constant Prices, Series M, No. 64, United Nations, New York,

Statistics Canada - 136 - Collected Articles of Kishori Lal

1979. The 1993 SNA differs from both the 1968 UN SNA and the 1979 UN Constant Price Manual in its recommendation to use chain indices.

It is self evident that the more remote the base year becomes, the less relevant are its prices for purposes of deflating the value of current flows of goods and services. When the base year is changed, there are two ways in which the series on either side of the new base year may be linked. The first method is to revalue not only the series for all years subsequent to the new base year at new prices, but also the series for all the years preceding the new base year in order to have an unbroken series expanding on either side of the new base year. The second method is to leave the data for all years up to and including the new base year unchanged and simply use the new base year prices for valuing all flows of goods and services from the new base year onwards. In order to develop an unbroken series expanding on either side of the new base, one would simply link the series through a chain index. We reject the first method in favour of the second on conceptual grounds. A new base year is required because the old one is not relevant anymore: thus one can not effectively argue for the first method. Both the 1968 UN SNA and the 1979 UN Constant Price Manual opted for the first method whereas the 1993 SNA recommends the use of chain indices for linking the series.

The 1993 SNA recommends: "In general, the constant price series should not be allowed to run for more than five, at the most, ten years without rebasing" (paragraph 16.76). It further recommends: "The preferred measure of year to year movements of GDP volume is a Fisher volume index, changes over longer periods being obtained by chaining: that is, by cumulating the year to year movements" (paragraph 16.73a). This is a very significant recommendation. Its implications for all countries, particularly developing countries, are very significant. There is a lot of debate in the media, in the legislatures, and in the scholarly journals about growth rates. A fair amount of this debate is meaningless because the growth rates measured with reference to fixed weights established a decade or so earlier are misleading. As the relative prices have become very volatile, fixed weight growth rates are not reliable. A five percent growth rate could easily be three or four percent when the base year is changed. A few years ago, Allan H. Young, then Deputy Director of the Bureau of Economic Analysis, US Department of Commerce published a paper, "Alternative Measures of Change in Real Output and Prices", Survey of Current Business, April 1992 in which he compared (constant price) average annual rate of change in the manufacturing sector ( a most significant part of industrial production) of the United States, using three different fixed weight periods. The average annual rate of change for total manufacturing for the full decade 1977-87 was 4.7% using 1977 weights, 2.6% using 1982 weights, and a mere 1.6% using 1987 weights. These are horrendous differences for any analysis. As both the overall rate of inflation and the relative price changes are much higher in India than they are in the United States, the growth rate changes due to a shift in base years will be even more dramatic. To get a sensible reading of a growth rate, one must change the fixed weight to as current a period as possible, or preferably use chain indices.

Statistics Canada - 137 - Collected Articles of Kishori Lal

Additivity problem

Extrapolation of the base year value added

Even in many advanced countries, the input-Output tables are not produced with an annual frequency. Even when the tables are produced annually as in Canada, the tables are produced with a time lag of about three years. Thus there is a need to project industrial production to the current period using techniques other than double deflation. One can extrapolate value added using constant price gross output, physical quantities of gross output, constant price intermediate inputs, employment etc. Should one prefer deflated output over physical quantities of gross output as a projector of base year value added? The UN Manual recommendation is: "The value at constant prices of the goods available in both years should be obtained by deflating the current year value by a price index, rather than by extrapolating the base year value by a volume index .... The justification for this recommendation is that price relatives generally display less variation than quantity relatives. The range of variation of quantity relatives may be anywhere from zero to infinity whereas that for price relatives is much narrower (pp. 46-47)."

If in both the base year and the current year (a) the values of all relevant transactions, (b) the quantities of all goods and services and (c) the prices of all goods and services are recorded, then both choices give the same results and one need not worry about which method one uses. In reality, such a complete recording does not exist i.e. a statistician usually faces the problem of incomplete price and quantity information. Statisticians may still prefer quantities over prices if in their judgement the quantity information is less incomplete than price information.

Statistics Canada - 138 - Collected Articles of Kishori Lal

Concluding Remarks

The value of the measure of Industrial production is greatly enhanced when it is situated in a time series context and is comparable across jurisdictions. This leads us to recommend the following: weights for industrial production should be value-added weights; the preferred double deflation approach should be used for constant price value added; and the base year for prices should be as current as possible, preferably no later than five years. Any national statistical agency must always take into account its own unique institutional features but must be aware of the cost of not producing the highest quality measure for such an important statistic as industrial production.

Reference c. This note, prepared in March 1998, is to form part of Indian Statistical System, a commemorative Golden Jubilee Volume, of the Department of Statistics, Government of India.

Statistics Canada - 139 - Collected Articles of Kishori Lal

TABLE OF CONTENTS

Introduction

1. Mutually exclusive sectors 2. Quasi corporations 3. Government sector 4. Household sector 5. NPISHs 6. Sector Accounts 7. Head Office activities 8. Units of Homogeneous Production 9. Production account for institutional sectors 10. The Production boundary in the System 11. Financial intermediation services indirectly measured (FISIM) 12. FISIM on Own Funds 13. Bank of Canada 14. Value of Consumption of fixed capital 15. Modified Basic price valuation of goods and services 16. Valuation of intermediate consumption 17. Primary distribution of income accounts 18. Reinvested earnings on direct foreign investment 19. Secondary distribution of income accounts 20. Use of income account 21. Final consumption expenditure and actual final consumption 22. Gross capital formation for valuables 23. Inventories of structures 24. Machinery and equipment and progress payments 25. Livestock 26. Plantations, orchards, etc. 27. Software produced for own use 28. Entertainment, literary or artistic originals 29. Acquisitions less disposals of non-produced non-financial assets 30. Net lending or borrowing 31. Valuation issues, National Balance Sheet Accounts 32. Holding gains/losses for bonds 33. Government Employees Defined Benefit Pension Plan 34. Supply and use tables and input-output 35. Chain indices

Concluding Remarks

Introduction

The international System of National Accounts 1993 (1993 SNA) was published in 1993. This voluminous document, spread over 760 pages, comprising 21 chapters and 6 annexes, was prepared under the auspices of the Inter-Secretariat Working Group on National Accounts (ISWGNA), consisting of the Statistical Office of the European Communities (Eurostat), the International Monetary Fund (IMF), the Organization of Economic Cooperation and Development (OECD), the Statistical Division (UNSTAT) and regional commissions of the United Nations Secretariat, and the World Bank. Adoption of the 1993 SNA was unanimously recommended to the United Nations Economic and Social Council by members of its Statistical Commission at its twenty-seventh session, held in New York from February 22 to March 3, 1993. Canada implemented the 1993 SNA in 1997, and is one of the first countries in the world to do so. Australia is scheduled to implement the 1993 SNA later in 1998. Member countries of the European Union are planning to implement the 1993 SNA in the first quarter of 1999. Over the next couple of years, most countries will implement this new standard, keeping in view their institutional structure and statistical development. A new SNA questionnaire based on the 1993 SNA is scheduled to be ready in 1999 for international reporting purposes.

The 1997 Canadian System of National Accounts (CSNA) has incorporated the 1993 SNA guidelines, but with some modifications. Our occasional departures from the 1993 SNA guidelines are primarily prompted by pragmatic considerations, such as institutional structure, statistical data sources, availability of resources and their cost- effective use. We fully recognize the importance of promoting international comparability, but it should also be recognized that the specific circumstances existing at a given time in different countries can vary, often substantially. This paper examines the 1997 CSNA and highlights the remaining differences from the 1993 SNA, thus providing a better understanding of the Canadian system vis-a-vis that of other countries. It needs to be emphasised that most of our departures affect the overall GDP only marginally and are primarily in the sector details. Like in Canada, other countries as they implement the 1993 SNA, will also examine, adapt and revise certain guidelines to fit their own circumstances. It would be very useful if other countries would also prepare a document similar to this one, comparing their system of national accounts with the 1993 SNA. This would provide rich material for the future version of the international SNA.

The recent CSNA document: AThe 1997 Historical Revision of the Canadian System of National Accounts- Record of Changes in Classification of Sectors and Transactions, Concepts and Methodology@, issued in March 1998, points out that the CSNA is identical, for most of the significant areas, to the 1993 SNA. The CSNA is a fully integrated and comprehensive system, as is the 1993 SNA. The CSNA carries the full slate of the sequence of interlocking accounts described and recommended in the 1993 SNA. The CSNA starts with the production account with value added as a balancing item; the income and outlay account with saving as a balancing item; the capital account with net lending as a balancing item. It then describes in its financial account all the financial instruments required for the capital account. Periodic investments accumulated over a period are shown in the balance sheet accounts. Only a handful of countries are in the same league as Canada in producing the complete system of accounts recommended in the 1993 SNA.

Statistics Canada - 143 - Collected Articles of Kishori Lal

The CSNA also corresponds to the 1993 SNA in another important sense, namely, that input-output tables form an integral part of the production accounts. Statistics Canada has produced annually, starting with the reference year 1961, product by industry input-output tables (also referred to as make and use matrices) in both current and constant prices. The most recent input-output tables are for the year 1993; the tables for the years 1994 and 1995 are scheduled to be completed in the third quarter of 1998.

Our comparisons are organized following the order of the 1993 SNA chapters. We highlight in this paper only the most important of the remaining differences between the CSNA and the 1993 SNA. The purpose is to convey a good understanding of what the differences are and why are they being maintained. Such an understanding should help us improve future versions of both the CSNA and the international SNA.

1. Mutually exclusive sectors

In the CSNA, the sector accounts follow a different aggregation. The financial corporations sector is combined with the non-financial corporations sector to form the Acorporations and government business enterprises sector@ or, in short, Athe corporate sector@. The government sector is very similar to the one in the 1993 SNA. The CSNA=s persons and unincorporated businesses sector is an aggregation of the 1993 SNA=s NPISHs and household sectors. These three sectors are used for the income and outlay account, the capital and financial account and the balance sheet account but not for the production account. The capital and financial account as well as the balance sheet account provide separately for the non-financial corporations sector and the financial corporations sector, but no such separation is available for the income and outlay account. In the CSNA=s production account, all the producing units in the household sector are combined with those in the corporate sector and this merged sector is called Athe business sector of the Canadian economy@. Note that the term business sector is not defined or used in the 1993 SNA.

Statistics Canada - 144 - Collected Articles of Kishori Lal

This modification in the CSNA derives from the fact that the surveys conducted for estimating production or shipments by product, and inputs by product, typically use the establishment as the smallest unit of observation. The legal consideration that the establishment is incorporated or unincorporated is of secondary importance for statistical data. The detail on outputs and inputs by sector, even when it is available for certain industries, is of quite inferior quality for use in the production account.

2. Quasi-Corporations

Quasi-corporations are unincorporated enterprises that function as if they were corporations. The 1993 SNA, like the 1968 UN SNA, recommends that for sectoring purposes, quasi-corporations be treated as if they were corporations. In the 1993 SNA (4.50), three main kinds of quasi-corporations are recognized: a. Unincorporated enterprises owned by government units which are engaged in market production and which are operated in a similar way to publicly owned corporations. In the CSNA, such quasi- corporations are called government business enterprises and are included in the corporate sector, a treatment similar to that of the 1993 SNA. b. Unincorporated enterprises, including unincorporated partnerships, owned by households which are operated as if they were privately owned corporations. In the CSNA, there is no recognition of quasi-corporations owned by household. All unincorporated enterprises owned by households are added to persons to form the persons and unincorporated businesses sector for the income and outlay account, capital and financial account and balance sheet account. In the production account of the CSNA, these entities are added to the corporate sector to form the business sector. c. Unincorporated enterprises which belong to institutional units resident abroad. Such units are included in the corporate sector in the CSNA, a treatment similar to that of the 1993 SNA.

3. Government Sector

In the 1993 SNA (paragraph 4.113), the government sector consists of the following group of resident institutional units:

(a) All units of central, state or local government; (b) All social security funds at each level of government; (c) All non-market non-profit institutions (NPIs) that are controlled and mainly financed by government units. The sector does not include public corporations , even when all the equity of such corporations is owned by government units. These corporations form part of the corporations sector.

Statistics Canada - 145 - Collected Articles of Kishori Lal

The CSNA follows the same rules as the 1993 SNA for allocating units to the government sector but with one modification relating to NPIs. We have not been able to differentiate NPIs controlled and mainly financed by government units from NPIs mainly financed by government units. Control is a nebulous phenomenon and we have no hard information with which to measure it. Thus we have classified all those NPIs mainly financed by government units in the government sector, irrespective of the level of control.

4. Household Sector

In the 1993 SNA (paragraph 4.151), Athe household sector consists of all resident households. Defined as institutional units, households include unincorporated enterprises owned by households, whether market producers or producing for own final use, as integral parts of those households.@ The persons and unincorporated businesses sector of the CSNA relating to the income and outlay account, the capital and financial account, and the balance sheet account approximates to the definition of the 1993 SNA with one exception that the CSNA sector also includes NPISHs. In the production account, as noted above, all producing units of the household sector are merged with the corporate sector to form the business sector of the Canadian economy. Thus in the Canadian production account, the household sector is not separately identified. Our departure from the 1993 SNA guideline is due to how our production surveys are currently conducted. The legal identification of the producing establishment as unincorporated or incorporated is of secondary importance for industrial statistics. Thus the detail on outputs and inputs by unincorporated sector, even when available for certain industries, is of quite inferior quality to produce the production account for the household sector.

5. NPISHs

The 1993 SNA (paragraph 4.161) states: A Non-profit institutions are legal entities created for the purpose of producing goods and services whose status does not permit them to be a source of income, profit or other financial gain to the units that establish, control or finance them@. Some NPIs charge prices and fees that are economically significant. The 1993 SNA defines significant prices as A prices which have a significant influence both on the amounts the producers are willing to supply and on the amounts purchasers wish to buy A ( paragraph 4.161). NPIs which charge significant prices are typically part of the corporate sector. The majority of NPIs, however, are likely to be non- market producers that provide goods or services to other institutional units either free or at prices that are not economically significant. NPIs which are non-market producers and are mainly financed by the government are allocated to the government sector; those that are not allocated to the government sector are called non-profit institutions serving households (NPISHs).

The definition of NPISHs in the CSNA is very similar to the one in the 1993 SNA. However, the NPISHs sector is not separated from the household sector for the income and outlay account, the capital and financial account, and the balance sheet account in the CSNA. In the production account, detailed estimates for outputs and inputs are made separately for all the three sectors of the Canadian economy-business, government, and NPISHs.

Statistics Canada - 146 - Collected Articles of Kishori Lal

6. Sector Accounts

This section summarises the sector accounts in the 1993 SNA and the CSNA. The 1993 SNA recommends establishing the following accounts by institutional sectors:

a) Production account b) Primary distribution of income account; c) Secondary distribution of income account; d) Use of income account. e) Capital account f) Financial account g) Other changes in assets account h) Balance sheet.

In the CSNA, the three accounts- b), c) and d)- of the 1993 SNA are combined into a single income and outlay account. This is primarily a presentation issue, as most of the detail needed for the 1993 SNA accounts is developed to produce the CSNA account. The 1993 SNA accounts e) and f) are presented under the heading capital and financial account in the CSNA, and this again is primarily a presentation issue.

There is a major difference in the production account of the 1993 SNA and that of the CSNA. The business sector of the CSNA production account is coterminous with the aggregation of all the producing units of three sectors in the 1993 SNA- the non-financial corporations sector, the financial corporations sector, and the household sector. In addition, there are two other sectors of the production account- government sector and NPISHs- in the CSNA as in the 1993 SNA. The CSNA production account appears in the input-output tables.

7. Head Office activities.

The 1993 SNA states: "When the production of an enterprise takes place in two or more different establishments, certain ancillary activities may be carried out centrally for the benefit of all the establishments collectively. For example, the purchasing, sales, accounts, computing, maintenance or other departments of an enterprise may all be the responsibility of a head office which is located separately from the establishments in which the principal or secondary activities of the enterprise are carried out. In such a case, the costs of the central ancillary activities must be distributed to the establishments which they serve, for example in proportion to the latter's outputs or costs, and added to the latter's own costs" (paragraph 5.29). The 1993 SNA does not recognise the provider of ancillary activities such as a head office as an establishment, thus it has no output. It states (paragraph 5.13):

A(a) The output of an ancillary activity is not explicitly recognised and recorded separately in the System. It follows that the use of this output is also not recorded;

Statistics Canada - 147 - Collected Articles of Kishori Lal

(b) All the inputs consumed by an ancillary activity-materials, labour, consumption of fixed capital, etc.- are treated as inputs into the principal or secondary activity which it supports;

(c) It is not possible to identify the value added of an ancillary activity because the value added is combined with the value added of the principal or secondary activity.@

At Statistics Canada, head offices are identified as separate units with a geographical location to which employment and capital expenditures can be assigned; for purposes of industrial classification, the whole unit is assigned to a single industry, the one in which the bulk of the value added of the establishments it serves is generated. This is how the head offices have also been handled in the Canadian national input-output tables for the 1997 CSNA Historical Revision. Though they are identified as separate units, they disappear when assigned to an industry; thus operationally, the result in the Canadian treatment at the national level is quite similar to the one proposed in the 1993 SNA.

The 1993 SNA recommendation becomes problematic when the national input-output tables or industrial statistics are produced at the provincial or regional level. At the regional level, when the head office is situated in a region different from that of the producer units it serves, the strict application of the 1993 SNA recommendation would imply no contribution of the head office to the value added of its region. This result is counter-intuitive. Hence there is a need to reexamine the 1993 SNA recommendation.

In the Canadian provincial input-output tables, we intend to modify the 1993 SNA recommendation as follows: Head office is recognised as a separate establishment and for purposes of industrial classification, the whole unit is assigned to a single industry. That industry is the one in which the bulk of the value added of the establishments is generated, as we have done for national industrial statistics. The head office produces output which is completely used up as intermediate consumption by its serving establishments, thus reducing the value added of each of its serving establishments by the amount of use of head office service. Value added for the country as a whole does not change but its provincial or regional distribution does change, reduced in some regions counterbalanced by an identical increase in the region of the head office.

The value of the output of the head office may be equated to its costs or costs plus profits. The share of profits allocated to the head office may be equated to its share of the total wages paid by the enterprise multiplied by the total profits of the enterprise. One may devise some other convention to distribute profits. In any case, our preference is that the value of the output of the head office be equated to its costs plus shared profits.

8. Units of Homogeneous Production

The 1993 SNA states: "For purposes of input-output analysis, the optimal situation would be one in which each producer unit were engaged in only a single productive activity so that an industry could be formed by grouping together all the units engaged in a particular type of productive activity without the intrusion of any secondary activities" (paragraph 5.46). It further states: "Although the unit of homogeneous production may be the optimal unit, ... it may not always be feasible to partition establishments... into a series of mutually exclusive units of homogeneous production. In situations

Statistics Canada - 148 - Collected Articles of Kishori Lal of this kind, it will not be possible to collect directly from the enterprise or establishment the accounting data corresponding to units of homogeneous production. Such data may have to be estimated subsequently by transforming the data supplied by enterprises on the basis of various assumptions or hypotheses" (paragraph 5.47).

In the Canadian input-output tables (called supply and use tables in the 1993 SNA), we do not subdivide establishments to create units of homogeneous production except in the case of construction. The proponents of pure commodity technology perhaps forget that such conceptual perfection would require a separate vector of inputs for each of the twenty thousand or so commodities identified in the market. It is completely unrealistic to seek to achieve such a target. Aggregating twenty thousand commodities into a manageable set of 500-1000 commodity groups can hardly be called generating homogeneous production units.

9. Production account for institutional sectors

The 1993 SNA differs from the 1968 UN SNA regarding a proposal to compile production accounts for establishments and industries as well as for institutional units and sectors. According to the 1968 UN SNA, separate production accounts are required for establishments and industries but not for individual institutional sectors; only one consolidated production account is recommended for the economy as a whole.

The 1993 SNA recommends, as noted above, that there should be full production accounts, for institutional units and sectors, Afull@ in the sense of reporting gross output, intermediate consumption, and value added (paragraphs 6:1-4). Full production accounts for an institutional sector are not necessary to analyse its income and outlay account or its capital and financial account. Their additional utility, in the case of Canada, is quite marginal, especially when the very significant resources that would be required to develop such estimates for the current period are considered. However, the forthcoming collection of all commodity inputs and outputs from unified enterprise units will permit us to reexamine our approach to the development of production accounts by institutional sectors.

For the present, it is worth noting that the CSNA produces full production accounts for all years for which input-output tables are compiled but the sector classification is different from that of the 1993 SNA. The CSNA produces annual input-output tables with a lag of 22 years after the reference year. A business sector is created, which comprises all the producing units of the two corporate sectors and the household sector. Two additional sectors, the general government and NPISHs, produce goods and services primarily not for sale in the market but for their own consumption. All producing units of the Canadian economy are thus included in the production accounts of the business sector and the two non-market sectors.

Statistics Canada - 149 - Collected Articles of Kishori Lal

10. The Production boundary in the System

An important issue is the demarcation of the boundary for valuing production for SNA purposes. The 1993 SNA lists the following activities that fall within the production boundary of the System (paragraph 6.18):

a) the production of all individual or collective goods or services that are supplied to units other than their producers, or intended to be so supplied, including the production of goods or services used up in the process of producing such goods or services; b) the own-account production of all goods that are retained by their producers for their own final consumption or gross capital formation; c) the own-account production of housing services by owner-occupiers; and domestic and personal services produced by employing paid domestic staff.

Note that the own-account production of services (other than housing services) by members of the household for their own final consumption is excluded from the production boundary in the 1993 SNA (paragraphs 6.19-22).

Included in the production boundary, however, is so-called Aillegal production@. Illegal production was always implicitly contained within the boundary, and in the 1993 SNA this has been properly clarified. Illegal production (paragraph 6.30) comprises. "a) the production of goods and services whose sale, distribution or possession, is forbidden by law; b) production activities which are usually legal but which become illegal when carried out by unauthorized producers; e.g. unlicensed medical practitioners". Further: "Examples of activities which may be illegal but productive in an economic sense include the manufacture and distribution of narcotics, illegal transportation in the form of smuggling ... and services such as prostitution" (paragraph 6.32). It has generally been assumed that concealed production and the underground economy form part of the production boundary. This, too, has been clarified (paragraphs 6.34-36) in the 1993 SNA.

The boundary of production in the CSNA is quite similar to the one in the 1993 SNA; however, we have not been able to include any value for certain well-known illegal activities such as prostitution and narcotics. This is a weakness in the estimate of Canadian production. There are simply no reliable data available to enable us to make any publishable estimates. We recognise that our implicit assumption of zero value is not correct as such activities certainly exist in the economy. We did make an estimate for international smuggling of import of cigarettes, which became quite rampant in the early 1990's.

11. Financial intermediation services indirectly measured (FISIM)

Statistics Canada - 150 - Collected Articles of Kishori Lal income receivable by financial intermediaries minus their total interest payable, excluding the value of any property income receivable from the investment of their own funds..." (paragraph 6.125). Apart from a few differences, noted below, the approach to measuring the value of FISIM in the CSNA is quite similar to the one recommended in the 1993 SNA.

The 1993 SNA recognizes that certain countries may not be able to allocate FISIM among the various users. Thus it states: "In principle, the total output should, therefore, be allocated among the various recipients or users of the services for which no explicit charges are made. In practice, however, it may be difficult to find a method of allocating the total output among different users in a way which is conceptually satisfactory from an economic viewpoint and for which the requisite data are also available. Some flexibility has therefore to be accepted in the way in which the output is allocated. Some countries may prefer to continue to use the convention proposed in the 1968 version of the SNA whereby the whole of the output is recorded as the intermediate consumption of a nominal industry@ (paragraph 6.126).

We do not support this flexibility, particularly not allocating any output to final users, as advocated in the 1993 SNA. In our judgement, this last minute insertion of flexibility in the 1993 SNA was ill- advised. One of the most glaring weaknesses of the 1968 SNA was its guideline regarding the allocation of FISIM. It is to the credit of the CSNA that it has always allocated this output to users, including final users even when the 1968 SNA recommended otherwise.

12. FISIM on Own Funds

The 1993 SNA suggests that financial intermediaries= own funds should not be included in the calculation of FISIM Aas such income does not arise from financial intermediation@ (paragraph 6.125).

Contrary to the recommendation of the 1993 SNA, the CSNA includes intermediaries= own funds in the calculation of FISIM. We believe that the borrowers of these funds receive a service from institutions using their own funds. Since the publication of the 1993 SNA, there is a general convergence to the view that own funds generate financial services. We recognise that the rate of FISIM should be lower in this case as there is no service provided to the depositor of the funds, since there is no depositor.

13. Bank of Canada

The CSNA treatment of the central bank does not agree with the one in the 1993 SNA. The 1993 SNA (paragraph 6.132) states: AThe services of financial intermediation provided by central banks should be measured in the same way as all other financial intermediaries@. However, since the central bank plays a very different role than other banks, a different treatment seems justified. The central bank in Canada, the Bank of Canada is unlike the rest of the Canadian financial industry. The main functions of the Bank of Canada are to: formulate and implement monetary policy, issue and replace bank notes, undertake central banking services, and manage the public debt. Only the activities associated with central banking services can generate FISIM. However, on an operating cost basis, this function was estimated to account for only 9% of total expenses.

Statistics Canada - 151 - Collected Articles of Kishori Lal

At our request, the Inter-Secretariat Working Group on National Accounts (ISWGNA) deliberated and issued a clarification on the valuation of central bank output. In its January 1996 issue of SNA News and Notes, it is stated: "where this approach leads consistently to inappropriate results, output could...be measured at cost as for other non-market producers". We expect that many countries, like Canada, will now use the clarification issued by the ISGWNA, and not the original formulation in paragraph 6.132 in the 1993 SNA.

Further, the entire output, calculated as the sum of costs, is allocated in Canada to the federal government sector.

14. Value of Consumption of fixed capital

In the CSNA, consumption of fixed capital for the government sector, housing and agriculture is calculated using current prices while for other industries we use what enterprises report in their financial statements. Our departure from the recommended treatment is due to our statistical sources. At present, consumption of fixed capital is calculated by the Investment and Capital Stock Division of Statistics Canada using current market prices of fixed assets, but this information is available only by industry, based on establishments rather than by enterprises and secondly, such information is not segregated by sector.

The information on corporate profits for the CSNA, on the other hand, is available by enterprises. Thus it is not feasible to connect the value of consumption of fixed capital based on establishments with profits based on enterprises. This connection is an essential requirement for the income and expenditure accounts in the CSNA. We need consumption of fixed capital estimate by sector to use in the capital and financial accounts. We will reexamine our situation as we develop the capacity to integrate establishment data with enterprise data, using the forthcoming approach of collecting information via a unified enterprise survey. The new treatment will not affect GDP, but it will change the value of consumption of fixed capital which will be counterbalanced by an equal change in the value of net operating surplus.

15. Modified Basic price valuation of goods and services

In the 1993 SNA, the preferred method of valuation of output of goods and services produced for the market is at basic prices, especially when a system of VAT, or similar deductible tax, is in operation (paragraph 6.218) and is defined as follows: "The basic price is the amount receivable by the producer from the purchaser for a unit of a good or service produced as output minus any tax payable and plus any subsidy receivable, on that unit as a consequence of its production or sale. It excludes any transport charges invoiced separately by the producer" (paragraph 6.205a).

Statistics Canada - 152 - Collected Articles of Kishori Lal

There are taxes on products and other taxes on production; similarly, there are subsidies on products and other subsidies on production. The definition in the 1993 SNA of basic prices does not clearly state which of the two taxes and subsidies are being referred. The definition written in the European System of Accounts, ESA 1995, (published by EUROSTAT, Brussels, 1996) clarifies it and it reads as follows: AThe basic price is the price receivable by the producers from the purchaser for a unit of a good or service produced as output minus any tax payable on that unit as a consequence of its production or sale (i.e. taxes on products) plus any subsidy receivable on that unit as a consequence of its production or sale (i.e. subsidies on products). It excludes any transport charges invoiced separately by the producer. It includes any transport margins charged by the producer on the same invoice, even when they are included as a separate item on the invoice A (paragraph 3.48).

The 1993 SNA further states: "When output is recorded at basic prices, any tax on the product actually payable is treated as if it were paid by the purchaser directly to the government instead of being an integral part of the price paid to the producer. Conversely, any subsidy on the product is treated as if it were received directly by the purchaser and not the producer" (paragraph 6.206).

In the CSNA, we have not incorporated the recommended basic price in the valuation of goods and services; instead, we have modified the 1993 definition of basic price for our input-output tables. In its input-output tables, the CSNA reports taxes on products levied as a consequence of production or sale separately, as recommended; however, subsidies on products are not added to the prices; in other words, the prices recorded are the subsidised prices, not the prices plus subsidies as recommended in the 1993 SNA. The CSNA definition of modified basic price reads as follows:

AThe modified basic price is the price receivable by the producers from the purchaser for a unit of a good or service produced as output minus any tax payable on that unit as a consequence of its production or sale (i.e. taxes on products). It excludes any transport charges invoiced separately by the producer@.

The modified basic price used in the CSNA input-output tables is equivalent to the price, say for manufactured goods, as reported at the factory gate. In the CSNA, subsidies are recorded in the accounts of those who initially receive the money, not who eventually benefit from the subsidy program. The 1993 SNA recommends, as noted above, allocating subsidies on products as if they are received directly by the purchasers. This information is not available from the records as the subsidised product is purchased by many producing units and final consumers. Had we adopted the 1993 SNA recommendation, we would report the transactions, not at prices prevailing in the market but at assumed prices, a feature that is not very convincing. GDP at market prices, per our presentation in the input-output tables is identical, both on the income side and the expenditure side, as it is in the 1993 SNA. The industrial distribution in the CSNA is, however, different from the 1993 SNA . The advantage of the CSNA approach is that the valuation of transactions is transparent and verifiable from the enterprise records, a feature not available in the 1993 SNA.

16. Valuation of intermediate consumption

The 1993 SNA recommends: "Expenditures on goods and services intended to be used for intermediate consumption should be valued at purchasers' prices" (paragraph 6.220), and it defines purchasers=s price as follows: AThe purchaser=s price is the amount paid by the purchaser, excluding

Statistics Canada - 153 - Collected Articles of Kishori Lal any deductible VAT or similar deductible tax, in order to take delivery of a unit of a good or service at the time and place required by the purchaser. The purchaser=s price of a good includes any transport charges paid separately by the purchaser to take delivery at the required time and place@ (6.215).

We have no fundamental problem with this recommendation; this is how the records are kept in the industrial surveys. However, for deriving gross value added by industry at market prices (which in total is equal to total expenditure on GDP at market prices), one needs to value both outputs and intermediate consumption at the same prices. Otherwise, incomes and expenditures in the economy will not balance. Thus, one would need to add an aggregate of taxes less subsidies on products to the value added (derived through output at basic prices minus intermediate consumption at purchasers' prices) to balance it with GDP at market prices.

In the CSNA input-output tables, we start with the intermediate consumption at purchasers= prices. However, we then delineate each purchaser price into the following components: price charged at the factory gate; taxes on products; trade and transport margins. This formulation permits us to deflate all the components, both production and use, of a good or service in a very efficient way. We need one deflator for one commodity rather many for the same commodity at purchaser prices. Each purchaser has potentially a different price depending upon its status- intermediate consumer or a final consumer- its distance from the factory gate and whether it buys from a retailer or a wholesaler. Our delineation of the purchaser=s price, for both intermediate and final users, into the various components is an improvement over the one in the 1993 SNA, which limits itself only to presentation at purchaser=s price. In the 1993 SNA, there is no discussion of issues relating to an efficient and cost-effective deflation procedure in the context of input-output tables; thus the 1993 SNA does not discern the efficacy of recommending delineating purchaser=s price into its components of basic price, tax margins, trade margins and transport margins. This is a weakness in the 1993 SNA.

17. Primary distribution of income account

The 1993 SNA states: "The general purpose of the primary distribution of income account is to show how primary incomes are distributed among institutional units and sectors. Primary incomes are incomes that accrue to institutional units as a consequence of their involvement in processes of production or ownership of assets that may be needed for purposes of production. They are payable out of the value added created by production" (paragraph 7.2). Note that primary income comprises compensation of employees, taxes on production and imports, subsidies and operating surplus. The primary distribution of income account "... consists of two consecutive accounts: the generation of income account and the allocation of primary income account" (paragraph 7.1). The generation of income account "...represents a further extension or elaboration of the production account in which the primary incomes accruing to governments units and to the units participating directly in production are recorded. Like the production account, it may be compiled for establishments and industries as well as for institutional units and sectors" (paragraph 7.3).

Statistics Canada - 154 - Collected Articles of Kishori Lal rather than as producers whose activities generate primary incomes. In the CSNA, the primary distribution of income account is not disaggregated into generation of income account and allocation of primary income account ; instead, it becomes part of the income and outlay accounts. This is primarily a presentation issue, as most of the detail needed for the 1993 SNA is developed to produce the CSNA account.

18. Reinvested earnings on direct foreign investment

According to the 1993 SNA, both systems (SNA and Balance of Payments) "... require the saving or retained earnings of a foreign direct investment enterprise to be treated as if they were distributed and remitted to foreign direct investors in proportion to the ownership of the equity of the enterprise and then reinvested by them. In other words, two additional entries are required in the accounts of the enterprises and their foreign owners, one of which is the imputed remittance of retained earnings while the other is the imputed reinvestment of those earnings" (paragraph 7.120).

Starting with the first quarter of 1994, reinvested earnings based on ownership have been incorporated in the Canadian balance of international payments in both the current and the capital and financial accounts, as per the recommendations in both the 1993 SNA and the IMF=s Balance of Payments Manual, 5th edition 1993.

The recommendation of the 1993 SNA has not been adopted in the rest of the CSNA-the income and expenditure accounts and financial flows accounts, primarily due to statistical difficulties. The amount of detail required to adjust the financial account, across 20 institutional sectors and for almost 40 years on a quarterly basis for the back period was simply not available without incurring huge costs. Instead, both the balance of payments and the rest of the CSNA provide separate information on reinvested earnings on direct investment to allow users to go from one approach to the other.

19. Secondary distribution of income account

The 1993 SNA carries forward the balancing item from the primary distribution of income accounts to the secondary distribution of income account. Items in the secondary distribution of income account consist of current transfers such as: (i) current taxes on income, wealth, etc., (ii) social contributions and benefits, or (iii) other current transfers. As noted earlier, the CSNA income and outlay accounts are a consolidation of three accounts in the 1993 SNA , namely the primary distribution of income account, the secondary distribution of income account and the use of income account. This is primarily a presentation issue, as most of the detail needed for the 1993 SNA is developed to produce the CSNA account.

20. Use of income account

The purpose of the use of income account is to show how households, government units and non-- profit institutions serving households allocate their disposable income between final consumption and saving. The separate use of income account in the 1993 SNA is part of the income and outlay account in the CSNA. This is primarily a presentation issue.

Statistics Canada - 155 - Collected Articles of Kishori Lal

21. Final consumption expenditure and actual final consumption

An important development, both in relation to the CSNA and the 1968 UN SNA, is the specification in the 1993 SNA of the interrelationship between final consumption expenditure and actual final consumption for the three sectors (general government, NPISHs and households) in which final consumption takes place (paragraphs 9.93-99). On a practical level, it may be noted that each of the aggregates, whether referring to final consumption expenditure or actual final consumption, has to be derived from data on expenditures. Final consumption expenditure consists of the following components:

a) Household final consumption expenditure; b) Final consumption expenditure of NPISHs; c) Government final consumption expenditure: c.1 Government expenditures on individual consumption goods and services; c.2 Government expenditures on collective consumption services.

Actual final consumption of households is measured by the value of all the individual consumption goods or services acquired by resident households. There are three sets of goods and services entering into household actual final consumption:

d) those acquired through expenditure by households themselves: their value is given by item (a) above; e) those acquired as social transfers in kind from NPISHs: their value is given by item (b) above; f) those acquired as social transfers from general government: their value is given by item (c.1) above.

The value of actual final consumption of general government is given by item (c.2) above. NPISHs have no actual final consumption.

It should be emphasised that actual final consumption for the whole economy is exactly equal to final consumption expenditures.

We welcome this recommendation as it would help international comparability of household consumption. However, due to lack of estimates for government expenditures on individual consumption of goods and services, this recommendation has not been implemented in the CSNA.

22. Gross capital formation for valuables

The 1993 SNA states: "Gross capital formation is measured by the total value of the gross fixed capital formation, changes in inventories and acquisitions less disposals of valuables" (paragraph 10.32). Acquisitions less disposals of valuables is a brand new item in the definition of gross capital formation, for both the CSNA and the 1968 UN SNA. The 1993 SNA further states: "Valuables are assets that are not used primarily for production or consumption, that do not deteriorate over time under normal conditions and that are acquired and held primarily as stores of value....Valuables consist of: a) precious stones and metals such as diamonds, non-monetary gold, platinum ... held by

Statistics Canada - 156 - Collected Articles of Kishori Lal any units including enterprises provided that they are not intended to be used as intermediate inputs into processes of production; b) paintings, sculptures, etc. recognized as works of art and antiques; c) other valuables, such as jewellery fashioned out of precious stones, metals and collections" (paragraph 10.116).

This is an important additional element in both the capital formation and the capital account. We have not been able to estimate acquisitions less disposals of valuables as a separate item in the CSNA. Expenditures on valuables such as jewellery by the household sector are included in the personal expenditures, not in the capital account. Due to data problems, such expenditures do not form part of capital formation in the CSNA; instead, they residually form part of the balancing item, net lending or borrowing of the sectors.

Over the next few years, there are plans to construct and include estimates of valuables in the balance sheet as memorandum items.

23. Inventories of structures

In the CSNA, all structures, completed or unfinished, with or without contract of sale, are classified as fixed capital formation. However, in the investment in residential structures series , published by the Income and Expenditure Accounts Division, value of new housing construction is disaggregated in three parts as follows: Change in work-in-progress inventory, change in inventory of completed units, and sales of new dwelling excluding land. This additional information is quite useful for analytical purposes.

24. Machinery and equipment and progress payments

The 1993 SNA states:"Gross fixed capital formation is not recorded until the ownership of the fixed assets is transferred to the unit that intends to use them in production. Thus, new machinery and equipment that has not yet been sold forms part of additions to inventories of finished goods held by the producers of the assets. Similarly, imported machinery and equipment is not recorded as gross fixed capital formation until it is acquired by the unit that intends to use it" (paragraph 10.81). When progress payments are made for equipment such as ships, aircraft and rolling stock which take a long time to complete, the ownership, per the 1993 SNA guidelines, is not assumed to be transferred in stages as the payments are made, even if there is a contract of sale that was agreed in advance. Progress payments made under a contract of sale or otherwise are to be recorded as a financial transaction in the financial account, not as fixed capital formation. The same rule applies to progress payments made on imported machinery and equipment.

In the Canadian balance of payments, progress payments made on imported machinery and equipment are treated as a financial transaction, as recommended both in the 1993 SNA and the

Statistics Canada - 157 - Collected Articles of Kishori Lal

IMF=s Balance of Payments Manual 5th edition 1993. The CSNA has made an adjustment to exclude progress payments from capital formation in machinery and equipment when such payments involved transactions between residents and non-residents. Due to data constraints, the 1993 SNA recommendation for other progress payments has not been implemented. Where there is a contractual sale and progress payments, the CSNA has continued to treat work-in-progress for machinery and equipment as capital formation when both parties are domestic enterprises.

25. Livestock

The 1993 SNA recommends (actually it repeats the recommendation of the 1968 UN SNA) that livestock used in production year after year, such as breeding stock, dairy cattle, sheep reared for wool and draught animals, be treated as fixed assets. On the other hand, animals raised for slaughter, including poultry, are not fixed assets but are included in inventories (paragraph 10.83).

Due to data problems, the CSNA has not adopted this recommendation; instead the value of acquisitions less disposals of livestock during an accounting period is allocated to inventories.

26. Plantations, orchards, etc.

The 1993 SNA recommends (again, it repeats the recommendation of the 1968 UN SNA) that trees cultivated in plantations for the products they yield year after year - such as fruit trees, vines, rubber trees, palm trees, etc. - be treated as fixed assets (paragraph 10.83). The value of such fixed assets may be approximated, if necessary, by the costs incurred in their production during the period (paragraph 10.88).

Due to data problems, the CSNA has not adopted this recommendation. Instead, it treats these costs as intermediate consumption.

27. Software produced for own use

The 1993 SNA notes: "Computer software that an enterprise expects to use in production for more than one year is treated as an intangible fixed asset. Such software may be purchased on the market or produced for own use. Acquisitions of such software are therefore treated as gross fixed capital formation" (paragraph 10.92).

The Private and Public Investment (PPI) survey at Statistics Canada asks companies to report all expenditures on software along with their purchases of computers and other associated hardware. Any software purchased is included with office machinery in the PPI survey. The PPI survey has tried to get companies to report software expenditures separately from hardware expenditures. However this did not seem to be possible for them. Software expenditures that are most likely to be included are those purchased at the same time as the computer hardware, as well as any other large software expenditures.

Contrary to the recommendation of the 193 SNA, the CSNA, due to data problems, has not been able to capitalise the amount of expenditure made on the development of software, produced for own use.

Statistics Canada - 158 - Collected Articles of Kishori Lal

28. Entertainment, literary or artistic originals

The 1993 SNA recommends treating original creations - such as original films, sound recordings, manuscripts, etc. - as capital formation (paragraph 10.94). Due to data problems, the CSNA has not been able to capitalize the value of such originals. In some cases they are added to inventories, and in others, they are treated as intermediate consumption.

Over the next few years, there are plans to construct and include estimates of the above original creations in the balance sheet as memorandum items.

29. Acquisitions less disposals of non-produced non-financial assets

The 1993 SNA recommends that acquisitions less disposals of non-produced non-financial assets be reported in the capital account. These assets consist of land, sub-soil assets that may be used in the production of goods and services and intangible assets such as patented entities, leases, other transferable contracts, etc. (paragraphs 10.120-130). In the CSNA capital and financial account, transactions in land are included. Due to data problems, the CSNA has not been able to record explicitly expenditures on other non-produced assets in the capital finance account; instead they form part of the balancing item, net lending or borrowing of the sectors.

30. Net lending or borrowing

The CSNA capital and financial account by sectors include acquisitions less disposals of land and some identifiable produced capital assets; however, due to data problems, transactions in other existing assets , notably non-produced assets, are not identified; instead, they form part of the balancing item, net lending or borrowing of the sectors. At the level of the total economy, acquisitions less disposals of these assets would cancel out but they could be of significant value for individual institutional sectors. We recognize that this is a weakness in the capital and finance accounts of the CSNA.

31. Valuation issues, National Balance Sheet Accounts

The 1993 SNA states: " For the balance sheets to be consistent with the accumulation accounts of the System, a particular item in the balance sheet should be valued as if it were being acquired on the date to which the balance sheet relates,...This implies that assets and liabilities ( and thus net worth) are to be valued using a set of prices that are current on the date to which the balance sheet relates and that refer to specific assets" (paragraph 13.25).

In the CSNA, the following treatment has been implemented:

Produced tangible assets are valued at current prices, using a perpetual inventory method to obtain depreciated replacement cost estimates.

Statistics Canada - 159 - Collected Articles of Kishori Lal

Non-produced tangible assets, such as land surrounding structures and agricultural land, are reported on the basis of current valuations. Other non-produced assets are valued at current prices (typically net present values).

Financial data are reported at book value or at cost. Foreign currency denominated items are revalued for unrealized, or holding, gains and losses. Loans are shown net of allowances (or accumulated provisions less recoveries and write-offs).

Financial assets reflect a mixture of valuations though, generally, these are considered to be at book value, which could be either cost, equity or market value. Liabilities are reported at book or par values. The corporate share liability is equal to shares outstanding plus retained earnings (essentially, equity).

Practical difficulties of revaluing at current prices all financial assets and liabilities of all institutional sectors (there are 31 in the CSNA balance sheet) remain substantial. In the case of bonds, see item#32, we are very reluctant to restate the debt position of their issuers in the event of interest rate changes. Therefore, the CSNA has not been able to fully implement the 1993 SNA recommendations.

32. Holding gains/losses for bonds

With reference to bonds, the 1993 SNA states: "A bond is a security that gives the holder the unconditional right to a fixed money income or contractually determined variable money income over a specified period of time and also the right to a fixed sum as repayment of principal on a specified date or dates..." (paragraph 12.109). "The prices of marketable bonds change, however, when the market rates of interest change, the prices varying inversely with the interest rate movements. The impact of a given interest rate change on the price of an individual bond is less, the closer that bond is to maturity. Changes in bond prices that are attributable to changes in market rates of interest constitute price, and not quantum changes. They therefore generate nominal holding gains or losses for both the issuers and the holders of the bonds" (paragraph 12.111). As interest rates decline, the market prices of bonds increase. Should higher values of assets be reported for the holders of these bonds and higher values of liabilities and a higher debt position for the issuers of the bonds when the face value of the debt has not changed? The CSNA was not comfortable changing the value of debt due to a change in interest rates. Thus the CSNA has not implemented this recommendation of the 1993 SNA. See also item #31.

33. Government Employees Defined Benefit Pension Plan

The 1993 SNA states: ADefined benefit pension plans are those in which the level of pension benefits promised to participating employees is guaranteed . Benefits are related by some formula to participants= length of service and salary and are not totally dependent on the assets in the fund....The liability of a defined benefit pension plan is equal to the present value of the promised benefits@ (paragraph 13.78).

Statistics Canada - 160 - Collected Articles of Kishori Lal

The government of Canada and many governments in the other OECD countries have defined benefit pension plans for their employees. In the case of the government of Canada, all employer and employee contributions are deposited in the superannuation fund. An actuarial estimate is generally used to determine the liability of the government to the plan. Actuarial liability of the fund is published and is recognised in the public accounts of Canada as a public debt. However, the government uses these funds to finance its operations and there is no trustee other than the government to administer the operations of this fund.

Following the 1993 SNA, the CSNA had planned to include the liability of the government of Canada pension plan as a liability of the federal government and as an asset of the household sector. The amount is significant, more than 100 billion Canadian dollars, and the issue of public debt is important internationally. Thus we have delayed its implementation until other OECD countries implement the 1993 SNA so that the Canadian public debt position remains comparable with that of other countries.

34. Supply and use tables and input-output

The 1993 SNA includes an integrated set of supply and use tables as well as symmetric input-output tables. In the 1993 SNA symmetric input-output tables, the number of rows and columns are identical as well as the same classifications or units are used in both rows and columns- such tables are inter-industry or commodity by commodity. The 1993 SNA states: @ The System recommends that the statistical supply and use tables should serve as the foundation from which the analytical input-output tables are constructed@ (paragraph 15.7).

The CSNA produces the statistical supply and use tables or output and input tables(in short, called Ainput-output tables@, but not in the sense of analytical input-output tables mentioned in the 1993 SNA) similar to the ones recommended in the 1993 SNA. The dimensions of the Canadian tables are rectangular, meaning that the number of products is larger than the number of industries.

The Canadian statistical input-output tables have three broad sectors of the Canadian economy- business sector, government sector, and non-profit institutions serving households (NPISHs) sector. The business sector is coterminous with the aggregation of producing units of three 1993 SNA sectors, namely non-financial corporations sector, financial corporations sector, and the household sector. The business sector is disaggregated by industry based on the Canadian Standard Industrial Classification. The outputs and inputs of the other two sectors-government and NPISHs- are disaggregated not by industry but by broad functions, such as education, health , recreation, administration, etc. The 1993 SNA supply and use tables are disaggregated by industry and there is no automatic link between the sector and their underlying producing units or establishments. The Canadian inputs and outputs of the business sector provides a reasonable way to link sector-based and establishments-based statistics for the whole business sector.

In the CSNA, we do not have commodity by commodity analytical input-output tables, as recommended in the 1993 SNA. Such tables require homogenous production units which by and large do not exist. Hence they must be imputed, with the result that such tables are based on very artificial assumptions. See also item# 8 above on homogenous production unit.

Statistics Canada - 161 - Collected Articles of Kishori Lal

In the CSNA, we do not produce supply or output tables at basic prices as recommended in the 1993 SNA, but at modified basic prices. The Canadian modified basic price has the advantage that it is observed (and can be verified) in the transaction records of the producing units. The 1993 SNA basic price requires information which the producing unit does not have; hence it must be imputed for all the users of such a product. Our preference to connect our information with the accounting records of the institutional and producing units brings transparency to our statistical output. See also item #15 on modified basic price valuation.

35. Chain indices

The 1993 SNA states: A The preferred measure of year to year movements of GDP volume is a Fisher volume index; changes over longer periods being obtained by chaining: i.e., by cumulating the year to year movements@ (paragraph 16.73). Linking the series using chain indices has the well known property that the components do not add up to the aggregate. The 1993 SNA states this problem as follows: AIn order to preserve the volume movements at each level of aggregation, components have to be linked as well as the aggregates... The problem that emerges with this method is that the constant price values for the components do not add up to the constant price values of the aggregates after the series have been linked...In other words when every series at each level of aggregation is individually linked, the resulting constant price data are not additively consistent after the linking has taken place@ (paragraph 16.37).

Constant price GDP and their components are measured in the CSNA using fixed base volume indices. The base year for constant price series changes about once every five years. When the base year is changed, the CSNA, for its macro GDP and its components (in contrast to the details in the input-output tables) series does not recalculate the movements of volume in the previous series using the new base year values but chains them keeping the old growth rates fixed. Thus we have chain volume indices but the chain changes only occasionally. The constant price values of components add up to their aggregates for the current period (beginning from the latest base year) but such components do not add up for the earlier periods.

The problem of additivity assumes a much larger importance in the context of input-output tables, and the 1993 SNA has not provided any operational guidelines. The CSNA produces annual constant price detailed input-output tables which are benchmarks for the monthly real GDP by industry program. When the base year is changed , the previous years= tables are not chained. Instead, they remain produced in their earlier period=s base year. Thus we have input-output tables for 1986 to 1992 period at 1986 prices, and input-output tables for 1981 to 1986 period at 1981 prices, and so on. We have not been able to devise a system to link input-output matrices using chain indices, given that there would be statistical discrepancy for every aggregate in both commodity and industry space, and the tables must also be shown in a matrix format.

For GDP and its components, chaining the previous period but not changing the chain every year but every five years, or so, was considered a sensible modification in our implementation of the 1993 SNA recommendation of chaining on a yearly basis. As well, the CSNA continues to produce annual and quarterly chain indices for GDP and its components as supplementary information. Most of the CSNA users, particularly model builders, find our presentation of constant price series on GDP and its components quite convenient.

Statistics Canada - 162 - Collected Articles of Kishori Lal

Concluding remarks

In Canada, we have implemented the 1993 SNA but with some modifications and the resulting statistical series were released in 1997. Our occasional departures from the 1993 SNA are primarily prompted by pragmatic considerations, such as our institutional structure, our statistical sources as well as the availability of resources and their cost- effective use. This paper has provided a brief description of some of the most important differences still remaining between the CSNA and the 1993 SNA. It will be very useful to know if other countries have faced or are facing similar problems in their implementation of the 1993 SNA. A collection of such reports from many OECD countries will provide us very useful and rich source material for possible changes in the future version of the international SNA. If the problems faced by many countries are similar, then the Inter-Secretariat Working Group on National Accounts may consider issuing some clarifications as soon as possible so that other countries who have not yet implemented the 1993 SNA may benefit.

Note:

I am grateful to many colleagues in the System of National Accounts Branch at Statistics Canada, particularly Abe Tarasofsky, Lucie Laliberté, Yusuf Siddiqi, and Patrick O=Hagan, for their very useful comments and suggestions in the preparation of this paper. This paper, was completed on June 16, 1998, was presented at the OECD Meeting of National Accounts Experts, September 22-25, 1998 in Paris. A slightly revised version of this paper, called “Certain Problems in the Implementation of the International System of National Accounts 1993 - A Case Study of Canada” is to be published in the Review of Income and Wealth, June 1999.

Statistics Canada - 163 - Collected Articles of Kishori Lal

TABLE OF CONTENTS

Introduction

1. Public Sector Universe 1.1 Structure of the Government Sector 1.2 Universities 1.3 Colleges 1.4 Newfoundland School Boards 1.5 Hospitals 1.6 Residential Care Facilities 1.7 Other Health and Social Service Institutions 1.8 Deposit Insurance Corporations and Agencies 1.9 Housing Authorities 1.10 Universe Changes

2. Harmonization of Classification of Public Sector Transactions 2.1 Harmonization of Rules and Guidelines 2.2 Tax Credits 2.3 Profits of Government-Owned Liquor Boards and Gaming Activities 2.4 Licences, Privileges and Permits 2.5 Premiums Paid to Deposit Insurance Corporations 2.6 Fines, Penalties and Interest on Overdue Accounts 2.7 Payments in Respect of Losses of Government Business Enterprises 2.8 Forgiveness and Write-off of Loans 2.9 Transfers from Provincial to Federal Government 2.10 Revenues Netted Against Expenditure 2.11 Royalties 2.12 Supplementary Labour Income (SLI) 2.13 Gain (or Loss) on Foreign Exchange 2.14 Gain (or Loss) on Sale of Financial Investments 2.15 Supplementary Period Adjustments - Federal Government 2.16 Seigniorage 2.17 Federal Consolidated Special Purpose Accounts 2.18 Adjustment to Income Taxes of Life Insurance Companies 2.19 Québec Youth Allowances 2.20 Federal Payment to Quebec for the Purchase of Frigates 2.21 Adjustments to Employment Insurance Premiums 2.22 GST Low Income Refundable Tax Credits 2.23 Canadian Dairy Commission 2.24 Doubtful Taxes in Québec 2.25 Consumption Taxes Levied and Paid by the Same Government 2.26 Alberta Royalty Tax Credit 2.27 Refundable Credit for Losses on Mining Operation in Québec

2.28 Rent Supplements Paid by Provincial Housing Authorities to Landlords 2.29 Lot Levies 2.30 Local Capital Transfers 2.31 Miscellaneous Revenue - Local Government 2.32 Sales of Goods and Services to Other Governments 2.33 Pension Adjustments, British Columbia 2.34 Property Tax Credits B British Columbia, Alberta and Manitoba 2.35 Education Sales to Other Levels of Government B Quebec 2.36 Grants in Lieu of Taxes 2.37 Allowances for Losses on Government Loans 2.38 Liabilities of the Federal Government in Respect of Losses of Government Business Enterprises 2.39 Government Liabilities in Respect of Employee Pension Plans: Autonomous Plans with Invested Assets 2.40 Government Liabilities in Respect of Employee Pension Plans: Non-Autonomous Plans with Invested Assets 2.41 Treasury Bills 2.42 Notes Payable to the IMF, Outstanding Cheques and Bank Overdrafts 2.43 Foreign Subsidiaries of Government Business Enterprises 2.44 Individual Funds of the Caisse de dépôt et de Placement du Québec 2.45 Manitoba Telephone Investments for its Pension Obligations

3. Labour Income 3.1 Retirement Allowances 3.2 Employer Payroll Tax B Ontario and Quebec 3.3 Workers' Compensation B Hospital and Medical Expenditures 3.4 Workers' Compensation B Unfunded Liabilities 3.5 Interest on Overdue Employers= Contributions to Pensions 3.6 Fishing B Adjustment for Undercoverage 3.7 Insurance Agents= Commissions B Adjustment for Undercoverage 3.8 Tips B Adjustment for Undercoverage 3.9 Provincial Distribution of Wages and Salaries 3.10 Canadian Employees in Foreign Embassies and Consulates 3.11 Insurable Earnings of Status Indians

4. Capital Formation 4.1 Transfer Costs on the Sale of Existing Dwellings 4.2 National Defence Purchases of Non-Military Equipment and Structures 4.3 Capital Items Charged to Operating Expenses 4.4 Expenditures on Mineral Exploration 4.5 Progress Payments for Machinery and Equipment 4.6 Financial Leasing 4.7 Petroleum Incentive Payments Program 4.8 Capital Consumption in the Government Sector

Statistics Canada - 167 - Collected Articles of Kishori Lal

5. International Trade 5.1 Point of Valuation of Trade in Goods 5.2 Progress Payments on Imported Capital Goods 5.3 Trade in Goods for Processing 5.4 Reconfiguration of the Travel Account 5.5 New Breakdown of Commercial Services

6. Household Consumption Expenditure 6.1 Classification of Individual Consumption by Purpose (COICOP) 6.2 Leasing in the Household Sector

7. Financial Services and Insurance 7.1 Financial Intermediation Services Indirectly Measured (FISIM) B Allocation 7.2 FISIM B Reference Rates 7.3 FISIM B International Trade 7.4 FISIM B Own Funds 7.5 FISIM B Consumer Loan Companies 7.6 FISIM B Bank of Canada 7.7 Output of Insurance Industries

8. Input-Output – Classifications and Valuation 8.1 Industry Classification 8.2 Resource Royalties Industry 8.3 Presentation of Non-Market Producers 8.4 Commodity Classification 8.5 Valuation of Imports and Exports

9. Deflation 9.1 Rebasing of GDP from 1986 to 1992 prices 9.2 Output of Non-Market Services B Education 9.3 Quality Adjustment of the Volume of Wheat

10. Financial Flows and Balance Sheets 10.1 Seasonally Adjusted Financial Flow Accounts 10.2 Other Changes in Asset Account 10.3 Subsoil Assets 10.4 Timber Assets 10.5 Monetization of Gold 10.6 Leasing 10.7 Asset-backed Securities

Statistics Canada - 168 - Collected Articles of Kishori Lal

Introduction

The 1997 Historical Revision of the Canadian System of National Accounts had several goals:

(i) To implement the System of National Accounts 1993 (1993 SNA) released under the auspices of five international organizations - EUROSTAT, IMF, OECD, UN, and the World Bank. ii) To implement the revised International Monetary Fund=s Balance of Payments Manual, 5th edition, 1993 (BPM-5).

(iii) To harmonize the standards used for compiling public sector statistics as presented in two separate systems, the Canadian Financial Management System (FMS) and the Canadian System of National Accounts (CSNA), and reconcile both with the public accounts of the federal, provincial and local governments.

(iv) To rebase the constant price series of the CSNA from 1986 to 1992.

These goals were successfully met by the end of November 1997, as planned, with the release of the CSNA Historical Series. The simultaneous accomplishment of these goals has been a tall order. It has required the development of a process involving the senior staff and professionals from the System of National Accounts Branch to jointly develop and implement a detailed task-oriented plan.

The senior management of the SNA Branch established 10 working committees from all the concerned areas, each to deliberate on one of the 10 major projects and to reach decisions to be ratified by the Steering Committee, consisting of the Director General of the SNA Branch, as chairman, and all the senior managers of the SNA Branch. A complete list of members of the Steering Committee and members of the 10 working committees is provided at the end of this introduction.

More than 50 professionals, mostly from the SNA Branch, were involved in the undertaking. The involvement of so many professionals might appear cumbersome but this mode was deliberately chosen so that the decisions would be Aowned@ by the very persons responsible for their implementation. Some 150 decisions were made, a number much higher than was expected at the beginning of the exercise. These decisions were issued in August 1996 in a draft report entitled AThe 1997 Historical Revision of the Canadian System of National Accounts, Detailed Record of Issues, Discussion Notes and Decisions@. This early report was probably the most important one, as it documented all the decisions to be implemented for the 1997 Historical Revision of the Canadian System of National Accounts.

Statistics Canada - 169 - Collected Articles of Kishori Lal

The report was widely distributed internally at Statistics Canada as well as internationally. The SNA News and Notes (Number 5, January 1997), jointly issued by European Union, IMF, OECD, United Nations, and World Bank, announced its release and its availability to all our colleagues. The National Accounts directorates of all OECD countries and many other countries requested and received the document. The purpose of such a wide distribution was to share our experience with colleagues of other countries, to make them aware of our examination and resolution of several vexing problems in the 1993 SNA and to learn from them their solutions of similar problems. This, we believe, was a good distribution of labour.

We deeply appreciated the very encouraging comments (quoted below) made by Dr. Ivan P. Fellegi, Chief Statistician of Canada, on the report as well as his gracious recognition of the contributions made by all the members of the various committees:

"This is a landmark achievement in more ways than one. First, the substantive accomplishment speaks for itself: you collectively found a way to implement the 1993 SNA, together with the historical revision cycle. While I am no expert, I do understand the thousands of vexing problems that you must have found a way to solve within our context.

But second, I am really proud of your achievement in finding a way to work together in discovering solutions to problems that, within an integrated system, could only be solved that way. I salute both your inventiveness and perseverance.

Third, I am very pleased that you took the trouble to document your decisions.

I am grateful to all of you."

As we started implementing the decisions, the need for some changes in methodology, valuation and presentation became evident. Some decisions could not be implemented due to lack of source data, shortage of financial and personnel resources or time constraints. In the interest of transparency and integrity of the statistics, it was decided to fully record and explain all changes. Documentation is one area that is regrettably left undone at most institutions and, in my judgement, this is not a wise thing. I read a report some time ago in which I noted an African saying: AWhen an old citizen dies, a library is lost@. How elegantly put!

Four documents are completed relating to the changes made in this Historical Revision. The first, the present one, AThe 1997 Historical Revision of the Canadian System of National Accounts, Record of Changes in Classification of Sectors and Transactions, Concepts and Methodology@, elaborates on the changes, over one hundred in all, made in the accounts. It updates the earlier report issued in August 1996 in several ways: decisions which did not involve any change in the accounts were pruned out; several new issues were added and many decisions were rewritten.

Statistics Canada - 170 - Collected Articles of Kishori Lal

The second report, entitled AThe 1997 Historical Revision of the Canadian System of National Accounts, Current Price Gross Domestic Product, 1961-93, A Statistical Representation from the Old to the New@, presents the changes made to one of the most important series, the current price GDP. It includes 13 tables, one for each of the following topics:

$ Personal Expenditure on Consumer Goods and Services; $ Government Current Expenditure on Goods and Services; $ Gross Fixed Capital Formation, Residential Structures; $ Gross Fixed Capital formation, Non-Residential Structures; $ Gross Fixed Capital Formation, Machinery and Equipment; $ Exports and Imports of Goods and Services; $ Wages, Salaries and Supplementary Labour Income; $ Net Income of Unincorporated Business; $ Indirect Taxes; $ Subsidies; $ Current Price GDP Expenditure; $ Current Price GDP Income; $ GDP Changes, Significant Items.

The third report, entitled AThe 1997 Historical Revision of the Canadian System of National Accounts, Industry, Final Demand and Commodity Classification Systems@, deals with the industry classification used in the Input-Output tables, the provincial GDP and the monthly GDP by industry; the commodity classification used in the Input-Output tables and the final demand classification used in the Income and Expenditure Accounts and the Input-Output tables. This report on classification systems is crucial for inter-temporal comparisons. To develop the most detailed and consistent time series requires a careful review of the various versions of the industry and commodity classifications in order to adopt common nomenclatures applicable to a long period, in the case of Canada from 1961 to date. It gives great pleasure to report that at the link level for the entire historical period, we have delineated 167 industries, 122 final demand categories and 476 commodities, an enviable degree of detail anywhere in the world.

The fourth report, entitled ARemaining Differences between the 1997 Canadian System of National Accounts and the 1993 International System of National Accounts@, examines the 1997 CSNA and highlights the remaining differences from the 1993 SNA, thus providing a better understanding of the Canadian System vis-à-vis that of other countries. Our occasional departures from the 1993 SNA guidelines are primarily prompted by pragmatic considerations, such as institutional structure, statistical data sources, availability of resources and their cost-effective use. It would be very helpful if other countries would also prepare a document similar to this one, comparing their system of national accounts with the 1993 SNA. This would provide material for the future version of the international SNA.

Statistics Canada - 171 - Collected Articles of Kishori Lal

The four reports are the first set of documents related to the 1997 CSNA Historical Revision. We are also planning to produce several others during the next two years:

1) A report explaining the remaining differences between the international standards for Balance of Payments and the current Canadian practices.

2) A text proposing changes to the 1993 SNA to be reflected in its future version. This will partly borrow from previous reports, but also will probe into certain issues still unresolved or whose resolution is deemed to be unsatisfactory in the 1993 SNA.

3) Subject-specific sources and methods documents, to be prepared by the concerned Divisions of the SNA Branch.

The first four reports, noted above, are available upon request from my office.

Kishori Lal Tel. No.: 613 951-9157 Fax No.: 613 951-9031 E-mail: [email protected]

Statistics Canada - 172 - Collected Articles of Kishori Lal

Members of the Steering Committee

Kishori Lal , SNA Branch (Chairman), Lucie Laliberté, Balance of Payments Division; Darryl Rhoades, Industry Measures and Analysis Division; Art Ridgeway, Public Institutions Division; Yusuf Siddiqi, Input-Output Division; Claude Simard, National Accounts and Environment Division; Abe Tarasofsky, SNA Branch; Karen Wilson, National Accounts and Environment Division.

Members of the Working Committees

Public Sector Universe Public Institutions Division - Graham Lyttle, Art Ridgeway; National Accounts and Environment Division - Dan Finnerty, Cynthia Haggar-Guénette, Charles Wright; Input-Output Division - Dave Leblanc.

Harmonization of the Public Sector Financial Management System and the Canadian System of National Accounts Public Institutions Division - Paul Blouin, Jacinthe Bourdeau, Aldo Diaz, Andy Gareau, Margot Greenberg, Rob Loggie, Terry Moore, Art Ridgeway, James Temple; National Accounts and Environment Division - Dan Finnerty, Gylliane Gervais, Pat O'Hagan, Michel Pascal; Input-Output Division - Louis David, Dave Leblanc, Krishna Murty, Christian Nicol, Dave Van Luven.

Labour Income National Accounts and Environment Division - Ed Bunko, Katharine Fraser, Cynthia Haggar- Guénette, Anna MacDonald, Michel Pascal, Chris van Millingen, Karen Wilson; Public Institutions Division - Ferhana Ansari; Input-Output Division - Dave Leblanc.

Capital Formation National Accounts and Environment Division - Ian Cope, Katherine Findlay, Katharine Kemp, Luc Provençal, Patrick O'Hagan; Investment and Capital Stock Division - Dave Collins, Justin Lacroix, Richard Landry, Ben Marois; Input-Output Division - Tom Dempsey, Dave Leblanc, Steve O'Brien; Public Institutions Division - James Temple.

International Trade National Accounts and Environment Division - Ian Cope; Balance of Payments Division - Hugh Henderson; Input-Output Division - Farida Pira; International Trade Division - Marlene Sterparn.

Household Consumption Expenditure National Accounts and Environment Division - Anna MacDonald, Michel Vallières; Input-Output Division - Steve O'Brien, Yusuf Siddiqi.

Statistics Canada - 173 - Collected Articles of Kishori Lal

Financial Services and Insurance National Accounts and Environment Division - Conrad Barber-Dueck, Louise Jones, Dave McDowell, Michel Vallières; Input-Output Division - George Haydu, Mehrzad Salem, Yusuf Siddiqi; Industry Measures and Analysis Division - Michel Girard; Balance of Payments Division - Hugh Henderson.

Input-Output Classifications and Valuation Input-Output Division - Larry Bolduc, George Haydu, Steve O'Brien, Yusuf Siddiqi; Industry Measures and Analysis Division - Michel Girard, Richard Martel; Balance of Payments Division - Hugh Henderson; National Accounts and Environment Division - Michel Vallières.

Deflation Input-Output Division - George Kitchen, Nugent Miller, Dick Richards, Mehrzad Salem; Industry Measures and Analysis Division - Michel Girard; National Accounts and Environment Division - Katharine Kemp, Michel Pascal, Michel Vallières.

Financial Flows and Balance Sheets National Accounts and Environment Division - Jacques Delisle, Joel Diena, Dan Finnerty, Joan Forbes, Gerry Gravel, Patrick O'Hagan, Rashmi Shukla, Karen Wilson, Charles Wright; Public Institutions Division - Andy Gareau, Richard Sauriol; Input-Output Division - Dave Leblanc; Industrial Organisation and Finance Division - Cathy Bakker, Robert Moreau.

Statistics Canada - 174 - Collected Articles of Kishori Lal

1. Public Sector Universe

1.1 Structure of the Government Sector

Issue: Should we follow the 1993 SNA to delineate the structure of the Government sector and organize it such that for both the Canadian Financial Management System (FMS) and the Canadian System of National Accounts (CSNA), the same set of institutional units forms a sector, a sub-sector or a component?

Pre-Revision Treatment: The universe for the Government sector was not the same in the FMS and the CSNA. The organization of the sector also differed between these systems. Neither the FMS nor the CSNA fully covered the Government sector.

Discussion Notes: It was agreed that a common universe was essential both for operational reasons and for improving the integration of the data sets. It would facilitate the classification of government expenditure on a functional basis in the CSNA. The existing differences have caused considerable additional work in the various Divisions of the SNA Branch.

The 1993 SNA proposes two approaches for sub-sectoring the Government sector. One approach is to have a separate sub-sector for social security funds. The other is to associate social security funds with the appropriate level of government. It was felt that having a distinct sub-sector for social security schemes, which would show separately the Canada Pension Plan (CPP) and the Quebec Pension Plan (QPP), was the most appropriate choice for Canada.

The 1993 SNA does not delineate components below the sub-sectors of government (Federal, Provincial, Local and Social Security Funds), but it was felt important to maintain greater detail in both the CSNA and the FMS.

In both systems, the components estimated separately under Federal Government are General Government and Non-Autonomous Pension Plans. However, Non-Autonomous Pension Plans should not be shown separately in the CSNA until other OECD countries implement the 1993 SNA.

There are four components of the Provincial/Territorial Government sub-sector:

$ General Government $ Non-Autonomous Pension Plans $ Universities and Colleges $ Health and Social Services Institutions, which include hospitals as well as other health and social service institutions.

Statistics Canada - 175 - Collected Articles of Kishori Lal

As in the case of Federal Government, Non-Autonomous Pension Plans should not be shown separately in the CSNA until other OECD countries implement the 1993 SNA.The Local Government sub-sector has two components, General Government and School Boards.

Change: A new structure of the Government sector, which satisfies the 1993 SNA guidelines and harmonizes the FMS and the CSNA universes, has now been put in place. The only exception is that the decision to have a separate Non-Autonomous Employee Pension Plans sub-sector in the CSNA is delayed until other OECD countries implement the 1993 SNA.

1.2 Universities

Issue: a. Should universities be transferred to the Government sector? b. If so, should they be transferred as a block?

Pre-Revision Treatment: Universities were not part of the Government sector in the CSNA or the FMS. They formed part of the household sector in the CSNA, as they were considered >associations of individuals= serving the community.

Discussion Notes: The 1993 SNA recommends that institutions mainly financed by government be classified to that sector. The majority of universities, 88 out of 143, in Canada continue to be mainly financed by governments. These universities account for over 98% of total revenues of universities. The remaining ones are mostly theological universities attached to a larger parent university. The transfer of universities to the Government sector thus seems warranted.

It was also recommended that, for analytical ease, all universities be placed in the same component. In Canada, the universities that are not mainly financed by government account for a very small proportion of the activity. Public Institutions Division should monitor them annually so that this recommendation could be reexamined if the situation changes in future.

Change: All universities have now been moved from the household sector to a component of the Provincial/Territorial Government sub-sector. For analytical ease, all universities have been placed in the same component.

1.3 Colleges

Issue: a. Should non-profit colleges be a separate component of the Government sector? b. Should all these colleges be in the same component? c. Should all other private colleges remain in the business sector?

Pre-Revision Treatment: All non-profit colleges were in the Provincial Government sub-sector in both the FMS and the CSNA. Other private colleges were in the business sector.

Statistics Canada - 176 - Collected Articles of Kishori Lal

Discussion Notes: It was considered important to classify non-profit colleges as a separate component of the Provincial Government sub-sector to provide an estimate of total expenditures of post-secondary education. There are only a handful of non-profit colleges that are not mainly financed by provincial governments and their contribution to total expenditures is less than 2%. It is sometimes argued that institutions not controlled by government, even if the latter provides most of their financing, should not be classified in the Government sector; some 26 colleges in Quebec fall in this category. Specific control information for the other provinces is not available. The notion of control is always somewhat nebulous. Thus the fact of being mainly financed by the government was considered a sufficient criterion to classify an institutional unit in the Government sector.

All colleges that are separate institutional units should be in the same component for analytical ease since those not mainly financed by government account for a very small proportion of the activity. Public Institutions Division should monitor them annually so that this recommendation could be reexamined if the situation changes in future. All private (other than non-profit) colleges should remain in the business sector.

Change: Non-profit colleges have now been placed in a separate component of the Provincial Government sub-sector. For statistical quality reasons, particularly of quarterly data, colleges and universities are combined into a single component for dissemination purposes. For analytical ease, all non-profit colleges have been placed in the same component. All other private colleges remain in the business sector.

1.4 Newfoundland School Boards

Issue: Should Newfoundland school boards be transferred to the Government sector?

Pre-Revision Treatment: These school boards were part of the household sector in the CSNA.

Discussion Notes: School boards in Newfoundland continue to be separate institutional units and are mainly financed and increasingly controlled by the provincial government. Following a referendum in Newfoundland in September 1997, the Constitution of Canada has been amended to give the provincial government more control over the school boards.

Change: School boards in Newfoundland have now been moved from the household sector to the school board component of the Local Government sub-sector.

1.5 Hospitals

Issue: Should all non-profit hospitals, mainly financed by government, be placed in a component of the Provincial Government sub-sector?

Statistics Canada - 177 - Collected Articles of Kishori Lal

Pre-Revision Treatment: Hospitals formed a distinct sub-sector of Government in both the Income and Expenditure Accounts and the Input-Output tables. For international reporting purposes, however, the hospitals were combined with the Local Government sub-sector in the Income and Expenditure Accounts. The so-called private hospitals were classified to the business sector.

Discussion Notes: The 1993 SNA recommends that all non-profit institutional units mainly financed by government be classified to the Government sector. As hospitals are mainly financed by provincial governments, it was recommended to move the existing Hospitals sub-sector to a component of the Provincial Government sub-sector in the Income and Expenditure Accounts. The so-called private hospitals were actually non-profit organizations mainly financed by provincial governments that were wrongly allocated to the business sector.

Change: All non-profit hospitals (public and private) mainly financed by government have now been transferred to the Provincial Government sub-sector.

1.6 Residential Care Facilities

Issue: Should all non-profit residential care facilities (RCFs) be allocated to the Government sector?

Pre-Revision treatment: Residential care facilities were allocated to a sector based on the >ownership= variable on the RCF survey. Most of them were in the household sector, with others in the government or the business sector.

Discussion Notes: There are a significant number of proprietary RCFs that must remain in the business sector. The remaining RCFs are non-profit organizations with the vast majority mainly financed by government. The high percentage of funding from provincial governments plus the regulatory authority the latter exercise on these entities points to their allocation to the Provincial Government sub-sector. It is sometimes difficult to draw the line between residential care facilities and hospitals. Having all non-profit residential care facilities in the Provincial Government sub- sector ensures that any shift between these two components would have no impact on the sub-sector data. However, RCFs are a very heterogeneous group with many providing very little in terms of medical or social assistance other than housing.

Change: All non-profit residential care facilities now form a sub-component of the Provincial Government sub-sector. For statistical quality reasons, particularly of quarterly data, residential care facilities are combined with other sub-components for dissemination purposes. The proprietary RCFs remain in the business sector.

1.7 Other Health and Social Service Institutions

Issue: Should there be a separate sub-component for other health and social service institutions within the Provincial Government sub-sector?

Statistics Canada - 178 - Collected Articles of Kishori Lal

Pre-Revision Treatment: These institutions were not separately identified in the Provincial Government sub-sector. Discussion Notes: There are several sub-components of the Health and Social Service Institutions component. The most important ones are hospitals and residential care facilities. Remaining institutions should be classified to another sub-component, >Other Health and Social Service Institutions, n.e.s.=. There are a number of such health and social service agencies in Quebec that should be classified in this sub-component.

Change: A separate sub-component has been created for other health and social services within the Provincial Government sub-sector. For statistical quality reasons, particularly of quarterly data, this sub-component is combined with other sub-components for dissemination purposes.

1.8 Deposit Insurance Corporations and Agencies

Issue: Should all deposit insurance corporations and agencies be allocated to the Government sector?

Pre-Revision Treatment: There are several provincial deposit insurance corporations, and a federal one. Some were classified as business enterprises, while others were classified as special funds of government.

Discussion Notes: Two aspects of these agencies were examined, their regulatory nature and their financing. Financial institutions must join these agencies, purchase deposit insurance and must do so from the specified agency. The examination of the records of deposit insurance corporations also revealed that their capital was often insufficient to meet their obligations when financial institutions failed. Thus governments have had to provide them with additional funds so they could pay the claims of the financial institutions. This also indicates that they are not truly operating as businesses.

Change: All the deposit insurance corporations and agencies have now been transferred to the Government sector. See also Item 2.5.

1.9 Housing Authorities

Issue: There are a number of entities that are commonly referred to as housing authorities. These entities finance, construct or manage buildings that primarily provide housing for persons with low incomes. To what sector should they be classified?

Statistics Canada - 179 - Collected Articles of Kishori Lal

Pre-Revision Treatment:

Federal Housing Authority

The Canada Mortgage and Housing Corporation (CMHC) is the only federal entity in this group. It was classified as a Government Business Enterprise (GBE).

Provincial Housing Authorities

The majority were treated as GBEs. In the province of Quebec, the housing authority was a special fund of the government. In Nova Scotia, it was part of the Housing Ministry. In BC, there were two entities, one a special fund of the government and the other a GBE.

Local Housing Authorities

Some of these entities were in Local Government, while the remainder were in the Business Sector. There were 18 entities classified to the Local Government sub-sector in both the FMS and the CSNA. These were entities that were linked directly to municipal governments. The >Offices municipaux d'habitation= (OMHs) in Quebec were also classified to Local Government in the FMS, but not in the CSNA.

Discussion Notes:

Federal Housing Authority

The financial statement of CMHC includes three accounts: the Minister=s Account, the Corporate Account and the Administered Funds Account. The Minister=s Account was treated as part of departmental expenditures, for which CMHC is only an agent. The remaining two covered the mortgage lending and mortgage insurance activities of CMHC. These accounts generally showed a positive net income. This entity (without the Minister=s Account) had been classified as a GBE and it was recommended that it remain so, as the CMHC fees for mortgage lending and mortgage insurance services are comparable to those of other market-oriented institutions.

Provincial and Local Authorities

The Public Institutions Division has had access to financial statements of all of the provincial entities, but financial statements of local entities have proven much more difficult to obtain. From the information available, all provincial and local entities should be classified as non-profit organizations. As such, they may be in any of the three sectors of the economy - non-profit organizations serving households, government or business. Looking at the sources of revenue for these entities, we found that rents accounted for less than 50% of the revenues of all provincial entities operating housing units. In many cases, rental income covered less than 30% of the expenditures. Given that governments provide most of the funding for these agencies and set

Statistics Canada - 180 - Collected Articles of Kishori Lal regulations governing their day-to-day operations, it was recommended that they be classified to the Government sector.

A review of the mandate of these entities made it quite clear that they have a role in the general delivery of social assistance. In fact, they are largely an alternate delivery vehicle for social programs rather than a real competitor to private housing. The increasing integration of health and social welfare delivery was likely to further blur the line between social housing and other social assistance programs. In several provinces, the provincial housing authority was active in supporting not only housing for the poor, but also specialized housing that would fall in the category of residential care facilities, such as senior citizen homes, nursing homes and shelters for women and children.

Provincial Government Housing Authorities

These entities, listed below, should be classified as special funds of provincial governments. Most of them were classified as GBEs.

1. Newfoundland and Labrador Housing Corporation 2. Prince Edward Island Housing Corporation $ and its 9 Local Housing Authorities 3. Nova Scotia Housing Development Fund (already a Special Fund - SF) and its 18 Regional Authorities 4. New Brunswick Housing Corporation 5. Société d=Habitation du Québec (already a SF) 6. Ontario Housing Corporation $ and its 54 Local Housing Authorities 7. Manitoba Housing and Renewal Corporation $ Manitoba Municipal Housing Authority (agent) 8. Saskatchewan Housing Corporation $ and its 271 Local Housing Authorities 9. Alberta Social Housing (formerly Alberta Mortgage and Housing Corp.) 10. Provincial Rental Housing Corporation (BC) 11. BC Housing Management Commission 12. Yukon Housing Corporation $ Whitehorse Housing Authority (agent) 13. NWT Housing Corporation $ and its 47 Local Housing Organizations

Statistics Canada - 181 - Collected Articles of Kishori Lal

Local Housing Authorities

For most of these entities, there was very little data available. Those entities that are part of a municipal government should belong to Local Government. Other local housing organizations that are non-profit institutional units mainly funded by government should also be part of Local Government, in the same way as other boards and commissions, etc. Finally, there are co-ops and other non-profit agencies which might be part of the non-profit institutions serving household sector or the Government sector, but for which we lacked the data to differentiate.

It was proposed that the Offices municipaux d'habitation= (OMHs) in Quebec and other institutions identified as part of a municipality (see list below) remain part of the Local Government sub-sector . The inclusion of OMHs in Quebec would be a change for the CSNA:

1. St. John's Non-Profit Housing 2. Saint John Non-Profit Housing 3. Offices municipaux d'habitation (653 entities) 4. Metro Toronto Housing Corporation 5. Brantford Non-Profit Housing 6. Hamilton Non-Profit Housing 7. Ottawa Non-Profit Housing 8. Regional Municipality of Durham Non-Profit Housing 9. Regional Municipality of Peel Non-Profit Housing 10. Thunder Bay Non-Profit Housing 11. City of Toronto Non-Profit Housing 12. Windsor Non-Profit Housing 13. Yorkton Housing Corporation 14. Edmonton Non-Profit Housing 15. Calgary Non-Profit Housing 16. Capital Region Housing Corporation (Victoria) 17. City of Vancouver Housing Corporation 18. Greater Vancouver Public Housing Corporation

Other Housing Authorities

It was proposed to continue classifying all other local organizations to the business sector, in the absence of adequate data.

Change: i) All provincial government housing authorities have now been moved from the business sector to the Provincial Government sub-sector. ii) In the CSNA, Quebec=s OMHs (653 entities) have now been moved from the business sector to the Local Government sub-sector; all local housing authorities identified as part of a municipality have remained in Local Government in the FMS; iii) In the absence of data, other local housing authorities remain in the business sector.

Statistics Canada - 182 - Collected Articles of Kishori Lal

1.10 Universe Changes

Issue: How far back should the universe changes be made?

Discussion Notes: Both the CSNA and the FMS must produce time series on a consistent basis. This requires that the institutional units making up a sector of the economy must be consistently defined. But there is always a tussle between the need for long time series and the cost of producing them. Common sense must prevail. It was the collective judgement that it would not be cost effective to go back prior to 1961 for each institution that has changed sector.

The attached table, Universe Changes in the Public Sector of Canada, lists all the universe changes made to implement the 1993 SNA guidelines and harmonize the FMS and the CSNA universes. There are three types of change:

1) Transfer of an entire class of institutional units from one sector to another, such as universities which go from the household sector to the Government sector. These changes apply equally in the FMS and the CSNA.

2) Transfer of individual institutional units from one sector to another. These instances are listed by level of government (federal and provincial). Several new entities are also listed under the two levels of government. These changes apply equally in the FMS and the CSNA.

3) Changes to the Local Government sub-sector. Several individual institutional units are transferred to the new sub-sector. This applies only in the CSNA, as these units were already classified properly in the FMS.

Change: The effective date for change is the inception date of the institutional unit or 1961, whichever comes later. In incorporating the changes, every effort has been made to use the accounting records of the institutional units.

Statistics Canada - 183 - Collected Articles of Kishori Lal

UNIVERSE CHANGES IN THE PUBLIC SECTOR OF CANADA

Effective ENTITIES Previous New Date for Sector Sector Change

QUEBEC Société du Palais des congrès de Montréal ...... Government GBE 1983 Société québécoise des transports ...... GBE Government 1988 Société québécoise d'initiatives agro-alimentaires (SOQUIA) ...... Government GBE 1975 - Société québécoise des pêches (consolidated by SOQUIA) Société parc-autos du Quebec métropolitain ...... Government GBE 1980

New Entity Société du Centre des congrès de Quebec ...... GBE 1993

ONTARIO Metropolitan Toronto Convention Centre Corporation...... Government GBE 1984 Ontario Housing Corporation...... GBE Government 1964 - Local Housing Authorities (54) ...... GBE Government 1964 Ontario Place Corporation...... Government GBE 1971 ORTECH Corporation (Ontario) ...... Government GBE 1989 Ottawa Congress Centre...... Government GBE 1984 Ontario Cancer Treatment and Research Foundation ...... Households Government 1961 Province of Ontario Savings Office ...... GBE Government 1961

New Entities Ontario Casino Corporation...... GBE 1994 Ontario Clean Water Agency ...... Government 1993 Ontario Financing Authority...... Government 1993 Ontario Realty Corporation...... Government 1993 Ontario Training and Adjustment Board ...... Government 1993 Ontario Transportation Capital Corporation ...... Government 1993

MANITOBA Manitoba Centennial Centre Corporation...... Government GBE 1968 Manitoba Housing and Renewal Corporation...... GBE Government 1967 - Manitoba Municipal Housing Authority ...... GBE Government 1967 Manitoba Lotteries Corporation ...... Government GBE 1984 North Portage Development Corporation ...... Government GBE 1983 Manitoba Cancer Treatment and Research Foundation ...... Business Government 1980

SASKATCHEWAN 1970 Sask Sport Inc...... Government Business 1970 - Sask Sport Distribution Inc...... Government Business

- Administration Centres for Sports, Culture and 1989 Recreation (1989) Inc...... Government Business 1970 - Western Canada Lottery B Saskatchewan Division Inc...... Government Business 1974 Saskatchewan Housing Corporation ...... GBE Government 1974 - Local Housing Authorities (271) ...... GBE Government 1988 Saskatchewan Research Council...... Government GBE

Statistics Canada - 184 - Collected Articles of Kishori Lal

UNIVERSE CHANGES IN THE PUBLIC SECTOR OF CANADA

Effective ENTITIES Previous New Date for Sector Sector Change

New Entity Saskatchewan Lotteries Trust Fund for Sport, Culture and Recreation...... Government 1991

ALBERTA Alberta Social Housing...... GBE Government 1967

New Entities ARCA Investments Inc...... Business 1984 Pension Plan Administration Fund ...... Government 1992

BRITISH COLUMBIA B.C. Housing Management Commission ...... GBE Government 1967 B.C. Pavilion Corporation ...... Government GBE 1984

New Entity B.C. Transportation Financing Authority ...... Government 1993

YUKON TERRITORY Yukon Housing Corporation ...... GBE Government 1972 - Whitehorse Housing Authority ...... GBE Government 1982

NORTHWEST TERRITORIES Northwest Territories Housing Corporation ...... GBE Government 1974 - Local Housing Organizations (47) ...... GBE Government 1974

The following changes apply to the System of National Accounts only.

LOCAL

QUEBEC Business Government 1968 Offices municipaux d=habitation ...... X

NORTHWEST TERRITORIES Prov Govt. School Brds 1989 Yellowknife Education Districts 1 and 2 ...... Prov Govt. School Brds 1995 . . Sir John Franklin Territorial High School ......

X = Not implemented for Assets and Liabilities (FMS)

Statistics Canada - 185 - Collected Articles of Kishori Lal

2. Harmonization of Classification of Public Sector Transactions

2.1 Harmonization of Rules and Guidelines

Issue: Should there be a harmonization of the rules and guidelines in the two systems, FMS and CSNA, for compiling Public Sector financial statistics?

Pre-Revision Treatment: Identical transactions (as reported in financial statements and public accounts of the various governments) were sometimes recorded differently in the two systems: classifications were different and so were conventions on the recording of transactions on a gross or net basis. This was confusing to the users.

Discussion Notes: In the 1996 Budget Plan of Canada, as submitted to Parliament on March 6, 1996, it was noted: AThere are currently three systems of government financial accounting - each serving a different purpose. A working group has been set up - on the Auditor General=s recommendation - charged with the task of generating a better understanding of, and access to, these. Progress is being made in reducing the differences in concepts, where possible. The revision to the System of National Accounts in 1997 will reduce some of the current discrepancies between the measure of the deficit on a National Accounts and a Public Accounts basis. In addition, the Financial Management System - which gathers information on all three levels of government using a common framework - is being revised to reduce differences with the National Accounts definitions@ (page 119).

Change: By and large, both the FMS and the CSNA have now adopted the same rules for classifying and recording transactions. Any exception to the rules is fully spelled out and approved by the senior management of the SNA Branch. Every effort is made to harmonize the universe, the classification criteria and the valuation principles.

2.2 Tax Credits

Issue: Should income taxes be recorded on a net basis, after deduction of tax credits, or should they be recorded on a gross basis and the tax credits treated as an expenditure?

Pre-Revision Treatment: Many more tax credits were treated as expenditure in the FMS than in the CSNA.

Discussion Notes: A distinction was made between refundable tax credits, which have value even for persons and businesses without taxable income, and non-refundable tax credits, that only have value in the context of the tax system by reducing tax liability. As well, the refundable tax credits actually paid out are reported in financial statements and public accounts. Information on refundable tax credits deducted from tax liability and on non-refundable tax credits is much less readily available.

Statistics Canada - 186 - Collected Articles of Kishori Lal

Refundable tax credits could be considered as expenditure programs being delivered via the tax administration. Further, the recording of such programs as expenditure would improve inter- provincial comparability, as they can be delivered through direct payments or the tax system, depending on the jurisdiction.

One must be careful in how the term >refundable= is used, particularly with respect to the corporate tax, as it may really refer to the ability to carry forward or backward credits which may not be truly refundable. Some tax credits may be refunds and should be treated as such; for an example, see Item 2.26, Alberta Royalty Tax Credit. The treatment of each refundable tax credit is shown in the table below.

REFUNDABLE TAXES $ Millions Previous Treatment in of Corresponding Tax LEVEL NAME 1993/94 FMS SNA

Personal Income Tax Federal GST Low IncomeTax Credit 2,685.00 Gross Gross Child Tax Credit (Gone in 1994/95) 1,612.00 Gross Net Provincial NS Home Ownership Plan 0.20 Gross Net

QC Real Property 263.00 Gross Net Sales Tax 188.70 Gross Net Adults Who Lodge Their Direct Ascendants 8.30 Gross Net

ON Property and Sales Tax 879.00 Gross Net Ontario Home Ownership Savings Plan 25.50 Gross Net

MB Property 77.20 Gross Net Cost of Living 66.20 Gross Net

SK Family 0.03 Gross Net

AB Renters Assistance 0.07 Gross Net Royalty 0.30 Gross Net

BC Venture Capital 7.60 Gross Net Sales Tax 50.00 Gross Net

Corporate Income Tax QC Losses Other Than Capital Losses 82.20 Gross Net Scientific Research and Experimental Development 312.00 Gross Net Manpower Training 24.10 Gross Net Cinematographic Productions 27.30 Gross Net Taxi Companies 0.30 Gross Net Small Business Capitalization 1.00 Gross Net Property Taxes for Forest Products 3.10 Gross Net

Statistics Canada - 187 - Collected Articles of Kishori Lal

Change: All refundable tax credits are now treated as expenditures in both the FMS and the CSNA. In other words, the corresponding income taxes are recorded on a gross basis. The treatment of the other tax credits remains unchanged.

2.3 Profits of Government-Owned Liquor Boards and Gaming Activities

Issue: Should profits of government-owned liquor boards and gaming activities be treated as indirect taxes or as profits?

Pre-Revision Treatment: Profits of liquor boards were always treated as government investment income in the FMS. In the CSNA, they were treated as indirect taxes starting with the 1971 Historical Revision. Profits of government gaming activities such as lotteries, casinos and video lottery terminals were treated as government investment income in both systems.

Discussion Notes: The 1993 SNA notes: "Fiscal monopolies are public corporations, public quasi- corporations, or government-owned unincorporated enterprises that have been granted a legal monopoly over the production or distribution of a particular kind of good or service in order to raise revenue and not in order to further the interests of public economic or social policy. Such monopolies are typically engaged in the production of goods or services which may be heavily taxed in other countries, for example, alcoholic beverages, tobacco, matches, petroleum products, salt, playing cards, etc. The exercise of monopoly powers is simply an alternative way for the government to raise revenue instead of the more overt procedure of taxing the private production of such products. In such cases the sales prices of the monopolies are deemed to include implicit taxes on the products sold. While in principle only the excess of the monopoly profits over some notional "normal" profits should be treated as taxes, it is difficult to estimate this amount, and, in practice, the value of the taxes should be taken as equal to the amount of the profits actually transferred from fiscal monopolies to government." (Paragraph 7.69e)

Treating profits of liquor boards and government gaming activities as part of operating surplus would mean that such income results from the use of a factor of production. This is not the case, since this income accrues to government as a result of its fiscal monopoly. It was felt that such revenues should always be considered indirect taxes.

At the 1996 Federal-Provincial Conference on Provincial Economic Accounts and Public Sector Statistics, some provincial focal points objected to the decision to classify revenues of Government- owned gaming activities as indirect taxes and we agreed to re-examine it. After re-examination, we remained convinced that the profits of these units were not a return on capital but were due to government monopoly. Hence, they are more like taxes than investment income. However, that should not preclude provincial governments to present them otherwise.

Statistics Canada - 188 - Collected Articles of Kishori Lal

In the CSNA, total profits (remitted and unremitted) of government-owned liquor boards were treated as indirect taxes. The same logic should apply to gaming activities. In the FMS where only remitted profits are treated as government revenue, the unremitted profits should be shown as a reconciliation item between the two systems.

Change: Total profits of government-owned liquor boards and gaming activities are now treated as indirect taxes in the CSNA, while their remitted profits are treated as such in the FMS. The difference between the two will be shown as a reconciliation item. The profits are presented separately, thus permitting users to re-aggregate them differently than in the standard presentation of the CSNA or the FMS.

2.4 Licences, Privileges and Permits

Issue: Should licences, privileges, concessions and permits be considered indirect taxes, transfers or sales of services?

Pre-Revision Treatment: They were variously classified in the CSNA as indirect taxes, transfers from persons to governments or as sales of services. In the FMS, they formed a separate category.

Discussion Notes: The 1993 SNA notes: "Households make payments to government units to obtain various kinds of licences, permits, certificates, passports, etc., and in some cases it is not clear whether government units actually provide services in return, such as testing or inspection, or whether the payments are de facto taxes. As explained in chapter VIII, paragraph 8.54 (c), the treatment of certain controversial borderline cases has been decided by the following convention, based on the practices followed in the majority of countries: payments by households for licences to own or use vehicles, boats or aircraft and also licences to hunt, shoot, or fish are treated as taxes, while payments for all other kinds of licences, permits, certificates, passports, etc. are treated as purchases of services and included in household consumption expenditure." (Paragraph 9.62). Licences purchased by businesses are to be considered taxes on production unless the government renders some service in issuing the licence, such as an inspection (paragraph 7.70).

In developing the FMS classification of revenues and expenditures by function and sub-function, Statistics Canada consulted representatives from the Departments of Municipal Affairs across Canada. This is why all local governments identify privileges, licences and permits as a separate class in their administrative documents.

It was recognized that, in some instances, it was open to interpretation whether the licence, permit or concession was an indirect tax (unrequited payment, as the government provides nothing in return) or a service. However, the 1993 SNA distinction should be followed. It was therefore decided make a complete review of licences in order to ensure that their treatment in the CSNA complies with the 1993 SNA recommendation. The conclusion of the review was that although some local government licences should be treated as services, data limitations prevented the split of this category. Thus all local licences continue to be classified as indirect taxes.

Statistics Canada - 189 - Collected Articles of Kishori Lal

Change: The 1993 SNA recommendation to treat payments by households for some specific licences such as those to own or use vehicles, boats or aircraft and also to hunt, shoot or fish as indirect taxes has been adopted. All other licenses are treated as sales of services. Note that previously, payments by households for licenses to own and operate vehicles were treated as transfers, thus more like direct taxes. The change results in a large upward revision to household expenditure on services. All licences purchased by business are treated as indirect taxes. All revenues from concessions and those licences classified as sales of services in the CSNA are now also treated as sales in the FMS. All local licences continue to be treated as indirect taxes.

Separate sub-classes for licences and concessions within sales of goods and services= are being maintained in the FMS to allow an aggregation of all licences.

2.5 Premiums Paid to Deposit Insurance Corporations

Issue: Should premiums paid by financial corporations to deposit insurance corporations be treated as indirect taxes?

Pre-Revision Treatment: These premiums were treated as insurance revenues, just like those paid to other insurance companies.

Discussion Notes: Deposit insurance corporations were previously classified to the business sector. However, it was decided to shift them to the Government sector because they did not behave as commercial insurance companies in three crucial respects: (a) premiums charged consistently failed to cover claims; (b) claims were paid on uninsured assets; and (c) deposit insurance was required by the same governments holding the monopoly on the service (see Item 1.8). As a result, deposit insurance is more like an indirect tax than a payment for insurance services.

Change: Deposit insurance premiums are now treated as indirect taxes. See also Item 1.8.

2.6 Fines, Penalties and Interest on Overdue Accounts

Issue: Should fines, penalties and interest on overdue payments be classified in the same category as the income to which they relate (indirect taxes, sales of goods and services or transfers)?

Pre-Revision Treatment: Fines and penalties paid on indirect taxes, even when separately identified, were combined with indirect taxes. Other fines and penalties (including those paid by business), when they could be separately identified, were classified as other transfers from persons. Interest identified separately usually was not included with the underlying category.

Discussion Notes: The 1993 SNA notes: "Fines and penalties imposed on institutional units by courts of law or quasi-judicial bodies are treated as compulsory current transfers. However, fines and other penalties imposed by tax authorities for the evasion or late payment of taxes cannot usually be distinguished from the taxes themselves and are, therefore, grouped with the latter in practice and not recorded under this heading." (Paragraph 8.96).

Statistics Canada - 190 - Collected Articles of Kishori Lal

It was recognized that there was no quid pro quo for these payments since the fine was merely a payment for an offence and no service was obtained from it. Therefore, fines should not be considered payment for a service but rather as a transfer from persons, as was the existing practice. Fines paid by business should be recorded as transfers from business, but data limitations did not allow the splitting of fines between persons and businesses in a straightforward manner.

The 1993 SNA further notes: "In principle, interest charged on overdue taxes or fines, or penalties imposed for the attempted evasion of taxes, should be recorded separately and not as taxes. However, it may not be possible to separate payments of interest, fines or other penalties from the taxes to which they relate, so that they are usually grouped with taxes in practice." (Paragraph 8.51). Where interest was charged on overdue accounts, it should be recorded separately, i.e., not as taxes.

Change: The 1993 SNA recommendation has been adopted. Fines and penalties paid by persons are now treated as compulsory current transfers. Fines and penalties imposed on overdue taxes of businesses are now treated as taxes. Interest is recorded separately, whenever possible, as part of financing charges. See also Item 3.5.

2.7 Payments in Respect of Losses of Government Business Enterprises

Issue: Should payments to cover losses of a GBE by the parent government be treated as a subsidy or as a capital transfer?

Pre-Revision Treatment: Payments to cover losses of a GBE by the parent government were treated as negative trading profits remitted to government.

Discussion Notes: The 1993 SNA recommends that transfers paid to public corporations to compensate for persistent losses incurred on their productive activities as a result of charging prices lower than their costs of production, as a matter of government policy, be treated as subsidies (paragraph 7.78c).

On the other hand, payments to public corporations made to cover large and unanticipated operating deficits accumulated over two or more years should be treated as capital transfers (paragraph 10.141). These were previously treated as negative remitted trading profits.

Change: The 1993 SNA recommendation to treat payments to cover losses on current production as a subsidy and those to cover large operating deficits accumulated over two or more years as a capital transfer has been adopted.

2.8 Forgiveness and Write-off of Loans

Issue: Should the write-off and forgiveness of government loans be treated as a current transfer, a capital transfer or a valuation adjustment?

Statistics Canada - 191 - Collected Articles of Kishori Lal

Pre-Revision Treatment: Write-off and forgiveness of loans were handled on a case by case basis. In some instances, they were treated as a subsidy, in others as a capital transfer or as a balance sheet adjustment. Decisions were not always the same in the two systems. In the CSNA, many cases were treated as balance sheet adjustments, while in the FMS, the majority were treated as transfers. Public Institutions Division was splitting only transfers to business into subsidies and capital transfers; transfers to other sectors were not so divided.

Discussion Notes: The 1993 SNA makes a distinction between loans that are forgiven or cancelled by mutual agreement (paragraph10.139) and loans that are unilaterally written off by the creditor (paragraph10.140). The first case is to be treated as a capital transfer from the creditor to the debtor. In the second case, there is no entry in either the capital or the financial accounts. The adjustment to assets and liabilities is recorded in the Other Changes in the Volume of Assets Account.

It would be difficult to identify mutual agreement write-offs versus unilateral ones. In practice, all write-offs by the Government sector were treated as mutual agreement ones (unless otherwise specified) and all write-offs by the business sector as unilateral (unless otherwise specified). The majority of the cases in Canada would appear to fall into the first group and should be treated as capital transfers. This was the treatment in the FMS. Each case should be treated the same in both the FMS and the CSNA.

Change: The 1993 SNA recommendation to treat debts forgiven by governments as capital transfers was adopted in both the FMS and the CSNA, on the assumption that they are being forgiven by mutual agreement. As for debts written off by the business sector, they are considered unilateral write-offs of debt by the creditor, and are treated as balance sheet adjustments in the Other Changes in the Volume of Assets Account.

2.9 Transfers from Provincial to Federal Government

Issue: Should transfers from provincial to federal government be recorded on a net or gross basis?

Pre-Revision treatment: The FMS was recording these transactions on a gross basis, showing the full value of these transfers. In the CSNA, such transactions were netted against the federal transfers to provincial government and the latter presented on a net basis.

Discussion Notes: The consequence of the CSNA treatment was that the flows from a lower to a higher level of government were not shown. It was deemed important to record the gross flows for analytical purposes. Examples of transactions netted out included provincial payments to the federal government for the Crop Reinsurance Fund Account and the Agricultural Commodities Stabilization Accounts.

Change: Provincial transfers to the federal government are now recorded on a gross basis in the CSNA, as has been the practice in the FMS.

Statistics Canada - 192 - Collected Articles of Kishori Lal

2.10 Revenues Netted Against Expenditure

Issue: Should revenues of Government sector units, generally sales of goods and services, be netted against expenditures of these units?

Pre-Revision Treatment: The FMS always showed gross expenditures and gross revenues of Government sector units, whereas the CSNA typically presented expenditures on a net basis. Government sector units include departments and agencies as well as autonomous organizations, Boards, Commissions and Funds.

Discussion Notes: The FMS aims to provide information on all the activities of government and, thus records gross revenues and expenditures. Since the net data can always be derived from the gross, it was agreed to record gross flows.

Change: The CSNA now follows the FMS practice of recording revenues and expenditures on a gross basis.

2.11 Royalties

Issue: Should revenues from royalties be classified as sales of services, as indirect taxes or as investment income?

Pre-Revision Treatment: Royalties on books, recordings, etc., were treated as sales of services. Royalties on land and extraction of resources were treated as investment income. This treatment was introduced in the 1971 CSNA Historical Revision, before which such royalties were treated as indirect taxes.

Discussion Notes: There are essentially 2 types of royalties:

(1) Royalties on books, recordings, films, etc. There was agreement that these royalties are sales of services, as recommended by the 1993 SNA (paragraphs 6.146 and 7.92). This type of royalty is not usually received by government and is not the one at issue here.

(2) Royalties on land and subsoil assets. In the 1993 SNA, the revenues from the ownership of land and sub-soil assets are treated as rent, that is, as property income:

"Owners of land and subsoil assets may put them at the disposal of other units by arranging contracts or leases under which the tenants, or users of the assets, agree to pay the owners property incomes in the form of rents. The regular payments made by the lessees of subsoil assets are often described as royalties, but they are treated as rents in the System." (Paragraph 7.87).

Discussion centred on whether such revenues are investment income (property income in the terminology of the 1993 SNA) or sales of services. Since they derive from the government=s ownership of the land, rather than from its power to tax, it was decided to continue treating them as investment income.

Statistics Canada - 193 - Collected Articles of Kishori Lal

In the 1993 SNA, the payments for the use of land and sub-soil assets are part of the value added of the industry using such assets (also called >industry of origin=). This treatment does not require creating an industry to produce resource royalties. As a result, it would lead to the elimination of the "Royalties" industry in the Canadian Input-Output tables. Royalties would now be paid out of the surplus of the using industry. GDP would not change, but the value added of the using industries (mining, forestry...) would increase with the reallocation of rent from intermediate consumption to primary input, with an identical offset in the output and the value added of the major industry group >Finance, insurance and real estate=, to which the >Royalties= industry belonged.

Change: Royalties on books, recordings, films etc. remain treated as sales of services. Royalties on land and subsoil assets remain treated as investment income. However, they are now part of the value added of the using industry and the "Royalties" industry in the Canadian Input-Output tables is eliminated.

Unfortunately, rents paid by farmers for land leases have remained treated as payments for services. This is an oversight that will be corrected at the next opportune time.

2.12 Supplementary Labour Income (SLI)

Issue: How should social contributions in cash (designated as >SLI= in the CSNA) paid by a government on behalf of its employees into government accounts, and those paid by one government for another employing government, be treated?

Pre-Revision Treatment: SLI paid on behalf of one=s own employees into government accounts was consolidated as an inter-fund transaction in the FMS. Often these payments for departments were recorded in the accounts of a central agency. FMS has shown these under the function >General Services=. There were also instances, particularly in education, where one government paid some portion of SLI on behalf of another government unit.

Discussion Notes: The practice in the CSNA, in accordance with international standards, has been to record employer social contributions separately from wages and salaries. Given the decision to harmonize both systems, it was agreed that the FMS would record these payments on a gross basis and not eliminate them as inter-fund transactions. Also, the SLI paid by central agencies should be distributed by function.

There were instances where the SLI was paid by a government other than the employing government. For example, some provincial governments made the contributions to the teachers' pensions, while the school boards paid their salaries. Such contributions were recorded as SLI paid by the provincial government in the CSNA, according to the practice of attributing expenditures to the government that made them. However, for analytical purposes, the wages and salaries and the SLI should be recorded in the same sub-sector. This can be rectified by treating pension plan contributions on behalf of employees in another sub-sector as transfers to that sub-sector. The SLI would subsequently be deemed to have been paid by the sub-sector receiving the transfer, thereby aligning the wages and salaries and the related social contributions.

Statistics Canada - 194 - Collected Articles of Kishori Lal

Change: SLI paid on behalf of one=s own employees into government accounts is no longer treated as an >inter-fund= transaction in the FMS and thus is not eliminated in consolidation. As well, the FMS now distributes the SLI by function. Social contributions made by one government on behalf of another employing government are now treated as a transfer to that government, which is deemed to make the payments. Both these changes also apply in the CSNA. See also Item 2.33.

2.13 Gain (or Loss) on Foreign Exchange

Issue: Should gains (or losses) on foreign exchange be treated as revenue (or expenditure) or as an adjustment on the balance sheet?

Pre-Revision Treatment: The FMS recorded realized gains (or losses) as revenue (or expenditure), and unrealized gains (or losses) on the balance sheet, whereas the CSNA recorded both realized and unrealized gains (or losses) on the balance sheet.

Discussion Notes: The 1993 SNA notes: "However, as holding gains are recorded on an accrual basis in the System, the distinction between realized and unrealized gains, although useful for some purposes, is not so important in the System and does not appear in the classifications and accounts". (Paragraph 12.72).

In the 1993 SNA, these gains (or losses) are recorded in the revaluation account. The IMF=s GFS Manual (paragraph 5.5) has them in the financing account (schedule of sources and applications of funds). They are not part of income (or expenditure) in either system.

Change: The FMS will record both realized and unrealized gains (or losses) in the forthcoming Financing Account, a treatment similar to the one applied in the CSNA. See also Item 2.14.

2.14 Gain (or Loss) on Sale of Financial Investments

Issue: Should gains (or losses) on sale of financial investments be recorded as revenue (or expenditure) or on the balance sheet?

Pre-Revision Treatment: The FMS recorded these gains (or losses) as revenue (or expenditure), whereas the CSNA recorded them on the balance sheet.

Discussion Notes: As in the case of gains (or losses) on foreign exchange, the 1993 SNA recommends to record these on the Balance sheet, and the IMF=s GFS Manual, in the financing account. It was agreed that the FMS should follow the 1993 SNA guidelines.

Change: FMS will record such gains (or losses) on investments in the forthcoming Financing Account, a treatment similar to the one applied in the CSNA. See also Item 2.13.

Statistics Canada - 195 - Collected Articles of Kishori Lal

2.15 Supplementary Period Adjustments B Federal Government

Issue: Supplementary period adjustments are transactions recorded after March 31, but attributed to the previous fiscal year in the Public Accounts of Canada. How should they be treated?

Pre-Revision Treatment: The CSNA recorded all supplementary period transactions not dealing with sales or purchases of goods and services in the first quarter of the next fiscal year. Sales and purchases of goods and services were split 50%-50% between the fourth quarter of one fiscal year and the first quarter of the next. The FMS followed the Public Accounts.

Discussion Notes: The existing practice in the CSNA is not documented. Given that the Public Accounts are audited financial statements of government transactions, it was agreed that one would require considerable justification to move transactions to another fiscal year. It was recognized that on a conceptual basis there could be an argument for spreading some transactions over the fiscal year but this would be very difficult to do in practice.

Change: All supplementary period transactions are now recorded in the last quarter of the fiscal year (first quarter of the calendar year) in which they appeared in the Public Accounts, both in the FMS and the CSNA.

2.16 Seigniorage

Issue: Should seigniorage be treated as a revenue or as a liability? Pre-Revision Treatment: The revenues derived by the Department of Finance from seigniorage, i.e., the difference between the face value and the production costs of coins, was treated as a revenue in the FMS, but not in the CSNA.

Discussion Notes: The 1993 SNA treats both coins and currency notes as a liability of the government. The IMF's GFS manual treats seigniorage as revenue. This inconsistency in treatment was raised by Statistics Canada with the International Inter- Secretariat Working Group on National Accounts (ISWGNA). In answer to Statistics Canada, ISWGNA recommended that the 1993 SNA treatment be adopted in the forthcoming IMF=s GFS Manual (see SNA News and Notes, #4, July 1996). Note that coins are already classified as a liability of the government in the Financial Flows and the Balance Sheet of the CSNA.

Change: In accordance with the recommendation of the 1993 SNA, seigniorage is no longer treated as a revenue, and coins are recorded as a liability in the FMS.

2.17 Federal Consolidated Special Purpose Accounts

Issue: Should all special purpose accounts be recorded in the FMS and the CSNA?

Statistics Canada - 196 - Collected Articles of Kishori Lal

Pre-Revision Treatment: The revenue and expenditure of several special purpose accounts (listed below) were not recorded in the CSNA, though they were in the FMS. Neither the FMS nor the CSNA included these accounts in the federal government balance sheet.

Discussion Notes: These accounts are not significant and it is difficult to get their quarterly data, as required for the CSNA. Some of them were not even considered part of the public sector universe under the CSNA. With the 1997 Historical Revision, the two universes are harmonized. Consequently, while these accounts have a minimal importance, they should be recorded in the CSNA. The accounts in question are:

Canadian Heritage Alexander Graham Bell National Historic Site Claudia de Hueck Bequest Account Fort Langley Legacy Foundation - Donations Marconi Celebration Trust Fund National Archives of Canada - Donations National Battlefields Commission - Trust Fund Account National Library - Special Operating Account

Environment Endangered Species - Donations Fish Habitat Restoration Account

Finance Canadian Commercial Bank and Northland Bank Holdback Account

Fisheries and Oceans Supplementary Fines Fish Account

Governor General Friends of Rideau Hall Account

Health Sioux Lookout Zone Hospital Medical Research Council - Donations for Research

Human Resources Development Canadian Centre for Occupational Health and Safety - Donations

Indian Affairs and Northern Development Environmental Studies Research Fund

Statistics Canada - 197 - Collected Articles of Kishori Lal

Industry National Research Council of Canada - H.L. Holmes Fund Natural Sciences and Engineering Research Council - Donation Trust Fund

Natural Resources Environmental Studies Research Fund

Privy Council Canadian Transportation Accident Investigation and Safety Board – Flight Recorder Software Systems Account National Round Table on the Environment and Economy - Donations

Public Works and Government Services Seized Property Proceeds Account

Transport Fines for the Transportation of Dangerous Goods

Change: All federal consolidated special purpose accounts are now included in the federal government revenue, expenditure and balance sheet in both systems.

2.18 Adjustment to Income Taxes of Life Insurance Companies

Issue: Should the CSNA allocate any portion of corporate income tax collections from life insurance companies to personal income taxes?

Pre-Revision Treatment: The CSNA was allocating a small portion of the corporate income taxes collected from life insurance companies to the household sector. This adjustment was not done in the FMS. Discussion Notes: The assets associated with life insurance policies are deemed to be assets of households (see 1993 SNA, Annex IV, page 573, paragraph 22). The income from these assets was attributed to households and was recorded on a gross basis, before deduction of the corporate tax. This tax was then shown as being paid by households. It was argued that one could as easily record the income earned net of corporate taxes.

Change: The CSNA no longer allocates a portion of the corporate income tax paid by insurance companies to personal income taxes; instead, life insurance companies are deemed to pay the entire tax and the income attributed to the household sector is recorded net of that tax.

2.19 Quebec Youth Allowances

Issue: How should the compensatory payment by the Quebec Government to the Federal Government related to the defunct Quebec Youth Allowance Program be recorded?

Statistics Canada - 198 - Collected Articles of Kishori Lal

Pre-Revision Treatment: When the Quebec Youth Allowance Program existed, the federal government transferred 3.0 tax points to the Quebec government to support it. After it ended in 1974, there was no reason for the Federal Government to continue transferring the tax points. An arrangement was made whereby the Quebec Government would continue receiving the tax points, but would remit an equivalent amount to the Federal Government. Thus, in the Quebec Public Accounts, income tax revenue is shown as including these 3.0 tax points and the remittance to the federal government is recorded as a negative federal transfer. The federal Public Accounts show a negative transfer to the province. The CSNA recorded these transactions as reported in the Public Accounts of the two governments. The FMS recorded the Quebec provincial income and expenditure on a net basis, and the Federal income and expenditure on a gross basis. In the FMS, the 3.0 tax points were added to the federal personal income tax revenue and an equivalent expenditure added as equalization payment to Quebec. The result was that the amount corresponding to the tax points was counted twice, once as Federal personal income tax and once as Quebec personal income tax.

Discussion Notes: It was agreed that the grossing up of the revenue and expenditure of the Federal Government was not a neat solution. A simpler solution would be to treat the amount as a transfer from the Quebec Government to the Federal government. This would also be consistent with the earlier decision concerning transfers from lower levels of government (see Item 2.9).

Change: Both the FMS and the CSNA now record the Quebec remittance made in counterpart of the 3.0 tax points transferred by the Federal Government as a general purpose transfer from the Quebec government to the Federal Government. This change has no impact on the surplus or deficit position of either government.

2.20 Federal Payment to Quebec for the Purchase of Frigates

Issue: How should the federal government=s payment to MIL Davie shipyards through the Quebec government in 1992 be recorded?

Pre-Revision Treatment: This transaction was treated as a purchase of defence goods from the Quebec government in the CSNA. However, because the money actually went to the Quebec government=s general revenue fund, it was treated as a transfer to the Quebec government in the FMS.

Discussion Notes: At the time, MIL Davie was building part of the new frigates for the Canadian armed forces. The discussion hinged on whether it was possible for a government to subsidize its own purchase of goods and services. It was generally felt that the federal payment should not be considered a subsidy, but an increase in the price of the goods being purchased. Since the Quebec government was not purchasing anything from MIL Davie, its payment would be a subsidy.

Change: Both the FMS and the CSNA now treat the federal payment as a purchase of goods from the Quebec government by the Department of National Defence, rather than as a transfer to that government. The Quebec government=s payment to MIL Davie remains classified as a subsidy.

Statistics Canada - 199 - Collected Articles of Kishori Lal

2.21 Adjustments to Employment Insurance Premiums

Issues: Should the EI premiums be (a) adjusted to an accrual basis and (b) benchmarked on the Public Accounts?

Pre-Revision Treatment: The CSNA adjusted monthly EI premiums to an accrual basis, while the FMS adhered to the reported values. Neither of the two systems benchmarked premiums to the ones in the Public Accounts.

Discussion Notes: The Public Accounts show EI premiums after adjustments for over- or under- payments related to previous years. In recent years, these adjustments have been getting larger. The CSNA had been reversing these prior-period adjustments to better reflect employer/employee contributions in the current period. The resulting estimates were therefore closer to an accrual basis. The FMS recorded the monthly premiums as reported. In addition, neither of the two systems benchmarked the monthly premiums on a cash basis to the fiscal year figures in the Public Accounts.

Change: The CSNA and the FMS now benchmark monthly premiums on a cash basis to the figures in the Public Accounts. In addition, the CSNA continues to reverse the prior-period adjustments, in order to show the premiums on an accrual basis.

2.22 GST Low Income Refundable Tax Credits

Issue: Should the monthly data on GST low income refundable tax credits be benchmarked on the Public Accounts?

Pre-Revision Treatment: Both the FMS and the CSNA add back these credits to the GST in order to show the latter on a gross basis. The FMS source is the Public Accounts. The CSNA source is the monthly data Revenue Canada, which do not add up to the figures in the Public Accounts used in the FMS.

Discussion Notes: It was agreed that the monthly data should be benchmarked on the Public Accounts.

Change: The monthly data on GST low-income refundable tax credits, like those on the GST, are now benchmarked to the Public Accounts.

2.23 Canadian Dairy Commission

Issue: Should the quarterly data from the Canadian Dairy Commission (CDC) used in the CSNA be benchmarked to the figures in the Public Accounts?

Pre-Revision Treatment: The FMS incorporated the data from the Public Accounts on a gross basis. The CSNA used the net quarterly data supplied by the CDC, without benchmarking them to the Public Accounts.

Statistics Canada - 200 - Collected Articles of Kishori Lal

Discussion Notes: It was agreed that the quarterly data used in the CSNA should be benchmarked on the Public Accounts.

Change: The CSNA now benchmarks the quarterly data from the Canadian Dairy Commission to the figures in the Public Accounts.

2.24 Doubtful Taxes in Quebec

Issue: The Quebec Government reports gross income tax liabilities and makes a provision for doubtful taxes. Actual losses are charged against that provision. Should the income tax receipts be recorded on a gross or a net basis, that is, before or after deduction of that provision?

Pre-Revision Treatment: The CSNA had been recording the gross taxes as reported in the Quebec Public Accounts. The FMS had been recording the net taxes.

Discussion Notes: Taxes should be recorded on a gross basis according to the 1993 SNA. However, all provincial governments, except that of Quebec, report income taxes on a net basis in the Public Accounts. In the interest of comparability between provinces, it was deemed preferable to record the Quebec income taxes on a net basis. Change: Income tax receipts for Quebec are now shown on a net basis, after deduction of the provision for doubtful taxes, in both systems.

2.25 Consumption Taxes Levied and Paid by the Same Government

Issue: Should consumption taxes be recorded on a gross basis or a net basis, that is, before or after deduction of consumption taxes paid by the government levying the tax?

Pre-Revision Treatment: These taxes were recorded on a net basis in the FMS, and on a gross basis in the CSNA.

Discussion Notes: Only two provincial governments record the taxes they pay to themselves. In the FMS, consumption taxes are reported on a net basis for these two provincial governments and on gross basis for the other ones. In the interest of harmonization and inter-provincial comparability, it was deemed preferable to show all consumption taxes on a gross basis. The GST paid by federal departments is fully refunded to them and thus not affected by this decision.

Change: Consumption taxes are now recorded on a gross basis, that is, before deduction of taxes paid by governments to themselves, in both systems.

2.26 Alberta Royalty Tax Credit

Issue: The Alberta corporate tax is reported in the Alberta Public Accounts net of the royalty tax credit. How should the credit be recorded?

Statistics Canada - 201 - Collected Articles of Kishori Lal

Pre-Revision Treatment: The FMS grosses up the corporate tax by the amount of the tax credit and imputes an equivalent expenditure under >Resource conservation B industrial development=. The CSNA records the tax as reported.

Discussion Notes: The Alberta royalty tax credit is not tied to a corporation=s tax return. It depends on the average price of oil and is capped at a certain level of total royalties paid. Since it has nothing to do with the corporate tax, the latter should be recorded on a gross basis. The tax credit is really a royalty refund and should be deducted against royalties.

Change: Alberta royalty receipts are now recorded after deduction of this credit/refund in both the FMS and the CSNA. The Alberta corporate taxes have been grossed up by the value of the royalty tax credit, so that the surplus or deficit remains unchanged. See also Item 2.2.

2.27 Refundable Credit for Losses on Mining Operations in Quebec

Issue: Should the Quebec mining tax be recorded on a gross basis (that is, before deduction of the refundable credit for losses), and this credit be treated as a subsidy? Discussion Notes: The Quebec mining duties system, reformed in 1985 to allow mining companies to obtain a positive cash flow during their initial years of operation before they became profitable, did not work out as anticipated. The credits exceeded the mining tax collections each year from 1987-88 on. A new mining duties system was announced in the 1993-94 budget and described in the 1994-95 budget. Its aim was to lower the tax credit for losses, deemed too generous.

Up to 1987-88, the mining tax was presented in the Quebec Public Accounts as a revenue of the Department of Energy and Resources, and the credits for losses, as an expenditure under Program 10, Component 2, of that department. In 1988-89, the Quebec government decided to present the mining tax on a net basis. Starting in 1987- 88, however, FMS treated this expenditure as a capital transfer rather than as a production subsidy.

Change: In keeping with the decision to record taxes on a gross basis (i.e., before deduction of the related refundable tax credits) whenever possible, the Quebec mining tax is now recorded on a gross basis. The refundable tax credit is now treated as a subsidy in both systems, and shown under >Resource conservation B Industrial development= in the FMS classification by function.

2.28 Rent Supplements Paid by Provincial Housing Authorities to Landlords

Issue: Should rent supplements paid by Provincial Housing Authorities to landlords be treated as an expenditure on services or as a subsidy?

Pre-Revision Treatment: Provincial Housing Authorities, then part of the business sector, made payments directly to landlords to "supplement" the rent paid by the tenants living in subsidized housing. In this sense, these payments were akin to a purchase of rental services by the Provincial Housing Authorities.

Statistics Canada - 202 - Collected Articles of Kishori Lal

Discussion Notes: Provincial Housing Authorities have now been transferred from the Business sector to the Provincial Government sub-sector (see Item 1.9). It could be argued that the rent supplements they pay to landlords serve to lower the price of lodging and therefore could be considered production subsidies. However, since these payments by the Housing Authorities reduce the rent of a specific group of tenants and do not constitute a general reduction of rent, they should continue to be treated as an expenditure on services.

Change: Rent supplements continue to be treated as an expenditure on services, but by the Government sector instead of the business sector as the housing authorities have been transferred from the latter to the former.

2.29 Lot Levies

Issue: Should lot levies be treated as indirect taxes or sales of services? Pre-Revision Treatment: They were treated as sales of services in the FMS and as indirect taxes in the CSNA.

Discussion Notes: While lot levies could have been linked to specific services in the past, they now resemble taxes and should be treated as such.

Change: Lot levies are now treated as indirect taxes in both systems. In addition, they remain in a separate category in the FMS, due to the requirements of Federal-Provincial Fiscal Arrangements.

2.30 Local Capital Transfers

Issue: Capital transfers from local governments to their own GBEs, other businesses and persons are recorded in the FMS, but not in CSNA. Discussion Notes: It was agreed that there was no reason to leave out such capital transfers in the CSNA.

Change: The CSNA now records all local capital transfers.

2.31 Miscellaneous Revenue B Local Government

Issue: In the FMS, one of the items under local government (general administration component) revenue is >miscellaneous revenue=. The same item was split in the CSNA as follows: 70% to sales of goods and services, 10% to transfers from persons and 20% to inter-fund transactions.

Discussion Notes: These percentages were adopted at the time of the 1986 CSNA Historical Revision, on the assumption that the category >miscellaneous revenue= contained items that should be excluded from revenue. The problem occurs when municipalities report revenues under the heading >Other=, which sometimes includes a >surplus from previous year= or an >adjustment to previous year's figures=. Such items should be treated as inter-fund transactions in both the FMS and the CSNA. Yet, in the FMS, >Other= was recorded as miscellaneous revenue.

Statistics Canada - 203 - Collected Articles of Kishori Lal

A re-examination of available data has indicated that the percentages estimated at the time of the 1986 CSNA Historical Revision were reasonable. It was agreed that FMS should adopt the same percentages.

Change: The FMS has now adopted the CSNA practice of splitting miscellaneous revenue as follows: 70% to sales of goods and services, 10% to transfers from individuals and 20% to inter-fund transactions.

2.32 Sales of Goods and Services to Other Governments

Issue: Part of the revenue from sales of goods and services to other governments were recorded as transfers in the CSNA on the grounds that they were more like grants than sales of services.

Discussion Notes: Based on earlier research, the Input-Output Division reclassified these revenues from >sales of goods and services= to >transfers=. An inter-divisional working group reviewed the relevant documentation and concluded that the most of these revenues were sales of goods and services.

Change: The CSNA has now adopted the FMS practice to show all such revenues as sales of goods and services.

2.33 Teachers= Pension Plans B British Columbia

Issue: The British Columbia government used to make the employers= contributions to the teachers= pension plans. Since 1992/93, it makes a transfer to local school boards, which in turn make the contributions to the pension plans. FMS has continued to show the latter as paid by the provincial government, whereas the CSNA has shown them as made by local school boards beginning in 1992/93.

Discussion Notes: The Education, Culture and Tourism Division (ECTD) continues to show the contributions as coming from the Provincial Government in order to maintain inter-provincial comparability, even though the payments are made by local school boards. The ECTD data are used as is in the FMS, but are corrected in the CSNA. In effect, as of 1992/93, the Input-Output Division takes these contributions out of the provincial expenditure as recorded by ECTD and transfers them to local school boards.

Change: The FMS has now adopted the existing CSNA practice to show these contributions to pension plans in the >school boards= component of the Local government sub-sector and to record the provincial expenditure as a transfer to school boards. See also Item 2.12.

2.34 Property Tax Credits - British Columbia, Alberta and Manitoba

Issue: The British Columbia, Alberta and Manitoba governments have programs to lower the effective tax rate of designated homeowners and renters. How should these programs be treated?

Statistics Canada - 204 - Collected Articles of Kishori Lal

Pre-Revision Treatment: The FMS added back these tax credits to the property taxes, showing the property taxes on a gross basis, in other words, before deduction of the credits. In the CSNA, these tax credits were classified as a general transfer from provincial government to local government.

Discussion Notes: In all cases, the taxpayer does not actually pay the full property tax. In Alberta and Manitoba, the provincial government transfers the equivalent revenue to local governments. In British Columbia, the provincial government collects property taxes and transfers to school boards an amount equivalent to this >tax reduction=. In some other provinces, property tax assistance is granted by way of refundable tax credits handled through the income tax system.

Change: The FMS has adopted the CSNA practice of treating such credits as transfers from provincial government to local governments, and thus no longer adds them back to local property taxes.

2.35 Education Sales to Other Levels of Government B Quebec

Issue: Quebec school boards receive money from the federal and provincial governments for adult training. Are these other governments receiving services in return or are they simply giving grants (transfers) to the school boards?

Pre-Revision Treatment: The revenues of school boards from adult training were treated as transfers from the federal and provincial governments in the FMS, and >sales of goods and services= in the CSNA.

Discussion Notes: School boards in Quebec offer adult training courses and the revenues and expenditures associated with the program appear in their annual reports. In the FMS, federal and provincial government payments for adult training were considered grants to school boards. In the CSNA, they were treated as sales of services to other governments. Since school boards are in fact performing a service, the related revenue should be treated as sales of services. Similar services by school boards in other provinces are already treated as such..

Change: The FMS has now adopted the CSNA practice of classifying Quebec school board revenues from adult training as sales of services.

2.36 Grants in Lieu of Taxes

Issue: Should payments by the federal government and provincial governments to local governments in lieu of property taxes be considered as grants or indirect taxes?

Pre-Revision Treatment: Payments by other governments to local governments in lieu of property taxes were treated as grants (i.e. transfers).

Statistics Canada - 205 - Collected Articles of Kishori Lal

Discussion Notes: As all government in Canada is exercised in the name of the Crown, and the Crown cannot tax itself, a legal arrangement was established whereby local governments received grants in lieu of taxes on the land and buildings of other governments. A business owning the same land or property would pay property taxes. The economic reality was that both paid taxes, but in one instance it was called a tax and in the other, a grant in lieu of tax. The two systems should reflect the economic reality.

Change: Grants in lieu of taxes paid by the federal government and provincial governments are now treated as indirect taxes in both the CSNA and the FMS.

2.37 Allowances for Losses on Government Loans

Issue: Should government loans to government business enterprises (GBEs) and others (governments and businesses) be shown net of allowances for losses?

Pre-Revision Treatment: The FMS showed government loans to GBEs and others net of the allowances for losses, whereas the CSNA did not. The rationale for ignoring the allowances in the CSNA was that loan assets must be reconciled to loan liabilities and a loan liability should be reported as the amount for which a borrower is liable (that is, gross of allowances). However, both the CSNA and the FMS recorded loan liabilities of GBEs on a gross basis.

Discussion Notes: There is general agreement that if a government reports an allowance for loss against a loan to a GBE for which it does not expect to be paid back, then the loan asset would be best shown net of this amount in both the FMS and the CSNA. This reflects standard accounting practice.

Change: The CSNA has now adopted the practice of deducting allowances for losses on government loans to government business enterprises and others, thus showing these loans on a net basis. As a result, the corresponding liabilities are now also shown net of allowances in both the FMS and the CSNA.

2.38 Liabilities of the Federal Government in Respect of Foreign Loans of Government Business Enterprises

Issue: Should allowances for losses on foreign loan assets of the Export Development Corporation (EDC) and the Canadian Wheat Board (CWB) also be recorded as a liability of the federal government?

Pre-Revision Treatment: In the FMS, this liability was ignored. In the CSNA, the liability was deducted from the federal government's assets, specifically, from its investment (equity) in EDC and CWB.

Statistics Canada - 206 - Collected Articles of Kishori Lal

Discussion Notes: The balance sheet of the federal government shows loans to enterprises and to other domestic entities, as well as loans to foreigners (including loans administered by EDC). All these are shown net of allowances for losses in the FMS and the CSNA (see item 2.37).

EDC makes additional loans to foreign countries, and so does the CWB. However, EDC deducts allowances from the loan assets, while CWB does not, because of the federal guarantee attached to the loans. The FMS followed this accounting for the allowances. In the CSNA, the allowances were deducted from the loan assets of both EDC and CWB.

EDC and CWB borrow from financial institutions to finance the loans they make to foreign countries. The federal government guarantees this borrowing and debt. As a result, it records in the Public Accounts a liability on its balance sheet equal in value to the allowances on the foreign loans of EDC and CWB. This liability relates to the probability of having to finance the repayment of funds borrowed by EDC and CWB.

The CSNA excluded this liability, because it was viewed as a contingent liability. However, the CSNA did reflect the decline in value of the loan assets of the enterprises as a reduction in the value of the government=s investment in EDC and CWB.

Change: Both the CSNA and the FMS have now adopted the following presentation. A federal government liability is recorded in respect of the allowances for losses against the loan assets of EDC and the CWB; and a corresponding claim on the federal government, equal to the allowances, is shown on the balance sheet of those enterprises.

2.39 Government Liabilities in Respect of Employee Pension Plans: Autonomous Plans with Invested Assets

Issue: Should the actuarial deficiencies of employee pension plans with invested assets and that are managed by trustees be recorded as liabilities of the provincial governments?

Pre-Revision Treatment: Both the FMS and the CSNA ignored the actuarial deficiencies.

Discussion Notes: In Canada, employee pension plans with invested assets and that are managed by trustees independent of the employer are referred to as >trusteed pension plans=. Such plans are separate institutional units and are thus autonomous schemes. The 1993 SNA treats trusteed pension plans as private funded schemes (paragraph 8.63 b).

These funds consist of a reserve of money (invested in securities) that employees and their employer have contributed towards future payments of pension upon retirement of the employees. The reserve becomes trusteed when the money is deposited with, and managed by, an independent group or organization that acts as a trustee. An actuarial appraisal of the fund is done from time to time to determine if the fund is adequate to fulfil its present and future obligations. Actuarial deficits are liabilities of the employer.

Statistics Canada - 207 - Collected Articles of Kishori Lal

Most provincial governments guarantee any deficiency, in the event that the assets of the plan are insufficient to meet benefit payments or cover administration costs. Some governments do not explicitly recognize a liability or record it on their balance sheet. Since private sector employers are required to record these actuarial deficiencies as liabilities, it seems reasonable to treat governments in a similar fashion.

Change: The actuarial deficiencies of the trusteed pension plans of government employees will be recorded as liabilities for all provincial governments in both the FMS and the CSNA.

The implementation of this change will be delayed in the CSNA, until other OECD countries implement the 1993 SNA.

2.40 Government Liabilities in Respect of Employee Pension Plans: Non-autonomous Plans without Invested Assets

Issue: Should a liability of the federal government and some provincial governments be recorded for the accrued liability and/or actuarial deficiency in respect of employee pension plans without invested assets?

Pre-Revision Treatment: The FMS recorded government liabilities (representing the accumulated amount of contributions and investment income credited) in respect of employee pension plans without invested assets, but the CSNA did not. Both the FMS and the CSNA ignored any actuarial deficiency related to these plans.

Discussion Notes: In the government sector, all employer and employee contributions to pension plans without invested assets are deposited in the Consolidated Revenue Fund of the government. The government uses these funds to finance its current operations, but is responsible for all future pension payments. When pension payments are made, they come out of the Consolidated Revenue Fund. An actuarial estimate is generally made to determine the liability of the government to the plan.

Generally speaking, these are non-autonomous plans (under the control of the employer), with no special reserve (invested assets) from which to pay benefits. It seems reasonable to treat these employee pension plans of the federal and provincial governments in a similar fashion to autonomous plans with invested assets, since: (i) the governments are not likely to default; (ii) government employees think of these plans as part of their net worth.

Change: The FMS will record on the balance sheet of governments all liabilities (even actuarial deficiencies) in respect of their employee pension plans without invested assets. This covers cases where these liabilities may not be recorded on the balance sheet, but acknowledged in the notes of government public accounts.

The implementation of this change will be delayed in the CSNA until other OECD countries implement the 1993 SNA.

Statistics Canada - 208 - Collected Articles of Kishori Lal

2.41 Treasury Bills

Issue: Should Treasury bills be recorded at par, that is, at face value, on the federal government balance sheet? How should the unamortized discount on Treasury bills be treated?

Discussion Notes: The unamortized discount on a Treasury bill is the difference between the face value of the bill and the amount received from its sale, difference that remains to be amortized over the life of the bill. Both the FMS and the CSNA record Treasury bills at par. In the FMS, outstanding Treasury bills are recorded as a gross liability and the unamortized discount, as an asset on the balance sheet of the federal government. In the CSNA, outstanding Treasury are also shown as a gross liability, but the unamortized discount is recorded as a reduction of >other liabilities=.

However, the 1993 SNA recommends to record Treasury bills at market value, in other words, to deduct the unamortized discount from the face value of the Treasury bill. This is also the current treatment of debt securities that are issued at discount. The CSNA treatment is not consistent with the 1993 SNA, but in effect yields the same lower liability for the government. The current IMF Manual on Government Finance Statistics recommends recording the total government debt at face value (i.e., on a gross basis) and the unamortized discount as a memorandum item in reconciling the change in total debt.

Change: It was decided not to adopt the 1993 SNA recommendation with respect to Treasury bills. In addition, the FMS has now adopted the CSNA practice to show Treasury bills at par and the unamortized discount as a reduction of liabilities.

2.42 Notes Payable to the IMF, Outstanding Cheques and Bank Overdrafts

Issue: In the FMS, notes payable to the IMF, outstanding cheques and bank overdrafts are recorded as liabilities, whereas in the CSNA they are shown as a reduction of assets. Thus, the asset >cash= is reduced by the value of outstanding cheques, the asset >government enterprise claims= by that of notes payable to the IMF and the asset >currency and other deposits= by that of bank overdrafts.

Discussion Notes: Notes payable to the IMF are part of Canada's official reserves. According to IMF guidelines (Manual on Government Finance Statistics, 1986, page 61), the FMS should show these notes as a reduction of assets, as is done in the CSNA, rather than as a liability.

There are two kinds of cheques outstanding, namely cheques in transit and cheques that have never been cashed (and may never be). To the extent that data allow, deposits must be recorded net of cheques in transit. A cheque drawn on the payer's account may be credited to the recipient's account, but not yet cleared and debited from the payer's account. This inconsistency is corrected by deducting cheques in transit. In the case of the federal government, however, there are cheques outstanding that may never be cashed. Such cheques should not be deducted from >deposits= or >cash=. They remain accounts payable.

Statistics Canada - 209 - Collected Articles of Kishori Lal

Both the FMS and the CSNA agree that overdrafts should be treated as liabilities.

Change: The FMS has now adopted the CSNA practice to record notes payable to IMF and other international organizations as a reduction of assets. The CSNA has now adopted the FMS practice to record cheques that have never been cashed as accounts payable and overdrafts as liabilities.

2.43 Foreign Subsidiaries of Government Business Enterprises

Issue: In the FMS, foreign subsidiaries of government business enterprises are consolidated with the parent enterprise, while the CSNA looks at Canadian operations of GBEs and only the investment in foreign subsidiaries is relevant. As a result, the FMS includes all foreign earnings in the revenue of GBEs, while the CSNA only takes into account dividends received.

Discussion Notes: The foreign subsidiaries of GBEs are not part of the Canadian business sector as they are non-resident units. Therefore, their financial statements should not be consolidated with those of the parent enterprise in the CSNA.

The Industrial Organization and Finance Division records the investment of Canadian parent companies in foreign subsidiaries on the basis of their share of equity and includes a portion of the net income of the foreign subsidiary in the revenue of the parent, based on that same share. This income is also reflected in the value of the investment in the subsidiary on the balance sheet of the parent. This treatment would seem an appropriate one for the FMS.

The CSNA records in the revenue of the parent only the dividends received from the foreign subsidiary, which are usually not equal to the parent=s share of the net income of the subsidiary based on its share of the subsidiary=s equity. The value of the investment in the foreign subsidiary shown on the balance sheet of the parent, however, does correspond to its share of the equity. This inconsistency between flows and stock is handled by an adjustment in the >Other Changes in Assets Account= in the CSNA.

Change: The FMS no longer consolidates foreign subsidiaries of GBEs with the parent enterprise. Instead, it now includes in the revenue of the parent the portion of the net income of its foreign subsidiary corresponding to its share of the equity. The CSNA treatment is unchanged. The difference between the two estimates of revenue is now shown as a reconciliation item.

2.44 Individual Funds of the Caisse de Dépôt et de Placement du Québec

Issue: In the FMS, both the General Fund and the Individual Funds of the Caisse de dépôt et de placement du Québec are classified in the business sector. In the CSNA, only the General Fund belongs to the business sector. The Individual Funds are classified in the household sector or the government sector.

Statistics Canada - 210 - Collected Articles of Kishori Lal

Discussion Notes: The General Fund and the Individual Funds are managed by the Caisse de dépôt et de placement du Québec. The CSNA shows the investment portfolio of the Individual Funds on the balance sheet of the government or the household sector. In the FMS, all the funds are shown on the balance sheet of the Caisse.

The Caisse=s General Fund includes diversified investments and is a pooled fund. The participating deposit holders include the Régie des rentes du Québec, the Société de l'assurance automobile du Québec and six other unit holders. These agencies/groups deposit money with the Caisse which, as any other deposit-taking institution, invests it in various securities to provide a return to the depositor.

The Individual Funds of the Caisse, such as the supplemental pension plan of the construction industry, also include diversified investments, but each >Fund= has only one >depositor=. The balance sheets of the depositors show deposits with the Caisse as assets, while the Caisse=s balance sheet shows corresponding deposit liabilities.

The CSNA views the Caisse as the trustee of the Individual Funds. Accordingly, it considers that the deposits of these Funds are not assets of the Caisse and that the investments of each fund should be reported by the depositors. The latter can only be obtained from the Caisse, however, since the financial statements of the depositors show only a deposit asset. The portfolios of the Individual Funds relating to government are recorded on the balance sheet of the government sector, and those of the other individual funds, on that of the household sector.

Change: The FMS has now adopted the CSNA practice to treat only the General Fund as an asset of the Caisse de dépôt et de placement du Québec and to show all Individual Funds on the balance sheet of their depositors.

2.45 Manitoba Telephone Investments for its Pension Obligations

Issue: The investment portfolio of Manitoba Telephone relating to its pension obligations and the corresponding pension liability are recorded on the balance sheet of the household sector in the CSNA, but on the balance sheet of the enterprise in the FMS.

Discussion Notes: Manitoba Telephone holds investments for its pension obligations. On its balance sheet, these assets are equal to the pension liability. The CSNA treats this arrangement as a trust account and thus records the value of the investment portfolio and the pension liability on the balance sheet of the household sector. The FMS would also accept this as a trust account.

Change: The FMS has now adopted the CSNA practice to show the investment portfolio and the pension liability of Manitoba Telephone on the balance sheet of the household sector.

Statistics Canada - 211 - Collected Articles of Kishori Lal

3. Labour Income

3.1 Retirement Allowances

Issue: Are retirement allowances an element of labour income? If so, should they be included in wages and salaries or supplementary labour income (SLI)?

Pre-Revision Treatment: Retirement allowances, designated as >retiring allowances= by Revenue Canada, were left out of labour income. It was not until Revenue Canada began capturing data from the T4A Form in 1990 that any hard data became available on retirement allowances.

Discussion Notes: Retirement allowances made by employers to employees are definitely labour costs and should be included in labour income. The question was whether to include them in wages and salaries or SLI. The argument leaned toward SLI (see 1993 SNA, paragraph 7.35c) but not without reservations. Another option was to develop a third category of labour income that could be described as unfunded benefits. This category would include retirement allowances and other lump sum payments such as maternity benefits and special leaves, all paid by the employers. Given the data limitations, it was recommended not to create this new category but rather to treat retirement allowances as a new element of SLI.

Change: Retirement allowances are now included in SLI beginning with 1990. Prior to that year, source data are not available.

3.2 Employer Payroll Tax B Ontario and Quebec

Issue: Should employer payroll taxes be included in SLI or treated as indirect taxes?

Pre-Revision Treatment: The welfare component of SLI included the payroll taxes paid by employers in Ontario and Quebec.

Discussion Notes: It was felt that any employer contribution not giving a specific economic benefit to the employee or his dependents should be considered an indirect tax, and not SLI. This would be in line with the 1993 SNA recommendation of treating payroll taxes on production the same way as taxes on buildings, land and other assets used in production (paragraph 7.21).

Change: The Ontario and Quebec employer payroll taxes are now excluded from the welfare component of SLI from the inception of each province's payroll tax scheme and classified as taxes on production.

3.3 Workers' Compensation B Hospital and Medical Expenditures

Issue: Should hospital and medical expenditures made by the workers= compensation boards on behalf of workers be deducted from their assessment revenue in estimating employers= contributions to workers= compensation?

Statistics Canada - 212 - Collected Articles of Kishori Lal

Pre-Revision Treatment: These expenditures were deducted from assessment revenue.

Discussion Notes: Given that the provincial workers= compensation boards are wholly funded by employers= contributions, hospital and medical expenditures made by the Boards on behalf of workers should not be deducted from assessment revenue in estimating employers= contributions to workers= compensation. Recent documentation received from the Ontario Workers= Compensation Board confirmed that the Board was responsible for the costs of all services provided to an injured worker. Payments for medical and hospitalization costs are made directly out of assessment revenue and employers are not billed separately.

Change: Hospital and medical expenditures are no longer deducted from the assessment revenue of workers= compensation boards in estimating employer contributions, starting in 1961.

3.4 Workers' Compensation B Unfunded Liabilities

Issue: Should surcharges levied on employers for unfunded liabilities of workers= compensation boards be treated as SLI?

Pre-Revision Treatment: Surcharges of this type were not treated as SLI.

Discussion Notes: In recent years, provincial workers= compensation boards have levied surcharges on employers for unfunded liabilities. It was recommended that these amounts be included in SLI.

Change: Surcharges levied on employers for unfunded liabilities of workers= compensation boards are now included in SLI.

3.5 Interest on Overdue Employers= Contributions to Pensions

Issue: Should interest on late payments of employers= contributions to pension funds be treated as SLI?

Pre-Revision Treatment: The interest levied by pension funds on late payments of employers= contributions has not been included in SLI since 1985.

Discussion Notes: Research has shown that only a few employers have late payment charges levied on them by the pension funds on a regular basis. Among them are the governments of Ontario and Quebec. These interest charges, although not legitimate contributions, are included in the reported contributions. It may be advantageous for an employer to pay the interest charge levied by the pension fund, instead of borrowing money to meet its obligations. This practice is not prevalent among employers. Those operating in this fashion are simply using this practice to finance their operations.

Change: Interest on late payments of employers= contributions to pension funds has now been excluded from SLI for the whole historical period. See also Item 2.6.

Statistics Canada - 213 - Collected Articles of Kishori Lal

3.6 Fishing B Adjustment for Undercoverage

Issue: Should an undercoverage adjustment be made to the T4 Control total for wages and salaries earned in fishing?

Pre-Revision Treatment: The adjustment for fishing was the difference between the estimate of wages and salaries in the fishing industry (called the >industry benchmark=) and the wages and salaries reported on the T4 file by employers classified in the fishing industry. The industry benchmark was based on the value of fish caught and price indexes for fuel and fish.

Discussion Notes: The T4 Control total adjustment represented the earnings of workers on owner- operated boats who were not issued a T4, but a T4F instead by the owner, designated as employer. If the owner was not a designated employer, no T4F was issued. Fishermen not receiving a T4, whether they received a T4F or not, were instructed to declare themselves as >self employed= with fishing income on their T1 declaration. The estimate for net income of unincorporated businesses from fishing comes from the T1 file. As a result, the adjustment for undercoverage was actually a double count, the same amount being recorded as labour income and net income from fishing.

Change: The T4 total has now been accepted as fully representing the wages and salaries in the fishing industry and the T4 Control total adjustment has been eliminated from 1981 onward.

3.7 Insurance Agents= Commissions B Adjustment for Undercoverage

Issue: Should an undercoverage adjustment be made to the T4 Control total for commissions of life insurance agents?

Pre-Revision Treatment: Such an adjustment was made to the T4 Control total.

Discussion Notes: This adjustment represented the commissions of life insurance agents who are not issued a T4. Only agents whose employer deducts CPP/QPP and/or income tax receive a T4. The others receive a T4A and declare themselves as self-employed (i.e. independent agents) with commission income in their T1 declaration. According to documentation dating from 1976, independent agents= earnings should not be part of wages and salaries, but were included since they appeared to be not covered elsewhere. The adjustment ($617 million for 1993) was based on a dated study from the Superintendent of Insurance indicating that 50% of agents had >self-employed status=, while the other 50% had their CPP contributions remitted by their employer. Commission income is part of the net income of unincorporated business. As a result, the adjustment for undercoverage was actually a double count, the same amount being recorded as labour income and net income from insurance commissions.

Change: The T4 Control total adjustment for undercoverage of insurance commissions has now been eliminated back to 1965, date of inception of the T4 Supplementary File.

Statistics Canada - 214 - Collected Articles of Kishori Lal

3.8 Tips B Adjustment for Undercoverage

Issue: Should an undercoverage adjustment be made to the T4 Control total for tips?

Pre-Revision Treatment: Estimates for tips on meals outside the home, accommodation as well as hairstyling and beauty services were added to the T4 Control total, as the tips were deemed unreported to taxation authorities.

Discussion Notes: There was no concern with the rationale of the adjustment. However, it was found that the earlier estimate for tips on hairstyling and beauty services was too high in light of the T4 total for wages and salaries in this industry. It was also recommended expanding the adjustment to include tips received by porters and baggage carriers at railway stations.

Change: The adjustment for tips on meals, accommodation and hairstyling has been expanded to include porters and baggage carriers at railway stations. As well, there was a major downward revision in the estimate of tips on hairstyling and beauty services.

3.9 Provincial Distribution of Wages and Salaries

Issue: How far back should the provincial distribution of wages and salaries in the production account (i.e. in provincial GDP) be based on the province of employment as reported in the T4 file?

Pre-Revision Treatment: Beginning with 1985, the provincial distribution of wages and salaries has been based on the province of employment as reported on the T4 file. For the earlier period, however, it was based on industrial surveys and several other sources, but none as reliable as the >Province of employment= box on the T4 form.

Discussion Notes: Conceptually, estimates of wages and salaries for the production account should reflect where the production takes place, that is, the province of employment. Statistically, the most reliable source for that information is the T4 file. It was agreed to take back the estimates based on province of employment to 1981. For the period from 1961 to 1980, the provincial distribution of wages and salaries should remain based on industrial surveys and other sources. Resource and time constraints make it difficult to reconstruct the series prior to 1981. The break will remain between 1980 and 1981.

Change: The provincial distribution of wages and salaries for the production account is now based on the province of employment as reported on the T4 file back to 1981.

3.10 Canadian Employees in Foreign Embassies and Consulates

Issue: Should wages and salaries paid to Canadian workers in foreign embassies and consulates in Canada be included in labour income and thus, in Canadian GDP and, if so, under which sector?

Statistics Canada - 215 - Collected Articles of Kishori Lal

Pre-Revision Treatment: These wages and salaries were included in labour income and were recorded under a sub-component of the Government sector.

Discussion Notes: Wages and salaries paid to Canadians employed by foreign embassies and consulates in Canada were estimated at $132 million for 1993. Recent research has indicated that the estimate was about $100 million too high. There was also doubt whether these factor payments should be (a) included in GDP and (b) under a sub-component of the Government sector, since they were made by non-resident units. It was agreed to continue including them in labour income (and thus in Canada=s GDP) and treating them as exports of services. The workers involved should be deemed hired by a domestic institutional unit that sells their services to foreign embassies for an amount equivalent to their wages and salaries. This notional unit should be classified in the business sector.

Change: Wages and salaries paid by foreign embassies and consulates are still considered part of labour income and thus of Canada=s GDP and still treated as export of services. However, they are now deemed paid by a notional unit classified in the business sector. Finally, the estimates have been substantially lowered.

3.11 Insurable Earnings of Status Indians

Issue: Should insurable earnings of Status Indians be added to the T4 Control total, since employers of Status Indians are not obliged to report their employment income?

Pre-Revision Treatment: No adjustment was made to the T4 Control total in this case.

Discussion Notes: Indian Band Councils issued T4 slips for wages and salaries totalling about $445 million in 1993. This amount represents the employment income of Status Indians who have elected to contribute to CPP, as well as that of other employees. However, Indian Band Councils did report for the same year $347 million in insurable earnings for the purpose of employment insurance by Status Indians for whom they reported no employment income.

Change: Insurable earnings of Status Indians working for Indian Band Councils are now added to the T4 Control total to represent a minimum estimate of labour income of First Nations peoples on reserves.

4. Capital Formation

4.1 Transfer Costs on the Sale of Existing Dwellings

Issue: Should a new benchmark be constructed for transfer costs on the sale of existing dwellings?

Statistics Canada - 216 - Collected Articles of Kishori Lal

Pre-Revision Treatment: Transfer costs on the sale existing dwellings had been projected from a 1971 benchmark using multiple listing service sales, blown up to represent total sales of existing houses. Due to lack of documentation, it was unclear whether the benchmark included all transfer costs (real estate commissions, legal fees, inspection fees, etc.).

Discussion Notes: The 1993 SNA notes: "New fixed assets acquired by purchase are valued at purchasers' prices: that is, including not only all transport and installation charges but also all costs incurred in the transfer of ownership in the form of fees paid to surveyors, engineers, architects, lawyers, estate agents, etc., and any taxes payable on the transfer" (paragraph 10.37). The same logic applies to used assets. Therefore all transfer costs, not only real estate commissions, should be capitalized.

At the time of the 1986 CSNA Historical Revision, it had been decided to capitalize only real estate commissions on residential construction and to include the legal fees and other fees related to the transfer of used residential fixed assets in household consumption expenditure.

Change: A new benchmark for transfer costs on the sale of existing dwellings was constructed. It includes real estate commissions, legal fees and all other fees related to the transfer of used residential fixed assets.

4.2 National Defence Purchases of Non-Military Equipment and Structures

Issue: Should the CSNA treat non-military equipment and structures purchased by the Department of National Defence as gross fixed capital formation, as recommended in the 1993 SNA?

Pre-Revision Treatment: Department of National Defence spending on equipment and structures was treated as current spending.

Discussion Notes: The 1993 SNA states that military destructive weapons and their supporting systems should not be treated as fixed assets since "they are not in fact used repeatedly or continuously in production. Although durable, they are single-use goods." Vehicles and equipment such as warships, military aircraft, tanks, missile carriers and launchers, etc., used to deliver such weapons should not be treated as fixed assets either. Other structures and equipment used by the military such as airfields, docks, roads, hospitals, transport aircraft, etc., that can also be used for civilian purposes should be considered fixed assets. (Paragraphs 10.65-10.68)

Change: The 1993 SNA recommendation of treating non-military equipment and structures used for defence purposes as fixed assets has been adopted by the CSNA and the Investment and Capital Stock Division.

Statistics Canada - 217 - Collected Articles of Kishori Lal

4.3 Capital Items Charged to Operating Expenses

Issue: Should the CSNA continue to make an upward adjustment to the capital spending on machinery and equipment reported by businesses for the capital items they charge to operating expenses (abbreviated as CICOE)? Pre-Revision Treatment: CICOE estimates have been included in investment in machinery and equipment starting with 1947. The estimates were calculated by the Investment and Capital Stock Division as a percentage of total new investment in machinery and equipment in three industry groups (manufacturing 10%, utilities 1.3%, and trade 8%). In 1990, they amounted to $2.2 billion for manufacturing, $0.2 billion for utilities and $0.2 billion for trade. This adjustment of $2.6 billion in 1990 compared to a total of $46.5 billion for investment in machinery and equipment excluding CICOE.

Discussion Notes: The 1993 SNA does not explicitly address the issue of adjusting reported capital expenditures for items which should be capitalized, but may appear as current expenses on company books. It does state however: ASome goods may be used repeatedly, or continuously, in production over many years but may nevertheless be small, inexpensive and used to perform relatively simple operations. Hand tools such as saws, spades, knives, axes, hammers, screwdrivers and spanners or wrenches are examples. If expenditures on such tools take place at a fairly steady rate and if their value is small compared with expenditures on more complex machinery and equipment, it may be appropriate to treat the tools as materials or supplies used for intermediate consumption. Some flexibility is needed, however, depending on the relative importance of such tools. In countries in which they account for a significant part of the value of the total stock of an industry's durable producers' goods, they may be treated as fixed assets and their acquisition and disposal by producers recorded under gross fixed capital formation.@ (Paragraph 10.64).

Documentation on the existing estimate was hard to find. A recommendation made in the 1950's was apparently the only basis for this continuing adjustment since then. Given the lack of justification and the difficulty of making reliable estimates, it was recommended to discontinue this adjustment for CICOE and to undo it for the entire period subject to the historical revision.

Change: The adjustment for capital items charged to operating expenses has now been removed from the investment and capital stock series for the entire historical period. 4.4 Expenditures on Mineral Exploration

Issue: Should expenditures on mineral exploration be capitalized, as recommended in the 1993 SNA?

Pre-Revision Treatment: Most spending on mineral exploration was already capitalized in the CSNA. However, general exploration expenditures undertaken by a mining company on own account or by a mineral exploration company on contract, as well as geological and geophysical expenses incurred in the exploration of petroleum and natural gas were not capitalized.

Statistics Canada - 218 - Collected Articles of Kishori Lal

Discussion Notes: The 1993 SNA notes: "Mineral exploration is undertaken in order to discover new deposits of minerals or fuels that may be exploited commercially. Such exploration may be undertaken on own account by enterprises engaged in mining or the extraction of fuels. Alternatively, specialized enterprises may carry out exploration either for their purposes or for fees. The information obtained from exploration influences the production activities of those who obtain it over a number of years. The expenditures incurred on exploration within a given accounting period, whether undertaken on own account or not, are therefore treated as expenditures on the acquisition of an intangible fixed asset and included in the enterprise's gross fixed capital formation." (Paragraph 10.90).

"The expenditures included in gross fixed capital formation include not only the costs of actual test drillings and borings, but also the costs incurred to make it possible to carry out tests, for example, the costs of aerial or other surveys, transportation costs, etc. The value of the resulting asset is not measured by the value of new deposits discovered by the exploration but by the value of the resources allocated to exploration during the accounting period." (Paragraph 10.91).

Change: All exploration expenditure as well as geological and geophysical expenditures are now capitalized.

4.5 Progress Payments for Machinery and Equipment

Issue: Should progress payments towards the purchase of machinery and equipment be treated as fixed capital formation as they occur, or should they be treated as work-in-progress by the producer and recorded as capital formation by the purchaser only when the equipment is delivered? In particular, should progress payments received from (or paid to) non-residents be recorded in the current account (i.e., be treated as an export or import) or in the financial account (acquisition of a financial claim)?

Pre-Revision Treatment: Progress payments were originally treated as gross fixed capital formation throughout the System. In effect, ownership of the equipment was deemed to be transferred with each payment. However, since 1992, the Balance of Payments Division has considered the production financed by progress payments as an addition to work-in- progress, and the progress payments from non-residents to Canadian producers (and vice versa) as financial claims. Such payments are now recorded in the financial account of the Balance of Payments as they occur. The value of the export (or import) is recorded in full in the current account when the good being purchased crosses the border (i.e., when the financial claim is extinguished).

Thus, there has been an inconsistency between the Income and Expenditure Accounts and Balance of Payments since 1992. The discrepancy has usually been small for exports, as few Canadian goods are purchased through progress payments, but has been large at times for imports of machinery and equipment.

Discussion Notes: The 1993 SNA notes that work-in-progress is treated in the System as one component of inventories of outputs held by producers. However, the borderline between inventories of partially completed structures and equipment on the one hand and gross fixed capital formation on

Statistics Canada - 219 - Collected Articles of Kishori Lal the other is not clear. Gross fixed capital formation occurs when the ownership of the assets is transferred from their producers to their users. The legal transfer of ownership may actually take place in cases where stage payments are made by the purchaser, and these can often be used to approximate the value of the gross fixed capital formation. In the absence of a contract of sale, the output produced must be treated as additions to the producer's inventories, i.e., as work-in-progress, however large the partially completed structure may be. (Paragraph 6.74).

The CSNA is in agreement with the above recommendation. However, its implementation is not practical in the case of domestic transactions.

Change: The CSNA has made an adjustment to exclude the progress payments from capital formation in machinery and equipment when they involve transactions between residents and non- residents for the whole period under review. See also Item 5.2. Due to data constraints, progress payments from one resident producer to another continue to be treated as capital formation.

4.6 Financial Leasing

Issue: Should the 1993 SNA recommendation of allocating capital goods bought under finance leases to the industry of the user (lessee) rather than that of the owner (lessor) be adopted?

Pre-Revision Treatment: Capital goods acquired under finance leases were allocated to the industry of the lessor. The rental receipts of lessors (usually classified to the >other finance= industry) consisted mainly of interest received and depreciation on the leased capital equipment.

Discussion Notes: Financial leases are characterized by the fact that all the risks and rewards of ownership are, de facto, transferred from the legal owner of the good, the lessor, to the user of the good, the lessee (1993 SNA, paragraphs 6.118 and 6.119). In order to capture the economic reality of such arrangements, a change of ownership from the lessor to the lessee is deemed to take place. The lessor is treated as making a loan to the lessee, which enables the latter to finance the acquisition of the equipment.

The 1993 SNA recommendation on financial leasing is difficult to implement with respect to the business and government sectors because of lack of data. There are two potential sources of information, the lessor and the lessee. The lessors generally report their purchases of equipment in the Capital Expenditure Survey. The lessees generally do not, because they do not consider leased equipment as part of their capital expenditure. For instance, the response to a question added in the 1992 Capital Expenditure Survey on the acquisition of buildings and equipment through a financial lease was very low. As a result, the total capital formation financed through financial leases is accurate, but because it is reported by the lessors instead of the lessees, its industrial distribution can be misleading. Thus, purchases of aircraft tend to be reported under the financial industry instead of the transportation industry, which leads to inadequate measures of multifactor productivity by industry.

Statistics Canada - 220 - Collected Articles of Kishori Lal

However, there is enough information to allocate to the lessees vehicles purchased through financial leasing in the personal sector. In this case, the 1993 SNA recommendation can be implemented. See also Item 6.2.

Change: Motor vehicles acquired under finance leases have now been allocated to the using sector. See also Item 6.2. All other capital goods acquired under finance leases in the business and government sectors continue to be allocated to the industry of the owner.

4.7 Petroleum Incentive Payments Program

Issue: Should depreciation be estimated on expenses for the drilling of unproductive wells financed with grants from the Petroleum Incentive Payments (PIP) Program?

Pre-Revision Treatment: Investment on unproductive wells drilled with the help of PIP grants was assumed to depreciate over ten years in the CSNA. The estimate was made even though neither the government (the payer) nor the company receiving the grant was depreciating the value of the investment. This approach yielded an estimate of value added even when there was no production.

Discussion Notes: The PIP program provided grants (capital transfers) to encourage companies in drilling wells. Many of the wells drilled were unproductive and were written off.

The depreciation of unproductive investment that had been financed through PIP grants yielded a gross surplus where none existed. There were two options: to record depreciation and a corresponding loss, with value added equal to zero, or to record no depreciation and no loss, with value added still zero. The second alternative was recommended as it better reflected how businesses viewed this type of transaction.

Change: The estimate of depreciation on investment supported by PIP grants (capital transfers) for drilling wells which turn out to be unproductive has been eliminated.

4.8 Capital Consumption in the Government Sector

Issue: Should the CSNA follow the 1993 SNA recommendation to impute capital consumption allowances (CCA) on the capital stock of the government sector?

Pre-Revision Treatment: The CSNA has always imputed CCA on the capital stock of the government sector, except on that of the Department of National Defence.

Discussion Notes: Some argue that it is a mistake to impute CCA on government capital, as such an imputation artificially increases the value of GDP.

The 1993 SNA, however, recommends not only continuing to impute CCA on the capital stock of government sector civilian establishments, but also extending the imputation to the non-military equipment and structures purchased by the Department of National Defence. This recommendation was implemented in the last historical revision of the US national accounts in 1995. Fixed assets

Statistics Canada - 221 - Collected Articles of Kishori Lal should be treated consistently, whether purchased by the government sector or the private sector. One would not imagine ignoring the cost of capital consumption in the valuation of private sector output. It should not be ignored either in the valuation of government sector output. The addition of CCA on non-military equipment and structures does not >artificially= increase GDP. Rather, it should be seen as the correction of a previous >artificial= undervaluation of GDP. The inclusion of an imputation for CCA in the current expenditure of the government sector has no impact on the sector=s net lending or borrowing.

CCA should now also be imputed on non-military equipment and structures purchased by the Department of National Defence, since they are now treated as part of gross fixed capital formation and capital stock (see Item 4.2 above). However, imputations for CCA should be shown explicitly to allow users to remove them if desired.

Change: The imputation of CCA on the capital stock of civilian establishments of the government sector has now been extended to non-military equipment and structures of the Department of National Defence.

5. International Trade

5.1 Point of Valuation of Trade in Goods

Issue: Should the recommendation of the1993 SNA and the IMF=s Balance of payments Manual, 5th edition (BPM-5), to value trade in goods at the border of the exporting country for both imports and exports be adopted?

Pre-Revision Treatment: The basic customs data on goods compiled by the International Trade Division were valued at the plant (place of direct shipment to Canada) for imports, and at the border for exports. The Balance of Payments Division (BOPD) then made adjustments to the customs data to express the value of exports at the plant (place of direct shipment from Canada) and to reflect conceptual differences with respect to the timing and coverage of transactions or to the change of ownership of goods.

Discussion Notes: Both the 1993 SNA (paragraphs 14.36-14.38 and 15.35) and the BPM-5 (paragraphs 219 ff.) recommend valuing the trade in goods at the border of the exporting country. Adopting a valuation at the border represents a significant shift between the goods and services accounts for both exports and imports. The transportation charges from plant to border (known as inland freight) were always recorded as trade in transportation services. For exports of wheat and natural gas, however, freight charges were recorded as part of the valuation of the trade in these goods. Shifting inland freight from >services= to >goods= would align Canada=s practice with that of other countries (including the U.S.).

Statistics Canada - 222 - Collected Articles of Kishori Lal

Change: The recommendation of the 1993 SNA and the BPM5 to record the value of exports and imports of goods f.o.b. at the border of the exporting economy has been adopted. See also Item 8.5.

5.2 Progress Payments on Imported Capital Goods

Issue: Should progress payments on imported capital goods be treated as gross fixed capital formation or as financial claims?

Pre-Revision Treatment: Progress payments on both domestically-produced and imported capital goods were treated as part of gross fixed capital formation in the Income and Expenditure Accounts, as well as in the Input-Output Tables. They were recorded as financial claims in the Balance of Payments, except for government payments.

Discussion Notes: Progress payments represent legal claims in respect of financial assets and should be recorded in the financial account. To record them otherwise would give rise to difficulties in the reconciliation of Canada=s trade estimates with those of other countries. The gross capital formation should be recorded when the goods are effectively acquired by the unit that intends to use them (1993 SNA, paragraph 10.81), that is, when the capital is put in service.

Change: Progress payments are now treated as financial claims, as recommended by the 1993 SNA. See also Item 4.5.

5.3 Trade in Goods for Processing

Issue: The 1993 SNA and the BPM-5 recommend, on practical grounds, recording all goods imported for processing and then re-exported as trade in goods. Should this treatment be adopted?

Pre-Revision Treatment: Two commodities, uranium and gold, were treated differently from others in the Canadian Balance of Payments when they entered Canada for processing. Uranium was valued in imports and exports of goods at its >pre-processing price=, while charges for upgrading it were recorded in exports of services. The value of gold imported and later re-exported was deducted from trade in goods as reported on Customs documents, because the foreign client typically retained ownership of the gold. Only the upgrading charges on the imported gold were recorded in exports of services. Due to insufficient data, the processing of other commodities was not recorded separately in services. The original import of the commodities was recorded in goods and their subsequent re- export after processing was also recorded in goods, at a higher value.

Discussion Notes: The adoption of the treatment recommended in the 1993 SNA (paragraphs 14.61- 64, 11.65 and 14.93) and the BPM-5 (paragraphs 197, 198, 436, 438, 439), namely that a change of ownership is deemed to occur when goods are cleared at the border, would streamline the recording of Canada=s trade in goods.

Change: The recommendation of the 1993 SNA and the BPM-5 with respect to goods for processing has been adopted.

Statistics Canada - 223 - Collected Articles of Kishori Lal

5.4 Reconfiguration of the Travel Account

Issue: The BPM-5 =s new standard on international travel (1) continues to include a broad range of travellers' goods, (2) includes all outlays related to travel for educational and medical purposes, (3) transfers passenger fares to the transportation account, and (4) makes a new distinction between business and personal travel. Should the CSNA adopt all or some of its recommendations?

Pre-Revision Treatment: The BOPD recorded under >travel= (a) travellers' spending on hotels, restaurants, etc. (this overall estimate was broken down into several categories of goods and services on the payments side, but not on the receipts side), (b) passenger fares paid by Canadians to foreign carriers, and vice versa, and (c) certain out-of-pocket expenses of individuals indicating education or health as the purpose of travel. It excluded from >travel= (a) tuition and living expenses of students, recorded under >other services=, and (b) medical expenses of Canadian travellers covered by provincial insurance plans, recorded under >government services=, as well as payments from non-residents to Canadian hospitals. Finally, it made no distinction between business and personal travel.

Discussion Notes: The reconfiguration of the travel account according to the BPM-5 (paragraphs 241-251) is essentially a data issue and most of it can be accomplished. The Education, Culture and Tourism Division, which collects and publishes the data on international travel, has agreed to adopt the new definition of spending on travel. This avoids the risk of two sets of estimates on international travel being published by Statistics Canada.

Change: The reconfiguration of the travel account according to the BPM-5 was carried out in two stages. With the release of the BOP for the first quarter of 1996, revisions were incorporated from 1994 onwards. The changes were made for the entire historical period with the release of the BOP Historical Revision in June 1997.

5.5 New Breakdown of Commercial Services

Issue: To what extent should the new breakdown of commercial services suggested in the BPM-5 be implemented?

Pre-Revision Treatment: The Canadian Balance of Payments has in recent years published annual estimates for some 25 categories of services other than >travel= and >government=. The 25 categories comprised three headings under >freight and shipping=, two under >miscellaneous services= and the remainder, under >business services=.

Discussion Notes: Under the new presentation recommended in the BPM-5 (paragraphs 158-168; 230-240; 252-266; Appendix 3), >transportation= would gain >passenger fares= from >travel= and >transport-related services= (mainly airline support), from business services; at the same time, it would lose >inland freight=, shifted to goods. The addition of new estimates developed in the U.S. for truck transportation should be considered. >Ocean freight=, >inland water freight= and some other items would need to be combined for reasons of data availability and confidentiality (the BPM-5 calls for ocean freight to be broken down between passenger, freight and other). Construction would become

Statistics Canada - 224 - Collected Articles of Kishori Lal a new category, requiring separate estimation; >information services= would need greater coverage. Estimates for the new category >merchanting= would be difficult to construct, not only on account of the treatment of merchanting itself (e.g., netting) but because some merchanting commissions may already be included in goods trade. New estimates would have to be developed for legal and management services. The >other services= category would lose >student expenditures= to >travel=, and >organized sports and performing arts activities= to the new category >personal, cultural and recreational services=. A significant data problem would remain in the case of auto tooling.

Change: A modified version of the new classification was implemented at the time of the release of the BOP for the first quarter of 1996, with the major exception of the shifting of inland freight from the transportation account to the goods account. This was carried out in June 1997, along with the extension of the coverage of ocean freight and insurance and the addition of new estimates for (a) truck transportation beyond the border, (b) legal and management services, and (c) starting in 1990, construction services. Many trade commissions were left in goods, and auto tooling continues to be treated as a service rather than a good. The new classification facilitates international comparisons and can be more easily matched with the revised input-output classification of goods and services.

6. Household Consumption Expenditure

6.1 Classification of Individual Consumption by Purpose (COICOP)

Issue: Should the CSNA implement the international Classification of Individual Consumption by Purpose (COICOP)?

Pre-Revision Treatment: The Classification of Household Goods and Services (CHGS) of the 1968 SNA was used to classify household consumption expenditure.

Discussion Notes: The COICOP referred to in the 1993 SNA (paragraphs 9.63 and 18.1) was further developed by the OECD and its revised draft was submitted to the National Accounts Experts Meeting held in Geneva, April 29 - May 2, 1996. The COICOP would replace the CHGS of the 1968 SNA used to classify household consumption expenditure (referred to as >personal expenditure= in the CSNA). Note that the COICOP is really a functionalization of the actual individual consumption of households and not just their consumption expenditures.

Change: The CSNA has adopted the COICOP, keeping in mind the need for historical continuity in the categories of household expenditure in both the Income and Expenditure Accounts and the Input- Output Tables.

6.2 Leasing in the Household sector

Issue: Should the CSNA follow the 1993 SNA recommendation to allocate capital goods purchased by households under financial leases to the household sector?

Pre-Revision Treatment: Household expenditure on vehicles was equal to total motor vehicle sales, less >fleet purchases= (i.e., purchase of several vehicles at a time), presumably all made by business,

Statistics Canada - 225 - Collected Articles of Kishori Lal less the estimated business use of vehicles purchased by persons, plus the estimated personal use of vehicles purchased by business. Since the survey on retail sales of motor vehicles covered financial leasing, part of the value of vehicles leased to persons under a financial lease was implicitly included in household expenditure. The value of these vehicles was also included in the capital formation of the finance leasing companies, leading to a double count.

Household consumption expenditure also included, as a current service expense, rental payments made under an operating lease along with lease payments made under a financial lease. These estimates of combined spending on motor vehicle renting and leasing were based on the Family Expenditure Survey.

Discussion Notes: It was agreed to apply the same convention to all leases: those greater than one year would be treated as financial leases (as a capital loan to the consumer buying the good); those less than one year would be considered as operating leases. The full value of vehicles purchased by the household sector under financial leases is to be included in household consumption and deducted from capital formation.

The decision to record the full value of vehicles purchased under financial leases in household consumption requires that the lease payments on these vehicles no longer be treated as a current service expenditure. The payment for operating leases would still be considered as current service expenditure.

Note that the calculation of the value-added taxes (GST and HST) on leased vehicles must be based on the lease payments, not on the total value of the vehicles recorded in household expenditure under financial leases.

Change: The CSNA has adopted the 1993 SNA recommendation. See also Item 4.6.

7. Financial Services and Insurance

7.1 Financial Intermediation Services Indirectly Measured (FISIM) B Allocation

Issue: Should the CSNA adopt the 1993 SNA recommendation of allocating a FISIM output to all users?

Pre-Revision Treatment: The CSNA has always tried to allocate the FISIM output produced by the various financial corporations to the sectors using the services. However, FISIM provided by trust companies, credit unions and consumer loan companies was all allocated to the household sector. No FISIM was calculated for security and investment dealers that are subsidiaries of banks, and therefore no FISIM was allocated to the users of their services, on the belief that they charged explicitly for all their services. While trust and mortgage companies have dealings with the government sector, their intermediation services were not allocated to that sector.

Statistics Canada - 226 - Collected Articles of Kishori Lal

Discussion Notes: The 1993 SNA recommends that FISIM output be allocated to all sectors using financial intermediation services, that is, businesses, households, governments and the rest of the world (paragraph 6.125). The FISIM attributed to businesses represents an intermediate expense for them and therefore does not affect the level of GDP, while that attributed to the household and government sectors, which are final demand sectors, is part of GDP. This recommendation is controversial, as it represents a major departure from the 1968 SNA. For this reason, the 1993 SNA shows flexibility in proposing several modalities of application, including the option of not allocating any FISIM output to final demand (paragraph 6.126), a treatment equivalent to that of the 1968 SNA. The CSNA does not support this option.

It is unclear why the FISIM output of credit unions, trust companies and consumer loan companies was entirely allocated to the household sector, as these financial institutions also deal with the other sectors. Credit unions and trust companies, in particular, make substantial mortgage loans, which by definition are transactions with business. Likewise, the assumption that investment and security dealers always charge explicitly for their services must be reviewed, as banks and investment dealers have become more integrated. Although investment dealers do not take deposits, they take part in activities similar to financial intermediation, much like consumer loan companies.

Change: The CSNA has continued using the methodology outlined in 1993 SNA for estimating the FISIM of banks, and has now extended it to all other financial corporations, i.e. trust, mortgage and consumer loan companies, as well as credit unions. The FISIM on mortgages and other business transactions of credit unions has now been shifted from household consumption to intermediate consumption, leading to a reduction in GDP. The total output of credit unions was not affected. FISIM is now calculated on government deposits in all financial corporations and on government borrowing from all financial corporations.

7.2 FISIM B Reference Rates

Issue: Should the CSNA adopt the 1993 SNA recommendation to use reference rates for distributing FISIM between borrowers and depositors and further allocating the resulting amounts among sectors (households, government, business and non-residents)?

Pre-Revision Treatment: FISIM output was split between borrowers and depositors on the basis of outstanding loans and deposits. The FISIM attributed to each was further broken down by sector according to each sector=s share of the total loans and deposits. The method relied on the assumption that a dollar borrowed or lent resulted in an identical service and that all sectors paid the same service rate for FISIM.

Discussion Notes: The 1993 SNA defines a reference interest rate as one which "represents the pure cost of borrowing funds - that is, a rate from which the risk premium has been eliminated to the greatest extent possible and which does not include any intermediation services" (paragraph 6.128).

Statistics Canada - 227 - Collected Articles of Kishori Lal

The reference rate would lie in between the >average= interest rates given to borrowers and depositors. For example, if the average rate paid on deposits was 5% and the one demanded on loans was 7%, the reference interest rate would lie in between at, say, 6%. In this case, the margin earned would be 1% on both deposits and loans. To simplify the calculation, a single rate could be applied to all instruments (different types of loans and deposits). However, this could result in the service portion of the interest (or the margin earned) being negative in some cases. This could occur when a long- term interest rate was lower than the current one. The problem with a single rate was that it only represented one point in time. Therefore, it might be necessary to use several reference rates to accurately estimate FISIM.

It should be noted that banks engage in numerous cross-term transactions. Thus, a bank may be willing to break even or even lose money on one instrument so that larger profits can be made elsewhere. This practice influences the levels of various reference rates.

For chartered banks, calculations were done (a) with a single reference rate (the midpoint between the rate of return on all assets and the one on all liabilities) and (b) three different rates for each instrument (e.g. mortgages, consumer loans, etc.), one for short-term loans and deposits, one for medium-term, and the other for long-term instruments. In the case of deposits, the single and multiple rates yielded similar results with respect to the split of FISIM by sector. In fact, the results were very similar to those obtained under the old method, where the percentage of assets in each sector was used to determine the split, except for securities. The split of FISIM on loans by sector was very sensitive to the choice of rate. The single rate method was easy to calculate, a useful feature when estimates need to be calculated quarterly in a very short time, but resulted in negative values and in large swing between sectors, especially when long-term rates on deposits and loans changed quickly. The multiple rate method resulted in few negative values and in smaller sectoral shifts. In conclusion, the use of more than one reference rate increases the complexity of calculations but seems to yield more sensible results.

These results were obtained using data on chartered banks, since there was no detailed information for other financial corporations. However, they could serve to split the FISIM of other financial corporations by sector.

Change: The split of FISIM output between depositors and borrowers is now based on the total value of deposits and loans outstanding. The FISIM attributed to depositors is split by sector according to each sector=s share of deposits. The split of FISIM on loans is obtained by the multiple reference rate method.

7.3 FISIM B International Trade

Issue: Should the CSNA adopt the 1993 SNA recommendation to allocate FISIM to international trade?

Pre-Revision Treatment: No FISIM output was allocated to international trade in the CSNA.

Discussion Notes: It is not known why no FISIM was attributed to international trade. The 1993

Statistics Canada - 228 - Collected Articles of Kishori Lal

SNA is clear on the issue: A...an allocation should also be made to non-resident borrowers and lenders. These allocations together appear as exports of FISIM in the rest of the world account. Similarly, an element of FISIM should be shown as payable to non-resident financial corporations by resident borrowers and lenders. Together, these elements should appear as imports of FISIM in the rest of the world account.@ (Annex III, paragraph 4).

The IMF=s BPM-5 states that "such imputations are equivalent to reclassifying a portion of interest as financial services. As a reflection of the views of national balance of payments compilers, this procedure is not recommended in this Manual. As a result, these implicit services are reported indistinguishably under investment income (interest)" (BPM-5, footnote to paragraph 258). This is an instance where the 1993 SNA and BPM-5 diverge.

Change: The CSNA has implemented the 1993 SNA recommendation of allocating FISIM to international trade (both exports and imports). The Canadian Balance of Payments, on the other hand, follows the IMF=s BPM-5 recommendation not to do so. The difference between the two is shown as a reconciliation item.

7.4 FISIM B Own Funds

Issue: Should the CSNA adopt the 1993 SNA recommendation not to include financial intermediaries= own funds in the calculation of FISIM?

Pre-Revision Treatment: The value of FISIM was estimated as total property income received (excluding income from securities) less interest paid. This approach automatically entailed the inclusion of income earned on own funds.

Discussion Notes: The 1993 SNA suggests that financial intermediaries' own funds should not be included in the calculation of FISIM "as such income does not arise from financial intermediation" (paragraph 6.125). Although the lending of own funds by a financial corporation cannot be precisely defined as FISIM, the borrowers of these funds still receive a service from a financial corporation. As well, the view that own funds generate financial services is gaining acceptance, as demonstrated in the various papers on FISIM prepared by Eurostat.

As is the case with other intermediated funds, loans made from own funds consist of a >transfer= portion (the pure rate of borrowing money) and a service portion. Therefore the global estimate of FISIM (i.e., interest received minus interest paid), which includes interest received from own funds but no interest paid, should be adjusted for this transfer portion. For example, if a financial intermediary lends $1000 of his own funds for which he receives $100 in interest, only a portion of this interest can be considered a service, and therefore the output of a financial intermediary. If the value of the service is estimated at $20, the remaining $80 should be removed from the FISIM total.

Statistics Canada - 229 - Collected Articles of Kishori Lal

Change: The CSNA has continued to include own funds in the calculation of FISIM, and thus rejects the 1993 SNA recommendation. However, the CSNA has amended its methodology so that only the borrower=s portion of service charges remains in the FISIM calculation. This new treatment of own funds now applies to all financial corporations, including consumer loan companies, to avoid an overestimation of the output of financial corporations.

7.5 FISIM B Consumer Loan Companies

Issue: Consumer loan companies differ from other financial corporations in that they do not accept deposits. Should a FISIM output be attributed to them?

Pre-Revision Treatment: FISIM was attributed to consumer loan companies, although the methodology was inadequate.

Discussion Notes: One may interpret the 1993 SNA as suggesting that consumer loan companies engage in financial intermediation even though they do not take deposits and are not loaning their own funds, as they "incur liabilities on their own account in order to mobilize funds" (paragraph 6.133). This is why the CSNA has attributed FISIM to them.

The old methodology overestimated the FISIM output of consumer loan companies. The general approach for the estimation of FISIM is to make it equal to the net interest received. FISIM output of consumer loan companies should be calculated in the same manner. However, since consumer loan companies do not accept deposits, FISIM is to be attributed only to borrowers. Of that amount, the portion allocated to the household sector should be based on the ratio of the companies= consumer loans to their total loans. A further adjustment is needed for the fact that consumer loan companies get a higher return on consumer loans than other loans. The remainder of the FISIM output should be allocated to the business sector.

Change: The CSNA has continued to attribute FISIM to consumer loan companies, but has changed the methodology to reflect their net interest received.

7.6 FISIM B Bank of Canada

Issue: Should the CSNA adopt the 1993 SNA recommendation of treating the central bank as any other financial corporation?

Pre-Revision Treatment: The Bank of Canada has been classified as a financial corporation. However, its output has always been measured as the sum of expenses. One third of the output was allocated to the household sector, and the rest, to the government sector.

Discussion Notes: The treatment of the Bank of Canada in the CSNA is not in conformity with the recommendation of the 1993 SNA with respect to central banks (paragraph 6.132). However, since the Bank of Canada plays a very different role in the sector than other banks, a different treatment seems warranted. Its main functions are to formulate and implement monetary policy, issue and

Statistics Canada - 230 - Collected Articles of Kishori Lal replace bank notes, manage the public debt and provide other banking services. It was felt that only its activities associated with financial intermediation could generate FISIM. However, on the basis of operating costs, this function was estimated to account for 9% of total expenses.

At Statistics Canada=s request, the Inter-Secretariat Working Group on National Accounts (ISWGNA) deliberated and issued a clarification on the valuation of central bank output. In the January 1996 issue of SNA News and Notes, it states that where the approach recommended in the 1993 SNA "leads consistently to inappropriate results, output couldY be measured at cost as for other non-market producers".

No documentation was found explaining the existing split of services provided by the Bank of Canada (one-third to the household sector and two-thirds to the government sector). The household sector=s share was presumably based on the "cost of administering Canada Savings Bonds by the Bank of Canada". Since this cost is estimated to be quite small, it was recommended to stop allocating any Bank of Canada FISIM to the household sector.

Change: The CSNA continues to classify the Bank of Canada as a Government Business Enterprise (GBE) and to calculate its output as the sum of expenses. However, this output is now entirely allocated to the government sector.

7.7 Output of Insurance Industries

Issue: Should the CSNA implement the 1993 SNA recommendation of including in the output of insurance industries the investment income from their technical reserves?

Pre-Revision Treatment: The output of property and casualty insurance was deemed equal to premiums less claims. The output of life insurance companies was deemed equal to the operating expenses plus the dividends paid by the stock insurance companies.

Discussion Notes: In the 1993 SNA (Annex IV, paragraphs 14-18, 22-23), the output of property and casualty insurance is defined to include investment income from the technical reserve plus net premiums (premiums less claims). The technical reserve is the sum of prepaid premiums and reserves against future claims and other contingencies. The output of life insurance is defined as equal to premiums less claims, plus investment income from the technical reserve, less increase (plus decrease) in actuarial reserves. This formulation of output of insurance corporations approximates the sum of their administrative costs and the profits distributed to policyholders.

For both life and non-life insurance, total claims paid often exceed the premiums receivable. The income earned by insurance companies from prepaid insurance premiums and reserves against future claims affects the rate of premiums. An adequate measure of the service provided by insurance companies must take into account this income, as well as premiums and claims.

Statistics Canada - 231 - Collected Articles of Kishori Lal

Change: The CSNA has now adopted the 1993 SNA recommendation of including the investment income from technical reserves in the output of insurance companies.

8. Input-Output B Classifications and Valuation

8.1 Industry Classification

Issue: Should the industry classification used in the Input-Output tables as well as the monthly and the provincial GDP by industry reflect the structure of the 3-digit classes of the 1980 SIC, both for the business and non-business sectors?

Pre-Revision Treatment: The industry classification employed in the CSNA was established at the time of the 1986 Historical Revision and was based on Statistics Canada=s Standard Industrial Classification 1980 (1980 SIC) at the 3 and 4 digit levels. Several industries could not be mapped into the 3-digit classes of 1980 SIC. The non-business sector industries were presented only in final demand. Aggregation parameters were different for the business and the non-business sectors.

Discussion Notes: The industry classification employed in the CSNA has shortcomings. In several cases, industries could not be mapped into 3-digit 1980 SIC classes; several industries encompassed one or more 3-digit SIC classes as well as one or more four-digit classes of another 3-digit SIC class. This made it difficult to reconcile the estimates of wages and salaries by industry in the I-O tables with those based on the T-4, since the latter were available only at the 3-digit level. For some industries not covered by annual surveys, the estimates at the 3-digit level had to be split arbitrarily. Finally, the Industrial Monitor program of the Industrial Measures and Analysis Division required an industry classification better aligned with the 3-digit level of the 1980 SIC.

The issue of incorporating the draft 1997 North American Industry Classification Structure (NAICS) was examined in this context. It was concluded that the incorporation of the 1997 NAICS would require a major restructuring of the present industry classification employed in the CSNA, an undertaking that could not be completed given the time and resource constraints in the current Historical Revision. Therefore the 1997 NAICS was left aside and only the 1980 SIC was taken into consideration.

To remedy the problems in the present industry classification, several existing industry definitions needed to be changed and new industries to be created in the business sector at the worksheet level. The increase in the number of industries delineated in the CSNA did not, however, go far enough to allow a one to one correspondence between these industries and the 3-digit classes of the 1980 SIC, as required by the Industrial Monitor project, for the following reasons. (a) There were no output data for all industries at the 3-digit level, examples being agriculture and construction industries. (b) Input data were also lacking for several industries; for instance, in the case of wholesale and retail trade, there would be data on inputs and profits beginning only from 1993. (c) For other industries, data weaknesses did not allow further disaggregation, two examples being finance and personal services.

Statistics Canada - 232 - Collected Articles of Kishori Lal

Further, the industries in the non-business sector were numbered differently than their counterpart industries in the business sector, which was a source of confusion. It was therefore recommended that such industries no longer be identified with a separate number, but instead be assigned the same number as their counterpart industries in the business sector in a separate work file. Under such an arrangement, value added could be presented for a whole industry or separately for the business and non-business components of that industry. This change should be carried back to 1961.

Change: The CSNA has revised the industry classification starting with the reference year 1981, both for the business and non-business sectors, to facilitate work on the Industrial Monitor and the reconciliation of estimates of wages and salaries. In addition, a work file has been created for the non-business sector industries so that the valued added can be presented for a whole industry or separately for its business and non-business components, back to 1961.

8.2 Resource Royalties Industry

Issue: Should the 1993 SNA recommendation of treating resource royalties as an investment income rather than a service output be adopted, thus eliminating the Resource Royalties industry?

Pre-Revision Treatment: Resource royalties were treated as a service produced by an industry called >Government royalties on natural resources=.

Discussion Notes: According to the 1993 SNA, revenues from the lease of land and the use of subsoil assets, called >royalties= in the CSNA, are to be treated as property income, not as production income. This treatment does not require the creation of an industry producing the royalties and thus entails the elimination of the >Royalties= industry in the Canadian Input-Output tables, as royalties will now be paid out of the surplus of the industries using the land and sub-soil assets. GDP will not change, but the value added in the using industries will increase with the reallocation of rent from intermediate consumption to primary input, with an identical offset in the output and the value added of >Finance, insurance and real estate=, to which the >Royalties= industry belonged.

Change: In accordance with the 1993 SNA, the industry >Government royalties on natural resources= is eliminated and so is the payment of such royalties out of the intermediate consumption of the industries using these assets.

8.3 Presentation of Non-Market Producers

Issue: Should the CSNA follow the 1993 SNA recommendation with respect to the presentation of non-market producers?

Pre-Revision Treatment: In the Canadian input-output tables, non-market producers were not shown either in the supply table or the use table. Instead, their inputs were directly allocated to final consumption.

Statistics Canada - 233 - Collected Articles of Kishori Lal

Discussion Notes: According to the 1993 SNA, the output of non-market producers (such as establishments belonging to general government or non-profit institutions) must be shown in the Supply tables and their inputs, in the Use tables (paragraphs 15.62-67). It was recommended that the CSNA follow this presentation, but with one modification, namely that non-market producers be classified by sector rather than by industry, due to the lack of detailed basic statistics by industry. However, the value added of non-market producers would be disaggregated by industry.

Change: The 1993 SNA recommendation to present non-market producers in the Supply and Use tables has been accepted, but these producers are classified by sector, not by industry. However, their value added has been disaggregated by industry.

8.4 Commodity Classification

Issue: Should the classification of goods and services (called >commodities= in the Input-Output tables) employed in the CSNA be harmonized with the revised classifications for (a) industries in the CSNA and (b) services in the Balance of Payments? Should the revised commodity classification be carried back to 1986 to facilitate the deflation of the Input-output tables?

Pre-Revision Treatment: The commodity classification reflecting the Harmonized System/Standard Classification of Goods (HS/SCG) system was introduced in the reference year 1988.

Discussion Notes: When a new commodity classification reflecting the HS/SCG system was introduced in the Input-Output tables in 1988, some desirable changes could not be made on account of restrictions imposed by the medium level of commodity aggregation. Furthermore, because of the CSNA=s revision policy, the new commodity classification could not be applied before 1988. Since the base year of the constant price series was 1986, this created problems and the deflation for subsequent years had to be carried out at the commodity link (L) level. There was also a need to align the commodity classification with that of the J series (the most detailed in the Income and Expenditure Accounts) so that the two systems could be harmonized without having to split commodities or the J series.

Since it was important for trade negotiations, a number of commodities would have to be split to accommodate as much as possible the new service categories in the Balance of Payments based on the IMF classification.

The commodity classification employed in the CSNA was reviewed in the light of the revision to the industry classification described above. Some commodity definitions would have to be changed to harmonize them with industries, especially where former industries combining 3- and 4-digit classes were changed to 3-digits. Other commodities would need to be split to provide distinct outputs for some of the new industries created in both the non-business and business sectors. However, the commodity classification was already sufficiently detailed to allow the estimation of specific commodity outputs for the majority of new industries.

Statistics Canada - 234 - Collected Articles of Kishori Lal

The commodity >Government royalties on natural resources= would be eliminated, in line with the elimination of the industry >Government royalties on natural resources=, explained above.

As a counterpart to the non-business sector producers, new commodities would have to be created, such as services of welfare organizations, university services, defence services, and other government services.

In the case of household consumption, there were more categories in the Input-Output tables than in the Income and Expenditure Accounts (IEA), so that several of the former could be made to correspond to the latter. However, in a few cases, a single I-O category had to be split between two IEA categories. These problem cases could be resolved either by splitting Input-Output categories or combining those of the IEA.

Change: A revised commodity classification has been adopted and carried back to 1986. The commodity >government royalties on natural resources= has been eliminated. The number of commodities at the most detailed level has been increased from 627 to 679.

8.5 Valuation of Imports and Exports

Issue: Should the 1993 SNA recommendation to value both exports and total imports f.o.b. at the border of the exporting country, and value detailed imports c.i.f. in the Input-Output tables be adopted?

Pre-revision treatment: The Canadian Input-output tables differed from the 1993 SNA with respect to the valuation of imports, both at the total level and the detailed level. The detailed imports by commodity were valued c.i.f. at the Canadian border inclusive of import duties. Total imports were recorded c.i.f., without customs duties. The valuation of exports was in conformity with the 1993 SNA.

Discussion Notes: The 1993 SNA recommendation results in two valuations of imports in the input- output tables, one for the total and the other for detailed imports by commodity. It also entails abandoning the Canadian practice of including import duties in the valuation of detailed imports. For these reasons, it was deemed preferable not to adopt it.

It should be noted that the inclusion of import duties in the detailed imports by commodity in the input-output tables was, and continues to be, offset by an adjustment at the aggregate level such that the total value of imports of goods and services is the same as in the Balance of Payments.

Statistics Canada - 235 - Collected Articles of Kishori Lal

With the Historical Revision, new information has become available on goods carried by Canadian carriers in the US and vice versa. In the Canadian Balance of Payments, the transportation of Canadian imports by Canadian carriers in the US is now recorded as an export of transportation services, and the transportation of Canadian exports by American carriers in Canada, as an import of transportation services. This practice does not conform to the recommendation of the 1993 SNA and the BPM-5, which is to record them as entries in the transportation account to offset the valuation of exports and imports of goods. The treatment adopted in the Canadian Input-output tables is the same as in the Balance of Payments.

Change: The 1993 SNA recommendation to value both exports and imports f.o.b. at the customs border of the exporting country has been adopted in the Balance of Payments. The Input-Output tables continue to record detailed imports c.i.f. including import duties, with an offset of these duties at the aggregate level, so that the total value of imports of goods and services is the same as in the Balance of Payments.

9. Deflation

9.1 Rebasing of GDP from 1986 to 1992 Prices

Issue: Should we rebase the constant price series of GDP from 1986 to 1992? How should we link the constant price series for the period prior to the base year?

Pre-Revision Treatment: In the CSNA, GDP and its components were measured at constant prices using fixed base volume indexes. Since 1961, the base year for constant price series has changed as follows:

1961 for the period from 1961 to 1971; 1971 for the period from 1971 to 1981; 1981 for the period from 1981 to 1986; 1986 for the period from 1986 to date.

Thus, since 1981, the base year has been changed in the CSNA about once every five years to reflect the evolution of prices in the economy. On this occasion, the series prior to the base year were not recalculated using the new base year prices; instead they were linked and growth rates originally calculated were maintained. The result was a chain of volume indexes linked only occasionally. In addition, annual and quarterly chain indexes of the growth of expenditure aggregates are presented in the Income and Expenditure Accounts, as a supplement to the fixed base indexes.

Statistics Canada - 236 - Collected Articles of Kishori Lal

Discussion Notes: Given the continually changing economic environment, the structure of prices, over time, moves away from that of the base year. When the base year is far removed from the present, the rate of growth of the economy is likely somewhat overstated. This is because the goods and services registering the strongest real growth tend to be those experiencing the weakest increase in prices. Valued at the prices of a distant past, they make up a larger proportion of GDP and thus carry more weight in the growth of GDP than they would at the prices of a more recent time. As a result, the estimate of GDP at constant prices for the current period becomes less reliable. It was thus necessary to shift to a more recent one. The year 1992 was chosen, as it was the first >normal= year after the introduction of the GST in Canada in 1991. It was also the latest year for which a final set of statistics was available. Finally, since the United States were also using 1992 as the base year of their National Accounts, the change would facilitate comparisons between the two countries.

It was suggested not to abandon the existing practice, that is, of linking each previous period while maintaining the growth rates calculated during that period. Chain linking the previous period every five years or so was deemed an acceptable departure of the 1993 SNA recommendation of chaining on a yearly basis (paragraph 16.73).

It was decided that the present practice of chaining value added by industry every five years was also adequate. However, the annual and quarterly chain indexes in the Income and Expenditure accounts should continue to be produced as supplementary information.

Change: The base year of the constant price series of GDP was changed from 1986 to 1992 prices. The pre-1992 period was linked while maintaining the earlier growth rates. As well, the annual and quarterly chain indexes continue to be produced as supplementary information in the Income and Expenditure Accounts.

9.2 Output of Non-Market Services B Education

Issue: Should output volume indicators be used to estimate the output of public education at constant prices?

Pre-Revision Treatment: Constant price labour inputs were estimated on the basis of an index of the number of full-time teachers employed, and constant price capital consumption allowances, by deflating the current price series with a price index of capital stock. Constant price output was obtained in the Input-Output tables as the sum of these two primary inputs, plus deflated intermediate inputs and deflated net indirect taxes.

Discussion Notes: The 1993 SNA recommends to produce volume indexes of output for education on the basis of the volume of education services delivered to households (paragraph 16.134).

Statistics Canada - 237 - Collected Articles of Kishori Lal

One method of implementing this recommendation could be to use student enrolment broken down into relatively homogeneous categories and, for aggregation purposes, to apply a base year weight to each category or type of education - primary, secondary, special, college, university. University education should be further decomposed by field of study. The weight of each category could be equivalent to the relative cost of training, approximated by student fees plus government grants. The new method should be applied as far back as possible, given basic statistical sources. It was suggested that, in the future, the weights be updated more frequently and efforts be made to obtain similar sub-annual data.

Change: The 1993 SNA recommendation has been adopted, but only in the Input-Output tables, where output volume indicators based on student enrolment are now used to measure the output at constant prices of universities back to 1981, and of other public education back to 1986.

9.3 Quality Adjustment of the Volume of Wheat

Issue: Should the quality adjustment of the volume of wheat, taking into account variations in the grades of wheat produced, be identical throughout the CSNA?

Pre-Revision Treatment: The Income and Expenditure Accounts Division and the Industry Measures and Analysis Division adjusted the volume estimates for variation in the grades of the wheat crop, while the Input-Output Division (IOD) did not. Discussion Notes: From 1981 to 1987, the Input-Output Division deflated wheat with a unit value index based on farm cash receipts. From 1988 on, this was replaced by an unweighted volume index of tonnes of wheat produced. The Division argued that the data on wheat grades were weak and that an adjustment was difficult to make. Representatives from other Divisions insisted that the adjustment was indispensable given the wide variation in the quality of wheat from one year to the next.

Change: An identical quality adjustment is now made to the volume of wheat throughout the CSNA.

10. Financial Flows and Balance Sheets

10.1 Seasonally Adjusted Financial Flow Accounts

Issue: Should the financial accounts be seasonally adjusted in order to further the integration of the CSNA quarterly accounts?

Pre-Revision Treatment: A subset of seasonally adjusted data (the Financial Market Summary Table) has been released at the same time as other CSNA components since the first quarter of 1987.

Statistics Canada - 238 - Collected Articles of Kishori Lal

Discussion Notes: The Financial Flow Accounts are part of an integrated set of economic accounts. It is generally recognized that there is a need to remove the effects of seasonality from these accounts. The Income and Expenditure Accounts are seasonally adjusted. The financial accounts (the second set of accumulation accounts) extend the income and expenditure sector accounts by (i) yielding complete statements on sources and uses of funds and (ii) showing the financing of economic activity. The data on financial transactions have been analysed for seasonality at different levels of aggregation (over instruments and sectors). The Financial Flow Accounts do embody seasonality and should therefore be seasonally adjusted and be made available in seasonally adjusted form.

It should be noted that many analysts of financial data use only the unadjusted data and link such data to interest rate and exchange rate fluctuations. Therefore, the Financial Flow Accounts should also continue to be available in unadjusted form.

Change: Seasonally adjusted Financial Flow Accounts are now produced at a level of sector aggregation sufficient to tie in with the seasonally adjusted data in the sector estimates of the Income and Expenditure Accounts.

10.2 Other Changes in Assets Account

Issue: Should the CSNA develop the >Other Changes in Assets Account= (also called >stock-flow reconciliation account=) that explicitly links the balance sheet accounts and the flow accounts (capital and financial accounts) by recording changes in the value of assets and liabilities that do not result from transactions?

Pre-Revision Treatment: A stock/flow reconciliation account was produced in 1985 for the period 1984-85. This work followed the 1977 United Nations Provisional Guidelines for Balance Sheets and Reconciliation Accounts. The recommendations in the 1993 SNA are quite similar to the 1977 guidelines.

Discussion Notes: The >Other Changes in Assets Account= (1993 SNA, Chapter 12) is concerned with the recording of changes in the value of assets, liabilities and net worth between opening and closing balance sheets that do not result from the transactions recorded in the capital and financial accounts. This account has two components -- the >revaluation account= and the >other changes in the volume of assets account=.

Change: An >Other changes in assets account= (stock-flow reconciliation account) based on the 1993 SNA, and thus explicitly integrating flows and stocks, is in the process of being developed.

10.3 Subsoil Assets

Issue: Should subsoil assets be included in the balance sheet account?

Statistics Canada - 239 - Collected Articles of Kishori Lal

Pre-Revision Treatment: These assets were left out of in the balance sheet account.

Discussion Notes: The 1993 SNA states: "... natural assets must not only be owned but capable of bringing economic benefits to their owners, given the technology, scientific knowledge, economic infrastructure, available resources and set of relative prices prevailing on the dates to which the balance sheet relates... Thus, known deposits of minerals that are not commercially exploitable in the foreseeable future are not included in the balance sheets..." (paragraph 10.11).

Subsoil assets include reserves of oil, gas, coal and metallic and non-metallic minerals. Most countries have not been able to include such assets in their balance sheets because of (a) lack of physical data on the assets and (b) lack of agreed upon valuation rules. Yet, these assets constitute an important part of a nation's wealth and should not be ignored. At Statistics Canada, we have developed, as part of the Environment Satellite Account, both physical and monetary accounts for oil, gas, coal and other minerals. Estimates have been calculated according to three methods, namely (i) the net price method, (ii) the net present value method and (iii) the replacement cost method. The method of net present value is the one recommended by the 1993 SNA.

Changes in reserves of subsoil assets due to discovery or depletion are to be reflected in other changes in the volume of assets account. Changes in the monetary values due to price changes are to be recorded as holding gains (or losses) in the revaluation account.

Change: The 1993 SNA recommendation to include subsoil assets in the balance sheet account has been adopted. Those subsoil assets for which data are available are included in national wealth under >non-produced assets= beginning in 1998. A breakdown of these assets by sector may also be available at a later date.

10.4 Timber

Issue: Should timber be included in the balance sheet account?

Pre-Revision Treatment: These assets were left out of in the balance sheet account.

Discussion Notes: The 1993 SNA recommendation on timber is similar as the one on subsoil assets. Standing timber is a renewable natural resource. It can be viewed either (i) as a produced asset, classified under the heading >work in progress on cultivated assets= or (ii) as a tangible non-produced asset, classified under the heading >non-cultivated biological resources= (paragraph 12.26.). Given the methodology employed to derive the estimates, no distinction could be made between timber that is produced and timber that is not. It was therefore decided to treat all timber as a tangible non- produced asset.

Both physical and monetary accounts have been calculated for timber. The value of timber (the present value of expected future rent, calculated as the selling price of timber less costs of bringing it to market) is the value of timberland, since the values of timber and land could not be separated.

Statistics Canada - 240 - Collected Articles of Kishori Lal

Additions, reductions and changes in the value of timber are to be shown in the >Other Changes in Assets Account= (the >stock-flow reconciliation account= in the CSNA).

Change: The 1993 SNA recommendation to include timber in national wealth has been adopted. Timber will be shown on the national balance sheet beginning in 1998 as a non-produced asset, under the heading >non-cultivated biological resources=. A breakdown of these assets by sector may also be available at a later date.

10.5 Monetization of Gold

Issue: Should the CSNA adopt the 1993 SNA recommendation of transforming the commodity gold into gold held as foreign exchange reserves by the monetary authorities, via reclassification of such gold in the Other Changes in the Volume of Assets Account?

Pre-Revision Treatment: Gold reserves of the monetary authorities (recorded in the Exchange Fund Account) were treated as financial assets on their balance sheet. When the Bank of Canada, on behalf of the Exchange Fund Account, increased gold reserves (in other words, >monetized= gold) by transacting either with other monetary authorities or with suppliers of gold, an export of commodity gold was deemed to take place by the CSNA and the Balance of Payments, even when there was no export.

Discussion Notes: This treatment was required in order to bring the overall supply and disposition of commodity gold into balance. In practice, monetization activity gave rise to gold >available for export= and this amount was classified as receipts. This reflected transfers of gold to the Exchange Fund Account, equivalent to an increase in Canada's external assets. Thus there was an artificial treatment of gold whenever it was monetized.

The 1993 SNA demystifies the treatment of gold: "Transactions in monetary gold consist of sales and purchase of gold among monetary authoritiesY Only gold that is held as a financial asset and as a component of foreign reserves is classified as monetary gold. Therefore, except in limited institutional circumstances, gold can be a financial asset only for the central bank or central government. Purchases (sales) of monetary gold are recorded in the financial account of the domestic monetary authority as increases (decreases) in assets, and the counterparts are recorded as decreases (increases) in assets of the rest of the world. Transactions of other sectors in gold (including non- reserve gold held by the authorities and all gold held by financial institutions other than the central bank) are treated as acquisitions less disposals of valuables (if the sole purpose is to provide a store of wealth) and otherwise as final or intermediate consumption, and/or change in inventories." (paragraph 11.64)

"If authorities add to their holdings of monetary gold by acquiring commodity gold, i.e., newly mined gold or existing gold offered on the private market, or release monetary gold from their holdings for non-monetary purposes, i.e., for sale to private holders or users, they are deemed to have monetized or demonetized gold, respectively. When the authorities acquire gold, the transaction is recorded in the capital account as a positive entry under acquisition less disposals of valuables or change in

Statistics Canada - 241 - Collected Articles of Kishori Lal inventories and counterpart entries are recorded in the accounts of the institutional units or the rest of the world supplying the gold... Monetization or demonetization itself does not give rise to entries in the financial accounts." (Paragraph 11.65)

"The creation or disappearance of monetary gold (referred to as monetization or de-monetization of gold) is - unlike transactions in existing monetary gold B recorded in the other changes in volume account." (Paragraph 11.21).

Acquisition or disposition of official gold reserves can result from transactions between the Bank of Canada and other monetary authorities. However, the Bank of Canada also transacts with suppliers/demanders of commodity gold. Given the lack of information on gold as a store of value, it is assumed that all such transactors hold commodity gold as inventory. With the adoption of the 1993 SNA treatment, new reserve assets are created in the other changes in assets account with commodity gold flowing into this account as inventory and moving out as reserve gold.

Change: Non-monetary gold is now treated as a commodity, and transactions in non-monetary gold are recorded as inventory change. The transformation of gold into a financial asset by the monetary authorities (the so-called monetization of gold) is now recorded in the Other Changes in Assets Account. This treatment was extended back to 1968. No changes were made prior to that date. See also Item 10.2.

10.6 Leasing

Issue: What are the implications of the earlier decisions on financial leasing (Items 4.6 and 6.2), particularly of vehicles, for the financial and balance sheet accounts? Pre-Revision Treatment: In the financial account and the balance sheet account, leases of machinery and equipment were generally treated as financial leases, with leases receivable recorded on the books of the lessors and leases payable recorded on the books of the lessees.

For vehicle leases, the treatment was uneven, since the accounting of lease contracts was not consistent across the leasing branches of the vehicle manufacturers. Some companies accounted for such contracts as financial leases, showing a lease receivable on the balance sheet and interest income on the income statement. Others used operating lease accounting, showing fixed assets (the vehicles) on the balance sheet and rental receipts and depreciation on the income statement. As a result, liabilities in respect of vehicle leases were somewhat understated in the financial account and the balance sheet account, especially those of the household sector.

Discussion Notes: The 1993 SNA states: "A financial lease is a contract between lessor and lessee whereby the lessor purchases a good that is put at the disposal of the lessee and the lessee pays rentals that enable the lessor, over the period of the contract, to cover all, or virtually all, costs, including interest. Financial leases may be distinguished by the fact that all the risks and rewards of ownership are, de facto, transferred from the legal owner of the good (the lessor) to the user of the good (the lessee). The System's treatment of financial leasing is designed to move away from the

Statistics Canada - 242 - Collected Articles of Kishori Lal legal arrangements to capture the economic reality of such arrangements, by treating goods under a financial lease as if they were purchased by the user." (Paragraph 13.23)

Because financial leasing and operating leasing are fundamentally different activities, the 1993 SNA attempts to assign them specific characteristics, while recognizing that it may be difficult to do so: "It is therefore essential to distinguish between the two types of leasing, even though financial arrangements may be devised which are hybrids of the two and which are consequently difficult to classify." (Paragraph 6.119)

It is evident that rules must be established in order to properly classify leasing, especially vehicle leasing, in the accounts. A major objective is to treat leasing as consistently as possible in all the accounts of the CSNA. Another important issue is the existence of factoring (contract sales) and securitization in the leasing market (see Item 10.7), a practice which argues for treating all lease contracts as financial assets under lease receivables=.

Machinery and equipment leases are treated as financial leases in the CSNA. The sector accounts generally allocate the leased capital to non-financial corporations (lessees) and the lease receivables as financial assets of financial corporations (lessors). In addition, a payable to the lessor is recorded as a liability of the lessee.

Vehicle leases to individuals are more complicated. One view is that they are nothing more than rentals, in other words, as operating leases. The other view is to treat them as conditional sale contracts under which the vehicle is being sold to the lessee, in other words as financial leases. Financial leases are essentially a form of financing under which a lessee acquires a property right concurrently with a liability.

Vehicle leases are characterized by the following:

$ the lessee normally keeps the vehicle, and has exclusive use of it, during a significant portion of the asset's economic life; $ the lessee assumes the risk and responsibility for the vehicle; $ the lessee has the option to purchase the vehicle at the end of the lease for the pre-determined residual value.

In short, a lease that transfers the benefits and risks of ownership from the lessor to the lessee should be accounted for as a financial lease.

It was agreed that, in household consumption expenditure, vehicle leases greater than one year (almost all such leases) would be treated as financial leases. This implies that borrowing and debt in the form of leases should be allocated to the household sector. As a result, subsequent principal payments would be shown in the financial account along with changes to the corresponding liability in the balance sheet account. Specifically, there would be additional >other loan= liabilities (lease payables) in the household sector, equal to the remaining principal payments plus the residual value. These liabilities would be matched by receivables on the books of the lessors. The lessors are found

Statistics Canada - 243 - Collected Articles of Kishori Lal in the following sub-sectors in the financial and balance sheet accounts: (i) sales finance and consumer loan companies, (ii) other private financial institutions, (iii) non-financial private corporations and (iv) issuers of asset-backed securities.

Change: The 1993 SNA recommendation to treat vehicle leases to individuals as financial leases has been implemented. It has led to an increase in the borrowing and debt of the household sector as measured in the financial account and the balance sheet account. The treatment of vehicle leasing is now in line with that of leasing of machinery and equipment.

10.7 Asset-backed Securities

Issue: Asset-backed securities result from the pooling of various types of loan assets and the corresponding issue of a >security=, whose income is derived from the assets. In most cases, the securitized assets are removed from lenders' balance sheets. Should this off-balance sheet activity (i.e., asset-backed securities, and the resulting securitized assets) be included in the financial and balance sheet accounts, as recommended in the 1993 SNA? If so, how?

Pre-Revision Treatment: The treatment of this activity was partial and inconsistent. The focus was on accounting for securitized assets (in order not to understate credit market debt and borrowing), rather than on the security itself. For example, National Housing Act mortgage-backed securities (NHA-MBS) were treated not as securities, but as mortgage assets of the household sector. Other asset-backed securities were accounted for in a hit and miss fashion. Unanticipated drops in loan assets were verified and, if survey respondents indicated securitization activity, the appropriate amounts were added back to the assets of the lenders. However, the securities themselves were ignored.

Discussion Notes: The 1993 SNA states: "New negotiable securities are often issued backed by existing assets such as loans, mortgages, credit card debt, or other assets (including accounts receivable). This re-packaging of assets is often referred to as securitization. The creation of the new assets gives rise to entries in the financial account and the new assets should be classified as securities other than shares." (Paragraph 11.75)

First, asset-backed security issues must be recorded as liabilities of the issuer, as they are assets of other institutional sectors. Second, the securitized assets must also be integrated into the financial accounts, in order to correctly measure credit market debt, both mortgage and non-mortgage.

Common types of instruments securitized are credit card receivables, vehicle loans, lease receivables, commercial loans and residential mortgages (conventional and insured under the NHA).

Since the process of securitization involves the creation of a special purpose trust that (i) holds securitized assets in trust and (ii) issues asset-backed securities, a new institutional sector must be established. Statistics Canada has recently implemented a survey of such trusts.

Statistics Canada - 244 - Collected Articles of Kishori Lal

Change: The recommendation of the 1993 SNA to include asset-backed securities in the financial account and the balance sheet account has been adopted. They are recorded under >securities other than shares=. Most of the activity is shown in a new institutional sector called >Issuers of Asset- backed Securities=, which holds most of the securitized assets.

Statistics Canada - 245 - Collected Articles of Kishori Lal

This paper deals with a problem in internationally comparable economic statistics, namely, the fact that countries measure value added by industry differently. The economic measure, value added, is important both in its own right and because it is a component of other economic measures such as productivity. Value added by industry measures the additional value created by a production process. This additional value, created by factors of production such as labour and capital, may be calculated either before or after deducting the consumption of fixed capital used in production. Thus, gross value added by industry is the value of its output of goods and services less the value of its intermediate consumption of goods and services and net value added as the value of output less the values of both intermediate consumption and consumption of fixed capital.

Different Ways of Measuring Value Added by Industry

Although the above definition of value added by industry is unambiguous, expressing it in monetary terms is neither conceptually straightforward nor uniform among countries, as is indicated by the fact that three different ways of measuring value added by industry are currently prevalent among OECD member countries. Consequently, their respective measures of value added by industry are different and non-comparable. Further, or more importantly, the incorporation of these different measures into other economic measures renders them different and non-comparable as well. The OECD publication, Services Statistics on Value Added and Employment (1997 edition) lists value added by industry in the services sector and reports that, in each of its member countries (except Poland and Korea), value added is measured in terms of basic prices, factor cost, or market prices. Canada, Denmark, Greece, Iceland, Ireland and the United Kingdom measure value added at factor cost; Finland, Hungary and Sweden measure it at basic prices, and the other 18 countries measure it at market prices.

The 1993 SNA=s Treatment of Value Added by Industry

Let us situate the present practices of the OECD countries in the context of the 1993 SNA. The 1993 SNA recognises three alternative measures of valuation of value added: gross value added at basic prices, gross value added at producer=s prices and gross value added at factor cost, the last one derivable from either of the two former ones. These measures depend upon how the output is valued; the valuation of intermediate consumption is in all cases at purchasers= prices. Two methods of valuing output are specified: basic price and producer=s price. These methods are defined as follows:

AThe basic price is the amount receivable by the producer from the purchaser for a unit of a good or service produced as output minus any tax payable, and plus any subsidy receivable, on that unit as a consequence of its production or sale. It excludes any transport charges invoiced separately by the producer@ (paragraph 6.205a). AThe producer=s price is the amount receivable by the producer from the purchaser for a unit of a good or service produced as output minus any VAT, or similar deductible tax, invoiced to the purchaser. It excludes any transport charges invoiced separately by the producer@ (paragraph 6.205b). Output valued at producer=s price thus includes taxes like excise taxes and other non-deductible taxes on products. The concept of purchaser=s price is quite straight-forward and is defined as follows: AThe purchaser=s price is the amount paid by the purchaser, excluding any deductible VAT or similar deductible tax, in order

Statistics Canada - 249 - Collected Articles of Kishori Lal to take delivery of a unit of a good or service at the time and place required by the purchaser. The purchaser=s price of a good includes any transport charges paid separately by the purchaser to take delivery at the required time and place@ (paragraph 6.215).

Although the 1993 SNA recognises two methods of valuation of output, its preference is clear: AThe preferred method of valuation is at basic prices , especially when a system of VAT, or similar deductible tax , is in operation, although producer=s prices may be used when valuation at basic prices is not feasible@ (paragraph 6.218). Valuation at basic prices may not be feasible if a country=s establishment surveys do not separate the excise and sales taxes on products in the revenues earned from the sale of products. In that case, valuation at producers= prices is the only alternative.

Some comments on these alternative valuation methods, using Canadian statistics, are in order. Typically, Canada=s establishment surveys report the value of production or sales at the boundary of the producing units, an approach that excludes taxes on products at time of sale as well as separately invoiced transport charges. Canada=s valuation of output is at modified basic prices; the main difference between it and the 1993 SNA definition of valuation at basic prices relates to subsidies on products. Canada=s output is valued at the subsidised prices, the prices actually paid by purchasers, while the 1993 SNA recommends valuing output at actual prices plus subsidies. The latter approach results in purchasers reporting payments for products at values higher than their actual prices, counterbalanced by their assumed receipt of subsidies, as subsidies in the 1993 SNA are allocated to users of products. The 1993 SNA=s approach distorts the records of the reporting units, which is why Canada prefers actual prices appearing on invoices. This is not a major issue, as the volume of product subsidization is quite small in Canada. In effect, Canada uses approximate basic prices.

As mentioned earlier, the 1993 SNA recognises three alternative measures of value added by industry, derivable from two valuation measures of output and from the purchasers= price valuation of intermediate consumption. These are: gross value added at basic prices, gross value added at producers= prices and gross value added at factor cost. According to the 1993 SNA: a. AGross value added at basic prices is defined as output valued at basic prices less intermediate consumption valued at purchasers= prices@ (paragraph 6.226). b. AGross value added at producers= prices is defined as output valued at producers= prices less intermediate consumption valued at purchasers= prices@ (paragraph 6.227). In the absence of VAT or similar deductible tax, this is equivalent to the Ameasure of gross value added at market prices@ (paragraph 6.227). c. AGross value added at factor cost can ... be derived from gross value added at basic prices by subtracting other taxes less subsidies on production@ (paragraph 6.229).

The 1993 SNA prefers gross value added by industry at basic prices and accepts value added by industry at producers= prices when valuation at basic prices is not feasible.

Statistics Canada - 250 - Collected Articles of Kishori Lal Statistics Canada publishes gross value added at factor cost by industry. In its surveys, revenues from output or sales are typically recorded at modified basic prices and purchases of goods and services to be used for intermediate consumption are typically reported at purchasers= prices. The Canadian value of output at modified basic prices minus intermediate consumption at purchasers= prices approximates the 1993 SNA=s value added at basic prices. However, Statistics Canada publishes value added by industry at factor cost, by subtracting from the modified basic price calculation of value added Aother taxes less subsidies@ on production. In terms of magnitude, the most important other taxes on production are real estate and payroll taxes and other subsidies on production are subsidies for manpower training. Note that most subsidies paid in Canada are subsidies on products such as wheat and other grains, subsidies on intra-urban local bus passenger fares, subsidies on the movement of grains by railways etc. Statistics Canada prefers to measure value added by industry at factor cost, as this measure, in its judgement, better reflects the intuitive meaning of the concept. As indicated, this concept has been extensively used both for productivity analysis and, more generally, in the economic literature for a long time.

It was noted above that many OECD countries measure value added by industry at market prices, with the result that, in their statistics, excise and other sales taxes on products levied at point of sale are added to revenues of producing units. In other words, both the values of output and of value added, of a given industry, are increased by the amount of sales taxes collected by that industry, although they are transmitted to the government which levied them. For example, in the cases of trading enterprises and hotels and restaurants, all collected sales taxes increase their respective value added, even though they do not retain these funds. Although recognised in the 1993 SNA, this approach is counter-intuitive. In addition to its other consequences, the higher the sales taxes levied by the government, the higher are the corresponding amounts of value added of trade, in spite of the fact that the enterprises have not used their own factors of production to generate the increase in value added.

To summarise, all three concepts of value added by industry: at factor cost, at basic prices, and at market prices are prevalent in OECD member countries. Of these, value added by industry at factor cost produces the lowest total. Value added by industry at basic prices raises the measure of value added by industry, principally of real estate and owner-occupied dwellings, although the value of output is unchanged. On the other hand, measuring value added by industry at market prices further raises the values of both output and of value added by industry, principally of trade and hotels and restaurants. In the presence of VAT or similar deductible taxes, the very meaning of value added by industry at market prices is questionable. As the 1993 SNA puts it: Ain the presence of VAT, the producer=s price excludes invoiced VAT, and it would be inappropriate to describe this measure as being at market prices@ (paragraph 6.227). Since value added by industry, measured by one or another of these three valuations, is published by the OECD, the unwary user is apt to consider the series comparable when, in fact, they are far from comparable. The differences serve to render problematic, if not distinctly misleading, measures which incorporate them, such as measures of industry shares and productivity.

Some Statistical Evidence

The magnitude of differences between the three concepts may be illustrated by statistics pertaining to the United States and Canada. GDP at market prices for the United States in 1992

Statistics Canada - 251 - Collected Articles of Kishori Lal was US$6,234 billion, as reported in its 1992 input-output tables (summary Table 2.2, page 20). In terms of value added by industry, the shares of the components were as follows: $3,645 billion for labour compensation, $506 billion for indirect business tax and non tax liability (equivalent to the 1993 SNA=s taxes on production less subsidies on production) and $2,083 billion for other value added (equivalent to the 1993 SNA=s mixed income of unincorporated business enterprises and other operating surplus). As GDP at factor cost removes all net taxes on production, it would equal $6,234 billion less $506 billion or $5,728 billion. Indirect taxes contributed 8.1% of GDP at market prices at the total economy level. However, two thirds of the indirect taxes, $328 billion, were levied just in two sectors, trade (wholesale and retail) and real estate (including owner occupied). These two sectors together contributed $1571 billion or 25.2% of value added at market prices, but their contribution dropped to $1244 billion, or 21.7% of value added at factor cost. This represents a huge drop in industrial share.

GDP at market prices for Canada in 1992 was Canadian $698 billion-$604 billion for value added at factor cost, $57 billion for net taxes (taxes less subsidies) on products and $37 billion for other net taxes (taxes less subsidies) on production. Other taxes on production refer, as noted above, to such taxes as real estate property taxes and payroll taxes. Net other taxes on production were allocated to industries paying such taxes, but $22 billion of $37 billion were paid by the finance insurance and real estate sector. Thus, if Statistics Canada published value added by industry at basic prices, total value added would increase to $641 billion (604+37), and the share of the real estate sector would be much higher, as a disproportionate share of property taxes are paid by it. With our existing surveys, Statistics Canada cannot produce value added by industry at market prices, as respondents are asked to report value of sales of products excluding any taxes on products imposed at time of sale.

In the United States, the allocation of taxes on products by industry for the value added by industry calculation closely reflect the regulations and practices that determine which industry is liable for tax collection during the periods the taxes were imposed. (Additional details may be found in a paper by Robert E. Yuskavage, AImproved Estimates of Gross Product by Industry@ published in the August 1996 issue of Survey of Current Business (page 140), the monthly publication of the US Bureau of Economic Analysis). There is no generally accepted economic theory behind this allocation of taxes on products. If the regulations change, the industrial allocation of taxes changes, hence the value added also changes.

Different Tax Regimes

In many other OECD countries which also publish value added by industry at market prices, the tax regime is quite different from that of the United States. VAT is quite prevalent in these countries, thus industries do not pay taxes on products and per the1993 SNA guidelines (or per general economic reasoning) their producer prices should not include taxes on products. Many of these countries are members of the European Community and have now implemented the 1993 SNA in their accounts. Will they continue to publish value added by industry at market prices? If yes, will their statistics on value added by industry at market prices include VAT? If they do, what is the rationale? If they do not, where are such taxes allocated?

Statistics Canada - 252 - Collected Articles of Kishori Lal Value Added by Industry at Constant Prices

We have seen how different and non-comparable are the different methods at current prices. The differences are exacerbated at constant prices. Even with the same current price values of output and of value added by industry, constant price measures and growth rates of value added by industry will differ for many reasons. Some of them are noted here. Constant price measures may be calculated with reference to a fixed base year or an annually changing base year and the results are different. Constant price value added by industry may be calculated residually using a double deflation method or the base year values may be projected using such measures as constant price gross output, employment by industry or any other proxy. Again, the results are different. Constant price calculations are very sensitive to the choice of index number formulas, base year, method of derivation, etc. Thus, one needs to be doubly cautious in the analysis of such an important constant price series as value added by industry.

Concluding Remarks

It has been shown that the measures of value added by industry published by the OECD countries using three different methods of calculation are non-comparable. As or more importantly, the incorporation of these different measures into other economic measures, such as productivity, renders them different and non-comparable as well. But data users do not necessarily know this. They generally expect such measures to be comparable because they are published under one title and without caveats; consequently, their analysis could and would be unsound.

Therefore, I appeal to the OECD and other international organisations to start adding bold faced caveats on non-comparable data published in their name. Of course, it would be better to encourage member countries to provide data using internationally recommended guidelines and conventions. However, given their unique statistical structures and statistical data sources, it may not be possible for some countries to completely follow the international guidelines. In such circumstances, and for such an important series as value added by industry, it is incumbent upon all of us to develop macro adjustments, by industry, to Acorrect@ these country series before they are reported in international publications. Such macro adjustments can never be perfect, but their inclusion will vastly improve the international comparison of value added by industry and the analyses following from it.

Note:

I am grateful to many colleagues in the System of National Accounts Branch at Statistics Canada, particularly Abe Tarasofsky, Lucie Laliberté and Yusuf Siddiqi for their very useful comments and suggestions in the preparation of this paper. This paper was presented at the OECD Meeting of National Accounts Experts, Paris, 21-24 September 1999.

Statistics Canada - 253 - Collected Articles of Kishori Lal

Input-Output Accounts – The Rectangular Format

By Kishori Lal Statistics Canada, Ottawa

Input-Output tables are like recipe books. Inputs are the ingredients to produce outputs. Like a good chef, one lists the ingredients used by each and every industry to produce output(s). One arranges these recipes in some order so that these can be easily located and the retrieval is efficient. No large scale database base can be satisfactorily operated without an efficient retrieval system. The arrangement of the database should make economic and institutional sense or one can say that the database should portray the realities of the economy. In addition to the industries, there are other transactors in the economy who use similar ingredients but they don't produce any marketed output. These are called categories of final demand. The household sector, government sector, business sector for capital formation (including inventories), and the rest of the world sector (exports less imports) use inputs as final consumers. Thus one has an input table by industry and an output table by industry and an input table by category of final demand.

In the rectangular format of the Input-Output tables, inputs and outputs of industries are presented in separate tables; both inputs and outputs are classified by commodity. Commodities are clearly distinguished from industries, the number of commodities exceeding the number of industries.1 The final demand matrix has the dimensions of commodities by categories of final demand.

In the traditional square Input-Output (inter-industry) table, industries are not distinguished from commodities. No separate tables are presented for inputs and outputs. Only one table is presented, whose dimensions are industry by industry. The final demand matrix here has the dimensions of industries by categories of final demand. The column and row totals of the inter-industry table are equal (actually are manipulated to be equal).

The inter-industry table cannot be assembled directly from the data derived from census or survey records. In the real world, an industry produces not only the commodities which are primary to, or characteristic of, the industry, it also produces commodities which are characteristic of other industries (secondary commodities). Industries can report how much of each commodity they produced and they can report the commodities they consumed but they usually cannot report separately the commodity inputs required for each commodity output. Consequently it is not possible to measure directly how much one industry uses of another industry's output.2

Recently, Professors Carter and Leontief mentioned two very important advantages of the rectangular format of the Input-Output tables over the traditional square Input-Output tables 3:

(1) It admits as much detail as is available in the basic census or survey records; and

(2) The meaning of each entry is straightforward because observed transactions are not combined with fictitious transfers (a feature of inter-industry square tables).

Statistics Canada - 257 - Collected Articles of Kishori Lal

In addition to the above two advantages, there are other perhaps even more important advantages which need to be emphasized. The rectangular format can be used as an integration tool for statistical development. Indeed, it forces integration of the various statistical series in the system. This derives from its commodity dimension and commodity balances.

So many heterogeneous commodity classifications prevail in most economies that it is difficult to find a common denominator while retaining reasonable detail. Just look around at the export commodity classification, import commodity classification, commodity classification for industrial censuses, commodity classification used in national sample surveys on consumer expenditures, economic classifications of governments expenditures, and try to find a reasonable size common denominator. Chances are that these various classifications won't match. We will have more to say on classifications later on in this article.

Input-Output tables present the most detailed accounting of a particular economy. In the preparation of rectangular tables, there are many items which are not available from census or surveys. One has to estimate them. This gives the management of a statistical agency a statistical tool to identify and rectify the weaknesses or gaps in the statistical series. These weaknesses are not as well exhibited in the inter-industry table.

Despite the above, many countries continue adopting the traditional approach to produce Input-Output tables. The UN System of National Accounts 4 does not recommend it. The database in the real world does not encourage its adoption. Interpretation of each entry is rather difficult. One avoids at least these problems in the rectangular format. Then why hang on to the traditional square inter-industry tables? Economists and statisticians are not noted for discarding traditions even when real world database does not support them!

Input-Output tables, as we noted above, present the most detailed accounting of a particular economy. One must examine all the industries and all the categories of final demand in a given economy. One must note each and every input used by each industry. Some of these inputs will be available from census or surveys. Others might have to be estimated from other sources. Even engineering data should be used. Industries will be using commodity inputs and some non-commodity or what we call primary inputs. Primary inputs are: indirect taxes less subsidies; wages and salaries; supplementary labour income; net income of unincorporated business and other operating surplus. Note that the primary inputs are those inputs which are not current outputs of other industries. On the other hand, a commodity is defined as a good or service normally intended for sale on the market at a price designed to cover the cost of production. Primary inputs are not commodities.

One should count all the inputs and all the outputs for each and every industry. Difference between the outputs of an industry and the inputs is "other operating surplus" which could even be negative. All the inputs (commodities plus primary inputs including operating surplus) equal commodity outputs of the same industry. A similar detailed listing of inputs must be done for categories of final demand.

Statistics Canada - 258 - Collected Articles of Kishori Lal

Classifications

Detailed listing of outputs of industries and of inputs of industries and categories of final demand require certain classification constraints. One has to develop, if not already available, a Standard Industrial Classification, which uniquely allocates a particular establishment to a single industry only. And unless the structure of the establishment has changed, it should be given the same industrial classification year after year.

An establishment is formally defined as the smallest unit that is a separate operating entity capable of reporting all elements of basic industrial statistics. It is typically a factory, mine, store, farm, airline, hotel or similar unit.

An industry is defined as a group of operating units (establishments) engaged in the same or similar kind(s) of economic activity.

Many countries don't have good Standard Industrial Classification. The detail of a particular class of industries will be different in different countries. This definitely makes international comparisons rather difficult.

The Industrial classification should be carefully examined. One should not accept in total ready-made international classification. Each economy needs a particular emphasis in the industrial classification. For example, in a highly developed industrial economy, agriculture might contribute 5 per cent of gross domestic product but in India its share is roughly half. One has to disaggregate agriculture into many sub-industries which may not be necessary in advanced countries. One might disaggregate agriculture by crops. Quite useful classification was developed by Professor G.S. Bhalla for his Haryana Input-Output tables.

Classifications are expensive to produce. They should be stable over long periods. One should not only look at the importance of certain industries now but the potential importance in the foreseeable future too. For example, energy is going to remain a serious problem for India at least in the foreseeable future. Hence, one should build many sectors by type of energy even if their present importance in terms of contribution in minimal. Also one should leave room in the classification scheme to add new industries. Statistical system in India probably will be extensively computerized in future. Thus, one should keep computer systems dimension in mind. Computer programs are expensive to write and untidy programs are costly to run. Note that in the computer system one is billed by seconds of CPU (Central Processing Unit) time and the present rate is Rs 3 per second or ten thousand rupees per hour.

Categories of Final demand usually have to be developed. But they must in part relate to the industrial classification used. For example, business capital formation might have as many sub-categories as industries (or aggregation of industries) so that one could relate capital to output. The government sector should be disaggregated in such a way as to make it easy to study the regional, functional and economic dimensions of the government expenditures. For example, expenditures on education must always be segregated from other government expenditures. These are big expenditures and have different economic effects than expenditures on say, defence.

Statistics Canada - 259 - Collected Articles of Kishori Lal

Disaggregation of the household sector should relate to other macro aggregates usually produced in the Income and Expenditure Accounts.

The biggest problem in classification is the commodity classification.5 Goods and services produced and used (or sold and purchased) by all transactors should be uniquely listed. Even in the economically and statistically advanced countries, the same commodity is labelled differently by the exporter, importer, producing industries, using industries and other users. Without a unique code assigned to a commodity by all transactors, it will be impossible to balance supply with demand, a requirement and a very useful feature of rectangular tables. This is the most time consuming and troublesome part of the production of Input-Output tables.

It will be very useful to classify primary products of industries in the same sequential order as industries. In the process of classifying primary products, it is quite possible that one will be forced to re-examine the already decided sequential order of industrial classification. This is an iterative process. As an industrial classification, one should always make sure that enough space is left to add commodities later; furthermore, it is important to provide for commodities which are not very important now but which might become important in the future. One must create at least one commodity for each industry, but it is our experience that, at least in the initial stages, the total number of commodities should be two or three times the number of industries. Some industries will produce only one primary product whereas others will produce more than one. It is easier to aggregate than to disaggregate. Each economy is quite often required to develop its own commodity classification for input-output purposes because the existing classifications are hardly suitable for purposes of commodity balances. Without a proper classification one can't produce Input-Output tables. Indeed, for any statistical compilation, the fundamental element is classification.

Inputs and Outputs v/s. Purchases and Sales

Transactions as recorded in census or surveys don't always refer to outputs or inputs but to sales or purchases by transactors. In the input-output analysis, we are concerned with the technological relationships of production processes. Hence, it is necessary to relate inputs to production rather than to sales. Production (output) equals sales adjusted for withdrawals or addition of finished goods and goods-in-process from the producers' inventories. Similarly, it is an industry's consumption or use (input) of a commodity rather than its purchase, which must be related to output. Amounts purchased but not yet used should show as an addition to raw material inventories. Here is a very serious statistical problem. This requires data on inventories by commodities. It is worthwhile to have occasional surveys of inventories of some important industries for certain commodities so that one has a "feeling" for adjusting sales and purchases. In the absence of that, one could only prorate raw material inventories over all purchases and finished goods inventories over all sales.

Valuation

All the above transactions (inputs by industries and by categories of final demand), being derived basically from accounting records, are in purchaser prices. Purchaser price is defined to be the cost of goods and services in the market to the point of delivery to the purchaser. Purchaser price can also be defined as equal to producer price plus transport, trade and commodity tax margins (where

Statistics Canada - 260 - Collected Articles of Kishori Lal applicable). Producer price is the selling price at the boundary of the producing establishment excluding sales and excise taxes levied after the final stages of processing. These definitions are not universal. Indeed they are not exactly the same as in the UN System of National Accounts.

Input-Output tables are finally recorded in producer prices though initially one estimates the commodity balances in purchaser prices.

In a producer price system every transactor pays the same unit price for the same commodity whereas in the purchaser price system it varies with the transactors. Let me elaborate this point. Suppose a consumer buys a shirt. The shirt is produced in Bombay, shipped to Amritsar for sale and the government levies some sales tax on it. The producer sold to the wholesaler for Rs. 20. The wholesaler added a margin of Rs. 5 and sold it to the retailer; then retailer added a margin of Rs. 10 and the taxes are Rs. 3. The transportation cost, say, is Rs. 2. Total purchaser price is Rs. 40. Unless one disaggregates the purchaser price transaction of Rs. 40 into the various margins, one can't relate to the price which the producer charges. Furthermore, without disaggregating the purchaser price, one can't account for such industries as trade and transportation. In both systems the total price does not change for the given shirt:

Producer price 20 Wholesale Trade Margin 5 Retail Trade Margin 10 Transport 2 Tax 3 ------Equals Purchaser Price 40

The consumer pays Rs. 40 or pays producer price Rs. 20 plus trade margins Rs. 15, plus transport cost Rs. 2, plus tax Rs. 3. The producer price system slices the purchaser price into the various components.

Purchaser price Input-Output tables are very difficult to manipulate. Let us elaborate it by giving an example of how to remove price change from Input-Output tables in current prices. Suppose one has a table comprising of 595 commodities, 191 industries, and 136 different categories of final demand (these are the dimensions of the Canadian Input-Output tables). Here there are 327 transactors. In a purchaser price system, for each commodity one could potentially need 327 deflators-frightful task indeed! Let us take the above example of a shirt. At purchaser price, a consumer pays Rs. 40. Suppose the same shirt is purchased by an exporter. He does not incur retail margins, perhaps no tax either. He pays only Rs. 27. We have two purchaser prices for two transactors. In the producer price system, the exporter pays Rs. 20 producer price, Rs. 5 wholesale margin and Rs. 2 transport margins. The same producer price for every transactor. Thus in the producer price system, one needs one deflator for one commodity by all the 327 transactors. Then one needs deflators for margins, transport cost and taxes to complete the system but there again for a particular margin say retail margin, one needs only one deflator for all the 327 transactors.

Statistics Canada - 261 - Collected Articles of Kishori Lal

In the Input-Output tables, trade sectors are defined as producing margins. In the above example of shirt, wholesaler purchased the shirt for Rs. 20 and sold it for Rs. 25, margin being Rs. 5. We have already accounted for Rs. 20 from the producer so we only need to add margin of Rs. 5 from the wholesaler to bring it to the doors of the retailer.

Suppose one would like to study the contributors to inflation. Purchaser price system does not distinguish the various intermediaries (sometimes, major cause of inflation) whereas producer price system does. Then one can look at all the stages through which our shirt has gone and find out who has contributed to inflation.

The Accounting Framework

Having made above the introductory remarks to the rectangular system, we will now introduce the accounting framework in Chart 1 that we have developed for the Canadian Rectangular Input-Output tables. Not many countries have experience in Rectangular Input-Output tables. Canada has now developed annual Input-Output tables in both current and constant prices for 14 consecutive years 1961 to 1974 in the rectangular format.6 Work is now in progress to produce Input-Output tables for the following 2 years, 1975-76.

The Rectangular Input-Output tables contain two sets of interrelated accounts, the Commodity Accounts and the Industry Accounts. The former details the supply and disposition of individual commodities. The latter details the commodity composition of the output of industries and the complete costs of production (including profits) of industries. These accounts are described below. While the Commodity and Industry Accounts treat the primary inputs and final demand, these entries are also described together in a separate section.

Commodity Accounts

As noted, the Commodity Accounts display the supply and disposition of each commodity. Matrix V displays the production of each commodity produced by each industry, valued at producers' prices. (The total supply of each commodity includes not only the production of the commodity by domestic industries but also three of the final demand categories - imports, withdrawals from inventories and government revenue from the sales of goods and services). As Input-Output analysis is concerned with the technological relationships of production processes, it is necessary to relate inputs to production rather than to shipments. Production equals shipments adjusted for withdrawals or additions of goods-in-process and finished goods from the producers' inventories.

The disposition of commodities by the various classes of transactors are shown in matrices U and F. Matrix U contains the use, on current account, by each industry of each commodity for the purpose of using it as an intermediate input in the production of other commodities. As implied above, it is an industry's consumption or use of a commodity which must be related to production. Thus entries here refer to amounts used and not to purchases. Amounts purchased but not yet used show as additions to raw materials inventories.

Statistics Canada - 262 - Collected Articles of Kishori Lal

The total laid down cost of a commodity at the using establishment (the purchasers' prices) will often include for-hire transportation charges, trade margins (if the commodity passes through one or more trading establishments) and commodity taxes such as sales or excise taxes; these margins are an addition to the producer's price.

It is necessary to have a uniform basis of valuation for entries as the amounts of transportation charges, trade margins and commodity taxes included in purchasers' prices vary considerably from one purchaser to another. Thus the entries in matrix U do not usually represent the actual transaction values (since these may include margins), but they represent estimates of the producer's value of the transactions. For-hire transportation charges and trade margins are treated as separate commodity inputs purchased from the industries producing those margins. Commodity taxes levied after producer's price stage and included in the purchaser's value are shown as inputs into the purchasing industry in the row entitled "Commodity taxes" (matrix Yl). Therefore the total costs of each industry include the full value of all transactions in respect of intermediate inputs.

Matrix F contains the demand of each commodity by final demand category - personal expenditure, fixed capital formation, additions to (the value of physical change in) inventories, gross government expenditure on goods and services and exports. The three other columns in matrix F are imports, withdrawals from (the value of physical change in) inventories and government revenue from the sales of goods and services. These three categories supplement the supply of goods and services produced by domestic industries. Matrices U and F include the use of commodities originating either as imports, as withdrawals from inventories, or as government-produced goods and services. Three adjustments are thus required - imports because, they are extraneous to the production of domestic industries, withdrawals from inventories because these were produced in the earlier period, and government revenue from the sale of goods and services because the costs have already been fully accounted for in government gross expenditure - to effect equality with the output of domestic industries.

The methods of valuation and the treatment of transport, trade and commodity tax margins for commodities purchased by transactors in the final demand categories are identical to those employed for the intermediate demand of industries.

It is possible to present the exogenous sources of commodities, imports, withdrawals from inventories and government sales, as additions to production (extension of the V matrix) rather than as adjustments to use (negative value in the final demand column vectors). The present formulation can be identified more easily with the traditional presentation in the Income and Expenditure Accounts, and this relationship will be developed below. In any case, for each and all commodities, total supply equals total disposition.

The Industry Accounts

The total output of each industry, classified by commodity, is shown in the industry's row of matrix V. The inputs of each industry are shown in the Industry's column which covers both matrix U for intermediate inputs and matrix Yl for primary inputs. Matrices U and V were discussed above under the Commodity Accounts.

Statistics Canada - 263 - Collected Articles of Kishori Lal

The primary inputs in matrix Yl consist of net indirect taxes (commodity indirect taxes, plus other indirect taxes, less subsidies), wages and salaries, supplementary labour income, net income of unincorporated business and surplus.

Chart 1 The Accounting Framework of Rectangular Input-Output Tables

Final Demand Commodities Industries Categories Total

Commodities U F q

Industries V g

Primary Inputs Y1 YF n

Total q/ g/ e/ Notations used in the above chart: V : is a matrix of the values of outputs U : is a matrix of the values of intermediate inputs F : is a matrix of the values of commodity inputs of final demand categories Yl : is a matrix of the values of primary inputs of industries YF : is a matrix of the values of primary inputs of final demand categories q : is a vector of the values of total commodity outputs and q/ is the transpose of q g : is a vector of the values of total industry outputs e : is a vector of the values of total inputs (commodities plus primary) of final demand categories n : is a vector of the values of total primary inputs (industries plus final demand categories)

Matrices defined: V : is a matrix of the values of commodity outputs. There are 595 commodities and 191 industries in the Canadian tables. In it, each row shows the distribution by commodity of the output of an industry; each column shows the distribution by industry of output of a commodity. The data relate to domestic output only. The gross output of an industry is the aggregate value of goods and services produced and work done by the industry. It is equal to the value of industry's sales plus any increase (less any decrease) in the value of physical change in stocks of finished products and work in progress. The matrix dimensions are 191 industries by 595 commodities. U : is a matrix of the values of intermediate commodity inputs. In it, each row shows the distribution by industry of the input of a commodity, each column shows the distribution by commodity of the input of an industry. The matrix dimensions are 595 commodities by 191 industries. F : is a matrix of the values of commodity inputs of final demand categories; personal expenditure on consumer goods and services; fixed capital formation, business and government; value of physical change in inventories, withdrawals and additions; gross government current expenditure on goods and services; exports, imports; and government

Statistics Canada - 264 - Collected Articles of Kishori Lal

revenue from the sales of goods and services. The matrix dimensions are 595 commodities by 136 categories of final demand.

Yl : is a matrix of the values of primary inputs of industries. Primary inputs are: indirect taxes less subsidies, wages and salaries, supplementary labour income, net income of unincorporated business and other operating surplus. Other operating surplus in the usual Income and Expenditure Accounts includes profits and other investment income, inventory valuation adjustments, capital consumption allowances and miscellaneous valuation adjustments. The matrix dimensions are 7 primary inputs by 191 industries. YF : is a matrix of the values of the primary inputs associated with final demand categories. These consist of indirect taxes, wages and salaries, supplementary labour income, and depreciation which is part of surplus. Wages and salaries and supplementary labour income are paid by the government and personal sectors. The estimate of surplus (depreciation) relates to the government sector and non-profit institutions in the personal sector. The matrix dimensions are 7 primary inputs by 136 categories of final demand.

The commodity indirect taxes represent the sum of commodity taxes levied beyond the producers' prices valuation level on the commodities used by each industry as intermediate inputs. (Commodity indirect taxes and other indirect taxes, paid by final demand categories and shown in matrix YF, are discussed below). Government subsidies received by industries are treated as revenues and are shown as negative entries in the input vectors of the receiving industries. Wages and salaries cover all payments, including payments-in-kind such as commissions, bonuses, tips, and taxable allowances. Wages and salaries are calculated on a gross basis, that is before tax deduction.

Supplementary labour income consists of other expenditures by employers on labour account which are regarded as payments for employees' services. Included here are employer's contributions to unemployment insurance, pension funds and other social insurance schemes.

The net income of unincorporated business includes the net earnings of working proprietors, single or in partnership, earnings from independent professional practice and the accrued net income of farm operators from farm production. It also includes the net rental income of persons from the rental of residential and non-residential property. Owner-occupants of housing are considered to be in business renting to themselves, and this imputed net rental income is also included here.

The entry surplus includes the profits generated by corporations and government business enterprises (before dividends), all other investment income (interest, royalties) except the rental income noted above, capital consumption allowances for both the corporate and unincorporated sectors, the inventory valuation adjustment and certain other valuation adjustments to correct the values reported from business accounting records to the requirements of the System of National Accounts.

As noted, surplus includes profits before distribution, that is dividends are attributed to the paying, not the receiving industry; the interest element in surplus is treated in like fashion: interest payments form part of the surplus of the paying industry, rather than the receiving industry.

For each and all industries total outputs equal total inputs, intermediate and primary.

Statistics Canada - 265 - Collected Articles of Kishori Lal

Primary Inputs and Final Demand

As is evident from the Chart of the Accounting Framework, the discussion of the Commodity Accounts and the Industry Accounts covers most of the elements of primary inputs and final demand. However, it is useful to treat these categories concurrently. They represent the measures of Gross Domestic Product, and can be directly related to other parts of the System of National Accounts.

There is one small matrix not yet described, YF. This contains elements of primary inputs purchased directly by final demand categories. These are discussed in turn.

It has already been emphasized that the output of commodities is valued at producers' prices, that is before the addition of indirect commodity taxes and other margins levied after the final stage of processing, and that entries are debited to the purchaser at this valuation, with the margins being accounted for separately. Thus when commodities are bought in final demand categories, the value of indirect taxes (and other margins) payable on a commodity must be added to that commodity's column in each of the final demand categories to yield the actual transaction value.

Custom import duty is added (if applicable) to the value of each imported commodity, to bring its valuation in line with the producers' prices. As this treatment inflates the value of total imports by the level of custom import duties, an offsetting adjustment equivalent to total value of import duties is made at the intersection of the import column and the commodity indirect taxes row.

Surplus in matrix YF reflects the depreciation on assets owned in the government sector and by non-profit institutions classified to the personal sector. Assets which are charged to fixed capital formation in these sectors are depreciated, such as buildings, roads, equipment, etc. All residential housing, even when owner-occupied, is classified to business, as noted, and thus depreciation on housing is part of the surplus of the appropriate industry in matrix Yl, not YF.

With the description of all the submatrices, certain identities become apparent.

The total outputs of industries equal the total intermediate inputs of industries plus their total primary inputs.

The total outputs of commodities of industries equal the total intermediate inputs plus the total commodity inputs into the final demand categories.

Intermediate inputs being common, primary inputs of industries equal commodity inputs of final demand categories.

Adding primary inputs of final demand categories to both sides, we get:

(a) primary inputs of industries and of final demand categories equal to

Statistics Canada - 266 - Collected Articles of Kishori Lal

(b) commodity inputs of final demand categories plus primary inputs of final demand categories. (a) is called Gross Domestic Product and (b) is called Expenditure on Gross Domestic Product.

Some elaboration of the final demand categories was left for development in this section.

Personal expenditure on consumer goods and services represents the purchase of commodities as well as commodity taxes applicable thereto and the wages and salaries, and supplementary income of persons employed by the personal sector. The personal sector is defined to include not only individuals and families but also private non-profit organizations such as labour unions, religious organizations, universities and private clubs. Personal expenditure on consumer goods and services is divided into column vectors according to the purpose or function of the purchase, e.g., foods, beverages and tobacco; clothing, footwear and accessories; medical care and health expenses; recreation entertainment, education and cultural services; etc., with additional commodity detail under each heading.

Gross government current expenditures on commodities and primary inputs are classified according to level of government-federal, and provincial and local. The accounting units of government, and therefore their expenditures, are grouped by main purpose, education, health, defence, and other government goods and services.

Gross fixed capital formation shows expenditures on construction commodities and on machinery and equipment commodities as well as the commodity taxes applicable and includes purchases by government and by business. Business capital outlays are further sub-divided into 40 industry groupings for construction and 39 groupings for machinery and equipment. It was not possible (due to the lack of data) to develop as many industrial sub-categories for business capital formation as the number of industries in the inputs table. There are 191 industries in the inputs table compared to only 40 industrial groupings for construction (and 39 for machinery and equipment). This compromise was quite acceptable since we were forced (due to confidentiality) to aggregate the 191 industries into 43 industrial groupings for publishing the inputs and output tables.

It was noted that where the production of a commodity exceeds its use in the accounting period, either as an intermediate input or in some final demand category, the balance shows as an addition to stocks, as raw material inventories when purchased as an intermediate input but not yet used, as goods in process or finished goods inventories, or as inventories of goods purchased for re-sale. Similarly when use exceeds production in the accounting period, the balance shows as withdrawals from stocks in the same categories. Withdrawals from stocks are a source of supply of a commodity, and are shown in a separate column.

The valuation of inventories presents special problems because firms employ a variety of accounting methods. In periods of changing prices, firms may report changes in the book values which do not reflect changes in the quantities held of a commodity. These gains or losses on inventory account are not appropriate when valuing the production or use of a commodity in the current period. Thus, where changes have occurred in the prices of the relevant commodities, the opening and closing stocks of these commodities in the period of account are revalued. The change in stocks of

Statistics Canada - 267 - Collected Articles of Kishori Lal commodities so valued is called the "value of physical change in inventories", and it is this value which is entered in the final demand categories. The difference between the value of physical change and the change in book value is called the "inventory valuation adjustment". This is an adjustment implicitly included in the item "surplus".

Exports are classified by commodity. As with other elements of commodity demand, the valuation of exports is at producer prices.

There is no precise value for imports which corresponds conceptually to the producers' values used for domestic products. International Input-Output practice suggests that imports should be valued at the border of the importing country inclusive of imports duties (i.e. including all charges involved in conveying the imports to the border plus import duties). These are measurable import values which, under a system of relatively competitive pricing, are likely to correspond to domestic producers' values. Hence, the producers' values of imports are defined to be imports c.i.f. the domestic border plus import duties. The value of import duties is shown with sign reversed, at the intersection of imports column and the commodity indirect taxes row.

Government goods and services represent payments by industries and final demand sectors for services produced within the government sector.

The Input-Output Model

The Input-Output model presented here (the same one we have used for Canada) is only one of the family of Input-Output models associated with Professor Wassily Leontief of the United States. Let us call this model Open Output Determination Model.

Let us explain the specification of the Open Output Determination Model in terms of the chart of the accounting framework of Rectangular Input-Output tables.

The specification of the model based on the information contained in this framework involves two sets of assumptions. The first has the function of allocating the production of commodities among industries. The second establishes the production functions of industries which in turn determine the requirements of industries for commodity inputs. In combination with the accounting balance between total demand (by industries and final demand categories) and domestic production, these assumptions establish the Input-Output model in which outputs are determined as a function of final demand.

The simplest assumption concerning the allocation of commodity production among industries is that industries will preserve their observed share of the market for each domestically produced commodity irrespective of the levels of commodity production. The mathematical expression of this assumption is the following matrix equation:

(1) g = Dq

Statistics Canada - 268 - Collected Articles of Kishori Lal

In this equation vector g represents the value of industry outputs and vector q the value of domestically produced commodity outputs. Matrix D is a matrix of coefficients which is calculated by dividing each element in a column of the output matrix V of the Chart by the corresponding total commodity output. It will be referred to as Domestic Market Share Matrix.

The simplest way of defining the production function of industries is to suppose that the values of the inputs of each industry are fixed proportions of the value of the total output of the industry and are thus independent of the composition of this output. This assumption, which has always been a basic assumption of Input-Output analysis, is now being referred to in the literature as the "industry technology" assumption. The mathematical expression of this assumption is the following matrix equation:

(2) Ui = Bg

In this equation matrix U is the matrix of commodity inputs of the Chart. Vector i is a column vector, equal in dimension to the number of industries, whose elements are all equal to 1. The matrix product Ui, therefore, represents vector containing the sum of the intermediate inputs of all industries classified by commodity. Matrix B is a matrix of coefficients which is obtained by dividing each element in a column of Matrix U by the corresponding total industry output. Matrix B will be referred to as the Industry Technology Matrix. Vector g represents, as before, industry outputs. The equation as a whole states that the total intermediate inputs classified by commodity can be calculated as the product of the Industry Technology Matrix and the vector of industry outputs.

The mathematical expression of the accounting balance between total supply and total disposition is as follows:

(3) q + m + a + v = Bg + e + X + R where m is a vector of the values of imports a is a vector of the values of government production v is a vector of the values of withdrawals from Physical Change of inventories X is a vector of the values of domestic exports R is a vector of the values of re-exports and e is a vector of the values of the following final demand categories: Personal Expenditures; Fixed Capital Formation; Values of Additions to physical change of inventories; Gross Government Current Expenditures.

(3a) q = Bg + e + X + R - m - a - v

Equations (1) and (3a) yield the following open output determination model.

(4) g = ( I - DB ) -1 D ( e + X + R - m - a - v)

Statistics Canada - 269 - Collected Articles of Kishori Lal

Equation (4) defines linear transformation of final demand categories into industry outputs.

Industry output can easily be transformed into commodity output by using equation (3a). The dimensions of the inverse in equation (4) are industry by industry, which are smaller in magnitude than a matrix of the dimensions commodity by commodity. It is cheaper (in terms of computer time) to run an industry by industry dimension matrix for inverse than a commodity by commodity matrix.

( I - DB ) -1 D post-multiplied by exogenously-specified final demand calculates industry outputs. This model does not account for any leakages from the domestic industries. To the extent that imports, and/or withdrawals from inventories and/or government production of goods and services share with the domestic industries in the supply of a commodity, the impact of an increase in the final demand on domestic industries will be reduced. These leakages are specified below:

(5) m = µ ( Bg + e + R ) where vector m represents imports and µ is a diagonal matrix whose elements are calculated as the ratios of imports to use, use defined as Bg + e + R. It is to be noted that this import share assumption implies that domestic exports of a commodity are supplied from domestic industries that produce the commodity. Of course, domestic exports may have imports indirectly embodied in them to the extent that producing industries import their intermediate inputs.

( 6) a = α ( Bg + e + X ) where vector a represents government production of goods and services and α is a diagonal matrix of coefficients whose elements are calculated as the ratio of government production to use, use defined as Bg + e + X

(7) v = β ( Bg + e + X ) where vector v represents withdrawals from inventories (VPC) and β is a diagonal matrix of coefficients whose elements are calculated as the ratio of withdrawals to use, use defined as Bg + e + X. It is possible to present the exogenous sources of commodities, imports, withdrawals from inventories and government sales, as additions to production (extension of the V matrix) rather than as adjustments to use (negative value in the final demand column vectors). The present formulation can be identified more easily with the traditional presentation in the Income and Expenditure Accounts, and this relationship will be developed below. In any case, for each and all commodities, total supply equals total disposition

Substitution of equations (5), (6) and (7) into (3a) yields the following:

(8) q= Bg + e + X + R - µ ( Bg + e + R ) - α ( Bg + e + X ) -β ( Bg + e + X )

(8a) q = ( I - µ - α - β) Bg + ( I - µ - α - β) e + ( I- α - β) X + ( I - µ ) R

Statistics Canada - 270 - Collected Articles of Kishori Lal

Equations (8a) and (1) yield the following open output determination model which makes an allowance for leakages out of intermediate demand as well as final demand.

(9) g= [I - D ( I - µ - α - β) B ] -1 D [ ( I - µ - α - β ) e + ( I - α - β) X + ( I - µ ) R]

This formulation which makes an allowance for leakages out of final demand not only gives a proper account of total leakage but also makes it possible for the model user to change the specifications for leakages in the exogenous demand.

The precise mathematical expression defining inverse used by us in the Rectangular framework is:

[I - D ( I - µ - α - β ) B ] -1 D

The above presentation where we have post-multiplied the inverse by the matrix D is not common. We have done it to simplify it for the user. Normally a user would like to know the impact on industrial output of a given (by the user) final demand. Final demand can only be specified, in any realistic sense, in dimensions of commodities. That specification needs the above form of inverse (which includes post-multiplication by D) otherwise the user has to pre-multiply his specifications with D. Furthermore, it may not be possible (due to confidentiality) for a statistical agency to publish D in as much detail as B. It is our experience that the confidentiality provisions start hitting the outputs matrix much earlier than the inputs matrix. For these reasons - simplification to use and avoiding confidentiality - we have decided to publish the inverse with post-multiplication by D.

In a simple language, what does an impact (or inverse) matrix contain? The inverse matrix contains the direct and indirect requirements for inputs per dollar of delivery to final demand. For example, with an increase in the final demand for automobiles, there will be a direct increase in purchases by the motor vehicle industry from industries producing steel, components, tires, etc. But the effect of the production of an automobile does not end with the purchase of steel, components, tires, etc. It begins a long chain of production since each of the purchase will require, in turn, various inputs. Two types of inputs are therefore identified: direct inputs are those purchased by the industry under consideration: indirect inputs are those purchased by all industries in which production is required in order to supply inputs to the first industry. The Impact table captures the total impact of these effects of demand throughout the economy.

Caution

Once the above model is operational, one can do all kinds of multiplier analysis such as calculating output and income multipliers, employment multipliers, etc. One can calculate the impact on imports. One can calculate the effect on employment. There are so many tricks that one can play with such a model.

This model (as all economic models) can only hope to approximate reality. All models are caricatures of reality. All models have their limitations. A user is warned to be completely familiar

Statistics Canada - 271 - Collected Articles of Kishori Lal with all the built-in assumptions and data weaknesses of an economic model before accepting the model results. Beware of "sophisticated" models based on weak databases.7

One must realize that the same database could give different results depending upon the level of aggregation the model builder has used. Aggregation is quite a serious problem but unfortunately one does not hear much of it in most economics text books. Aggregation forces a particular point of view, a particular analysis. Indeed, aggregation itself is a model.

Let me elaborate the above point using the Canadian Input-output tables. Input-output impact tables were published at two levels of aggregation for all the fourteen years 1961-74. Still another level of detail (which we do not publish so as to avoid disclosure of information on individual firms) is used internally for model building. We looked at a particular commodity and a particular industry whose definitions remained the same at the two published aggregations. The impact of this particular commodity on the particular industry was different depending upon which (level of aggregation) impact tables one looked at. This is an anomalous situation. However, this apparent anomaly can be explained by the fact that other commodities (other than the one under study) were aggregated differently in the two aggregations. But the fact remains that different aggregations give different results.

One might draw a lesson. It is necessary to avoid temptation, as much as possible, to aggregate data in the early stages of economic analysis. Aggregation is easier to do than disaggregation. The implications of this simple statement must be well understood. Micro database must be preserved as much as possible because it is directly relatable to the detail provided by the respondents. Only then, would one be able to arrange custom-made aggregations for economic analysis. We noted above in the introduction that the data retrieval system should be efficient. Without an efficient data retrieval system, one will remain a prisoner of dictated (by others) aggregations and one's horizon of economic analysis will remain narrow indeed.

Statistics Canada - 272 - Collected Articles of Kishori Lal

References

a. Excepting minor editing, this is a reproduction from the PSE (Punjab School of Economics) Economic Analyst, Amritsar, Vol. 1, June, 1980, pp. 1-23

1. Pioneering work in this area was done by Professor Stone of Cambridge University. Work in Canada, the first country to implement this scheme and further extend this into a rectangular framework was pioneered by Terry Gigantes of Statistics Canada and Professor Matuszewski of Université Laval of Québec.

2. Jack Alterman and Martin L. Marimont, "Prices and Price Analysis in the Framework of the National Accounts", The Review of Income and Wealth, series 16, No. 2, June 1970, pp. 145-146.

3. Anne Carter and Wassily W. Leontief, "Goals for the Input-Output Data System in the Seventies", the anniversary issue of Survey of Current Business, Vol. 51, No. 7, Part-II, July 1971, p. 31.

4. UN, A System of National Accounts, Series F, No. 2, Rev. 3, New York, 1968.

5. Statistics Canada, The Input-Output Structure of the Canadian Economy, 1961, Vol. 1, Ottawa, 1969, pp. 167-261.

6. Statistics Canada, The Input-Output Structure of the Canadian Economy, 1961-74, Ottawa, February 1979 and The Input-Output Structure of the Canadian Economy in Constant Prices, 1961-74, Ottawa, May 1979.

7. O. Morgenstern, On the Accuracy of Economic Observations, Princeton, 1963.

Statistics Canada - 273 - Collected Articles of Kishori Lal

Compilation of Input-Output Tables: Canadaa

by Kishori Lal Statistics Canada

Abstract: This paper provides a description of the annual Input-Output Tables for Canada. It describes the accounting framework and notes its close affinity to the one described in the United Nations report, A System of National Accounts. It demonstrates the ready derivation of GDP and Expenditure on GDP, both in current and constant prices, from the Input-Output Accounts as well as their relationship to the other subsystems of the Canadian System of National Accounts, particularly the Income and Expenditure Accounts and Real Domestic Product by Industry. Compatibility of basic accounting records of the transactors with the rectangular (commodity-industry) format of the Canadian tables is described. The need to have a consistent commodity classification and to develop a consistent valuation of all transactors in the economy is emphasized. The particular formulation of the Input-Output Impact tables is noted. The problem of deflating trade margins and the resolution of this problem is described. A strong plea is made for the economics profession to pay more attention to the problem of aggregation; all economic analysis is approached with blinkers but the aggregation problem is not even recognized as a blind spot in most analyses.

Introduction

The Input-Output Division of Statistics Canada prepares and publishes annual Input-Output Tables for Canada in both current and constant prices. The latest tables refer to the year 1977 and with this there is a completely consistent historical series of annual tables from 1961.

The accounting framework of the Canadian tables given in Chart 1 bears a close relationship to the one described in the United Nations report, A System of National Accounts.1 The inputs and outputs of industries are presented in separate tables; both inputs and outputs are classified by commodity. Commodities are clearly distinguished from industries, the number of commodities exceeding the number of industries. The Canadian Input-Output Tables are rectangular rather than square. The commodity industry format has important advantages over the traditional inter-industry square format: (i) it admits as much detail as is available in the basic economic records; (ii) the meaning of each entry is straightforward because observed transactions are not combined with fictitious transfers; and (iii) it provides a statistical audit of the consistency, integrity and comprehensiveness of economic statistics. The last advantage assumes particular importance when input-output statistics are developed in a central statistical organization like Statistics Canada.

Canada is one of the pioneering countries in both the theoretical and empirical development of rectangular input-output tables, a development pursued by practitioners in Statistics Canada and in other institutes. Preliminary Input-Output Tables for the year 1961 were published by Statistics Canada in 1969 in two volumes: Input-Output Structure of the Canadian Economy, 1961; Vol. I (catalogue 15-501), Vol. II (catalogue 15-502). In these volumes a comprehensive description of the

Statistics Canada - 275 - Collected Articles of Kishori Lal detailed accounting framework, analytic uses, the published tables, classification systems and selected definitions and an extensive section on the mathematical treatment of analytic uses were given. This document gives the most complete description of the development of the accounting framework and the classification systems that underlie the current Canadian tables, although subsequent work has led to refinements in the classification systems, particularly in the articulation of the categories of final demand, in comparison to the preliminary 1961 tables. Further evidence of the contemporary Canadian work on the development of rectangular input-output tables, applied in a provincial context, can be found in Rapport intérimaire sur le système de comptabilité économique du Québec, le système et son fonctionnement, Bureau de la statistique du Québec et Laboratoire d'économétrie, Université Laval: juillet 1967. Another illustration of the application of the Canadian input-output accounting framework in a provincial context can be found in Input-Output Study of the Atlantic Provinces, 1965, by Kari Levitt; Vol. I, Social Accounting Matrix and Models (catalogue 15-503E), Vol. II Structural Analysis and Data Sources (catalogue 15-504E), published by Statistics Canada.

Current price input-output tables: Three data tables are prepared and presented for each year: (1) Make Matrix, (2) Use Matrix, and (3) Final Demand Matrix. At the most detailed level, the system is balanced with 191 industries, 595 commodities and 136 categories of final demand. In addition there are seven primary inputs: wages and salaries, supplementary labour income, net income of unincorporated business, other operating surplus, commodity indirect taxes, other indirect taxes, and subsidies. At this level of detail, some of the entries in the various matrices are confidential under the provisions of the Statistics Act. Though all the tables are prepared at the L level of aggregation, it is only at aggregation level S and M that we publish these tables. The dimensions of these aggregations are given below:

At aggregation S, there are 16 industries, 14 categories of final demand and 49 commodities and primary inputs. At aggregation M, there are 43 industries, 29 categories of final demand and 100 commodities and primary inputs. At aggregation L, there are 191 industries, 136 categories of final demand and 602 commodities and primary inputs.

The three data tables are published both at aggregations S and M. As well, two impact (Inverse) tables - one at each aggregation S and M (see discussion of the definition of impact tables in a later section) - and a table on commodity margins at aggregation M are published annually. Thus, in total, nine table are published for each year.2 In addition to these public tables, we provide at request, additional detail or special aggregations, subject to confidentiality, of the input-output data at the L level of aggregation.

Statistics Canada - 276 - Collected Articles of Kishori Lal

Chart 1 The Accounting Framework of Rectangular Input-Output Tables

Final Demand Commodities Industries Categories Total

Commodities U F Q

Industries V G

Primary Inputs Y1 YF N

Statistics Canada - 277 - Collected Articles of Kishori Lal

unincorporated business and other operating surplus. Other operating surplus in the usual Income and Expenditure Accounts includes profits and other investment income, inventory valuation adjustments, capital consumption allowances and miscellaneous valuation adjustments. The matrix dimensions are 7 primary inputs by 191 industries. YF : is a matrix of the values of the primary inputs associated with final demand categories. These consist of indirect taxes, wages and salaries, supplementary labour income, and depreciation which is part of surplus. Wages and salaries and supplementary labour income are paid by the government and personal sectors. The estimate of surplus (depreciation) relates to the government sector and non-profit institutions in the personal sector. The matrix dimensions are 7 primary inputs by 136 categories of final demand.

Constant price input-output tables: As in current prices, the balancing of the constant price tables and the complete deflation process are carried out for each year for 191 industries, 595 commodities and three primary inputs (commodity indirect taxes, other indirect taxes, and subsidies) and 136 categories of final demand. Gross Domestic Product at factor cost is calculated residually through the use of double deflation.

Three data tables are prepared and presented for each year: (1) Make Matrix, (2) Use Matrix (the matrix of the values of intermediate commodity inputs as well as primary inputs), and (3) Final Demand Matrix. Though all of the tables are prepared at the L level of aggregation, it is only at the aggregation levels S and M that tables are published.3 The dimensions of these aggregations are identical to the ones in the current price tables except that the factor inputs are combined into Gross Domestic Product at factor cost. Impact tables are not presented in constant prices though we would do so at request.

The Input-Output Accounts and Their Relationship to the System of National Accounts

Input-Output Tables form part of the broader Canadian System of National Accounts which includes (in addition to Input-Output Tables) the Income and Expenditure Accounts, Real Domestic Product by Industry, Productivity Studies, the Financial Flows Accounts, the Balance of International Payments, and Balance sheets showing the assets and liabilities of the economy. The System of National Accounts was originally developed to satisfy the need for consistent and comprehensive measures of economic activity. As demands for data for economic analysis have grown over the years, the conceptual framework has been extended and refined; at the same time flows of data have been established to fill in the System through the exploitation of existing sources of information, the development of new sources, and the design of new estimating techniques.

The best known Accounts of the System are the Income and Expenditure Accounts which were designed, in the main, to provide current and comprehensive, though relatively aggregative, measures of the output of the economy in two ways: (1) as the value of the "final" expenditure on goods and services, and (2) as the income accruing to (or costs of) primary factors of production engaged in the production process plus certain non-factor costs such as indirect taxes.

Statistics Canada - 278 - Collected Articles of Kishori Lal

The items of final expenditure are identical in both the Income and Expenditure and the Input-Output subsystems. On the Gross Domestic Product side, the Input-Output breakdown of primary inputs is less detailed. The item operating surplus in the Input-Output Accounts encompasses profits and other investment income, capital consumption allowances, inventory valuation adjustments, etc.

The Input-Output Tables, and the estimates of Gross Domestic Product derived from them, use somewhat different statistical sources than the Income and Expenditure Accounts. Notwithstanding the different approaches used, the estimates of Gross Domestic Product for the economy as derived from the Input-Output Accounts very closely approximate those in the Income and Expenditure Accounts.

The notable difference in the two subsystems is in the estimation of operating surplus. In the Input-Output Accounts, the data are derived from establishment-based surveys, such as annual Census of Manufactures, and surplus is essentially calculated in the context of a balance between inputs and outputs. In the estimate of the components of surplus for Income and Expenditure Accounts, the data sources are surveys of companies and other administrative records. The use of different statistical reporting units yields differences in industrial allocations.

On the Expenditure side, there are some minor differences in the estimates of current government expenditure, exports and imports, inventories and personal expenditure. The difference in the estimates of current government expenditure and exports and imports are definitional. One example of a definitional difference is the treatment of research grants for defence which are treated as transfer payments in the Input-Output Accounts but as purchases of government services in the Income and Expenditure Accounts. In imports and exports, the transfers of funds by trade unions are treated as transfers in Input-Output Accounts and as sales/purchases of goods and services in Income and Expenditure Accounts.

The valuation of physical change of inventories presents difficulty in both the Input-Output and the Income and Expenditure Accounts. There is a lack of data on the valuation practices as well as on the commodity content of inventories. The two subsystems use different techniques for measuring the value of the physical change of inventories (VPC). For example, in the Income and Expenditure subsystem, the value of the change of inventories in 1977 is equal to closing inventories in 1977 less closing inventories in 1976. In the Input-Output subsystem, VPC of inventories equals closing inventories in 1977 less opening inventories in 1977. The closing inventories of 1976 may differ from the opening inventories of 1977 because of: classification changes; changes in the basis of valuation; and births and deaths of establishments.

Consumer expenditure makes up about three-fifths of the Gross Domestic Product at market prices, yet this is one sector not supported directly with regular surveys. In the process of commodity balancing (see discussion of this important tool in the next section) it is necessary to have estimates of the categories of consumer expenditure and their commodity content. Thus considerable care, using a range of data sources, is taken in estimating the various categories of personal expenditure and the commodity composition of each category.

Statistics Canada - 279 - Collected Articles of Kishori Lal

Occasional Family Expenditure Surveys provide information on the pattern of consumer expenditure. In addition, the Retail Trade Surveys when used with information on class of customer and commodity content of sales by type of store provide estimates of consumer expenditure. Balanced tables for preceding years do provide a reference point, and retail sales by type of store can be short run indicators of the consumption of related groups of commodities, although experience suggests that this approach be used with caution. In many cases, estimates can be verified against the trend in real consumption per person. Additional information on consumer expenditure is available in the form of data on direct selling by manufacturers, a detailed breakout of the disposition of energy from the Quarterly Report on Energy Supply and Demand in Canada, data from trade associations, and of course from the operating statements of some of the private non-profit organizations which form part of the Personal Sector. The outputs of certain commodities like imputed and paid rent and domestic services are allocated directly to consumers.

The Income and Expenditure Accounts show a separate entry called "Residual error of estimate," as a measure of the statistical discrepancy between the "income side" and the "expenditure side" of the accounts. In the Input-Output Tables, where each Industry and each Commodity Account must balance, there is no "residual error of estimate." The supply side is always equal to the disposition side for any commodity and all commodities. Deducting from the total supply (or disposition) a common intermediate expenditure total gives GDP or Expenditure on GDP. Thus GDP cannot be different from Expenditure on GDP.

The relationship between the constant price Input-Output tables and its counterparts in the other subsystems of the Canadian System of National Accounts is articulated in a later section on Constant Price Input-Output Tables.

Commodity Balancing in Input-Output Tables

The Introduction described the general accounting framework of the Canadian Input-Output Tables and noted that a feature of the rectangular framework was its compatibility with basic accounting records. However, considerable work is still required to transform data as reported by the economic transactors in the economy into the comprehensive and consistent economic accounting framework that is characteristic of Input-Output Tables, and the major elements of this process are outlined in the following paragraphs. Input-Output Tables relate the supply of each commodity to its disposition and this is done by constructing a commodity balance for each commodity. Producers report production of a commodity at producers' values while those acquiring the same commodity, be they industries purchasing inputs into the production process or final demand transactors, report the cost as incurred by them, at purchasers' values. A commodity balance necessitates a uniform valuation for a commodity for all transactors, and effecting such a system of uniform valuation requires the identification and measurement of the margins (trade, transport, tax, etc.) that make up the difference between the value that the purchaser pays and what the producer receives for a commodity.

There are also major areas - for example, the construction industry, government expenditures and capital formation - where the basic accounting records do not furnish the commodity information

Statistics Canada - 280 - Collected Articles of Kishori Lal necessary to construct the I-O tables. The approaches taken in the face of these problems are described in subsequent sections.

The fundamental element in statistical compilation is classification. Input-Output Tables classify transactions both by industry and by commodity. The industrial classification of transactors (the Standard Industrial Classification) has been in place for some time and presents no major problems. However, an accounting of economic activity requires a measure of what is done as well as who does it. A commodity balance is such a complete accounting, and as commodity balances are compiled only in the context of the Input-Output Tables, it is not surprising that serious problems are encountered in imposing this additional demand on the existing database.

There are three logically separable phases to constructing commodity balances. The obvious first requirement is a system of classification for commodities (goods and services) that can be consistently applied to all branches of economic activity. This requires that a commodity be coded consistently whether produced by a manufacturing industry, imported or exported, transported by boat, train or aeroplane, or purchased by some industry or in some final demand category. The classification system actually encountered in each of these spheres of activity tends to be peculiar to it, reflecting special interests and historical development. It is not surprising to find that export commodity classes reflect the importance of goods exported, and the same is true of import commodity classes, etc. But commodity balances require consistent commodity classes applied to all transactions, and in sufficient detail to be analytically meaningful. Thus the development and continued application of a consistent commodity classification system is a major undertaking by the input-output analyst.

The second requirement is to fill major holes in the basic databases. In spite of the vast array of statistics collected, there are significant and important areas affording very little direct knowledge, and, even that, often at very infrequent intervals. The database for manufacturing is more highly developed than for most other industries; yet the annual census provides no data on purchased services, and the commodity composition of inventories is unknown. Some industries are surveyed only at decennial census intervals, and then little detail on operating structures is obtained. In such case, the required series are estimated using related indicators, ad hoc studies and occasional surveys, etc. For example, taxation statistics for corporate and unincorporated businesses provide evidence on the gross output of some industries, and implicit input structures in other cases, even though the classification of multi-establishment companies is not consistent with the establishment-based concept of the Input-Output Tables. The point to be noted here is that various sources, some having direct reference but others providing only inferences and indicators, must be used in compiling the Input-Output Tables.

The diverse classification systems peculiar to the different spheres of economic activity are confronted and moulded into a common system of adequate consistency and the resulting data are then analysed. Varied and unpredictable differences immediately become apparent. This brings us to the final phase in commodity balancing. Supply must equal disposition for each of the 595 commodities. The under - or over-allocation of commodities must be examined and eliminated. The discrepancy may be due to several reasons - production may be unreported or misclassified; imports

Statistics Canada - 281 - Collected Articles of Kishori Lal and exports may be improperly valued, as well as misclassified; there may be timing inconsistencies, etc. There are no ready-made statistical approaches to solving such imbalances. The only approach is labourious investigation; one has to go back to the basic records to locate the sources of such imbalances. It is our experience that the commodity balance approach, with its detailed accounting of output by industries and of use by industries and final demand transactors, uncovers major problems in both statistics reported and classification.

Valuation

The Input-Output Tables are initially constructed in current prices with the entries valued at purchasers' prices, with subsequent calculation of the tables at producers' prices. Purchasers' prices represent what the purchaser pays while producers' prices are what the producer receives; the difference is the trade, tax and transportation margins. In producers' prices a commodity has the same valuation base throughout the system.

Note that what we call "Producers' Prices" are, in the UN SNA terminology, "approximate basic values". Our valuation in producer prices is readily understood by the industry experts and the data are easily comparable with other published series on industry statistics. We believe that it is quite important to keep our concepts as close as possible with those used by industry analysts and other users who are not familiar with the nuances of SNA terminology.

Our published data tables are always in producers' prices. The impact tables are manipulated from the producer price files and so are the constant price Input-Output Tables. But purchaser price tables are available to any user on request.

Margins

There are seven margins distinguished in the Input-Output Tables: retail margins, wholesale margins, tax margins, transport margins, gas margins, storage margins and pipeline margins. In the majority of cases, particularly trade and transport margins, data on the appropriate margin to be attached to each commodity are not directly available and must be estimated based on partial evidence and professional judgement.

Trade margins: Data on margins by commodities are generally not available. For wholesale and retail trade margins, gross trading margins by type of store are first calculated. Occasional commodity surveys provide a basis for allocating the type of store margins to groups of commodities. These estimates by type of store and by commodity group are fitted into a statistical framework and balanced. To complete the first stage the margins by commodity groups are expanded into the full detailed input-output commodity classes using proportionality assumptions. This exercise provides the first approximation of the margins on each commodity. In the next stage, i.e. at commodity balancing, margins may be modified in the light of the commodity balances and price spread studies. The industrial distribution of these margins is based on the Census data on class of customer and other relevant information and estimates.

Statistics Canada - 282 - Collected Articles of Kishori Lal

Transport margins: Transport margins are developed by mode of transportation (such as air, rail, water, services incidental to water, truck and other services incidental to transportation) giving the cost of transportation by commodity going from producer to purchaser. Most of the transport margins are generated by railways and trucks. Railways can distinguish freight revenue by about 300 commodity groups. This statistical and commodity coverage is not quite complete but useful to get a first approximation of the values of rail transport margins. The annual Truck Commodity Origin and Destination Survey gives revenue and tonnage by approximately 450 commodity groups. Certain ad hoc studies and surveys, with a big dose of professional judgement by industry experts, complete our first estimates for transport margins.

Gas margins are straightforward as all the gas margins apply to commodity Natural Gas. Pipeline distribution margins are directly allocatable to natural gas and crude oil. Storage margins apply to a very few primary outputs such as wheat and barley; most of the storage charges are not on commodity margins but charges for storing household effects.

Tax margins: Examples of tax margins are federal excise taxes, import duties, gasoline taxes, liquor gallonage taxes, profits of provincial government- run liquor commissions. Provincial sales taxes are allocated directly to categories of final demand as well as such intermediate industries as construction. All other tax margins are distributed by commodity. It is a painstaking exercise to go through the myriad tax levies and tax exemptions. This tax margin allocation is further subjected to an informal consultation with tax experts and other knowledgeable persons.

Use of Imports Versus Domestic Production

The value of inputs into the Use Matrix as well as the Final Demand Matrix includes (where applicable) imports as well as commodities produced by the domestic economy. The basic economic records such as the Census of Manufactures make no distinction between imports and domestic production for raw materials purchased for further processing. Similarly the Final Demand transactors such as the Household sector, Government sector, Business sector (for capital formation) do not record the purchase of commodities in terms of their origin - imports versus domestic production.

We note that some countries produce an Intermediate Transactions Matrix broken down into Domestic production versus Imports. They must have formidable basic records to produce such tables. Or, maybe, they are far more imaginative than we are. Recently OECD asked us to fill a standardized questionnaire wherein the origin of inputs between imports versus domestic production is maintained. We could not fill this part of the questionnaire.

In our tables, we show Imports as a negative column in the Final Demand Matrix. One could though assume that imports are a fixed share of total supply for each of the individual users, and prorate imports over all users. This way one could produce a matrix of Inputs for imports and domestic production separately. But this is not a statistical table, only a display of assumptions.

Statistics Canada - 283 - Collected Articles of Kishori Lal

There are a certain very limited number of commodities that are not produced at all in Canada; these include tropical fruits, coffee, tea, rubber, etc. Use of these commodities is shown in the Canadian I-O tables as rows just like use of other commodities.

We have assumed that the valuation of imports which is equivalent to the "producer price" of domestic production is imports c.i.f. to the border plus import duties.

Transfer of Secondary Outputs

The Canadian I-O tables are not inter-industry but have commodity- industry dimensions. The number of commodities is larger than the number of industries. We do not purify industries by transferring their secondary or subsidiary output except for own account construction (see below for construction). Industries by definition will produce their own principal products, but if there is a subsidiary output, it is left in the industry. We think that this reflects the transactions better and then we can easily relate our data to other statistics published in an industry framework. It is here that the role of Input-Output Tables as a statistical audit for other economic statistics becomes relevant and important.

Construction

Construction is defined in the Canadian I-O tables as work put-in- place both by establishments classified to the construction industry and by the own-account construction activities of other industries. Thus the total activity of construction is brought together in the I-O tables and transfers are accordingly made from the various industries to the Construction activity vectors of the I-O tables. We have broad "information" on the labour component and the material component of construction activity broken by own-account and contract but we have no regular surveys giving us input detail by commodity. It is somewhat easier to imagine a particular commodity going into a typical construction structure but it will be most difficult to estimate its allocation by own-account versus contract. The sheer lack of input detail forced us into putting the entire construction activity together. As well, there are some details available from ad hoc studies giving input detail of structures. These two considerations forced us to put the entire activity together.

Government Sector

The accounts published by governments usually follow a functional classification, with only minor consideration given to an economic object classification. The first major problem is to convert the functional detail into an economic object classification. The second problem is that governments are not asked to fill a structured questionnaire as are the industrial respondents. Governments keep their records in a way which is satisfactory for legislation needs, but not necessarily for statistical needs. Records of the various levels of government, federal, provincial and municipal, are unstructured and then the commodity detail is lacking. The available data are for very broad categories. The input-output analyst must estimate the detail by reference to detail buried behind the published records, and diligently code myriad information spread over tabular and textual forms. The commodity detail information is rather weak in the basic public accounts of the Government sector.

Statistics Canada - 284 - Collected Articles of Kishori Lal

Capital Formation

Statistics Canada collects information by industry on total capital formation in construction and in machinery and equipment. Hardly any information is collected on the types of commodity that an industry purchases on capital account. One has to study each commodity which could possibly be capitalized and then decide its allocation to a particular industry. Surprisingly a large number of commodities can quite easily be allocated. Some types of machinery are industry - specific, hence easily codeable to a given industry. Once one has exhausted the commodity supply side, then one checks the allocation totals to see if they are different from the survey ones. Going back and forth a couple of times will in most cases fix rather well the allocation by detail and total.

Impact Table

The definition of the Impact tables as published in the Input-Output Structure of the Canadian Economy involves two sets of assumptions. The first has the function of allocating the production of commodities among industries. The second establishes the production functions of industries which in turn determine the requirements of industries for commodity inputs.

(1) g = Dq

In this equation vector g represents the values of industry outputs and vector q the values of domestically produced commodity outputs. Matrix D is a matrix of coefficients which is calculated by dividing each element in a column of the output matrix V of the Chart by the corresponding total commodity output. It will be referred to as Domestic Market Share Matrix.

The simplest way of defining the production functions of industries is to suppose that the values of the inputs of each industry are fixed proportions of the value of the total output of the industry and are thus independent of the composition of this output. This assumption, which has always been a basic assumption of Input-Output analysis, is now being referred to in the literature as the "industry technology" assumption. The mathematical expression of this assumption is the following matrix equation:

(2) Ui = Bg

Statistics Canada - 285 - Collected Articles of Kishori Lal

will be referred to as the Industry Technology Matrix. Vector g represents, as before, industry outputs.

The mathematical expression of the accounting balance between total supply and total disposition is as follows:

(3) q + m + a + v = Bg + e + X + R

where m is a vector of the values of imports a is a vector of the values of government production v is a vector of the values of withdrawals from VPC X is a vector of the values of domestic exports R is a vector of the values of re-exports and e is a vector of the values of the following final demand categories: Personal Expenditures; Fixed Capital Formation; Values of Additions to Physical Change in Inventories; Gross Government Current Expenditures

(3a) q = Bg + e + X + R - m - a - v

Equations (1) and (3a) yield the following open output determination model.

(4) g = ( I - DB ) -1 D ( e + X + R - m - a - v)

Equation (4) defines linear transformation of final demand categories into industry outputs.

Industry output can easily be transformed into commodity output by using equation (3a).

(5) m = µ ( Bg + e + R )

where vector m represents imports and µ is a diagonal matrix whose elements are calculated as the ratios of imports to use, use defined as Bg + e + R. It is to be noted that this import share assumption implies that domestic exports of a commodity are supplied from domestic industries that produce the commodity. Of course, domestic exports may have imports indirectly embodied in them to the extent that producing industries import their intermediate inputs.

( 6) a = α ( Bg + e + X )

Statistics Canada - 286 - Collected Articles of Kishori Lal where vector a represents government production of goods and services and α is a diagonal matrix of coefficients whose elements are calculated as the ratio of government production to use, use defined as Bg + e + X

Substitution of equations (5), (6) and (7) into (3a) yields the following:

(8) q= Bg + e + X + R - µ ( Bg + e + R ) - α ( Bg + e + X ) -β ( Bg + e + X ) (8a) q = ( I - µ - α - β) Bg + ( I - µ - α - β) e + ( I- α - β) X + ( I - µ ) R

Equations (8a) and (1) yield the following open output determination model which makes an allowance for leakages out of intermediate demand as well as final demand.

(9) g= [I - D ( I - µ - α - β) B ] -1 D [ ( I - µ - α - β ) e + ( I - α - β) X + ( I - µ ) R]

The precise mathematical expression defining inverse used by us in the Rectangular framework is:

[I - D ( I - µ - α - β ) B ] -1 D

Aggregation

Another problem in economic accounting which is brought into sharp focus in the construction of Input-Output Tables is aggregation, a topic which has not received much attention in the literature.

As noted above, the Canadian Input-Output Tables are completely balanced at the L level of aggregation but are published at aggregation level S and aggregation level M. The levels of aggregation are hierarchical. Level S can be derived from level M and level M can be derived from level L. The hierarchical basis of aggregation is true for industries, commodities and categories of final demand.

Statistics Canada - 287 - Collected Articles of Kishori Lal

Gross output of industries and gross intermediate inputs of industries do not change with any aggregation. We do not eliminate intra-industry inputs at any level of aggregation.

But one must realize that the same database could give different results depending upon the level of aggregation the model builder has used. Aggregation is quite a serious problem but unfortunately one does not hear much of it in most economics textbooks. Aggregation forces a particular point of view, a particular analysis. Indeed, aggregation itself is a model.

Let me elaborate the above point using the Canadian input-output tables. Input-Output impact tables are published at two levels of aggregation for all the seventeen years 1961-1977. Still another level of detail (which we do not publish so as to avoid disclosure of information on individual firms) is used internally for model building. We looked at a particular commodity and a particular industry whose definitions remained the same at the two published aggregations. The impact of this particular commodity on the particular industry was different depending upon which (level of aggregation) impact tables one looked at. This is a seemingly anomalous situation. However, this apparent anomaly can be explained by the fact that other commodities (other than the one under study) were aggregated differently in the two aggregations. But the fact remains that different aggregations give different results.

One might draw a lesson. One must avoid the temptation, as much as possible, to aggregate data in the early stages of economic analysis. Aggregation is easier to do than disaggregation. The implications of this simple statement must be well understood. We strongly recommend to pay attention to keeping micro data sets. We say "keeping" micro data sets because data sets from the respondents are made available in this form. With them one would be able to arrange custom-made aggregations for economic analysis. It goes without saying that the data retrieval system should be efficient. Without an efficient data retrieval system, one will remain a prisoner of dictated (by others) aggregations and one's horizon of economic analysis will remain narrow indeed.

Constant Price Input-Output Tables

With the decision to establish a program of annual input-output tables in current prices, tables at constant prices were a logical corollary. In Canada one of the roles of input-output tables is to serve a supportive and integrating role for other parts of the system of national accounts and input-output tables at constant prices are necessary to completely fulfill this function. While, on the face of it, the construction of input-output tables in constant prices simply involves restating the commodity values from the current price tables in base year values, in fact a number of problems were encountered, and this section describes some of these problems, how they were approached and how they were resolved. It is important to note that there are a number of problems to which we do not yet have definitive solutions and these areas are continuing to be explored and developed as resources permit.

At the time the project was initially started the literature available to guide us was relatively sparse. More recent references, particularly the United Nations Manual on National Accounts at Constant Prices4 have reinforced most elements of the approach we adopted.

Statistics Canada - 288 - Collected Articles of Kishori Lal

Before discussing some of the problems and some of the features of the Canadian I-O tables a brief general description of the approach to deflation is necessary. The Canadian I-O tables in current prices are published at producers' values, but are in fact compiled as noted above at both producers' values and at purchasers' values. It might be possible to carry out deflation on either valuation basis, but deflation of producers' values offers a number of advantages, and this is in fact done. The Canadian Manufacturing Industry Selling Price Indexes are measured at the factory gate before inclusion of commodity indirect taxes or outward transportation, and these price indexes serve as the basis for calculation of the deflators for a majority of the commodities.

In valuation at producers' prices all commodity transactions are on a uniform basis of valuation thus permitting the use of a single price deflator across a commodity vector. In tables at purchasers' values, each transactions is potentially a unique amalgam of producers' value and various margins, and thus presents almost impossible demands on the construction of price deflators. In tables at any basis of valuation the problem of deflating the tax, trade and transport margins on commodities must be faced.

In the following paragraphs the development of the constant price input-output tables is examined from two points of view. The paragraphs immediately below note some of the aspects of relating the constant price input-output tables to other accounts in the Canadian System of National Accounts. Subsequent paragraphs discuss some of the problems peculiar to input-output tables, problems not encountered in constructing the tables at current prices. Of course all aspects of the tables are interrelated and these distinctions are made for the purpose of exposition.

The traditional measures of production published in Statistics Canada to which the input-output tables can be related are Gross National Product, Gross National Expenditure (the National Income and Expenditure Accounts) and Real Domestic Product by Industry. Only Gross National Expenditure (GNE) and Real Domestic Product by Industry (RDP) are published in constant prices.

The input-output tables encompass both GDP by industry and Expenditures on GDP and thus provide the only mechanism whereby the RDP measures can be reconciled with constant price GNE. While the theoretical concept of the RDP measures is similar to constant price GDP from the input-output tables there are of course differences in methodology. The input-output tables employ double deflation for all industries but the RDP indexes are constructed from single indicators or double deflation taking no express account of some service inputs. Where double deflation is employed in RDP, inputs are recorded at purchasers' values and the outputs at producers' values. Thus complete reconciliation of the RDP indexes with the input-output tables cannot be definitive but problem areas can be identified.

While the preparation of both annual constant price input-output tables and annual RDP indexes appears to have some elements of duplication, the RDP indexes do present a long historical series (from 1935 for total RDP, from 1919 for the Index of Industrial Production) or real output by industry measured on a consistent basis. The annual RDP indexes also serve as the basis for monthly estimates of real output by industry in the current period

Statistics Canada - 289 - Collected Articles of Kishori Lal

We have also observed that there can be a significant difference between the growth in production at factor cost and at market prices - such as when imports (which are taxed, but do not form part of domestic production) account for a changed share of domestic demand, or when the commodity composition of domestic demand shifts to more (or less) highly taxed commodities.

As was noted above, the estimates of Expenditure on GDP from the input-output tables can corroborate or refine the final demand categories in the GNE accounts. In constant prices, there is another dimension to the relationship between the final demand estimates from the input-output tables and GNE from the Income and Expenditure Accounts. Taking consumer expenditure as a particular example, final demand at constant prices from the input-output tables is calculated as the sum of commodities at producers' values, deflated with appropriate price indexes, and the various margins at constant prices.

The GNE account is estimated at purchasers' values, and deflated with purchasers' price indexes, principally elements of the Consumer Price Index (CPI). Thus the input-output tables present a powerful tool for reconciling different price indexes at different levels of valuation. Our experience with this exercise has been surprisingly good. The first add-ups revealed a few problems but analysis of these cases provided straightforward solutions in most instances, leaving only a few areas such as clothing, which is recognized as notoriously difficult to price, for further development.

There are also some problems in deflation which are largely peculiar to the internal characteristics of the input-output tables. The principal ones are noted below.

It was perhaps not surprising to discover that the commodity classification system developed for the current price input-output tables did not always result in commodity classes that were sufficiently homogeneous for deflation purposes - either from the point of view of the diversity of varieties within classes or of market differentiation. In the next review of the commodity classes attention will be directed to the needs for deflation, but considerable progress has been made. In the current tables imports of a commodity are routed with the domestic production of the same commodity. To the extent that import price indexes are available corresponding to the input-output commodity classes, imports in the input-output tables are deflated with import prices. Domestic output and exports are deflated with domestic prices, and all categories of domestic use, intermediate and final, are deflated with a weighted import and domestic price index. Similarly where exports attract a different price from the domestic market, exports are deflated with a special price index. Within the domestic market some commodities exhibit price discrimination, and the flow of these commodities to each market is deflated by the price appropriate to that market. For the commodity pipeline transportation it was observed that prices of pipelines carrying natural gas behaved differently from those of pipelines carrying petroleum products. This became two commodities for deflation. At present the price deflator for transportation margins is a weighted average for all modes of transport and for all commodities, but work is under way on developing price indexes for transport margins classified both by commodity and by mode of transport.

Probably the major problem in constructing constant price input-output tables is the deflation of trade margins; the measurement of the output (constant price margins) of trade industries is always

Statistics Canada - 290 - Collected Articles of Kishori Lal difficult whether within or outside the input-output framework. While a number of approaches were considered, two have been tried, with results that have been acceptable but not fully satisfactory. The suggestion of calculating margins in constant prices as the product of the margin rate in the base year and the volume of the commodity being traded was rejected on conceptual grounds. Constant margin rates imply a constant proportionality between the volume of a commodity being sold and the quantity of trade margin or service attached to that good. Observation of the market place in action suggests this is not so. Marketing techniques are constantly changing and the quantity of distribution service is generally becoming less. Filling stations require drivers to pump their own gasoline; free delivery becomes charged for; packaging and display are "economized" to reduce costs. The first approach adopted was to use the margin rates for each year from the current price tables. Thus the margin rates at current and at constant prices were the same, with the changing rates reflecting changes in the quantum of services. This approach worked reasonably well and is incorporated into our published tables for the year 1961-1971. It may also be noted that this was a period of relatively stable prices. This technique was originally carried into the 1970's with less agreeable results.

The use of current year trade margin rates exacerbated the effects of rapidly changing prices combined with a great diversity in mark-up practices. The next step was to emulate Goldilocks and the Three Bears in search of some formula that seemed "just right." The closest to Baby Bear's porridge proved to be using a margin rate calculated as the average of the given year's rates and the base year's rates from the current price tables. The two principal criteria by which different approaches were evaluated were reasonable productivity measures for the trade industries and an implicit price at purchasers' values that most closely approximated observed purchasers' prices. As mentioned above, this is one of the areas we plan to continue to investigate.

Indirect taxes and subsidies present few difficulties in constructing constant price input-output tables at both factor cost and market prices. Commodity indirect taxes at base year prices are calculated as the product of the base year tax rate and the quantum of the commodity taxed. Some slight imprecision can arise because it is not known just what margins should be added to the producers' value to calculate the tax. For highly taxed commodities such as alcoholic beverages, where in Canada at least, the tax is a large multiple of the producers' value of the commodity, any (say) timing error in the producer's value of the commodity as recorded is greatly inflated in the tax calculation. In the case of subsidies, in the majority of cases for the significant subsidies it has been possible to identify the commodity being subsidized; the procedure then becomes analogous to commodity indirect taxes. In the more trivial cases the subsidy is calculated as a fixed (base year) proportion of the constant price output of the industry receiving the subsidy.

One other problem, mentioned last because it arises less frequently, is the problem of rebasing constant price input-output tables. Periodic rebasing is necessary in constructing longer time series and the general Canadian practice has been to rebase at about ten-year intervals. These time segments are linked in a backward chained Laspeyres measure, with the chaining periods being the ten year intervals. Chained indexes are unfortunately not additive, and this non-additivity gives rise to adjusting entries which are implicit when the constant price series is expressed in index number form (the indexes of RDP) or explicit when given in actual currency units (as the constant dollar GNE). The matrix presentation of the input-output tables does not lend itself to the inclusion of

Statistics Canada - 291 - Collected Articles of Kishori Lal adjusting entries, and thus we cannot present chained rebased input-output tables in the same mode as the other constant price production accounts in the Canadian System of National Accounts. At the present time Canada has annual tables for seventeen years 1961-77, with the whole period based on 1971 prices; in addition the tables for years 1961-71 are available at constant 1961 prices. But the use of one base year becomes less defensible as the time period lengthens.

Conclusions

The Input-Output Accounting framework, when set in commodity-industry dimension, provides a powerful statistical audit mechanism for improving consistency, integrity and comprehensiveness of economic statistics. This aspect of the compilation of input-output tables has, in our mind, not been delineated properly in the literature. Compilation of input-output tables forces a re-examination of various and unintegrated commodity classification schemes prevalent in any central statistical organization; it brings to the forefront, as a matter of routine, any statistical holes in economic statistics; it brings to light the various and unintegrated valuation practices in economic series; it provides a meaningful, comprehensive and integrated database to reconcile the expenditure side of the national accounts with the income side in both current and constant prices.

The compilation of input-output tables serves several roles but the one which has so far dominated, at the cost of others, is the role in providing the framework for impact analysis with given shocks in the final demand. In Canada, our program, as an illustration, serves at least the following roles:

(i) Analysis of the structure and studies of the impact (direct and indirect effects) of exogenous factors on the economy in contrast to the macro economic analysis possible from the database of the other sub-systems of the Canadian System of National Accounts (CSNA). (ii) Market analysis in the context of a detailed, comprehensive and consistent database on the supply and demand of commodities. (iii) Basis for new frontiers in productivity analysis. (iv) A statistical audit of the consistency, integrity and comprehensiveness of economic statistics. (v) Benchmarks for the production accounts in the CSNA. (vi) Source of algorithms in the development of weighting patterns for price indexes, say, for capital formation.

When compilation is viewed in its broader role, the periodicity of input-output tables becomes an important question. Input-Output tables can serve better their broad role and be of high quality if they are produced annually. Time series improve quality of output and the experience of the staff remains relevant and useful.

We made a reference to aggregation problems in this paper. It is absolutely essential to develop an efficient data retrieval system. Furthermore, basic input-output tables must be balanced at as micro a level as possible and this micro data set must be preserved. It is worth emphasizing that the economics profession should pay more attention to the problem of aggregation of economic series and the errors which ensue with various levels of aggregation.

Statistics Canada - 292 - Collected Articles of Kishori Lal

References b. Excepting minor editing, this is a reproduction from the Review of Income and Wealth, December 1982, pp. 411-430. An almost identical version of this paper (plus statistical tables) was also published in the Proceedings of the International Association of Research in Income and Wealth, Gouvieux, France (August 16-22, 1981) titled Compilation of Input- Output Tables, Springer-Verlag, Berlin, 1982, pp. 8-36.

1. United Nations, A System of National Accounts, Studies in Methods, Series F, No. 2, Rev. 3, New York, 1968.

2. Statistics Canada, The Input-Output Structure of the Canadian Economy, (catalogue 15-201), Ottawa.

3. Statistics Canada, The Input-Output Structure of the Canadian Economy in Constant Prices, (catalogue 15-202), Ottawa.

4. United Nations, Manual on National Accounts at Constant Prices, Series M, No. 64, New York, 1979.

Statistics Canada - 293 - Collected Articles of Kishori Lal

Canadian Input-Output Tables and Their Integration With Other Sub-Systems of the National Accounts a

by Kishori Lal Statistics Canada, Ottawa

The System of National Accounts in Canada consists of several parts. The annual and quarterly Income and Expenditure Accounts were, historically speaking, the first set of statistics to be referred to with the title "National Accounts". The Balance of International Payments data, in their more summary form, are also part of the System of National Accounts and they, in fact, predate the Income and Expenditure Accounts. Greatly expanded structural detail on industries and on goods and services in current and constant prices is portrayed in the annual Input-Output tables of the System. The annual and monthly constant dollar measures of the contribution of each industry to total Gross Domestic Product at factor cost portray another part of the Canadian System of National Accounts.

Both the Input-Output Tables and the estimates of Gross Domestic Product by Industry use the establishment as the primary unit of industrial production. Measures of financial transactions (another sub-system of the National Accounts) use legal entity as the main unit of classification of transactors.

The Canadian System of National Accounts provides an overall conceptually integrated framework in which the various parts can be considered as inter-related sub-systems. At present, direct comparisons amongst those parts which use the establishment as the basic unit and those which use the legal entity can be carried out only at highly aggregated levels of data. This paper does not deal with the integration problems between establishment based and legal entity based statistics. Its focus is narrower; it is concerned with problems of integration of the annual production accounts derived from Input-Output tables (produced by the Input Output division), Income and Expenditure Accounts (produced by the Gross National Product division), and independent estimates of GDP by Industry (produced by the Industry Measures division), and how these data are integrated and what problems are faced in reconciliation. Intelligent understanding and appreciation of integration and reconciliation problems between the estimates derived from Input Output Tables and other sub-systems requires some background of the accounting framework of the Canadian Input-Output tables.

THE ACCOUNTING FRAMEWORK

The framework of the Canadian Tables bears a close relationship to the one described in the 1968 United Nations report A System of National Accounts. The inputs and outputs of industries are presented in separate tables; both intermediate inputs and outputs are classified by commodity. Commodities are clearly distinguished from industries, the number of commodities exceeding the number of industries. The Canadian Input-Output tables are rectangular rather than square. The

Statistics Canada - 295 - Collected Articles of Kishori Lal rectangular (Industry-Commodity) format has two very important advantages over the traditional square (inter-industry) format: 1) it admits as much detail as is available in the basic survey sources; and (2) the meaning of each entry is straightforward because observed transactions are not combined with fictitious transfers. Canada has a pioneering role in both the theoretical and empirical development of rectangular Input-Output tables, a development pursued by practitioners in Statistics Canada and in other Canadian institutes.

A schematic presentation of the accounting framework of the Canadian Input-Output Tables is given in Chart 1.

The Input-Output Tables for Canada contain two sets of interrelated accounts, the Commodity Accounts and the Industry Accounts; the former details the supply and disposition of individual commodities (goods and non-factor services); the latter details the commodity composition of the output of industries and the complete costs of production (including profits) of industries. These accounts are described below. While the Industry and Commodity Accounts treat the primary inputs and final demand, these entries are also described together in a separate section.

Commodity Accounts

As noted, the Commodity Accounts display the supply and disposition of each commodity. Matrix V shows the production of each commodity produced by each industry valued at producers' prices. The disposition of commodities by the various classes of transactors are shown in matrices U and F. Matrix U contains the use, on current account, by each industry of each commodity for the purpose of using it as an intermediate input in the production of other commodities.

Matrix F contains the demand for each commodity by final demand category - personal expenditure, fixed capital formation, additions to (the value of physical change in) inventories, gross government expenditure on goods and services, and exports. The three other columns in matrix F are imports, withdrawals from (the value of physical change in) inventories and government revenue from the sale of goods and services. These three categories supplement the supply of goods and services produced by Canadian industries in the period of account. It should be noted however that small amounts of goods and services are produced within the government sector and the production costs of these goods and services are already included in gross government expenditure. Matrices U and F include the use of commodities originating either as imports, as withdrawals from inventories, or as government-produced goods and services. Three adjustments are thus required - imports because they are extraneous to the production of Canadian industries, withdrawals from inventories because these were produced in an earlier period, and government revenues from sales of goods and services because the costs have already been fully accounted for in government gross expenditure - to effect equality with the output of domestic industries.

Statistics Canada - 296 - Collected Articles of Kishori Lal

Chart 1 The Accounting Framework of Rectangular Input-Output Tables

Final Demand Commodities Industries Categories Total

Commodities U F q

Industries V g

Primary Inputs Y1 YF n

Statistics Canada - 297 - Collected Articles of Kishori Lal

Industry Accounts

The total output of each industry, classified by commodity, is shown in the industry's row of matrix V. The inputs of each industry are shown in the industry's column which covers both matrix U for intermediate inputs and matrix YI for primary inputs. Matrices U and V were discussed above under the Commodity Accounts. The primary inputs in matrix YI consist of net indirect taxes (commodity indirect taxes, plus other indirect taxes, less subsidies) and GDP at factor cost (wages and salaries, supplementary labour income, net income of unincorporated business and other operating surplus).

For each and all industries total output equals total input, intermediate and primary.

Primary Inputs and Final Demand

There is one matrix not yet described, YF. This contains primary inputs purchased directly by final demand categories. Labour income in matrix YF represents the wages and salaries, and supplementary labour income paid to persons employed in the personal and government sectors. This includes the employees of non-profit institutions, such as universities and labour unions, and the income of domestic servants and baby sitters in the personal sector, and all public servants in the government sector. Surplus in matrix YF reflects the depreciation on assets owned in the government sector and by non-profit institutions classified to the personal sector. Assets which are charged to fixed capital formation in these sectors are depreciated, such as buildings, roads, equipment, etc. All residential housing, even when owner-occupied, is classified to business, and thus depreciation on housing is part of the surplus of the appropriate industry in matrix YI, not YF.

Economic Identities

With the description of all the matrices, certain identities become apparent.

The total outputs of industries equal the total intermediate inputs of industries plus their total primary inputs.

Statistics Canada - 298 - Collected Articles of Kishori Lal

The total outputs of commodities of industries equal the total intermediate inputs plus the total commodity inputs into the final demand categories.

Intermediate inputs being common, primary inputs of industries equal commodity inputs of final demand categories.

Adding primary inputs of final demand categories to both sides, we get: (a) primary inputs of industries and of final demand categories equal to, (b) commodity inputs of final demand categories plus primary inputs of final demand categories. (a) is called Gross Domestic Product at market prices and (b) is called Expenditure on Gross Domestic Product at market prices.

The above identity is true for both current price and constant price tables.

Gross Domestic Product (GDP) at factor cost (wages and salaries, supplementary labour income, net income of unincorporated business and other operating surplus) equals GDP at market prices less net indirect taxes (commodity indirect taxes plus other indirect taxes less subsidies).

Present situation

GDP in total and by Industry, and Expenditure on GDP in total and by category are produced annually in both current and constant prices through the Input Output tables but these estimates are produced with a lag of three years to the reference year. Statistics Canada, for example, completed the 1981 Input Output tables in the first quarter of 1985. Such tables and other ancillary estimates are available for all the years 1961-81. In addition to the annual estimates of GDP produced through the Input Output tables, other sub-systems also have produced GDP estimates for the entire period 1961-81 and onwards. For the common period 1961-81, there are, in total, three current price estimates of GDP in Statistics Canada; all give very similar results but still different, two cover full GDP at market prices and one GDP at factor cost; the difference between GDP at market prices and GDP at factor cost is the value of indirect taxes less subsidies. There are two current and two constant price estimates of Expenditure on GDP (actually one is GNE but the only statistical difference is an adjustment of net payments of investment income abroad). There are two constant price estimates of GDP (one at factor cost, the other at both factor cost and market prices) published annually by Statistics Canada. The details are different; industrial classes are differently aggregated; sectoring is not the same; valuation of specific commodities is not identical. This situation does require an integration and even more a rationalization.

The above differences were not, of course, deliberately planned but are the result of historical accidents. Of all the existing measures, the Input Output tables are the newest. As originally conceived, the Input Output tables were not meant to have primarily as benchmarks for the Canadian System of National Accounts but rather to serve as occasional analytic snapshots of the Canadian economy. The fact that the I-O tables were not initially constrained to match exactly the existing Canadian SNA structures has meant that the I-O tables provided a fresh reappraisal of existing measurement conventions. This has proven useful.

Statistics Canada - 299 - Collected Articles of Kishori Lal

What we expect in June, 1986

We are expecting that by June, 1986 we will complete an integrated and reconciled database for the annual production accounts for the years 1961-81 (also referred to as historical revisions) which will have for the first time the following series identical for the entire SNA; with this in hand we will make every effort to have the current period remain integrated and reconciled as well. The reconciled series for the historical revisions in current prices will be:

1) Wages and salaries in total and by Industry 2) Supplementary Labour Income in total and by Industry 3) Net Income 4) Indirect Taxes 5) Subsidies 6) Consumer Expenditure in total and by Category 7) Machinery and Equipment 8) Construction 9) Exports 10) Imports 11) Government Expenditures

What will not be identical on the GDP side will be operating surplus in the Input Output accounts and the equivalent in concept of the total of the following in the Income and Expenditure accounts: Profits and other investment income; inventory valuation adjustments; capital consumption allowances and miscellaneous valuation adjustments. The present practice of showing an explicit error of estimate in the Income and Expenditure accounts but not in the Input Output accounts (see section on Error of Estimate and the Input Output Accounts for further elaboration) will continue.

What will not be identical in Expenditure on GDP will be VPC (Value of Physical Change) in Inventories in the two sub-systems (Input Output Tables and Income and Expenditure Accounts). Again, an explicit error of estimate will be shown in the Income and Expenditure accounts but not in the Input Output accounts.

There will be two estimates of GDP, not the present three; the one at factor cost will be merged with the Input Output Accounts. There will remain two estimates of Expenditure on GDP, one produced through the Input Output Accounts and the other in the Income and Expenditure accounts but most of the individual estimates will be produced jointly.

Two annual estimates of GDP in current prices - why?

Only for the historical revisions period and the years for which Input Output tables are completed, will there be two estimates but for the current period, there will be only one estimate of GDP. This is primarily due to the fact that the Input Output accounts use establishment as the basic unit for estimating operating surplus whereas Income and Expenditure accounts must use company for estimating various components of operating surplus such as profits, investment income.

Statistics Canada - 300 - Collected Articles of Kishori Lal

On the expenditure side, as noted above, the two estimates will be identical except for VPC for the historical period and for the years for which Input Output tables are completed. For the current period, only one estimate of Expenditure on GDP will be prepared. In the delineation of Expenditure on GDP by category, VPC is a residual estimate in the Input Output accounts; on the other hand, in the Income and Expenditure Accounts, VPC is "calculated" independently with a liberal use of judgement. The rationale behind these two approaches is discussed below in the section on Error of Estimate and the Input Output accounts. Both approaches are used in various statistical bureaus. It is our plan to publish both these estimates. However many individual estimates other than VPC will be worked out, as mentioned above, jointly.

Error of Estimate and the Input Output Accounts

The question has occasionally been raised that Input Output tables hide information by not showing explicitly an error of estimate in the published tables; in the Income and Expenditure accounts such a statistical error is regularly exhibited. It is worth our while to go through this topic in a fairly detailed way to answer this nagging question.

Two independent sources are used to estimate the income side and the expenditure side of the GNP accounts. Conceptually these two estimates must be identical but in practice there is always a difference between the two. This statistical difference between the two is called the error of estimate. Similarly, two independent sources are used to estimate the current account and the capital account of the Balance of International Payments. The two balances must be conceptually identical but in practice there is always a difference between the two; this difference is called the statistical discrepancy.

In the GNP accounts, the error of estimate is divided by 2, one half being added to one side and the other half with the opposite sign being added to the other side; thus a statistically identical income and expenditure accounts are published in Canada. In the balance of payments, the entire amount of the statistical discrepancy is shown separately of the two accounts in the published balance of payments.

There are no set guidelines on how to exhibit the statistical error; furthermore, the UN SNA has no specific provision for explicitly showing error of estimate in the GNP type accounts. 16 out of 24 OECD countries do not show any error of estimate; 7 countries show an error of estimate on one side of the accounts and Canada is the only OECD country showing error on both sides.

Some think that explicitly showing the error of estimate is a good discipline; rather than allocating the error of estimate to one sector or to all sectors let the error be specifically shown and all sector(s) remain as estimated. Yes this is useful if independent sources are used for the two sides of the accounts and both sources are quite robust. In reality, they are not and a large number of adjustments are made to the independently recorded or surveyed estimates on both sides before an error of estimate is finalized. In these circumstances, the role of discipline imputed to the error of estimate is questionable.

Statistics Canada - 301 - Collected Articles of Kishori Lal

As noted, the UN SNA has no specific provision for an error of estimate in the accounts because the production accounts are built in the context of Input-Output tables. In the Input Output accounts, gross output less intermediate use measures GDP in total and the same gross output less intermediate use measures final demand in total when looked at in the context of disposition of gross output. GDP and Final Demand are identical in the Input-Output Accounts because the same gross output on both (output and disposition) sides is used with a common block of intermediate use.

The process defined so far simultaneously generates equivalent values of GDP and final demand and it provides the framework for all subsequent steps such as disaggregation of GDP and final demand.

All transactions are defined in terms of commodities, and this leads to the establishment of commodity balances. The output and intermediate use of industries is specified by commodity; furthermore, the commodity content is available from direct sources for such final demand sectors as exports, imports and government expenditures but there is no direct evidence of commodity content for most of the consumer expenditure categories and investment. Initial estimates are made of the commodity content (and totals) of consumer expenditure and of the commodity content of investment. For example, initial estimates of consumer expenditures are struck with several sources such as: commodity balances; retail trade establishment surveys supplemented by occasional (very occasional) trade commodity surveys and surveys of consumer expenditures (FAMEX). Expenditures on certain commodities such as alcoholic beverages and tobacco are scandalously underestimated in surveys of consumer expenditures, and by deduction other commodity expenditures may be overestimated. Furthermore, each category of consumer expenditures in the SNA represent expenditures in Canada whereas FAMEX commodity detail represents expenditures by Canadian. However, all these sources are used to establish initial estimates. Fixed investment totals as well as totals for exports, imports, government expenditures are provided independently and are identical in the various subsystems of the SNA.

While the value of the sum of the disposition, intermediate and final, of all commodities is equal to the sum of the supply of all commodities, this is not initially true when balances are struck for individual commodities. This leads to detailed investigation of the commodity balance of each commodity. All sources of apparent commodity imbalance are investigated, be they problems in classification, valuation, timing, or even improper reporting. The analysis of the commodity balances is an intensive process and one in which one can make use of any appropriate source of information to corroborate or modify the values found in the initial trial balance. In spite of these efforts it is recognized that a definitive resolution of small differences in the balance for an individual commodity is intractable. In practice, in the Input-Output accounts, these differences are reflected in Raw Material Inventory, the commodity content of which is not reported.

In the Input-Output framework where the commodity detail of an important part of the system is inferred from commodity balances, one cannot meaningfully and with confidence create a column of "pure" errors of estimates. For the sake of convenience only, the raw material inventory column has been chosen to reflect all the errors; however, every effort is made to examine any serious discrepancy between the Input-Output estimate in total and the GNP estimate and to assure that year to year change in inventories of each commodity does make some economic sense. For example, one

Statistics Canada - 302 - Collected Articles of Kishori Lal cannot withdraw continuously from inventories without once in a while restocking them. It is interesting to note that in Norway, a country recognized for having a highly developed System of National Accounts, changes in stocks of individual commodities are estimated in the context of commodity balances in the Input-Output framework.

As noted already, in the forthcoming historical revisions in the SNA, estimates of consumer expenditures, government expenditures, exports, imports, fixed capital formation will be identical in the Input-Output and GNP accounts. The only difference will be in the estimates of Inventory Change between the two series which will be readily available to our users. Though the total Expenditure on GDP in the two subsystems will not be identical, it will, most likely, be very close.

Annual Constant Price Estimates

We have noted above that for the historical revisions 1961-81 and for the other years for which the Input Output tables are completed, we will have identical totals for all the series except operating surplus, profits, investment income etc, VPC and error of estimate in current prices. However, we may not have the same identical results in constant prices. There is no definitive way of estimating "volume" or "quantum" except with reference to time and weights. Input Output tables express values in producer prices whereas the Income and Expenditure Accounts express values, say for consumer expenditures, in purchaser prices. There is no guarantee that deflating detailed expenditures in producer prices for commodities and margins separately will give the same results as deflating composite commodities (including margins as in retail trade). Chances are that the two series will not be very different but they will rarely be identical. There is another constraint in the Income and Expenditure Accounts that the current quarterly and annual series do not have the same database and data sources as the annual Input Output tables have, for the simple reason that the complete Input Output tables are produced with a lag of three years to the reference year and use detailed annual census/surveys not yet available for the current period.

By June 1986, the present two estimates of constant price GDP will be merged into one and that will be produced using double deflation method through the Input Output accounts. But for the Expenditure on GDP we will continue producing two series for two reasons: 1) there are two current price series on Expenditure on GDP which in detail are identical except for VPC and error of estimate; (2) even those areas which are identical in current prices may have different results in constant prices for reasons of "deflation" given above. Every effort will of course be made to monitor the differences in results and to reconcile any wide margins.

Monthly and Quarterly Estimates

Monthly GDP estimates in constant prices will be produced as at present but the industrial and sectoring classification will be reconciled and integrated with the ones in the annual Input Output benchmarks. Current price estimates of GDP (or GNP) by quarter will be benchmarked to the integrated annual estimates produced for the entire SNA; similarly, the quarterly constant price Expenditures on GDP (or GNE) will be benchmarked to the integrated annual estimates.

Statistics Canada - 303 - Collected Articles of Kishori Lal

Integration

Integration of the Production Accounts in the SNA requires the resolution of several important issues such as valuation, classification, revision cycle, deflation, documentation etc. But integration should not be equated with forcing or producing a single set of numbers. If two independent methodologies produce two sets, let them stand equally. Suppressing one set will serve poorly the whole approach to integration. We will deal below with the various aspects of integration issues faced by us in integrating the SNA series in Statistics Canada.

Valuation

Input Output tables in current prices are initially balanced in purchaser prices which are then converted into producer prices to have a consistent valuation for all users. Consistent valuation is crucial for relating use to supply of each commodity and more particularly for constant price GDP through double deflation. In the Income and Expenditure Accounts, all expenditure series are valued in "purchaser prices" because source data record only such prices. Deflation is done for the expenditure categories using such deflators as CPI (Consumer Price Index). Valuation is even more important for imports. Imports are valued in the Input Output tables equalling imports at c.i.f. (cost of insurance and freight) values plus import duties to make them consistent with the domestic producer values. In the GNP accounts, values of individual import commodities do not include duties and indeed are at FOB (free on board) values; however, these differences are clearly delineated. It is important that all the subsystems of the SNA delineate the various values to reconcile series using different valuation practices.

Classification

A fundamental element in statistical compilation is classification. At present the three divisions in the SNA preparing Production Accounts do not use identical sectoring. For example, GNP and Input Output divisions delineate each SIC industry into various sectors, whereas Industry Product division characterizes entire SIC industry as commercial or non commercial. Education is in three sectors in the Input Output and GNP accounts - Government, business and persons - but in the Industry Product it is allocated entirely to the non-commercial sector. The GNP accounts use major SIC divisions as aggregates whereas Input Output division aggregates three and four digit SIC into 191 industry classes and Industry Product has yet another aggregation. It is necessary to have one set of rules for sectoring and aggregation for the entire SNA. Series in one account may be more aggregated than in the other sub-system but one must have the capability to reconcile in a straightforward way one set of aggregation from another set.

There is a general issue of classification in the Bureau. We have one Standard Industrial Classification for establishments but several commodity classifications not concorded with each other. We have Industry commodity classification, Export commodity classification, Import commodity classification, Domestic Trade classification and so on. They are all useful and relevant but together they are unintegrated.

Statistics Canada - 304 - Collected Articles of Kishori Lal

Aggregation

Aggregation is quite a serious problem in economic analysis, yet this topic has not received adequate attention in the literature. One must realize that the same database could give different results depending upon the level of aggregation the statistician has used. Aggregation forces a particular point of view, a particular analysis. Indeed, aggregation itself is a model.

One must avoid the temptation, as much as possible, to aggregate data in the early stages of economic analysis. Aggregation is easier to do than disaggregation, The implications of this simple statement must be well understood. We strongly recommend to pay attention to keeping micro data sets. We say "keeping" micro data sets because data sets from the respondents are made available in this form. With them one would be able to arrange custom-made aggregations for economic analysis and integrate various series. This requires an efficient data retrieval system. Without an efficient data retrieval system, one will remain a prisoner of dictated (by others) aggregations and one's horizon of economic analysis will remain narrow indeed.

Revision cycle

The three divisions of the SNA have their own specific and unique revision cycle; GNP typically revises the data each June for the last four year whereas Input-Output and Industry Product often revise data for longer spans. For example in June 1984, GNP division revised the data for all the four years 1980-83. About the same time or a bit earlier, Input Output division produced the Input Output tables for the year 1980 and revised the data for several earlier years as well. Industry Product division has yet another revision cycle.

There is nothing wrong with incorporating the most up to date data but all must have the same revision cycle. Otherwise, in the above example, the years earlier than 1980 will remain different in the published series. The difference is explained only in terms of the revision cycle. This practice must be reexamined and consequent problems resolved. If the final data are not ready for say, 5 years, may be that should determine the revision cycle for every sub-system.

Easy retrieval

Integration requires that the data series be kept in machine readable format so that for reconciliation purposes or aggregation purposes one can manipulate from one set to another. To integrate GNP series with Input Output series will be a mind boggling exercise if the data are not available in easy machine retrievable format.

Documentation

To a large extent lack of integration and reconciliation is due to the absence of complete documentation of methodology, valuation, classification, revisions and other judgements made in the preparation of any statistical database. This straightforward and innocent looking statement is very

Statistics Canada - 305 - Collected Articles of Kishori Lal significant to resolve most of the problems that one faces in statistical series. Documentation is paid far less attention than it deserves; it requires patience and discipline. Computer programmers and systems analysts do this quite often; engineers and doctors do it regularly but our profession is rather lax. I can't emphasise enough the need for SNA subject matter analysts to start detailed documentation.

Constant Price Series

Integration assumes a rather analytical role when doing so in constant prices. One of the statistical roles for Input-Output Tables is that of statistical integration and the construction of constant price Input-Output Tables can serve to uncover anomalies among different price indexes. The Input-Output Tables are a balanced system valued at producers' prices. Thus the same price attaches to a commodity as output in the Make Matrix and as input in the Use and Final Demand Matrices. The addition of margins to producers' values renders them in purchasers' values and the synthetic price index so constructed can be compared with observed price indexes such as the Consumer Price Index used by the GNP accounts. In cases of disagreement between the observed and synthetic price indexes an investigation will usually reveal the probable cause and judgements can be made as to what and where adjustments are necessary.

Concluding remarks

Integration and reconciliation require a strong and long-term commitment by senior management, a willing cooperation by the various sub-systems of the SNA and a dedication to documenting the methodology, classification and aggregation parameters used, valuation practices, revision cycle and professional judgements made in the preparation of any statistical series. One will have to adhere to a discipline of developing and implementing common (and reconcilable) classification and aggregation principles for use of the entire SNA. Without this, the SNA can't claim to have an integrated database and more so an integrating mechanism for all the feeder series to the SNA. The current monthly and quarterly series are useful but they will become even more so if these are reconciled to the SNA base integrated annual series. We hope and expect that the SNA annual Series for the entire period 1961 and onwards will be fully integrated by June 1986 and that the monthly and quarterly series will soon afterwards be reconcilable to the annual series.

Reference c. Excepting minor editing, this is a reproduction from the Proceedings of an International Meeting, Baden near Vienna, Austria, (May 19-25, 1985) titled Problems of Compilation of Input-Output Tables, Wien, 1986.

Statistics Canada - 306 - Collected Articles of Kishori Lal

Canadian System of National Accounts – An Integrated Framework a

by Kishori Lal Statistics Canada, Ottawa

Abstract: Various features of the production accounts in the Canadian System of National Accounts (CSNA) are produced by three separate divisions. They were not fully integrated statistically or conceptually before the 1986 historical revisions of the national accounts. This paper reports on the steps taken to resolve the integration and reconciliation problems of such accounts in the CSNA. Integration requires a strong and long term commitment by the senior management, a willing cooperation by the various parts of the CSNA and a dedication to develop and implement a common set of classification and aggregation rules as well as consistent revision practices to be used for all the underlying statistical series both in the CSNA and the feeder system. Furthermore, integration requires documentation of the above as well as of the links that exist between the components of the CSNA. To be effective in the present environment of computerization, integration also requires that the software and hardware used for manipulating the database of the individual components are straight-forwardly linked. It is expected of the CSNA to develop a set of statistics which together produce an integrated, consistent and comprehensive profile of the economy as well as (and it is equally important) to upgrade the quality of the underlying statistical series from the feeder system.

The Canadian System of National Accounts (CSNA) consists of four principal components; these are: (a) Input-Output Accounts and their major derivatives, Real Gross Domestic Product Output by Industry, and Productivity Measures; (b) Income and Expenditure Accounts; (c) Financial Flows and the National Balance Sheets; and (d) Balance of International Payments and the International Investment Position. Their development has taken place over many decades, depending on the demands for policy and analytical purposes, the degree of computerization of the statistical system and the management commitment to use the CSNA not only to develop a set of statistics which together produce an integrated, consistent and comprehensive profile of the Canadian economy but as well to upgrade the quality of the underlying statistical series from the feeder systems.

Primary focus of the Input-Output Accounts is to measure output of goods and services by industry, use of goods and services by industry (also called intermediate inputs), use of goods and services by final demand sectors (also called Gross Domestic Product, expenditure based) and estimates of the costs of factors of production (also called GDP, income based). Growth of GDP by Industry expressed in constant prices relative to the growth of labour inputs (persons employed, hours worked) yields the present estimates of Labour Productivity. Labour is one of the many variables which affect real GDP; another important variable is the physical stock of capital. Combining two or more factors to explain productivity of an industry or of the total economy is called multifactor productivity analysis. Such a multifactor productivity program is being developed now in the CSNA.

Statistics Canada - 307 - Collected Articles of Kishori Lal

The Income and Expenditure Accounts component of the CSNA starts with the identical measures of current price GDP income based and current price GDP expenditure based. The Income and Expenditure Accounts, however, disaggregate the macro production accounts into institutional sectors of the economy providing estimates of current income and outlay and capital accumulation for each sector. These sector accounts encompass not only income arising from production but the redistribution of income which takes place through the process of inter-sectoral transfer payments and receipts. The capital finance accounts which measure saving and non-financial investment yield as a balancing item the lending capacity or borrowing requirements of the sector.

The Financial Flows component of the CSNA portrays the financing of economic activity. Transactions in financial assets and liabilities are shown as the counterpart to estimates of the balance between saving and non financial investment or net lending/borrowing obtained from the sector capital finance accounts of the Income and Expenditure Accounts. Sectors are further disaggregated to show financial intermediation by type of institution. Transactions in major types of financial instruments are recorded separately. The National Balance Sheet Accounts show the corresponding levels of financial assets and liabilities of Financial Flows Sectors, as well as estimates of stocks of physical asset and net worth.

Balance of International Payments deals with the economy's transactions with non-residents or the rest of the world. The basic statement is divided into two accounts, one covering current and the other capital transactions. The current account records payments and receipts for goods and services traded, investment income flows and transfer payments; the capital account covers transactions in financial assets and liabilities and records the inflow and outflow of capital funds. In condensed form, the balance of payments estimates are represented in the other components of the CSNA as the rest of the world sector, the current account being part of the Income and Expenditure (and Input-Output accounts) and the capital account forming the rest of the world sector in the Financial Flows. International Investment position, like National Balance Sheet, shows the level of outstanding assets and liabilities vis-a-vis non residents.

Linkage of the balance of international payments (and international investment position) to the other components of SNA is different in nature. The Balance of Payments does not fit sequentially into the system in the same way as the production accounts in the Input-Output component lead into Income and Expenditure Accounts and the capital finance accounts in the Income and Expenditure component lead into Financial Accounts.

Integration between (a) the capital finance accounts of the Income and Expenditure accounts and the Financial Flows, (b) current account of the Balance of Payments and the rest of the world account in the Income and Expenditure Accounts and (c) the international investment position between the Balance of Payments and Balance Sheets accounts for the rest of the world was largely in place before the 1986 historical revisions of the CSNA; however, this was not true for the production accounts as derived from Input-Output Tables (produced by the Input-Output Division), Income and Expenditure Accounts (produced by the Income and Expenditure Accounts Division), and estimates of Monthly GDP by Industry (produced by the Industry Measures and Analysis Division). This paper

Statistics Canada - 308 - Collected Articles of Kishori Lal focuses on the resolution of the integration and reconciliation problems of the production accounts of the CSNA.

Integration and reconciliation require a common set of classification and aggregation rules as well as consistent revision practices to be used for all the underlying statistical series in the CSNA. An equally important dimension to this is the documentation of data sources and methods and the establishment of links between the various components.

CLASSIFICATION

Fundamental to integration and reconciliation initiatives is a consistent classification of transactions and transactors. Before the 1986 CSNA historical revisions, the Input-Output component and its derivatives used an inconsistent sector classification and aggregation of industries. The sector classification - there are three domestic sectors in the economy, one is business, the other is government and the third is persons - used by the Input-Output tables and the Income and Expenditure Accounts was identical; however such was not the case between the Input-Output and the Productivity Measures or between the Input-Output and the Monthly GDP measures. For example, the industry "education" was allocated to all the three sectors - business, government, and persons - in the Input-Output and Income and Expenditure Accounts but for Productivity Measures and Monthly GDP it was entirely allocated to the non-business (government and persons) sectors. The Income and Expenditure Accounts used major SIC (Standard Industrial Classification) divisions as aggregates whereas Input-Output aggregated three and four digit SIC into 191 industry classes; Industry Measures and Analysis division used still another aggregation for the monthly measures. The aggregation rules adopted by the components of the CSNA were not hierarchically related. The result, for example, was that monthly measures of GDP were not serving their very important role as forerunners of annual GDP estimates by industry derived from the Input-Output tables.

The various components of the CSNA were not treating all transactions consistently. For example, urban transit systems started losing a lot of money beginning in the late 1960's and early 1970's. The governments began to subsidize them to cover such losses. Contributions by the governments were considered as operating subsidies in one component, and as grants in the other components of the CSNA with the result that estimates of GDP at factor cost were dramatically different for this industry amongst the CSNA components. This type of inconsistent treatment has now ceased.

Within the CSNA, classification of transactors and transactions is now consistent; however, this is not yet true in the various divisions of Statistics Canada. For example, CSNA allocates Electric Power, Gas and Water utilities to goods producing industries whereas SEPH (Survey of Employment, Payroll and Hours) and Labour Force Survey allocate this industrial group to services producing industries. This creates confusion for the users.

With the recent historical revisions, common SNA industry codes have been developed and implemented for all the CSNA integrated industry statistics series, be they national, provincial or sub-annual. A common set of codes both for detailed classes and aggregates is now used. This is an exceedingly important tool of integration.

Statistics Canada - 309 - Collected Articles of Kishori Lal

REVISION PRACTICES

The three divisions of the CSNA compiling production accounts had their own specific and unique revision practices. The Income and Expenditure Accounts division would typically revise data each June for the last four years whereas Input-Output and Industry Measures divisions often would revise data for longer spans. For example in June 1984, Income and Expenditure Accounts division revised the data for all the four years 1980-83. About the same time or a bit earlier, Input-Output division produced their tables for the year 1980 and revised the data for several previous years as well. Industry Measures and Analysis division had yet another revision cycle. The result was that at any given time, the finality of the series was never consistent amongst the various components.

The following rules and procedures have been put in place starting with the 1987 CSNA revision cycle. i) Annual benchmark estimates at current purchaser prices:

Final annual estimates for the year 1983 and almost final annual estimates for the year 1984 at current purchaser prices produced through the Input-Output accounting framework were produced in May 1987 for use by the National Income and Expenditure Accounts annual revision in June 1987. Note that final annual estimates have a specific meaning in our context; such estimates will not be reexamined till the next historical revision.

Because of the workload for the historical revision, 1982 final annual estimates were not completed in 1986 as per normal practices. Hence, there was a requirement to produce the 1982 estimates as well by May 1987. ii) Income and Expenditure Accounts estimates

Income and Expenditures Accounts division released the results of the annual revision with the first quarter estimates in July. In the second, third and fourth quarter releases, only estimates for the current year are subject to revision; i.e. at the time of the second quarter the first quarter could be revised; at the time of the third, the first and second quarters could be revised etc.

The span of years for a normal annual revision is four years, although in 1987, data for the years 1982 to 1986 were revised to incorporate the 1982, 1983 and 1984 input-output estimates. By 1988, Income and Expenditure Accounts division will be back to the usual four year cycle. Annual provincial accounts as well as more detailed national income and expenditures accounts follow the above revision consistently. iii) Annual benchmark estimates at constant producer prices:

Input-Output Division converted the current purchaser price estimates into current producer price and then constant price valuation by the end of July, 1987. As noted, Input-Output division produced current purchaser price annual estimates in May 1987. Between May and June some

Statistics Canada - 310 - Collected Articles of Kishori Lal fine-tuning of the Input-Output estimates was done in concert with the Income and Expenditures Accounts division. By the end of June the two divisions finalized the estimates for the three years 1982-84 such that the total GDP at market prices and all the components (except value of physical change in inventories and operating surplus) of GDP were identical in the two sub-systems. These estimates were converted in July 1987 into producer prices and then constant prices. iv) Monthly Real GDP at factor cost: Constant price final estimates for the years 1982 and 1983 and almost final estimates for the year 1984 became the annual benchmark estimates for monthly real gross domestic product by industry. Industry Measures and Analysis (IMA) Division released at the end of August 1987 benchmarked monthly estimates for the years 1982 to 1984 and projected estimates for 1985 and 1986 as well as the first six months of 1987. Normally, the 1982 final estimates would have been incorporated at the time of the release of the June 1986 estimates. The four years revision cycle will be reestablished at the time of the release of the June 1988 estimates of monthly GDP. In the monthly releases for July 1987 to May 1988, estimates for 1986 or earlier years are not subject to revision. v) Productivity Measures:

Revision practices for Productivity Measures are now completely consistent with the annual constant price GDP by Industry produced through the Input-Output program and the current period's estimation of the real annual GDP produced through the Monthly Program. In October 1987, Input-Output Division released the final Productivity Measures for the years 1982 and 1983 and preliminary measures for the years 1984-86. vi) Revision practices in the feeder system

There are no comprehensive rules, procedures or policies affecting in a consistent manner the revision practices for the series produced in the feeder system. Some series (for example, Consumer Price Indexes) are never revised; some (for example, Census of Manufacturers) remain open for revision for one year; others have varying practices. In the process of compilation of national accounts, certain (sometimes major) imbalances arise which are resolved by adjusting the basic sources and/or by macro adjustments. This resolution is quite often not incorporated into the series of the feeder system. Thus the link between the SNA revisions and the underlying series produced and published in the feeder system is not maintained.

STATISTICAL ESTIMATES, CURRENT PRICES

There used to be a fairly wide deviation of estimates even in current prices for the same concept amongst the various divisions of the CSNA. The aggregate totals were different and even more divergent were the underlying published details. The levels were different as well as the growth rates. The following are some examples to illustrate the statistical situation in the old (before integration) and new (after integration) series of the CSNA.

Statistics Canada - 311 - Collected Articles of Kishori Lal

In spite of the fact that the old series ( see Table 1) for the year 1981 were all released in mid-1985, the level difference of GDP at factor cost between Input-Output and Industry Measures was more than $9 billion. The new series were released in July 1986.

(i) Main components of GDP at Market Prices

Though the aggregate GDP at market prices in the Input-Output and Income and Expenditure accounts was not significantly different for the year 1981, the underlying main components recorded non-identical estimates. Even such series as indirect taxes and subsidies were not the same.

In the income based series (see Table 2 below for comparisons of old and new estimates by major components), it needs to be remembered that given the independent methodologies of estimating surplus in the Input-Output accounts and its corresponding elements (profits, investment income, inventory valuation adjustment and capital consumption allowances) in the Income and Expenditure accounts, this value remains different. Operating surplus in the input-output accounts is estimated from the establishment based surveys which provide information on shipments or production or revenues, use of raw materials etc. for such production, and wages and salaries. Operating surplus is the difference between production and the total of (a) all intermediate inputs, (b) labour payments and (c) net indirect taxes. It is a balancing item though closely monitored and analysed before its value is finalized. On the other hand, estimates of profits, and investment income in the Income and Expenditure accounts are, among other things, based on company and enterprise reports which provide direct estimates for book profits and investment income. Both sub-systems make conceptual and coverage adjustments to bring the estimates in line with the CSNA framework.

TABLE 1 GROSS DOMESTIC PRODUCT, 1981 ______Old New (Before (After Integration) Integration)

(a) GDP at Market Prices $ Millions $ Millions

Income and Expenditures Accounts 351,468 355,994 Input-Output Accounts 350,555 355,994

(b) GDP at Factor Cost

Input-Output Accounts 313,932 319,538 Industry Measures and Analysis 304,618 319,538

______

Statistics Canada - 312 - Collected Articles of Kishori Lal

At the present time, the two divisions can compare the surplus only at the total economy level because it is very difficult to properly allocate profits from enterprises to their operating establishments and head offices. Once the new central register (part of a major overhaul of business surveys now underway in Statistics Canada) is in full operation, the above problem of comparing establishment based and company based estimates of surplus by industry may be resolved. The two divisions, after a thorough examination and analysis, jointly agree on a common understanding of surplus in the sense that this item in the Input-Output accounts equals the aggregate of profits, investment income, inventory valuation adjustment, capital consumption allowances, and the statistical discrepancy in the Income and Expenditure accounts.

Labour income in the old series used to be identical in total, but it is only in the new series that such identity holds for industrial distribution as well.

It also needs to be noted that in the expenditure based series, (see Table 3 below for comparisons of old and new estimates by major components) the value of physical change of inventories (VPC) in Input-Output tables coincides with the aggregation of VPC and statistical discrepancy in the Income and Expenditures Accounts. Calculation of VPC is derived by two independent methodologies in the two estimates - one using direct estimation massaged with professional judgement and the other using residual methodology massaged with professional judgement.

There are 40 separately identified and annually published categories of consumer expenditures in the CSNA. As noted, the detail used to be quite a bit different in the old series. Even at the four main aggregate categories, three distinguished by durability of goods purchased and the fourth of services, the differences were not trivial. The following table illustrates the situation.

To recapitulate, the following statistical estimates in current prices in the new series are now identical both in total and in detail: wages and salaries, supplementary labour income, net income of unincorporated business, indirect taxes, and subsidies on the income side of the accounts; consumer expenditures, government current expenditures, fixed capital formation, exports, and imports on the expenditure side. In addition, total GDP at market prices is now identical for the base years 1961, 1971 and every year from 1980 onwards.

What is not identical on the income side of GDP is the operating surplus in the Input-Output accounts and the equivalent in concept of the total of profits, investment income, inventory valuation adjustment and capital consumption allowances in the Income and Expenditure accounts. VPC in inventories, part of the expenditure side of the accounts is not identical. It is important to emphasize, however, that these two series - surplus and VPC - are closely monitored and analysed by the two divisions before the estimates are finalized.

(ii) Methodological Changes, Consumer Expenditures

Integration of the current price estimates was helped partly by the implementation of common classification and consistent revision practices throughout the SNA branch. These procedural changes plus certain methodological modifications were made in order to realize full integration of

Statistics Canada - 313 - Collected Articles of Kishori Lal the estimates. Let me illustrate this with reference to the estimates of consumer expenditures for goods in the accounts.

TABLE 2 GDP AT MARKET PRICES - INCOME BASED 1981 $ MILLIONS

Old New IO I.E. IO I.E.

Labour Income 195,960 195,961 197,910 197,910 Net Income of 15,359 16,516 17,853 17,853 Unincorporated Business Indirect Taxes 45,790 46,431 45,956 45,956 Less Subsidies - 9,167 - 8,694 - 9,499 - 9,499 Surplus 102,613 102,003 103,774 103,601 Statistical Discrepancy - - 749 - 173

TOTAL 350,555 351,468 355,994 355,994

TABLE 3 GDP AT MARKET PRICES - EXPENDITURE BASED 1981 $ MILLIONS

Old New IO I.E. IO I.E.

Consumer Expenditures 194,388 193,280 196,191 196,191 Govt. Current Exp. 68,939 69,245 68,792 68,792 Fixed Capital 82,058 82,058 86,119 86,119 Exports 96,840 96,876 96,880 96,880 Less Imports - 92,782 - 92,782 - 93,001 - 93,001 Value of Physical 1,113 2,042 1,013 1,186 Change in Inventories Statistical Discrepancy - 749 - - 173

TOTAL 350,555 351,468 355,994 355,994

Statistics Canada - 314 - Collected Articles of Kishori Lal

TABLE 4 CONSUMER EXPENDITURES CURRENT PRICES 1981 $ MILLIONS

Before Integration After Integration IO I.E. IO I.E.

Durables 27,947 26,432 28,116 28,116

Semi-durable 22,139 22,792 21,947 21,947

Non-durable 59,438 60,188 59,423 59,423

Services 84,864 83,868 86,705 86,705

TOTAL 194,388 193,280 196,191 196,191

Input-Output division used to produce current price consumer expenditure for goods using primarily the commodity balance methodology whereas Income and Expenditure Accounts division used to produce such estimates using primarily the monthly and annual retail trade statistics. Retail trade statistics, once adjusted for national accounts purposes, have consistently undershot the level by more than 10 percent. The ongoing monthly and annual trade statistics program has not been benchmarked since 1971, the latest census of merchandising in Canada. SNA adjustments require, among other things, information on class of customer; retail sales to businesses are removed to derive consumer expenditures. The latest class of customer information was collected about a generation ago, for 1966. Another SNA requirement is to get information on details of commodities sold by the retail businesses. A general purpose retail trade store sells commodities which are allocated to many different categories of expenditures. The latest commodity survey was conducted for the year 1974. Given the above situation, the quality of estimates based on retail trade statistics became suspect, particularly the detail by categories.

The commodity balance approach starts with production statistics; to such data by commodity are added imports by commodity to derive its supply. Supply must equal disposition for each commodity. The under - or over - allocation of commodities must be examined and eliminated. The discrepancy may be due to several reasons: production may be unreported or misclassified; imports and exports may be improperly valued, as well misclassified; there may be timing inconsistencies; etc. There are no ready-made statistical approaches to solving such imbalances. The only effective approach is labourious investigation; one has to go back to the basic records to locate the sources of such imbalances. It is our experience that the commodity balance approach, with its detailed accounting of output by industries and of use by industries and final demand transactors, uncovers major problems in both statistics reported and classification.

Statistics Canada - 315 - Collected Articles of Kishori Lal

The estimates for consumer expenditures for the new (integrated) benchmark series are largely based on commodity balance approach. To further integrate the series and to produce high quality identical estimates for each and all categories, for the annual database (there are no sub-annual I-O accounts in Canada) officers of the two divisions fine-tune the series together as well as estimate the final retail sales taxes applicable to the estimates. A crucial element for fine-tuning the benchmark estimates is the expertise and the economic intelligence, developed in the estimation process for the current period by the officers of the Income and Expenditure Accounts Division.

STATISTICAL ESTIMATES, CONSTANT PRICES

For all the years for which Input-Output tables are completed, identical totals in current prices for all the series except operating surplus, VPC and the error of estimate exist. However, results in constant prices are not identical. There is no definitive way of estimating constant price level or growth except with reference to time and weights. Input-Output tables express values in producer prices whereas the Income and Expenditure Accounts express values, say for consumer expenditures, in purchaser prices. There is no guarantee that deflating detailed expenditures in producer prices for commodities and margins separately will give the same results as deflating composite commodities (including margins as in retail trade). The following table 5 for the latest benchmark year 1984 compares the main components of GDP in the two subsystems, the Input-Output (I-O) and Income and Expenditure Accounts (I.E.). The series in constant prices are not very different but they are not identical either. i) Implicit deflators

The differences between the I-O based and I.E. based valuation series in constant prices were not forced and will not be forced to be identical (in spite of the fact the current price series are identical) under present plans. It is worth noting, however, that several constant price series underlying the above main components are developed jointly and are indeed identical; these include, for example, construction (part of investment), factor inputs (wages and salaries, supplementary labour income, depreciation which together make more than 60 percent of total government current expenditures) in the government sector. All other elements are closely monitored to detect any major deviations. Such deviations are further investigated and are corrected, if necessary, on the basis of new evidence. Table 6 on Comparison of Implicit Deflators of Final Demand Categories shows how close such deflators are in the two components.

Statistics Canada - 316 - Collected Articles of Kishori Lal

TABLE 5 GDP AT MARKET PRICES 1984, $ MILLIONS

At Current Prices At 1981 Prices IO I.E. IO I.E.

Consumer Expenditures 252,030 252,030 206,616 205,986 Govt. Current Expenditures 89,989 89,989 72,198 72,491 Fixed Capital 84,700 84,700 77,212 77,328 Exports 126,334 126,334 119,283 119,846 Less Imports 110,163 110,163 101,049 100,249 Value of Physical Change in 2,714 4,322 2,436 3,710 Inventories Statistical Discrepancy - - 1,608 - - 1,357

TOTAL 445,604 445,604 376,696 377,755

Note that the Input-Output valuation of individual commodities comprises producer prices plus separate delineation of trade, transport and tax margins. On the other hand, I.E. (and CPI) valuation is at purchaser prices (which includes margins). In spite of this, the above series are almost identical. The price indexes for individual commodities (there are 600) which underlay the above deflators for final demand categories in the Input-Output system were, with very few exceptions, the same for intermediate use and production. Applying these price indexes and the double deflation methodology, Input-Output calculated constant price gross domestic product by industry. Constant price GDP by industry thus remains theoretically and statistically consistent with the final demand estimation in constant price. ii) Real GDP by Industry

The statistical differences for the total GDP and even more important for detailed industries were quite significant. The following two tables, table 7 on Gross Domestic Product at Factor Cost by Goods Versus Services Producing Industries (part A shows values and part B shows indices) and table 8 on 1984 GDP at Factor Cost by Industry illustrate the differences in the old and the new series.

The new series are based on benchmark methodology (see later section on methodology) and the old series were based on monthly projectors. The profile of the Canadian economy portrayed by the benchmark based series is quite a bit different from the one based on monthly projected estimates. For example, the share of goods producing industries in the total economy was 38.3% in the old series and 39.8% in the

Statistics Canada - 317 - Collected Articles of Kishori Lal

Table 6 COMPARISON OF IMPLICIT DEFLATORS OF FINAL DEMAND CATEGORIES BETWEEN INPUT-OUTPUT AND INCOME AND EXPENDITURES ACCOUNTS

Machinery and Consumer Construction Construction Equipment Expenditures Business Government Business I/O I.E. CPI I/O I.E. I/O I.E. I/O I.E.

1981 100.0 100.0 100.0 100.0 100.0 100.0 100.0 100.0 100.0 1982 110.4 110.2 110.8 105.4 105.4 106.9 107.3 107.9 106.8 1983 116.6 117.2 117.2 106.6 106.5 109.2 110.7 105.4 106.0 1984 122.0 122.4 122.3 110.8 110.6 113.6 114.5 106.7 106.2

Machinery and Net Equipment Government Government Exports Imports Expenditures Total GDP I/O I.E. I/O I.E. I/O I.E. I/O I.E. I/O I.E.

1981 100.0 100.0 100.0 100.0 100.0 100.0 100.0 100.0 100.0 100.0 1982 107.3 105.8 102.5 102.0 104.3 104.7 112.0 111.7 109.1 108.7 1983 105.2 105.2 103.5 102.6 104.4 104.4 119.0 118.4 114.3 114.1 1984 109.1 109.1 105.9 105.4 109.0 109.9 124.6 124.1 118.3 118.0

new series for the year 1984. The growth of goods producing industries was only 0.6% over the three years 1981-84; in contrast, the benchmark series show a growth of 5.8% over the same period. This very significant difference almost all occurred between 1983 and 1984. The new series show a growth over the three year period of almost equal rate for goods and services producing industries whereas the old series were showing hardly any growth for goods producing and over 6 percent for the services producing industries. The role of benchmarks for economic series becomes crucial in an economy going through cyclical movements.

Statistics Canada - 318 - Collected Articles of Kishori Lal

TABLE 7 GROSS DOMESTIC PRODUCT AT FACTOR COST BY GOODS VERSUS SERVICES PRODUCING INDUSTRIES IN 1981 PRICES, 1981-84

(A) $ MILLIONS Old New

Goods Services Goods Services Industries Industries Total Industries Industries Total

1981 126,991 192,547 319,538 126,991 192,547 319,538 1982 117,291 190,230 307,521 118,682 189,181 307,863 1983 121,853 195,348 317,201 123,607 194,251 317,858 1984 127,735 205,539 333,274 134,319 202,762 337,081

(B) INDICES (1981 = 100) Old New

Goods Services Goods Services Industries Industries Total Industries Industries Total

1981 100.0 100.0 100.0 100.0 100.0 100.0 1982 92.4 98.8 96.2 93.5 98.3 96.3 1983 96.0 101.5 99.3 97.3 100.9 99.5 1984 100.6 106.7 104.3 105.8 105.3 105.5

Table 8 provides comparisons of GDP by Industry at divisional level for the year 1984. Though the difference at the overall economy level was only 1.1% the difference at the industry level was quite significant. Even at the total goods versus total services producing industries level, the new series were lower by 1.3% for services and higher by 5.1% for goods. Note that these significant changes occurred over a short span of three year. Without benchmark adjustments, we would be giving a different portrayal of the economy. The industrial distribution in this table is at the divisional level without regard to sectoring into business and non-business industries iii) Methodological changes, Real Domestic Product by Industry

The most important methodological change was made in the integration of constant price series of GDP at factor cost by Industry for the business sector of the Canadian economy. The Input-Output division used to produce GDP by Industry in constant prices using double deflation methodology whereas the Industry Measures and Analysis division used to produce such estimates using a variety of approaches including double deflation. In the double deflation approach, one deflates commodity

Statistics Canada - 319 - Collected Articles of Kishori Lal

TABLE 8 1984 GROSS DOMESTIC PRODUCT AT FACTOR COST BY INDUSTRY IN 1981 PRICES, $ MILLIONS

Comparison Ratio Old New (New ) Old X 100)

Agriculture 10,181 10,677 104.87 Fishing & Trapping 469 536 114.29 Logging & Forestry 2,366 2,695 113.91 Mining, Quarrying & Oil Wells 19,029 20,753 109.06 Manufacturing 62,262 65,979 105.97 Construction 23,043 23,461 101.81 Transportation & Storage 16,703 16,924 101.32 Communication Industries 9,347 9,243 98.89 Other Utility Industries 10,672 10,462 98.03 Wholesale Trade 15,441 16,295 105.53 Retail Trade 20,800 21,363 102.71 Finance, Insurance & Real/Estate 49,005 46,631 95.16 Govt. Service Industries 22,933 23,043 100.48 Community Business & Personal 71,023 69,019 97.18 Services

Total Economy 333,274 337,081 101.14 of which Goods Producing Industries 127,735 134,319 105.15 Services Producing Industries 205,539 202,762 98.65

outputs of an industry and its intermediate use of commodity inputs, and the difference between the two deflated values equals GDP in constant prices. One may also produce constant price GDP estimates by industry by inferring them from the quantity measures of gross production or constant price intermediate inputs or by using employment as a proxy as well as by using double deflation wherever feasible. This was the situation in the Industry Measures and Analysis division in the period before integration of annual estimates. Now the Real Domestic Product by Industry is produced using double deflation for all the benchmark years for the annual industry statistics program. However, for the latest two years which do not yet have annual I-O tables, annual totals for RDP by Industry are derived summing monthly estimates produced using a variety of approaches as noted above.

Statistics Canada - 320 - Collected Articles of Kishori Lal

The double deflation approach is the only one which satisfies the requirement of an identity between GDP income based and GDP expenditure based in constant prices. However, there are certain important hazards using double deflation for deriving GDP of an industry whose value added makes a small (say 10%) proportion of total gross output. For such an industry estimate of GDP using double deflation could become very erratic and volatile. A small shift of relative prices for intermediate inputs and gross output will evidently translate into a very big change for value added. A solution to this problem, however, may simply be to consolidate industries with small ratios of value added to gross output with related industries at earlier or later stages of production. In other words, the problem of instability may be solved by aggregating into larger units whose value added are large enough in relation to gross output not to be too sensitive to the effects of changes in prices or technology.

In the 1982-84 benchmarks, for example, two industries, Motor Vehicle, and Refined Petroleum and Coal Products had very erratic value added estimates from the double deflation methodology. GDP in motor vehicle industry was aggregated to "motor vehicle parts" industries and this sub-aggregate was then redistributed to the two industries using primarily gross output as a proxy. Similarly Petroleum Refined Products industry was combined with Crude Petroleum and Natural Gas and again, the combined GDP was redistributed using primarily gross output as a proxy. The Canadian SNA approach of using double deflation with the above modification for specific industries is, in our judgement, a useful compromise.

The methodology underlying the monthly measures is not based on the preferred double deflation approach, for the simple reason that the required database does not exist. Monthly RDP uses projections based predominantly on gross production, (75% of the cases) and employees (20% of the cases). These projected series of course are step-adjusted when I-O benchmarks are available.

Another difference used to be in the classification of an industry. Industry Measures and Analysis division used to classify the entire 3 digit industry to either business or non business sector whereas the other SNA (I-O and I.E.) and Labour Division used to classify each establishment to a particular sector. Now the entire SNA uses the same sector classification.

It is important to note that the professionals of the Industry Measures and Analysis division brought forth their expertise and the economic intelligence, developed in the process of the current period, for jointly fine tuning the benchmark estimates. Without this effort, benchmark estimates would be of a lower quality.

CSNA DOCUMENTATION

Absence of full documentation of sources and methods, classification systems, valuation rules, revision practices and other judgements made in the preparation of any large scale and interlocking statistical database is always an impediment to the integration and reconciliation initiatives. CSNA is such a large scale and interlocking database; some people call it an economic labyrinth. As mentioned in the introductory paragraphs of this paper, final output of one CSNA component

Statistics Canada - 321 - Collected Articles of Kishori Lal becomes the starting point of the next component and the final output of the second component locks into the third one.

It is planned to have several documents prepared over the next while. The first and in some sense the most important one is the document describing the interconnections between the four components; how they hang together is the most important feature of this document. This document entitled AThe Canadian System of National Accounts or The Economic Labyrinth - A User Guide@ is planned to be completed by the end of 1987. This document is a guide to the various components of the CSNA, describing frameworks, major concepts, definitions and purpose of each component. It draws attention to accepted national accounting conventions and to the links that exist between the components of the system and explains how each provides a statement about an important aspect of the Canadian economy. The central theme carried by the report is the integrated nature of the overall system not only conceptually but also statistically. The recurring theme of the integrated nature of the system of accounts is a reflection of the high priority attached by Statistics Canada to improving the statistical linkage between the system's components.

This first report is intended to be a forerunner in a series of reports planned by Statistics Canada in the field of national accounting. It provides an overview of the entire CSNA - a sort of umbrella report describing the framework, general concepts, component links and uses of the accounts. Subsequently it is intended to publish separate reports on major individual components. These will emphasize the detail underlying the sources of data used and methodology employed in compiling the accounts.

INTEGRATION OF DATA PROCESSING IN THE CSNA

(i) Background

Data processing in the Canadian System of National Accounts has matured over the last two decades from a situation where each separate organizational entity had its unique and separate data processing operation to one of increasing logical and physical integration. A first requirement to automated integration of the accounts was, of course, that each component be automated with its data in machine processable form. This happened unevenly but has been accomplished over time. The needs for sharing working level data among the groups in the system of National Accounts were small and accomplished in an ad hoc manner (including computer tape, cards and paper).

In the late seventies there was a major trend in the data processing industry (and in the System of National Accounts) towards on-line systems as a means of increasing productivity and improving timeliness. A trend away from expensive computer service bureaux toward mini computer solutions began at that time. At the same time, newer versions of software allowing the end-users to build models and initiate computer operations with less and less reliance on scarce system development staff were designed.

Statistics Canada - 322 - Collected Articles of Kishori Lal

(ii) Current situation

There are three processors currently involved in the operation of the System of National Accounts. The IBM 3090 is the major Statistics Canada processor. Industry Measures and Analysis, and International and Financial Economics Divisions perform the majority of their processing on this central service.

The HP 3000 mini computer is housed in the Income and Expenditure Accounts Division to provide efficient time shared access and 24 hour operation for this division. It is linked electronically to the central IBM Processor.

A VAX 9600, centrally shared service, houses the structural accounts (Input-Output, productivity etc) using a sophisticated database and matrix language specific to the needs of these users. It is also linked electronically to the central IBM service.

Micro computers are becoming an increasingly effective part of our total service to professional and support staff. They are replacing paper spreadsheets and performing increasingly complex operations.

Our needs to share working level data have grown and faster, more mechanized tools (telecommunications) have been used but often still in an ad hoc manner.

(iii) Challenges for the future

Increasing emphasis on logical integration of all sectors of the National Accounts, the redevelopment of survey and census systems and the introduction of new classification systems (i.e. the harmonized commodity classification) and, of course, the ever changing available technology present challenges and opportunities for the future.

a) Higher speed data links and a direct link between the HP 3000 service and the VAX service will likely be needed in the near future.

b) Development of a concept of a common data dictionary for implementation on each of our three services is considered a viable option for better sharing of intelligence of our data resources among the various operational areas.

c) Micro to mainframe links will be further investigated to continue our effort at more effective data migration with these devices.

d) Opportunities to fully physically integrate our processing will be sought but not as a priority item. As our current suite of software becomes outdated due to changing demands for service, we will investigate the option of moving to one common processor to satisfy our diverse user needs. In the meantime our strategy will be to fully exploit the hardware and

Statistics Canada - 323 - Collected Articles of Kishori Lal

software in place and provide more intelligent networking and data sharing among all our devices.

INTEGRATION AND TIMELINESS

An environment which nurtures and fosters an integration and reconciliation of statistical series underlying a vast and interlocking database such as the CSNA routinely and rightly expects a higher quality of final product(s). However, it must also be emphasized that in the initial stages of its implementation, there is a possibility of a small degradation of timeliness of release of data. The CSNA management implemented the full integration with effect from the first quarter GDP release in 1987. This being the first time that the integration was in full vigour in the production mode, the release of the first quarter GDP estimates was delayed by a month. This delay was partly the result of the requirement of incorporating benchmark data for an additional year (a regular revision cycle would include four years' revision, but this time it spanned over five years); this situation would not arise next year. Also, inconsistent valuation of certain important export commodities forced the SNA and the feeder system professionals to reexamine the data. Although it took some time, this reexamination of data is the raison d'être of an integrated system. In the process of confronting and balancing the myriad series in the framework of SNA, inconsistencies are revealed which have to be resolved, fundamentally by going back to the feeder series and at times by using the expertise of the SNA professionals to adjust such series.

It is our judgement that over time not only will the quality improve but the previous timeliness will not suffer either. Integration like methodology is an investment. Pay-off of this investment must be judged over a reasonable time horizon.

CONCLUDING REMARKS

Integration requires a strong and long term commitment by the senior management, a willing cooperation by the various parts of the CSNA and a dedication to develop and implement a common set of classification and aggregation rules as well as consistent revision practices to be used for all the underlying statistical series both in the CSNA and the feeder system. Furthermore, integration requires documentation of the above as well as of the links that exist between the components of the CSNA. Sources and methods of each of the components link the overall framework for the entire SNA with the data supplying sources (feeder system). SNA or CSNA cannot fulfil its role or its potential unless it brings to the attention of the feeder system the imbalances revealed through its compilation. Equally important, commitment and cooperation of the CSNA and feeder systems are required to resolve these identified problems. To be effective in the present environment of computerization, integration also requires that the software and hardware used for manipulating the database of the individual components are straight-forwardly linked. It is expected of the CSNA to develop a set of statistics which together produce an integrated, consistent and comprehensive profile of the economy as well as (and it is equally important) to upgrade the quality of the underlying statistical series from the feeder system.

Statistics Canada - 324 - Collected Articles of Kishori Lal

Reference d. Excepting minor editing, this is a reproduction from the Proceedings of the Second International Meeting on Problems of Compilation of Input-Output Tables, March 13-19, 1988, Baden near Vienna, Austria, titled Compilation of Input-Output Data, Vienna, 1989.

Statistics Canada - 325 - Collected Articles of Kishori Lal

The 1993 International System of National Accounts and The Canadian Input-Output Tables a

by Kishori Lal Statistics Canada

Abstract: In this paper, the Input-Output tables are situated within the Canadian System of National Accounts (CSNA). The rectangular format of the accounting framework of the Canadian Input- Output tables and the integration rules established by the CSNA management are examined for their role in helping to integrate the various components of the CSNA. Lastly, a critical evaluation is made of the recommendations presented in the 1993 International System of National Accounts in its chapter on Supply and Use Tables and Input-Output from the perspective of the Canadian Input- Output tables. Reasons are stated in cases where we may not be able to implement the 1993 SNA recommendations.

The framework of the Canadian System of National Accounts (CSNA) bears a close relationship to the one described both in the 1968 UN SNA manual and the 1993 International System of National Accounts (1993 SNA) manual. The CSNA, like the 1993 SNA, is both a comprehensive and an integrated system, in the sense that all aspects of the economy are included and all of its components are interrelated. The CSNA comprises production accounts, income and outlay accounts, capital finance accounts, and financial flow and balance sheet accounts. The balancing item in the production accounts, value added, is the starting point of the income and outlay accounts; that of the income and outlay accounts, saving, is the take-off point for the capital finance accounts; and the balancing item of the capital finance accounts, net lending, is the point of departure for the financial flow accounts. Through these linkages, the CSNA is integrated both conceptually and statistically.

The CSNA corresponds to the 1993 SNA in the important sense that input-output tables form an integral part of the production accounts. Statistics Canada has produced annually, starting with the reference year 1961, product by industry input-output tables (also referred to as make and use matrices) in both current and constant prices. Such tables are produced with a lag of two and half years to the reference year. The latest input-output tables for the year 1990 were published in the fall of 1993. These tables provide benchmarks to current price, quarterly and annual, income and expenditure based gross domestic product (GDP) accounts, constant price monthly real gross domestic product by industry, and current and constant price annual provincial GDP by industry. The Canadian input-output tables successfully form an integral part of the production accounts because of two very significant factors. One is the accounting framework of the tables and the other is the integration rules established by the CSNA management. Both of these are elaborated below; in addition, the series which are integrated in the CSNA are noted. The last section of this paper comments on the recommendations and guidelines presented in the 1993 SNA in its Chapter XV Supply and use tables and input-output. These guidelines are examined from the perspective of the Canadian input-output tables.

Statistics Canada - 327 - Collected Articles of Kishori Lal

A. The accounting framework

The Canadian input-output tables are very similar to the ones described in the 1968 UN SNA. The outputs (make matrix) and inputs (use matrix) of industries are presented in separate tables; both intermediate inputs and outputs are classified by commodity (product). Commodities are clearly distinguished from industries, the number of commodities exceeding the number of industries: thus the Canadian input-output tables are rectangular rather than square. There are important advantages to the rectangular format over the square (or symmetric) format. As noted by Anne Carter and Wassily Leontief in their article "Goals for the Input-Output Data System in the Seventies", published in the US Department of Commerce Survey of Current Business July 1971, page 31: rectangular tables permit as much detail as is available in the basic census or survey records, and the meaning of each entry is straightforward because observed transactions are not combined with fictitious transfers (a feature of symmetric tables). Canada has played a pioneering role in both the theoretical and empirical development of rectangular input-output tables, in Statistics Canada and other Canadian institutions.

The input-output tables for Canada contain two sets of interrelated accounts: the commodity accounts and the industry accounts. The former details the supply and disposition of individual commodities (goods and services), the latter the commodity composition of the output of industries and their complete costs of production (including surplus). These accounts are described below in the schematic chart of the accounting framework of the Canadian input-output tables. While the industry and commodity accounts also treat primary inputs and final demand, these entries are described in a separate section.

1. Commodity accounts

As noted, the commodity accounts display the supply and disposition of each commodity. Matrix V (in the schematic chart) shows the value of production of each commodity produced by each industry. The disposition of commodities by the various classes of transactors are shown in matrices U and F. Matrix U contains the use, on current account, by each industry of each commodity as an intermediate input in the production of other commodities.

Matrix F contains the demand for each commodity by final demand category: personal expenditure, fixed capital formation, additions to (the value of physical change in) inventories, gross government expenditure on goods and services, and exports. The three other columns in matrix F are imports, withdrawals from (the value of physical change in) inventories and government revenue from the sale of goods and services. These three categories supplement the supply of goods and services produced by Canadian industries in the period of account. It should be noted, however, that small amounts of goods and services are produced within the government sector and the production costs of these goods and services are already included in gross government expenditure. Matrices U and F include the use of commodities originating either as imports, withdrawals from inventories, or government-produced goods and services. Three adjustments are thus required to effect equality with the output of domestic industries. They concern imports, because they are extraneous to the production of Canadian industries, withdrawals from inventories, because these were produced in an

Statistics Canada - 328 - Collected Articles of Kishori Lal earlier period, and government revenues from sales of goods and services, because the costs have already been fully accounted for in government gross expenditure.

2. Industry accounts

The total output of each industry, classified by commodity, is shown in the industry's row of matrix V. The inputs of each industry are shown in the industry's column, which covers both matrix U for intermediate inputs and matrix YI for primary inputs. Matrices U and V were discussed above in the context of the commodity accounts. The primary inputs in matrix YI consist of net indirect taxes (commodity indirect taxes, plus other indirect taxes, less subsidies) and GDP at factor cost (wages and salaries, supplementary labour income, net income of unincorporated business and other operating surplus). For each industry total output equals total input, intermediate and primary.

3. Primary inputs and final demand

There is one matrix not yet described, YF. It contains primary inputs purchased directly by final demand categories. Labour income in matrix YF represents the wages, salaries and supplementary labour income paid to persons employed in the personal and government sectors. This includes the employees of non-profit institutions, such as universities and labour unions, and the income of domestic servants and baby sitters in the personal sector, and all public servants in the government sector. Surplus in matrix YF reflects the consumption of fixed capital in the government sector and the non-profit institutions classified to the personal sector. Assets which are charged to fixed capital formation in these sectors are depreciated, such as buildings, roads, equipment, etc. All residential housing, even when owner-occupied, is classified to business, and thus depreciation on housing is part of the surplus of the appropriate industry in matrix YI, not YF.

Statistics Canada - 329 - Collected Articles of Kishori Lal

Chart 1 The Accounting Framework of Rectangular Input-Output Tables

Final Demand Commodities Industries Categories Total

Commodities U F q

Industries V g

Primary Inputs Y1 YF n

Statistics Canada - 330 - Collected Articles of Kishori Lal

4. Economic identities

The following identities emerge from the above matrices:

i) The total outputs of industries equal the total intermediate inputs of industries plus their total primary inputs. ii) The total outputs of commodities of industries equal the total intermediate inputs plus the total commodity inputs into the final demand categories. iii) Intermediate inputs being common, primary inputs of industries equal commodity inputs of final demand categories. Adding primary inputs of final demand categories to both sides, we get:

(a) primary inputs of industries and final demand categories equal to, (b) commodity inputs of final demand categories plus primary inputs of final demand categories. (a) is called gross domestic product at market prices and (b) is called expenditure on gross domestic product at market prices.

This is true for both current price and constant price tables.

Finally, gross domestic product (GDP) at factor cost (wages and salaries, supplementary labour income, net income of unincorporated business and other operating surplus) equals GDP at market prices less net indirect taxes (commodity indirect taxes plus other indirect taxes less subsidies). The CSNA term of commodity indirect taxes and other indirect taxes are called taxes on products and taxes on production respectively in the 1993 SNA.

B. Integration rules

Integration and reconciliation of the System's components require the resolution of several important issues such as valuation, a common set of classification and aggregation rules, as well as consistent revision practices to be used for all the underlying statistical series in the CSNA. An equally important consideration is the documentation of data sources and methods and the establishment of

Statistics Canada - 331 - Collected Articles of Kishori Lal links between the various components. But integration should not be equated with forcing or producing a single set of numbers. If two independent methodologies produce two sets, let them both stand. Suppressing one set will serve integration poorly. We will deal below with various integration issues that have been faced in Statistics Canada in integrating the SNA series. Within the CSNA management, there are three separate divisions, each headed by a Director, which produces some parts of the production accounts. The Input-Output Division produces input-output tables and annual productivity measures; the Industry Measures and Analysis Division produces monthly real gross domestic product by industry and annual provincial GDP by industry; and the National Accounts and Environment Division produces quarterly and annual income and expenditure accounts.

5. Valuation

Input-output tables in current prices are initially balanced in purchasers' prices, which are then converted into approximate basic prices, to have a consistent valuation for all users. Consistent valuation is crucial for relating use to supply of each commodity and more particularly, for calculating constant price GDP through double deflation. In the income and expenditure accounts, all expenditure series are valued in purchasers' prices because source data record only such prices. Deflation is done for the expenditure categories using such deflators as the CPI (consumer price index). Valuation is even more important for imports. In the input-output tables, imports are valued at c.i.f. (cost of insurance and freight) plus import duties to make them consistent with the domestic producers' values. In the income and expenditure accounts, values of individual import commodities do not include duties and indeed are at f.o.b. (free on board) values; however, these differences are clearly delineated. It is important that all the subsystems of the CSNA delineate the various values in order to reconcile series using different valuation practices. The series in the input- output tables are delineated such that one can move from one valuation to another without difficulty.

6. Classification

Fundamental to integration and reconciliation initiatives is a consistent classification of transactions and transactors. Before the 1986 CSNA historical revisions, the input-output component and its derivatives used an inconsistent sector classification and aggregation of industries. The sector classification - there are three domestic sectors in the economy: business, government, and persons or households - used by the input-output tables and the income and expenditure accounts was identical. However, such was not the case between the input-output and productivity measures or between the input-output and monthly GDP measures. For example, the industry "education" was allocated to all three sectors - business, government, and persons - in the input-output and income and expenditure accounts, but for productivity measures and monthly GDP it was entirely allocated to the non-business (government and persons) sectors. The income and expenditure accounts used major SIC (standard industrial classification) divisions as aggregates, whereas input-output aggregated three - and four-digit SIC industries into 191 industry classes. The Industry Measures and Analysis Division used still another aggregation for the monthly measures. The aggregation rules adopted by the components of the CSNA were not hierarchically related. One result was that

Statistics Canada - 332 - Collected Articles of Kishori Lal monthly measures of GDP were not serving their very important role as forerunners of annual GDP estimates by industry derived from the input-output tables.

Moreover, the various components of the CSNA were not treating transactions consistently. For example, contributions by the government to, say, the urban transit system were considered operating subsidies in one component and grants in other components, with the result that estimates of GDP at factor cost were dramatically different for this industry amongst the various components. This type of inconsistent treatment has now ceased.

With the 1986 historical revisions, common SNA industry codes have been developed and implemented for all CSNA integrated industry statistics series, be they national, provincial or sub-annual. A common set of codes, both for detailed classes and aggregates, is now used. This is an exceedingly important tool of integration.

7. Revision practices

Before the 1986 CSNA historical revisions, the three divisions of the CSNA which compile production accounts had their own specific revision practices. The National Accounts and Environment Division would typically revise data each June for the last four years whereas the Input-Output and Industry Measures and Analysis Divisions often would revise data for longer time spans. The result was that, at any given time, the finality of the series was never consistent amongst the various components.

The following rules and procedures have been put in place starting with the 1987 CSNA revision cycle, with an example from the revision cycle in 1992: a) Annual benchmark estimates at current purchasers' prices:

The final annual estimates for the year 1988 and almost final annual estimates for the year 1989 at current purchasers' prices, produced through the input-output accounting framework, were produced in May 1992 for use by the national income and expenditure accounts annual revision in June 1992. Note that final annual estimates have a specific meaning in our context; such estimates will not be reexamined till the next historical revisions, usually conducted after every five or ten years. b) Income and expenditure accounts estimates

The National Accounts and Environment Division released the results of the annual revision with the first quarter estimates in June, 1992. In the second, third and fourth quarter releases, only estimates for the current year are subject to revision; i.e. at the time of the second quarter estimates, the first quarter could be revised; at the time of the third quarter estimates, the first and second quarters could be revised etc. The number of years for an annual revision is normally four years. c) Annual benchmark estimates at constant producers' prices:

Statistics Canada - 333 - Collected Articles of Kishori Lal

The Input-Output Division converted the current purchasers' price estimates into current producers' price and, then, constant price valuation by the end of July, 1992. As noted, the Input-Output Division produced current purchaser price annual estimates in May 1992. Between May and early June, some fine-tuning of the input-output estimates was done in concert with the National Accounts and Environment Division. By the first week of June, the two divisions finalized the estimates for the two years 1988 and 1989, such that total GDP at market prices and all components (except value of physical change in inventories and operating surplus) of GDP were identical in the two sub-systems. These estimates were converted, in July 1992, into producers' prices and, then, constant prices. The term producer prices in the CSNA input output tables is similar to approximate basic prices in the 1968 UN SNA or basic prices in the 1993 SNA. d) Monthly Real GDP at factor cost:

Constant price final estimates for the years 1988 and almost final estimates for the year 1989 became the annual benchmark estimates for monthly real gross domestic product by industry. The Industry Measures and Analysis Division released, at the end of August, 1992, benchmarked monthly estimates for the years 1988 and 1989, and projected estimates for 1990 and 1991, as well as for the first six months of 1992. In the monthly releases for July 1992 to May 1993, estimates for 1991 or earlier years are not subject to revision. e) Provincial GDP by industry

We do not have, as of May 1994, a full coverage for the provincial GDP by industry program. About 90% of GDP by industry is covered at present and, by October 1994, we are hoping to have a full coverage. In any case, these estimates are fully consistent with those produced by national input- output tables and real GDP produced through the monthly program. f) Productivity measures:

The revision practices for productivity measures are now completely consistent with the annual constant price GDP by industry produced through the input-output program and the current period's estimate of the real annual GDP produced through the monthly program. g) Revision practices in the feeder system

There are no comprehensive rules, procedures or policies affecting in a consistent manner the revision practices for the series produced in the feeder system (the supplying divisions to the SNA). Some series (for example, consumer price indexes) are never revised; some (for example, census of manufacturers) remain open for revision for one year; others have varying practices. In the process of compilation of the national accounts, certain (sometimes major) imbalances arise which are resolved by adjusting the basic sources and/or by macro adjustments. This resolution is quite often not incorporated into the series of the feeder system. Thus the link between the SNA revisions and the underlying series produced and published in the feeder system is not maintained.

Statistics Canada - 334 - Collected Articles of Kishori Lal

8. CSNA documentation

The absence of full documentation of sources and methods, classification systems, valuation rules, revision practices and other judgements made in the preparation of any large scale and interlocking statistical data base - such as the CSNA - is always an impediment to integration and reconciliation initiatives. As mentioned in the introductory paragraph, the final output of one CSNA component becomes the starting point of the next component and the final output of the second component locks into the third one. It is planned to prepare several documents over the next while. Some documents have already been published. The first, and perhaps the most important one, entitled A User Guide to the Canadian System of National Accounts was published in 1989. This document is a guide to the various components of the CSNA, describing frameworks, major concepts, definitions and purpose of each component. It draws attention to accepted national accounting conventions and to the links between the components of the System, and it explains how each component provides a statement about an important aspect of the Canadian economy. The central theme of the report is the integrated nature of the overall System, not only conceptually but also statistically. This is a reflection of the high priority attached by Statistics Canada to improving the linkage between the System's components.

9. Statistical integration, present situation a) Current prices:

The statistical estimates in current prices in the following series are now identical, both in total and in detail:

Income side: wages and salaries; supplementary labour income; net income of unincorporated business; indirect taxes and subsidies.

Expenditure side: consumer expenditures; government current expenditures; fixed capital formation; exports and imports.

In addition, total GDP at market prices is now identical for every year from 1980 onwards.

What is not identical on the income side of GDP is the operating surplus in the input-output accounts and the equivalent concept of the total of profits, investment income, inventory valuation adjustment and capital consumption allowances in the income and expenditure accounts. Value of physical changes (VPC) in inventories, part of the expenditure side of the accounts, is not identical in the two subsystems. It is important to emphasize, however, that these two series - surplus and VPC - are closely monitored and analysed by the concerned divisions before the estimates are finalized.

Statistics Canada - 335 - Collected Articles of Kishori Lal b) Constant prices

As noted above, we have identical totals for all series except operating surplus, profits, investment income etc, VPC and error of estimate in current prices. However, we do not produce the same results in constant prices. It is our policy not to force an identical constant price total, even if current price total is identical, for the simple reason that there is no definitive way of estimating "volume" or "quantum" except with reference to time and weights. The input-output tables express values in producers' prices whereas income and expenditure accounts express values, say for consumer expenditures, in purchasers' prices. There is no guarantee that deflating detailed expenditures in producers' prices for commodities and margins separately will give the same results as deflating composite commodities (including margins as in retail trade). The two series are usually not very different but they are rarely identical. There is another constraint in the income and expenditure accounts namely, that the current quarterly and annual series do not have the same data base and data sources as the annual input-output tables. The annual input-output tables are produced with a lag of two and a half years to the reference year and use detailed annual census/surveys not yet available for the current period.

C. The 1993 SNA

The 1993 International System of National Accounts (1993 SNA) includes a chapter (Chapter XV) on Supply and use tables and input-output. Our comments in this section are made with reference to this chapter.

"The System includes an integrated set of supply and use tables, or matrices, as well as symmetric input-output tables, or matrices. They provide a detailed analysis of the process of production and the use of goods and services (products) and the income generated in that production. The concepts and definitions in the input-output tables of the SNA are the same as in the rest of the System" (paragraph 15.1). As in the 1968 UN SNA, "the integration of the input-output in the overall system of national accounts is an important feature of the SNA" (paragraph 15.2). We fully agree with the importance attached to the input-output tables in the 1993 SNA.

10. Basic and purchaser's prices

The 1993 SNA recognizes the following four components of the price paid by the purchaser of a product (paragraph 15.26):

Basic price of the product as output; Taxes on the product; Less subsidies on the product; Trade and transport margins in delivering the product to the purchaser.

The 1993 SNA defines the purchaser's price as "... the amount paid by the purchaser, excluding any deductible VAT or similar deductible tax, in order to take delivery of a unit of a good or service at the time and place required by the purchaser. The purchaser's price of a good includes any transport

Statistics Canada - 336 - Collected Articles of Kishori Lal charges paid separately by the purchaser to take delivery at the required time and place" (paragraph 15.28a).

The basic price is defined as "... the amount receivable by the producer from the purchaser for a unit of a good or service produced as output, minus any tax payable, and plus any subsidy receivable, on that unit as a consequence of its production or sale. It excludes any transport charges invoiced separately by the producer" (paragraph 15.28.c).

In the Canadian input-output tables, values in total at purchaser's price contain all the elements defined in the 1993 SNA. However, two elements - i) transport charges paid separately by the purchaser and ii) non-deductible taxes on products levied at the final stage of sale - are separately identified, such that the purchaser's price of any individual item in the CSNA input-output tables is lower by the value of these elements, compared with the 1993 SNA definition.

In the CSNA input-output tables, subsidies on products have not been articulated, such that the basic price and the purchaser's price are identical for products with no tax or trade and transport margins. Let us assume that a purchaser pays for a unit of a product $103 and that this value includes a product tax of $5 and a subsidy of $2. The basic price per the 1993 SNA is $100. Suppose that the tax of $5 is eliminated; the purchaser price will be $98 but the basic is still $100 per the 1993 SNA. However, in the CSNA, both the basic price and the purchaser's price is $100. This will need to be changed to conform to the 1993 SNA. GDP at purchasers' price (also called GDP at market prices) will not change but valuation at basic prices will change.

11. Valuation of product flows

a) Basic prices provide the most homogenous valuation for all the users. b) Basic prices are found most useful when a system of VAT or similar deductible tax is in operation. c) Basic prices record the amounts available to the producer.

"For exports and imports, the System adopts analogous price concepts: the free on board (f.o.b.) for exports and total imports and the cost insurance and freight (c.i.f.) price for detailed imports" (paragraph 15.35, emphasis mine). Note that the f.o.b. price is at the customs frontier of the exporting country both for exports and imports. Further, the 1993 SNA states: "The f.o.b. price is considered to be a special purchaser's price applied to flows of exports. ...The c.i.f. price is considered to be a basic price applied to flows of imports, equivalent to the basic price of a good or service produced by resident producers" (paragraph 15.36). This is a change from the 1968 UN SNA where "the protective duties form part of the basic value of imports" (paragraph 2.16 of the 1968 UN SNA). In contrast, in the CSNA input-output tables, the basic price of imports includes all import

Statistics Canada - 337 - Collected Articles of Kishori Lal duties. The valuation of imports in the CSNA input-output tables will need to be changed to conform to the 1993 SNA. This will not change GDP at purchasers' prices, but the valuation of individual imported products will exclude import duties. Also, there will be no need to list total import duties with the opposite sign at the intersection of the column vector of imports and the row vector of taxes on products; instead, the import duties will be articulated in the basic price use tables (both of intermediate consumption and final users), under each user of imports.

Data on detailed flows of imports from foreign trade statistics are most usually valued at c.i.f. prices. The c.i.f. price for detailed imports includes f.o.b. price for imports (valued at the customs frontier of the exporting country) plus the cost of insurance and freight between the frontier of the exporting country and the frontier of the importing country. Again, these services can be provided by both resident and non-resident producers of such services. In the balance of payments statistics, the System adopts f.o.b. price for valuation. One will need to add insurance and freight charges provided by both non-resident producers and resident producers to bring this valuation at c.i.f. price into the input-output tables at basic prices. The part charged by the non-resident producers should be removed from the total for services, as it is part of the c.i.f. price of imports of goods, such that the two together (goods and services) remain the same. But the insurance and freight charged by resident producers of the importing country to bring imports of goods from the frontier of the exporting country to the frontier of the importing country need to be properly delineated. In the CSNA input-output tables, the valuation of imports at c.i.f. prices ignores this component. The implementation of the newly recommended c.i.f. valuation will raise the total level of imports of goods by the amount of insurance and freight supplied by Canadian carriers and, at the same time, it

Statistics Canada - 338 - Collected Articles of Kishori Lal will require an imputation to raise the level of exports of services by an equal amount to balance the System. There will be no change in the net external balance in the exports and imports of goods and services, but the gross valuation will be higher at basic prices compared with the f.o.b. valuation in the balance of payments.

The 1993 SNA provides an alternative presentation which obviates the need to increase the gross flows of goods and services in the input-output tables at basic prices. Rather than imputing an addition to exports of insurance and freight services provided by resident carriers, the 1993 SNA recommends creating a new c.i.f./f.o.b. global adjustment on imports (see paragraph 15.69 and Table 15.1). This alternative presentation records imports and global adjustment as follows:

a) Imports of goods detailed by products are valued c.i.f.; b) All transport and insurance services on imports, provided by both resident and non-resident producers, which are included in the c.i.f. value of imports by products, are globally deducted. Thus, the total of imports of goods in the System is always recorded f.o.b. in the table; c) Those transport and insurance services on imports that are provided by non-resident producers, are recorded under imports of services.

Whether the provision of transport and insurance services to bring imports from the frontier of the exporting country to the frontier of the importing country by the resident producers is shown as an addition to exports of services or as a c.i.f./f.o.b. adjustment to imports, it requires additional information not at present collected in the CSNA.

12. Gross value added at factor cost

In the 1993 SNA, gross value added at factor cost "... is not recommended as a measure of value added in the System, since there are no observable prices such that output minus intermediate consumption equals gross value added directly ..." (paragraph 15.39). The measure recommended throughout the System is gross value added at basic prices, which is defined as "... output valued at basic prices less intermediate consumption valued at purchasers' prices" (paragraph 15.37a). It is noted, however, that gross value added at factor cost "... could be derived from gross value added at basic prices by subtracting other taxes less subsidies on production" (paragraph 15.39). It is further asserted: "Other taxes less subsidies on production, by definition, are in fact taxes or subsidies that cannot be eliminated from the prices of outputs and inputs. Therefore, gross value added at factor cost is essentially a measure of income and not output" (paragraph 15.39).

Information on other taxes less subsidies is no more difficult to obtain from industry records than information on intermediate consumption. Other taxes less subsidies cannot be eliminated from the prices of outputs and inputs but neither can compensation of employees, surplus or capital consumption. Gross value added at factor cost has been and remains a standard measure of output for analysis of productivity by industry. We do not see any logic in calling gross value added at basic prices an output measure, and gross value added at factor cost an income measure. Both are output measures, as GDP at purchasers' prices is an output measure in the System. Statistics Canada has

Statistics Canada - 339 - Collected Articles of Kishori Lal been producing and publishing GDP by Industry at factor cost for decades at both annual and monthly frequencies. In the CSNA input-output tables, the contribution of each industry to the aggregate gross domestic product at purchasers' prices is fully delineated in terms of the three prices - factor cost, basic prices and purchasers' prices. The flexibility of presentation in the CSNA may be emulated by other national statistical organizations.

13. Presentation of non-market producers

In the 1993 SNA, output of non-market producers (such as establishments belonging to general government or non-profit institutions) forms part of the supply tables and, similarly, their inputs form part of the Use tables (see paragraphs 15.62-67). In the CSNA input-output tables, non-market producers do not form part of either supply or use tables but, rather, their detailed inputs are directly allocated to final consumption (see the schematic chart, presented earlier). This is merely a presentational issue; however, the 1993 SNA's presentation is better and the CSNA will need to conform to it. When implemented, an additional benefit will occur, as there will be no need to have a separate column of revenues from the sale of goods and services in the final demand categories, and the value for government final consumption will be equal to the value noted in other parts of the CSNA.

14. Gross fixed capital formation, stocks of fixed assets and labour inputs

The 1993 SNA suggests that at least three rows be appended, as supplementary information, to the value added quadrant of the use tables, one each for gross fixed capital, stocks of fixed assets and labour inputs, classified by the same industry detail as the use tables (see paragraphs 15.96-105). Such a presentation will facilitate productivity analysis and provide useful additional information to a broad range of users.

15. Cross-classification of value added by sectors and industries

Statistics Canada - 340 - Collected Articles of Kishori Lal

16. Conversion of the supply and use tables into symmetric tables

Supply and use tables are typically rectangular in the sense that the number of products (commodities) is greater than the number of industries. "In the supply and use table, when the number of rows of products and columns of industries happens to be equal, we shall refer to a square (not symmetric) supply and use table" (paragraph 15.2). Symmetric tables are tables which have an identical number of rows and columns for either industries or products and which use "... the same classifications or units ... in both rows and columns" (paragraph 15.2). Symmetric tables are always product by product or industry by industry, but not industry by product. There are important advantages to the rectangular format over the symmetric format (see section A, The accounting framework).

In the 1993 SNA, several procedures are listed (see paragraphs 15.137-143) to convert the rectangular supply and use tables into symmetric input-output tables. One procedure suggested for producing symmetric tables is transferring secondary products from the activities in which they are produced to the activities for which they are principal products, such that all "off-diagonal" entries in the supply table are eliminated. Transferring the corresponding inputs within the industries in the use table is properly noted as being more complicated, as the basic data on inputs relate to industries and not to particular products produced in those industries. To overcome this difficulty, it is suggested that supplementary information be used. In the CSNA input-output tables, only construction activity is transferred from industries where it is produced as a secondary product to the construction activity. All other secondary products remain as they are reported. One would need to aggregate products produced or used to the same number as the total number of activities to have symmetric tables. Failing the use of supplementary information to generate symmetric tables, the 1993 SNA recommends using a mechanical approach by adding certain analytical assumptions (see paragraphs 15.144-149). It notes that the mathematical methods used when transferring outputs and associated inputs hinge on two types of technology assumptions (paragraph 15.144):

a) industry (producer) technology, assuming that all products produced by an industry are produced with the same input structure; b) Product (commodity) technology, assuming that a product has the same input structure in whichever industry it is produced.

The 1993 SNA opts for the product technology assumption. We have serious reservations about this recommendation. The product technology assumption may be valid if one can develop a vector of inputs for each of the twenty thousand or so products identified in the market. It is, however, completely unrealistic to seek to achieve such a data base. Aggregating twenty thousand products into a manageable set of 200-300, or even 500-1000, product groups can hardly be called a replication of the commodity technology. At this level of aggregation of products or industries, there is hardly any difference between the two sets. Furthermore, it is evident that at least some important expenditures such as R&D, management skills and management style, which are industry specific or even firm specific, affect similarly all the products produced within an industry. The commodity technology assumption is partly a hangover from the perception that one needs the same physical things to produce a given product, no matter who produces it.

Statistics Canada - 341 - Collected Articles of Kishori Lal

17. Double deflation

We agree with the 1993 SNA that "... the supply and use tables are the most complete consistent framework for constant price estimation and provide: a) Interdependent measures of prices and volumes; b) An important check on the numerical consistency and reliability of the entire set of such measures, interlinking values at constant and current prices, value and volume indexes and deflators" (paragraph 15.161). "Constant price measures for gross value added are possible in the input-output framework by using the double deflation method, as the difference between: a) The value of output deflated by a price index of output; b) The value of intermediate consumption deflated by a price index for these inputs" (paragraph 15.162).

Statistics Canada - 342 - Collected Articles of Kishori Lal sensitive to the effects of changes in prices or technology" (p. 53). In one year, value added in one industry may be erratic but in the next, a different industry might suffer. Ad hoc aggregation into large units would disturb the continuity of time series. Thus, one needs additional guidelines. In the CSNA, we have solved this problem as follows: Values added are combined as suggested by the UN Constant Price Manual. The combined value added is then redistributed using gross output or any other indicator as a proxy, but the combined value added for a given sub-aggregate remains unchanged. Without this constraint, the above noted GDP identity requirement will not be satisfied. These comments on the UN Constant Price Manual apply equally to the 1993 SNA recommendation on double deflation.

Concluding remarks

The rectangular accounting framework of the input-output tables is one of the most important factors for enabling them to play an integrating role in the production accounts in the SNA. Integration requires a strong and long term commitment by senior management and, a dedication to developing and implementing a common set of classification and aggregation rules, valuation principles as well as consistent revision practices used for all underlying statistical series. Furthermore, integration requires documentation of the above as well as of links between the components of the production accounts. The 1993 SNA supports the rectangular format in the supply and use tables. It further recommends manipulating the supply and use tables into symmetric input-output tables using a commodity technology assumption: it is this recommendation about which we have serious reservations. The commodity technology assumption is partly a hangover from the perception that one needs the same physical things to produce a given product, no matter who produces it. Also, the commodity technology assumption needs square tables, which, in our judgement, are not as useful for integration purposes.

Reference e. An earlier draft of this paper was presented at the Tenth International Conference on Input- Output Techniques, Seville, Spain, March 29 to April 3, 1993. This paper was completed in May 1994.

Statistics Canada - 343 - Collected Articles of Kishori Lal

Evolution of the Canadian Input-Output Tables 1961 to date

by Kishori Lal, Statistics Canada

Introduction

The work on Input-output (IO) tables in Canada started in the early 1960's with a small research staff of 6 to 7 persons under the kind, competent and visionary leadership of Terry Gigantes , a pioneer in input-output analysis. Right from the very beginning, it was decided that IO tables must fulfill several roles and provide: (a) an audit and management tool to improve economic statistics for their consistency, accuracy and comprehensiveness; (b) benchmarks for Gross Domestic Product (GDP), its income side and components, its expenditures side and components and GDP by industry estimates, both at current prices and constant prices and (c) a framework for structural analysis. About the same time, the 1968 international System of National Accounts, under the leadership of Professor Richard Stone of Cambridge University England, was being developed. It had incorporated an input-output framework as the foundation of the production accounts of the SNA. The Canadian IO framework turned out to be very similar to the one developed in the 1968 SNA. The first Canadian IO tables (use and make matrices), with industry by commodity dimensions, were published at the same time as the 1968 international SNA manual was released.

Given the roles assigned to the IO tables, it became obvious that they must also reflect industry information contained in the records of the producing establishments. In the real world, establishments produce both principal products and secondary products. Thus, it was decided to show both principal products and secondary products for industries in the make matrix. This was an innovation and, in our judgement, a very important one: it was no longer necessary to remove the output of secondary products and their inputs from the industry records in the IO framework. Further, there are many more commodities in the economy than industries. To reflect this information in a straight-forward manner, IO tables must be rectangular rather than square: that is, to have many more commodities (goods and services) than industries. This was another innovation of the Canadian tables and it closely resembled the thrust if the 1968 SNA and also closely reflected the real world. However, showing the detail of commodity output for each industry resulted in large number of commodities becoming confidential under the Statistics Act of Canada. Thus it was decided to establish a separate unit in the IO division to be responsible for developing IO models based on our own parameters, as well as those provided by outside users for simulations on confidential data. The users would receive useful results, while we at Statistics Canada protected confidential data. For IO tables to become benchmarks for other SNA series, it became necessary to incorporate the SNA conventions on pricing, sectoring, classifications, etc. These early decisions have served us well and continue to be in effect.

Statistics Canada - 345 - Collected Articles of Kishori Lal

Brief History

There have been four, not fully separable, phases in the evolution of the Canadian annual IO tables. We do not yet produce any sub-annual tables yet. The annual tables are: current price national IO tables, constant price national IO tables, current price provincial IO tables, and experimental projected IO tables. Statistics Canada started developing annual national IO tables for the year 1961 early in the 1960's, and it completed them in 1968. Soon thereafter, funding was provided to produce annual IO tables at current prices. Beginning with the reference year 1961, 37 annual national IO tables at current prices have been completed, the latest for the year 1997.

Early in the 1970's, funding was procured to produce constant price IO tables. So far, the following tables have been produced: annual tables for 1961 to 1971 at 1961 base prices; tables for 1971 to 1981 at 1971 prices; tables for 1981 to 1986 at 1981 prices; tables for 1986 to 1992 at 1986 prices and tables for 1992 to date at 1992 prices. We do not plan to produce constant price tables for the entire period since 1961 at one price base. We derive from the foregoing tables constant price value added by industry, using our preferred double deflation method. These value added series by industry are then chained or linked but retaining the growth rates produced during the respective periods. Thus, the growth rate of value added by industry for the 1961 to 1971 period remains the same, no matter which year we use to express the dollar values for the entire period since 1961. The 1993 SNA strongly recommends determining growth rates of GDP, or its components, or by industry, using chain indices.

Four occasional provincial -there are 10 provinces and since April 1999, three territories (previously two) in Canada- IO tables for the years 1974, 1979, 1984 and 1990 have been produced. In addition to the make, use and final demand matrices, inter-provincial trade flows have also been produced. These tables have not been used as benchmarks for provincial SNA series and principally only for provincial IO models, mostly funded by cost recovery from outside users. This program, as of the reference year 1997, has now been funded to produce annual provincial tables. They will now play the same roles as national IO tables. In addition, the information produced in the provincial IO tables from 1997 onwards will be used by the Governments of Canada and three participating provinces for the distribution of goods and services tax. This tax in Canada is quite similar to the value added tax in Europe. The latest provincial IO tables, which are quite similar to the planned tables for 1997 and onwards, were completed for 1996 in November 1999. Tables for 1997 were released in November 2000. Every November after that date, we will have one more year of provincial IO tables.

Complete national IO tables are produced with a lag of 28 months from the reference year. For the current 28 months, Canadian SNA series, both quarterly and monthly, continue to be produced but their reconciliation with fully balanced IO tables and commodity flows must wait till IO tables are completed. Should we project the national IO tables to the current period, with a view to further improving the accuracy and consistency of the current period SNA series? Prompted by this question, attempts have been made during the last few years to project national IO tables, on an experimental basis, for the current period. By necessity, only partial information could be used as the more comprehensive information is not available yet. Starting with the latest fully balanced IO tables, one can use such information as the timely shipments, sales and employment data, and many

Statistics Canada - 346 - Collected Articles of Kishori Lal other series normally used to produce the SNA series for the current period. Note that in Canada, we produce a full set of income and expenditure accounts at both annual and quarterly frequencies, as well as an economy-wide monthly GDP at constant prices by industry. This program is experimental and no funding has yet been requested.

Completion schedule

Statistics Canada=s computer facilities were relatively limited in the 1960's. Consequently, it took us about six years to finish the first IO table for 1961 in 1968. The dimensions in both industry space and commodity space have increased since the 1960's, but the interval between the reference year and the year of completion has been steadily reduced. Today, the current price national IO tables are completed after 28 months of the reference year; the constant price IO tables are completed after 32 months of the reference year; and the annual provincial IO tables are completed 34 months after the reference year. These are full set of tables with complete accounting of commodity flows with 680 commodities, 250 industries and 140 categories of final demand. We started with about 6 to 7 persons in the 1960's and the staff now is more than 50 persons, including a half-dozen of computer specialists.

Computing Environment

The first attempts at programming for the 1961 tables were done with MATOP, a mainframe software package. MATOP was a rigid spreadsheet-like language in which variables could not be referred to by name and the dimensions of a matrix were severely limited. The first inverse was calculated for the 1961 tables using a Fortran program, which took much longer to calculate than inverses calculated by today=s computers.

During the 1970=s, contracts for computing were entered into with outside firms (Computel, Systems Dimensions Limited (SDL), and Canada Systems Group (CSG)). For the first outside contract, it was necessary to go to Computel=s office in order to submit jobs; all programs, JCL (Job Control Language) and data were on punch cards. Any jobs for submission to the Statistics Canada mainframe computer had to be taken to the Statistics Canada main computer job submission area, printouts would be available for pickup 24 hours later. The division had several cardpunch machines for staff to use for their programs. Then, an IBM 2741 typewriter terminal was obtained, and from it, jobs could be submitted and printouts received. Later, access to the CSG computer was by means of a REMCOM card reader-printer. In the late 1970=s, a Digital PDP 11/34 computer was obtained for the division, along with terminals and a line printer. This allowed jobs to be created and stored online and then submitted in card image format to the Statistics Canada mainframe computer or to the CSG mainframe computer; job results were printed on the line printer. As a result, card punch machines and card readers became obsolete.

The first versions of in-house matrix software (FINDEM and UTILE) were developed, in the early 1970=s, to create the input, output and final demand matrices for one year. The input matrix contained values at purchaser prices, the margins (there are seven margins-transportation, pipeline, gas distribution, storage, wholesale trade, retail trade, and tax on products) and producer values of all

Statistics Canada - 347 - Collected Articles of Kishori Lal the commodities in the system. The first limited simulations were also done at this time. Most of the calculations to prepare the tables were done on worksheets, with the aid of calculators. To obtain data from other divisions, staff would have to go to the divisions to copy the information on worksheets. The first industry and commodity balancing at purchaser prices was done on cards. Generalized job submission routines were developed to allow analysts to perform calculations using the in-house matrix software, but without having to know JCL. FINDEM was a Fortran routine that performed rudimentary matrix operations on a restricted number of variables. UTILE, an extension of FINDEM, was the first attempt to introduce operations on sparse matrices; the number of matrix operations was limited as well as the sizes of the matrices.

The successor to FINDEM and UTILE was MOP (Matrix Operations Package), which performed a broad range of matrix operations on sparse matrices. It was the first attempt to integrate a random- access directory containing characteristics of matrices to allow error scanning, and it eliminated the need for definition statements for each variable referenced. Some of the matrix operations which could be performed included addition, subtraction, multiplication, inner products, transposition, normalization, printing, aggregations. This allowed the calculation of inverses, impact and multiplier tables. JCL was still required to run these programs.

In 1984, Input-Output Division=s computing was done on an in-house Digital VAX 780 mid-range computer (later a VAX 8600), other divisions in Statistics Canada also had access to it. In 1990, the division acquired its own Digital VAX 3900, later replaced by a VAX 4200. Staff accessed these computers by means of terminals, with a large line printer for printouts. This was a major advance in computing for the division, as the staff gained direct access to the matrix software with instant results shown on the screen. With these computers came the first access to spreadsheet software, 20/20 and LOTUS, as well as to the word processor WordPerfect. Interfaces were developed to transfer data between the matrix software and the spreadsheet software. In 1994, the division acquired a Digital ALPHA 3000-600 S computer (followed by an ALPHA 4100), personal computers connected to the ALPHA by a Local Area Network (Digital Pathworks), PC software Lotus and WordPerfect, later Microsoft Excel and Microsoft Word, Laser Jet printers. These computers operate 24 hours a day, seven days a week. Each computer acquisition has resulted in a marked increase in processing speed, computer memory and disk storage. The processing speed of the 4100 is considerably more than 10 times faster than the 300-600 S computer. These computers currently handle 50 and more users with no effect on processing speeds.

In moving to the Digital VAX computers, a new interactive matrix software package, TERF (Terminal Entry Review Facility), was developed. TERF is a fully interpretive, interactive language which allows multiple dimensions, sparse matrix operations, looping, run-time variables for indexing, error checking, and inheritance of characteristics for resulting variables. Confidentiality routines were also developed, allowing confidentiality checks on the data for the standard aggregation levels, as well as for the special aggregations demanded by customers. Special programs were developed for transferring the published data to the Statistics Canada mainframe CANSIM database. TERF is constantly being refined and enhanced to meet current data processing needs. The earlier matrix software was limited to two dimensions (e.g. industries and commodities), this

Statistics Canada - 348 - Collected Articles of Kishori Lal progressed to three dimensions (e.g. years, industries and commodities) and, now, to four dimensions (e.g. years, provinces, industries and commodities).

Until the mid 1990=s, the industry and commodity updating/balancing at purchaser prices was done with printouts and hand updates. The analysts would have printouts of their industries and commodities and would mark their changes on the printouts. In the 1970=s, these updates were punched on cards prior to entering into the master files. In the 1980=s and early 1990=s, the changes were entered using a screen-entry updating program via the terminals; this was very time consuming as all updates were proof-read before entering into the master files. For the 1993-1995 tables, divisional systems staff developed a multi-user Graphical User Interface (GUI) for the industry/commodity updating/balancing for the national tables, using Visual Basic. This program accesses directly and instantly the master files for the inputs, outputs and final demand, as well as the margins and the inter-provincial trade flows, and the user can then enter updates and see immediately whether or not the industry or commodity is balanced. Some analytical tools are also included in this software package. The updates can be processed in a few minutes and then stored in the master files, and this can be done several times in one day. For the 1995 final and 1996 preliminary tables, the updating/balancing program was extended to include the provincial dimension for 1996. For the 1996-1997 tables, extra flexibility was added for processing two years of national and provincial data.

The Input-Output Division collects and processes large amounts of data from many sources, from both inside Statistics Canada and outside. The majority of the data is available in machine-readable form, in a wide variety of formats, and is transferred to the divisional computer and stored in matrix format or spreadsheet software. In the mid 1970=s, the annual Census of Manufacturing data were put into matrix format and stored on the divisional computer for processing by the analysts; this was the first of the mechanized data available to the analysts. A single dimension of this matrix now contains approximately 50,000 establishments for a single year.

Over the years, the working levels for industries and commodities have increased. For example, the number of industry working levels has increased from 191 to 216 to the current level of 243; it will increase further with the implementation of NAICS (North American Industry Classification System) for the 1997 IO tables and onwards. The number of working level commodities has increased from 602 to 627 to the current level of 679, and will increase further because of the implementation of NAICS. The IO matrices have changed from single year matrices, to time series, to time series with the provincial dimension. The first in-house matrix software had only a handful of matrices whose number of rows or columns exceeded 1,000 elements; now there are some matrices with approximately 1,000,000 elements in one dimension. The international trade data are processed at the Harmonized System (HS) of commodities (over 21,000) and concorded to the IO working level of commodities.

The national input-output model was first developed in the early 1970=s on cards using the FINDEM software. The first versions of the national model used the inverse method; this was changed to the iterative method. The inter-provincial model was first developed for 1966 provincial data, again using cards and the FINDEM software. For both models, calculations are done at the worksheet

Statistics Canada - 349 - Collected Articles of Kishori Lal level, and the results aggregated to the level requested by a client. A GUI front-end has been developed for the National Model; the calculations are performed by TERF and the results are available in seconds, in a spreadsheet, or printout.

The publication environment has also changed dramatically over the years. For example, for the 1961 tables, MATOP was used at Computel, the data were on cards and the printouts were used for the publication. In later years, the text was typed separately and the data had to be specially formatted for sending to publication by outside firms. Now, with the desktop publishing software, the whole publication can be prepared by IO staff, data can be prepared, transferred into Excel and then into the desktop publishing software.

Integration

Current price estimates of GDP as well as the underlying detail such as labour income, mixed income, taxes and subsidies on products and other surplus are identical in the IO tables and other SNA series produced at Statistics Canada. Similarly, expenditures on GDP and the underlying detail such as household expenditures, expenditures by the governments. Gross fixed capital formation, exports and imports are all identical throughout the Canadian SNA. In addition, the classification of categories of final demand, say for households, (there are more than 40 categories based on the international Classification of Individual Consumption by purpose- COICOP) is identical. This permits users to navigate from one set of statistics to another. This, as noted above, was thought very important right from the beginning of IO work at Statistics Canada.

Concluding Remarks

The IO program in Canada today is vastly different from the one we started in the 1960's. Its speed of delivery has increased even when its dimensions both in industry space and commodity space have greatly expanded. Its coverage is now many-fold, from one national table to 12 provincial/territorial tables. It has developed a proud history of annual tables from 1961, a rare statistical feat anywhere in the world. Its constant price version provides an excellent framework for structural analysis. We would not have been able to achieve this without the availability and ever increasing computer capacity. All this marvellous expansion occurred because of the importance that we have attached, right from the very beginning, to the role of IO tables in the framework of economic production statistics in the Canadian SNA program. We have been very conscious of acquiring and retaining a highly dedicated and professional staff, without which such an expansion could not be possible.

Note:

I am grateful to many colleagues in the System of National Accounts branch at Statistics Canada, particularly Yusuf Siddiqi and Abe Tarasofsky for their very useful comments and suggestions in the preparation of this paper. A special acknowledgement is due to Marilyn Constantineau of the IO Division in the preparation of the computer environment section of this paper. An earlier draft of this paper was presented at the International Input-Output Association Meeting, Macerata, Italy, August 2000

Statistics Canada - 350 - Collected Articles of Kishori Lal

Introduction

The Canadian System of National Accounts (CSNA) strives to produce high quality statistics that are relevant, accurate, timely, consistent, and easily interpretable by users, both at home and abroad. Such a system must evolve to meet changing user needs and reflect institutional changes in the Canadian economy. At the same time, it needs to be sufficiently stable to produce consistent statistics within a coherent conceptual framework.

Over the last few years, a large number of countries have made efforts to achieve comparability of their national accounts statistics by adopting the System of National Accounts 1993 (SNA93), released under the auspices of five international organizations: EUROSTAT, IMF, OECD, UN, and the World Bank. To the best of our knowledge, this is the first time that a single conceptual framework has been released under the aegis of five international organizations.

Typically, every five years or so, a major review is made of the CSNA, and major changes are recorded, documented and widely disseminated to our users. The 1997 Historical Revision of the CSNA was the most comprehensive review. More than 100 changes were made. Some were classificatory, many were conceptual and reflected the new SNA93, and others were methodological and statistical. All changes were recorded and justified. They were then widely disseminated both in Canada and abroad in our document called : The 1997 Historical Revision of the Canadian System of National Accounts, Record of Changes in Classification of Sectors and Transactions, Concepts and Methodology, October 1998.

Canada implemented SNA93 in 1997, and since then many other countries, such as United States, Australia and member countries of the European Union, have also implemented it. Over the next couple of years, most countries will also implement it, in accordance with their institutional structure and statistical development. The CSNA incorporated the 1993 SNA guidelines, but with some modifications. These were primarily prompted by pragmatic considerations, such as institutional structure, statistical data sources, availability of resources and their cost- effective use. We fully recognize the importance of promoting international comparability, but it should also be recognized that the specific circumstances existing at a given time in different countries can vary, often substantially. To make our users aware of how fully we implemented SNA93 in our accounts, it was decided to prepare a paper highlighting the remaining differences between CSNA and SNA93, thus providing a better understanding of the Canadian system vis-a-vis that of other countries. It needs to be emphasised that most of our departures affected the overall GDP only marginally and were primarily in the sector details. This report, containing some 35 issues, was also widely circulated both in Canada and abroad, under the heading: Remaining Differences between The 1997 Canadian System of National Accounts and The International System of National Accounts, September 1998.

With the implementation of SNA93 in many countries, it was realised that certain of its recommendations needed clarification to ensure common interpretation. Thus a process was established to update SNA93 as new issues emerged. An inter-Secretariat Working Group on National Accounts (ISWGNA), was established jointly by the five international organisations (EUROSTAT, IMF, OECD, United Nations, and World Bank), to issue clarifications on any

Statistics Canada - 353 - Collected Articles of Kishori Lal issue raised by national statistical agencies. ISGWNA has already issued several clarifications, which have proven very useful in keeping SNA93 up to date.

Learning from our own experience and that of others who have implemented SNA93, and from the clarifications issued by ISWGNA, we have revisited some of the changes made in 1997 in our CSNA. There are certain conventions used in earlier periods which, on reflection, do not look right. Several changes were implemented with the May 2001 release of our macro national accounts and one change, relating to the government superannuation plan, was implemented in June 2000. Other changes, primarily for industry statistics and input-output tables (IO), are planned for implementation in September 2001. This report records the results of this revisiting exercise.

Government Employee Pension Plans

1. Government employee pension plans: Autonomous plans with invested assets

Issue: Should the actuarial deficiencies of government employee pension plans with invested assets, and managed by independent trustees, be recorded in the CSNA as liabilities of the government sector and as assets of the household sector?

Pre-revision treatment: Prior to the 1999 annual release of the balance sheet accounts, CSNA ignored the actuarial deficiencies of pension plans with invested assets. The Financial Management System (FMS) used for government financial statistics, however, recognized these deficiencies as liabilities of government in the 1997 historical revisions of their accounts.

Discussion notes:

In Canada, employee pension plans with invested assets managed by trustees independent of the employer are referred to as trusteed pension plans. Such plans are separate institutional units and, thus, autonomous. SNA93 treats them as privately funded schemes (paragraph 8.63 b).

These plans consist of a reserve of money (invested in securities) that employees and their employers have contributed towards future pension payments upon retirement of the employees. The reserve becomes trusteed when the money is deposited with, and managed by, an independent organization. Fund assets increase with contributions and investment income and decrease with benefit payments and withdrawals. An actuarial appraisal of the plan is done periodically to determine if the reserve is adequate to fulfil its present and future obligations. If not, the plan is deemed underfunded, with an actuarial deficit equal to the shortfall. Actuarial deficits are liabilities of the employer, and are met by way of special contributions.

Most provincial governments guarantee deficiencies, although some do not record them as liabilities on their balance sheets. Since private sector employers are required to record such liabilities, it seems reasonable to treat governments in a similar fashion. Prior to the 1997 historical revision, both the FMS and the CSNA ignored actuarial deficiencies.

Statistics Canada - 354 - Collected Articles of Kishori Lal At the time of the 1997 CSNA Historical Revision, it was recommended to recognize these deficiencies in respect of government employer-sponsored trusteed pension plans as liabilities of the government sector and as assets of the household sector. However, it was decided to delay implementation of this new treatment until other proposed pension changes (see issue # 2) were implemented.

Decision: With the 1999 annual release of the balance sheet in March 2000, recognize all outstanding actuarial deficiencies for government employee autonomous pension plans as liabilities of the government sector and as assets of the household sector. Report actuarial deficiencies in the Other Changes in Assets Account.

2. Government Employee Pension Plans: Non-autonomous plans without outside invested assets

Issue: Should the liabilities of government employee pension plans without outside invested assets, and under the control of government, be recorded in the CSNA as liabilities of the government sector and as assets of the household sector?

Pre-revision treatment: Prior to the 1999 release of the balance sheet and the first quarter 2000 release of income and expenditure accounts, CSNA ignored the liabilities of plans without outside invested assets. However, the Financial Management System (FMS) for government financial statistics has always reported them.

Discussion notes:

In Canada, government-sponsored employee pension plans usually are under the control of government. These non-autonomous plans do not have an accumulated pool of outside assets from which to pay benefits; the only assets noted, for example, in the federal government superannuation account are nominal federal government long term bonds.

The accrued liability of the employer with respect to these plans grows with contributions plus investment income and decreases with pension payments and withdrawals. An actuarial appraisal of the plan is done periodically in order to determine whether the funds are adequate to fulfil the plan’s present and future obligations. If not, the plan is deemed underfunded, with an actuarial deficit in the amount of the shortfall. Actuarial deficits are liabilities of the employer, and (while not always recognized/reported) are ultimately met by way of special contributions.

During the review process leading up to the 1997 CSNA Historical Revision, it seemed reasonable to treat employee pension plans of the federal and provincial governments in a similar fashion to autonomous plans with invested assets. The justification was that (i) the governments are not likely to default; and (ii) government employees think of these plans as part of their net worth. Therefore, it was recommended to recognize obligations under such pension plans as a component of government liabilities and of household sector assets. Such a decision would have had significant implications for the sector income and expenditure accounts.

Statistics Canada - 355 - Collected Articles of Kishori Lal At the time of the 1997 CSNA Historical Revision, however, it was decided to postpone the implementation of this recommendation until other OECD countries had implemented SNA93. (At that point, no other OECD country had utilized this concept and there was concern about the lack of international comparability of Canadian data, especially as the amount was substantial, at well over $100 billion.) Furthermore, there was concern that the new treatment did not strictly conform to SNA93 under the assumption that the federal employee superannuation plan was unfunded, as there were no outside invested assets.

Since then, two developments have occurred. First, the United States Bureau of Economic Analysis implemented SNA93 in 1999 and recognized obligations under US government employee pension plans both as liabilities of the government sector and as assets of the household sector. This is significant, as the institutional arrangements of the US government employee pension plan are similar to those of its Canadian counterpart. Thus, the national accounts of the nation with which most analysts make comparisons became inconsistent with the CSNA.

Second, the assumption that the superannuation account of federal government employees is unfunded was revisited. The interest accruing on the account is calculated on the basis of long term bonds of the Government of Canada. Thus, in essence, the nominal federal government long-term bond is the invested asset of the superannuation account. CSNA viewed the federal government as comprised of two sub-sectors: federal government administration, with a non- marketable long-term bond liability to the pension plan and, a non-autonomous pension plan, with this bond as an asset and a pension liability to employees in the personal sector. With these clarifications, the federal government employee pension plan looked very similar to a funded plan. The same view was adopted for similar provincial plans.

Given this reasoning and the impact of the US revisions, it was decided to implement this change in the spring of 2000. This revision affects the entire 1961 to 2000 period. Government sponsored plans are now treated consistently with autonomous or trusteed pension funds, yielding a more complete picture of personal saving and net worth. Further, the new pension accounting treatment is consistent with that of the Public Accounts of Canada, the FMS and the US national accounts. Government liabilities and personal sector assets in the balance sheet were substantially revised upwards. Government pension liabilities are valued at the higher of the accrued or actuarial values (i.e., at the accrued amount plus/minus actuarial adjustment). Actuarial deficiencies are reported in the Other Changes in Assets Account.

The accounting change has no effect on production or GDP but it has a significant effect on the sector accounts measures, in particular on saving (and net lending). In summary, employee contributions, which are included in personal income as part of wages and salaries, are no longer directed to government sector income through a current transfer; neither are employer contributions, which are included in personal income as part of supplementary labour income. In addition, the investment income of the pension funds is now counted as the investment income of persons rather than as government investment income. Current transfers from government to persons in the form of public service pension benefits disappear from personal income and, as with other pension plans, are now treated as a reduction in personal sector assets. The net effect of the change in treatment is increased personal saving, resulting from higher personal income

Statistics Canada - 356 - Collected Articles of Kishori Lal and reduced personal outlays. As a result, 1999 personal saving was revised upward by $11billion, pushing the saving rate to 3.6% (from 1.9%); likewise, government saving was revised downward by $11billion.

Note that our presentation of the Income, Outlay and Savings account of the household sector is different from that in SNA93 but the value of the resulting savings estimate is the same. CSNA disposable income is higher by the amount of net equity of households in the trusteed pension plans compared with that in SNA93, which reports this item explicitly and separately from personal disposable income. CSNA presentation will be reviewed in the future. In addition, the new treatment also requires an additional entry (beginning in 1990) to account for the transfer of the actuarially determined surplus from the pension plans to the federal government. Such transactions are classified as capital transfers and do not affect saving, but they do affect net lending.

Note that in the write-up of issues #1 and 2, the terms “household sector”, “personal sector” and “persons” have been used interchangeably.

Decision: With the 1999 annual release of the balance sheet in March 2000, and the first quarter 2000 release of income and expenditure accounts and financial flow accounts in June 2000, recognize all obligations of non-autonomous government employee pension plans as liabilities of the government sector and as assets of the household sector. Change accordingly both the income and outlay and the capital and financial accounts of the household and government sectors.

Valuation of Value Added

There are three interconnected issues which need to be resolved for valuation of value added, both at the industry and at the total economy level, to make our series as internationally comparable as possible, given our data sources. These are: valuation of output, value added by industry at basic prices or factor cost, and disaggregation of subsidies.

3. Valuation of output

Issue: Should we switch to valuation of output at basic prices, as recommended by SNA93 or continue with the Canadian version of modified basic prices which excludes subsidies on products from valuation of industry output?

Pre-revision treatment: In the CSNA IO tables for 1961 to 1997, value of output has always excluded both taxes on products levied and subsidies on products received as a consequence of production or sale. In other words, the prices recorded are the subsidised prices, not the prices plus subsidies. We call this mode of valuation “modified basic price”.

Discussion notes:

In SNA93 , the preferred method of valuation of output of goods and services produced for the market is at basic prices, especially when a system of VAT, or similar deductible tax, is in

Statistics Canada - 357 - Collected Articles of Kishori Lal operation (paragraph 6.218) and is defined as follows: "The basic price is the amount receivable by the producer from the purchaser for a unit of a good or service produced as output minus any tax payable and plus any subsidy receivable, on that unit as a consequence of its production or sale. It excludes any transport charges invoiced separately by the producer" (paragraph 6.205a).

There are taxes on products and other taxes on production; similarly, there are subsidies on products and other subsidies on production. There may be some ambiguity, for the uninitiated, regarding which taxes and subsidies are being referred to in the SNA93 definition. The definition in the European System of Accounts, ESA 1995, (published by EUROSTAT, Brussels, 1996) clarifies it and it reads as follows: “The basic price is the price receivable by the producers from the purchaser for a unit of a good or service produced as output minus any tax payable on that unit as a consequence of its production or sale (i.e. taxes on products) plus any subsidy receivable on that unit as a consequence of its production or sale (i.e. subsidies on products). It excludes any transport charges invoiced separately by the producer. It includes any transport margins charged by the producer on the same invoice, even when they are included as a separate item on the invoice “ (paragraph 3.48).

We have not incorporated in the CSNA the recommended basic price in the valuation of goods and services; instead, we have modified the 1993 definition of basic price for our IO tables. In its IO tables, taxes on products levied as a consequence of production or sale are excluded, as recommended; however, subsidies on products are not added to the prices. In other words, the prices recorded are the subsidised prices, not the prices plus subsidies as recommended in SNA93. The CSNA definition of modified basic price is as follows:

“The modified basic price is the price receivable by the producers from the purchaser for a unit of a good or service produced as output minus any tax payable on that unit as a consequence of its production or sale (i.e. taxes on products). It excludes any transport charges invoiced separately by the producer”.

The modified basic price used in the CSNA IO tables is equivalent to the price, say for manufactured goods, at the factory gate. The advantage of this price is that the valuation of transactions is transparent and verifiable from enterprise records, a feature not available in SNA93. Had we adopted the SNA93 recommendation, we would report the transactions, not at prices prevailing in the market but at assumed prices, a feature that is not very appealing.

Another advantage, which is even more useful, is the facility and efficiency that our modified basic price approach provide for calculating constant price IO tables. We produce current price supply tables at modified basic prices and we also produce current price use tables at both purchaser prices and modified basic prices. We start with use tables at purchaser prices as recommended by SNA93, and convert them to modified basic prices, the same prices as those used in supply tables. Once we have converted our purchaser price use tables into modified basic prices tables, both supply and use tables are based on the same prices of products. This additional calculation is not discussed in SNA93. In Canada, we collect prices of sale of products excluding taxes on products and without including subsidies on products. This set of prices is immediately applicable without any additional amendment to use tables, which are already at modified basic prices.

Statistics Canada - 358 - Collected Articles of Kishori Lal Producing constant price value added using SNA93 recommendations would be very difficult conceptually as basic prices are not observed in the market. One would need to invoke bold assumptions to deflate supply tables at basic prices. Further, deflating use tables at purchaser prices will be very expensive as, in principle, each cell in the use table at purchaser price has its own unique price deflator. As we produce both current and constant price supply and use tables as benchmarks for many series in the CSNA, we find SNA93 guidelines quite problematic.

Decision: Given that modified basic prices are transparent and more importantly, given their advantage for calculating constant price IO tables, it is reaffirmed that we continue with our present practice of valuing output at modified basic prices.

4. Value added by industry at basic price or factor cost

Issue: Should we continue to calculate value added by industry at factor cost or switch to its calculation at basic prices? Pre-revision treatment: In the CSNA, value added by industry at current prices is calculated in the context of IO tables. Value added by industry equals value of output at modified basic prices less value of intermediate consumption of goods and services at modified basic prices. Value added by industry separately specifies the following items: taxes on products paid by the producing units on their intermediate consumption of goods and services; other taxes on production (also called taxes on factors of production); all subsidies on production received by the producing units listed with negative sign; wages and salaries; supplementary labour income; mixed income (net income of unincorporated businesses); and other operating surplus. Other operating surplus, calculated residually, becomes automatically higher by the amount of subsidies received by the producing units. Since 1986, subsidies have been disaggregated into subsidies on products and other subsidies on production (also called subsidies on factors of production). The detail of our current price value added by industry for the IO annual tables,1986 to 1997, permits us to calculate value added at factor cost (equal to the sum of wages and salaries, supplementary labour income, mixed income and other operating surplus) and at basic prices (equal to the sum of value added at factor cost plus other taxes on production less other subsidies on production). However, we highlight only constant price value added by industry at factor cost in our releases. Furthermore, constant price value added by industry for the current period from the latest IO tables is only at factor cost. The SNA93 term, “Other taxes on production”, is called “taxes on factors of production” in the CSNA; similarly the SNA93 term “other subsidies on production” is called “subsidies on factors of production” in the CSNA.

Note that we have not disaggregated production subsidies into their two components for the IO years 1961 to 1985 and the years since the latest IO table for 1997. Such disaggregation is also not available for sub-annual series for any period.

Statistics Canada - 359 - Collected Articles of Kishori Lal Discussion notes: a. SNA93 recommendation

SNA93 prefers value added at basic prices (see paragraphs 6.226-6.232), which represents value of output at basic prices less value of intermediate use of goods and services at purchaser prices. Value of output at basic prices represents value of output at the gate of the producing unit excluding any taxes on products payable and including any subsidy on product receivable as a consequence of its production or sale. Note that in our industry surveys, the producing units are instructed to value sales excluding any taxes payable on products and also excluding any subsidy receivable on product. In Canada, taxes on products are far more prevalent and far more significant compared to subsidies on products. The Canadian value of output is not at basic prices as defined by SNA93 but at modified basic prices, the modification being the exclusion of subsidies receivable on products. The Canadian prices of output are observed prices as received by the producers and paid by the purchasers and listed on the invoices, hence verifiable from the records. The Canadian modified basic prices are always lower than the SNA93 basic prices by the amounts of subsidy on products receivable by any given producer.

Value added at factor cost, the one used in CSNA, is a very well known concept in economic literature, but it is not a concept used explicitly in SNA93. In SNA93, it is derived from value added at basic prices by subtracting other taxes less subsidies on production (see SNA93 paragraph 6.229). There is also a reference to value added at market prices (see paragraph 6.227) but this measure is not very meaningful in the presence of a value added tax (VAT) regime. Under VAT, prices charged by producers exclude invoiced VAT and such taxes are shifted to the final users. b. International comparison of value added by industry

A recent OECD report "Services- Statistics on value added and employment, Edition 2000", contains time series data on the services sector by industry detail, of all 29 OECD countries and it may be used to illustrate how comparable are such series. Many users would assume that both the contents of value added and the industry detail would be comparable across countries . But they are not.

There are two international industrial classifications: ISIC Rev 2 and ISIC Rev 3, not comparable in detail, which are used by OECD countries. ISIC Rev 2 was produced in the early 1980's and ISIC Rev 3 in the early 1990's. Six OECD countries (Iceland, Japan, New Zealand, Switzerland, Turkey, and United States) provide value added and employment data by industry using ISIC Rev 2. All the other 23 countries provide data using ISIC Rev 3. A few countries provide data on both SIC's. Canada provides data on ISIC Rev 3.

SNA93 prefers calculation of value added at basic prices which, as noted above, is equal to output at basic prices less intermediate expenditures at purchaser prices. The logic of using value added at basic prices rather than at traditional factor cost is that it better represents the cost of factors of production used in the production process. Value added at basic prices is higher than the value added at factor cost in Canada by the amount of other taxes on production (mostly

Statistics Canada - 360 - Collected Articles of Kishori Lal property and payroll taxes) less other subsidies on production (mostly labour training). Subsidies for labour training are not materially very important, amounting to just two-tenths of one percent of labour costs, and are prorated on wages and salaries by industry. Property taxes mostly affect the real estate industry, but payroll taxes are a proportion of wages and salaries in every industry.

Some countries use another concept of value added by industry, value added at market prices. This valuation is higher than the one at basic prices, mostly by product taxes collected by industries less subsidies on products received by them. Most of the product taxes, such as GST and PST, in Canada are collected primarily by trade establishments. Thus value added at market prices for trade industry will be much higher than its valuation at basic prices or at factor cost. The collection of such taxes by trade establishments on behalf of governments can hardly be called value added by trade. On the contrary, value added at market prices for highly subsidised industries, such as agriculture, will be lower than the one calculated at basic prices.

Both levels and growth rates are different depending upon valuation (basic price, market price or factor cost) and thus, productivity estimates are different. Amongst the 29 OECD countries, practices vary widely. Three countries (Canada, Iceland and Ireland) provide data at factor cost. Ireland as member of EU will shift soon to basic prices. Four countries (Japan, Switzerland,

Turkey, and United States) provide data at market prices. All the other 22 countries provide data at basic prices. c. Value added in SNA93 and CSNA

Let us demonstrate, as follows, the relationship of these three concepts of value added at the total economy level, both in SNA93 and CSNA by using the SNA93 example in Chapter 15, Table 15.1.

Statistics Canada - 361 - Collected Articles of Kishori Lal Total economy level value added, using three concepts a. Value of output at basic prices for total economy 3604 b. Value of intermediate use of goods and services at purchasers’ prices 1883 c. Value added at basic prices (a-b) 1721 d. Taxes less subsidies on products 133 d.1 Taxes on Products 141 d.2. Less subsidies on products -8 e. Total gross value added/ GDP at market prices (c+d) 1854 f. Other taxes less subsidies on production 58 f.1 Other taxes on production (see 93SNA, Annex 5, p.602) 94 f. 2 Less other subsidies on production (see 93 SNA, Annex 5, p.602) -36 g. Value added at factor cost (c-f) or (e-d-f) 1663 Let us redo the above information using Canadian SNA conventions: a*. Value of output at modified basic prices for total economy 3596 b. Value of intermediate use of goods and services at purchasers’ prices 1883 c*. Value added at modified basic prices (a*-b) 1713 d.1 Taxes on Products 141 e. Total gross value added/ GDP at market prices (c*+d.1) 1854 f. Other taxes less subsidies on production 58 f.1 Other taxes on production (see 93 SNA Annex 5, p.602) 94 f. 2 Less other subsidies on production (see 93 SNA Annex 5, p.602) -36 g. Value added at factor cost {c*-(f-d.2)} or (e-d-f) 1663 c. Value added at basic prices (g+f) 1721

Canada can continue providing data at factor cost and remain comparable only with Iceland, not a very useful scenario. We have all the details, both at current and constant prices, to provide value added data at basic prices for 1986 to 1997. We need to segregate subsidies on production into subsidies on products and other subsidies on production for the IO tables for 1961 to 1985 to value added at basic prices. Projectors from 1997 for monthly RDP series by industry need not change as the proxies currently used are employment or output.

Note that value added at market prices at 1854 and value added at factor cost at 1663 are the same in SNA93 and CSNA. We can easily calculate value added at SNA93 basic prices by adding to our traditional value added at factor cost by industry two items: other taxes on production less other subsidies on production.

Market price for the total economy is the same in the CSNA and SNA93. The Canadian market price for the total economy equals Canadian value added by industry at modified basic prices plus taxes on products paid by all transactors, whereas the SNA93 market price for the total economy equals its value added by industry at basic price plus taxes on products less subsidies on products by all transactors. Again, in both CSNA and SNA93, value added at factor cost by industry is the same and equals value added at market prices for the total economy less net taxes on production (taxes on products less subsidies on products, plus other taxes on production less other subsidies on production). In either case, value added by industry at basic prices falls

Statistics Canada - 362 - Collected Articles of Kishori Lal between valued added at factor cost and value added at market prices. Once we have all the details on taxes on products, other taxes on production, subsidies on products and other subsidies on production, value added at basic prices can be calculated, whether we start with output at basic or modified basic prices. d. Value added in CSNA Industry Measures and in Productivity Measures

At present, the concept of value added used in the CSNA industry measures ( both in IO tables and monthly GDP by industry) is at factor cost whereas the concept of value added used in the productivity measures is at basic prices. The productivity program takes CSNA value added by industry at factor cost and adds to it other taxes on production less other subsidies on production, and thus arrives at value added at basic prices. Once CSNA starts producing value added at basic prices, the productivity program will no longer need to make its own adjustments to the CSNA published estimates.

Decision: Switch to calculating for detailed industry statistics, value added at basic prices in September 2001. However, this requires developing estimates of subsidies on products.

5. Disaggregation of subsidies

Issue: Should we disaggregate, for historical periods as well as for the current period, and both at annual and quarterly frequencies, subsidies on production into subsidies on products and other subsidies on production?

Pre-revision treatment: Subsidies are already disaggregated into subsidies on products and other subsidies on production for annual IO tables for the reference year 1986 and onwards. No such disaggregation exists for the period 1961-85 and the period from the latest IO table to the current period. Further, no such disaggregation exists at sub-annual frequencies.

Discussion notes:

We seem to be reaching the conclusion, that starting with the release of CSNA industry series in September 2001, we should publish value added by industry at basic prices rather than at factor cost, but this requires development of estimates of subsidies on products. Our switch to value added at basic prices will bring us in line with the SNA93 recommendation and make our series internationally comparable.

It is worth repeating that value added at basic prices requires delineation of subsidies into subsidies on products and other subsidies on production. Value added by industry at basic prices equals value added by industry at our traditional factor cost plus other taxes on production less other subsidies on production. Other taxes on production are by and large on factors of production ( in Canada mostly property taxes and payroll taxes) and, similarly, other subsidies on production are provided by governments to enterprises to reimburse part of their expenditures on factors of production (in Canada mostly labour training and other labour related activities).

Statistics Canada - 363 - Collected Articles of Kishori Lal All subsidies are provided by governments to producing units to enable them to produce products which they otherwise might not be able to do. For example, governments subsidise transit systems so that they can charge lower ticket prices and still survive in the market. Governments subsidise farmers so that they can compete with international producers who are highly subsidised. Governments subsidise real estate developers so that they can charge lower rents from poor segments of society. Governments provided subsidies in the 1980's to petroleum refineries on their imports of crude oil to enable them to remain competitive. Governments provide subsidies to many producers for training employees. All subsidies are paid to producing units to fulfil some government policy objective.

In Canada, the amount of subsidies (subsidies on products and other subsidies on production) is quite small. For the period 1961 to 1973, subsidies amounted to about 1% of GDP. During 1974 to 1988, subsidies were higher and amounted to about 2% of GDP, for most of the years, and about 3% of GDP for a few years in the early 1980's. The much higher subsidies in the early 1980's were product subsidies mostly on inputs of crude oil imports. Subsidies since 1991 have continued to decline as a share of GDP, and now again amount to about 1% of GDP. Furthermore, most of the subsidies are subsidies on products. For example, in 1997, out of total subsidies of $9.4 billion, $8.3 billion were subsidies on products and $1.1 billion were other subsidies on production, primarily for labour training and other labour related activities. Labour subsidies were prorated on the basis of wages and salaries by industry and they accounted for just two-tenths of one percent of labour costs. Labour subsidies were even less important in the earlier periods. Labour subsidies are provided both by the federal government and the provincial governments; for example, in 1997, two-thirds of such subsidies were provided by the federal government and one-third by the provincial governments.

As noted, we disaggregated product subsidies from other subsidies only for annual frequencies and only for the IO tables for the years 1986 to 1998. Such disaggregation is also needed for the macro income and expenditure accounts, annual and quarterly, for them to switch to net domestic product at basic prices from net domestic product at factor cost. This means that the present indirect taxes less subsidies need to be separated into net taxes less subsides on products and net taxes less subsidies on factors of production. Net taxes on factors of production will be added to the present net domestic product at factor cost to bring its valuation to basic prices.

As subsidies represent a small share of GDP and other subsidies on production (subsidies on factors of production) represent even a smaller share, one of the issues we need to decide is whether it is worth spending resources to develop such disaggregation of subsidies for the periods 1961 to 1985, for which such disaggregation is not available at present. It will be quite expensive to develop such estimates for 1961 to 1985. Given that subsidies are such a small share of GDP, and other subsidies on production are a minuscule share, it may not be wise to spend resources distinguishing product subsidies from other subsidies on production. Thus, the recommendation is to classify all subsidies as subsidies on products for 1961-85 period.

Another issue concerns the approach to be taken to estimate labour subsidies for the current period from the latest IO benchmark year so that they can be confidently incorporated in the income and expenditure accounts. Federal labour subsidies, the more important ones in terms of value, are more easily identifiable and they are reported for very current periods. Even here,

Statistics Canada - 364 - Collected Articles of Kishori Lal however, numbers need to be verified, as is done now, by discussion with the concerned government department. Provincial public accounts reports are not as current, and it requires lot more work to identify labour subsidies but, luckily, they are less significant in terms of value. Some projection techniques need to be developed for provincial subsidies for the current period until IO benchmarks are produced. It should be noted that the use of the precise classification of SNA93 subsidies on products and other subsidies on production will result in some changes in the present estimates of these two subsidies in the IO tables for 1986 to 1997.

If we go along with the above recommendations, no additional articulation will be needed for current price value added at basic prices except that we will highlight it at basic prices rather than at factor cost. Value added at basic price for the period 1986 and onwards will be equal to our present factor cost value added plus other taxes on production less other subsidies on production, which are now calculated. For the period 1961 to 1985, value added at basic prices will be equal to our present factor cost value plus other taxes on production, as other subsidies on production are assumed to be zero.

For calculating constant price value added at basic prices by industry through double deflation for the business sector, one would not need to deflate other taxes on production and other subsidies on production, as they form part of basic prices. One will need to deflate other taxes on production for non-business sectors of the economy to add it to labour and consumption of capital to arrive at basic prices; other subsidies are not applicable. Taxes on products paid by producers on intermediate use of products and all subsidies received by producers will, of course, continue to be deflated.

Decision: Given that other subsidies on production are quite minimal now and were even less important in earlier periods, classify all subsidies as subsidies on products for the period 1961 to 1985. For the period 1986 and onwards, disaggregate subsidies following SNA93 classification into subsidies on products and other subsidies on production both for the IO benchmark years and for the current period, at annual and sub-annual frequencies.

Composite decision: As decisions for issues # 3,4,5 are implemented in 2001, produce value added by industry at basic prices, as recommended by SNA93, even though we will continue to produce value of output at modified basic price.

6. Employee Stock Options

Issue: When and how should stock options be valued and should their value be reported in the CSNA as part of labour compensation?

Pre-revision treatment: In the CSNA, stock options were valued at the time they were exercised by the employees and this value was included in labour compensation.

Statistics Canada - 365 - Collected Articles of Kishori Lal Discussion notes:

At the time of the drafting of SNA93, employee stock option plans were not very prevalent. Hence, this topic was not discussed therein, with the result that no international guidelines became available to national accounts compilers.

The issue of how best to report, in both corporate financial statements (income statements and balance sheets) and the national accounts, the phenomenon of compensating employees not only with cash but, also, with options to buy shares in their company was under discussion for some time, but it assumed importance in the mid to late 1990's. The appropriate income tax treatment of stock options has also been an ongoing subject of discussion.

In essence, and subject to a wide variety of terms and conditions, stock options allow holders to buy specific amounts of a corporation’s (usually common) shares at a specific price at any time during a specific future period. This phenomenon recently assumed greater importance than before, mainly for two reasons. One is the widespread and growing tendency for corporations (many of them in dynamic high-tech fields) to make stock options important, often very important, parts of the compensation packages offered to employees. The other is the dramatic increases (at least till 2000) in stock-market values of the shares of these corporations. It seems to be the general, though not necessarily the universal, case, both in Canada and the United States, that corporations issue the shares in question from treasury when stock options are exercised. Hence, issued shares appear in the equity section of the balance sheet valued at the exercise price.

In principle, there are three distinct times when the value of stock options can be calculated: when stock options are granted to employees, when they are vested to them, and when stock options are exercised by them. Events of definite economic, but of uncertain financial, significance occur when stock options are granted and vested to employees. At the time they are granted, the exercise price of the options is generally set equal to the market value of the shares. The intrinsic value of the options, which is the difference between the market value and the exercise price, would then be nil. At the time the options are vested (ranging from the granting dates to a few subsequent periods), their intrinsic value, which depends upon their market value, represents an unrealised gain. Another event occurs when the employee actually exercises the option by acquiring the shares. The employee still has an unrealised gain but the conversion shifts from options to share instruments. The market value of the shares when exercised less the exercise price granted to employees equals the unrealised capital gains. This value in most cases will be higher than the intrinsic value noted above. Neither value, the intrinsic value nor the employees’ unrealised capital gains, is expensed in the income statement of the corporations: thus their reported profits are overstated. This is troubling to both national accountants and the accounting profession in general. In July 2000, the accounting standard bodies of USA, Canada, United Kingdom, Australia and New Zealand deliberated on this issue and produced a report (available from the Australian Accounting Standards Board), called Accounting for Share- Based Payment: G4+1 Proposals, and invited comments from all concerned parties. The report is leaning towards valuing such options at the time they are vested and reporting them in the company records as labour compensation; hence a deduction from profits. Other than this report, there are no international guidelines.

Statistics Canada - 366 - Collected Articles of Kishori Lal

In addition to the valuation of stock options, the timing of tax thereon is also an issue, since it is generally considered inequitable to impose tax on unrealised gains. This inequity is the root of the government’s attempts to deal with stock options. These attempts inevitably affect the information that corporations provide to statistical agencies and which underpins the entries in their respective national accounts.

In Canada, the intrinsic value of stock options has not been reported in compensation of employees. However, the value of stock options, when exercised, has been included, up to taxation year 1999, as taxable benefits in the tax statement, called T4, which is submitted by employers to employees, In the latest (February 2000) budget, the federal government changed this long standing policy.

Pre-2000 budget tax treatment of options:

1. A taxable employment benefit equal to the difference between the market value of the shares at the time the option was exercised and the price paid by the employee to acquire the shares was generally included in income as taxable benefit and reported in the year the option was exercised. An exception to this rule was where the shares were of a Canadian controlled private corporation (CCPC): here, the taxable employment benefit was generally not included in income until the year of disposition of the stocks.

2. The individual was also taxed on the capital gains, if any, earned upon the later disposition of the shares.

February 2000 Budget

1. The Budget (February 28, 2000) introduced an amendment to the Income Tax Act to modify the way in which stock options are taxed. Generally, this proposal allows employees to defer unrealised gains from exercising stock options for publicly listed shares until the disposition of the shares. However, there is an annual limit of $100,000 subject to deferral.

Companies have few expenses associated with stock options. When they are exercised, the shares are usually issued from treasury. There may be issue costs but these are usually deemed overhead costs. There certainly is an opportunity cost associated with stock options, in that the company could have sold the shares on the open market for a price greater than it required its employees to pay. This opportunity cost, however, is not allowed as a deduction in the Income Tax Act.

Under the pre-revision treatment in the CSNA, labour income included an estimate of the value of employee stock options. T4 information provided by CCRA-Canadian Customs and Revenue Agency (previously called Revenue Canada)- was used to establish benchmarks. The reporting of stock options is as follows, following the budget changes in 2000:

1. The system for reporting continues to be a self assessment system.

Statistics Canada - 367 - Collected Articles of Kishori Lal 2. If an individual wishes to defer the taxable benefits of stock options, he/she must elect to inform the employer in writing. Both employer and employee must maintain a copy of the election document. The employer must report the value of the stock options on the T4; however, the value of the deferred benefit will be shown separately and not reported as employment income, as in the past.

3. The employer will calculate the benefit as was done in the past but it is now divided into two parts, one called “stock option benefit” and the other “deferred stock option benefit”. The deferred option benefit is not taxed until the disposition of the shares but the stock option benefit is taxable in the year it is reported.

4. CCRA will use this prescribed form to track the options and as an audit trail for future compliance testing. Thus, the tracking of the yearly value of stock options will change from the employment income portion of the T4 to a separate code; however we will still have the information available for our purposes.

There remains a concern with respect to the timing of the recording of stock options in the accounts. At present, they are recorded when exercised but the options could refer to other periods. It could be argued that the options were earned when granted and vested to employees. Closely related to the issue of timing is that of valuation. Recording the value of stock options in labour income when they are exercised automatically includes capital gains for a payment to a factor of production in the production accounts. This is troublesome. If one uses the intrinsic value of stock options when vested, there would be very few capital gains. The statistical system, however, does not easily permit us to address these concerns. Thus, as a default, they are allocated to the period in which the options are exercised and their value is calculated as the difference between market value and exercise price granted to employees.

No matter what we do, there remains a serious discordance a) between the time when employees actually earn the stock options, either by supplying services to the corporation or by promising to provide future services, and the time when they are compensated by means of them, and b) between the time when the stock options are granted and vested and the time when they are exercised. Our judgement is to rely upon the information available to us from tax records and defer any fundamental reexamination of these issues to a later period when new international or at least North American standards are in place. Thus it is recommended that the full value of stock options (both stock option benefits and deferred stock benefits) when exercised should continue to form part of labour compensation and also be reported in the period when exercised. Employees definitely look upon stock options as compensation. In the CSNA, our past and proposed treatment adds this value to labour compensation, thus residually reducing other surplus.

Decision: Continue to record the total value of stock options(stock option benefit and deferred stock option benefit) and to include it in labour compensation, even though the entire amount is not taxed, as well as expense the entire amount against profits.

Statistics Canada - 368 - Collected Articles of Kishori Lal 7. Software

Issue: Should development expenditures on all software (pre-packaged, custom design and own- account) be capitalized?

Pre-revision treatment: In the investment surveys conducted by the Investment and Capital Stock Division, enterprises reported that they capitalised some expenditures on software, though the amount was not reported up to1998. The same unknown amount was capitalised in the CSNA. A special request was made to the enterprises to report this amount for 1998. The amounts reported in response to this special survey showed that the enterprises were capitalising only about one quarter of total software expenditures.

Discussion notes:

As noted, the CSNA has capitalised whatever amounts the companies reported on our surveys. Some companies capitalize software and some do not; or a single company may capitalize some software expenditures but treat other software expenditures as current expenses.

SNA93 notes: “Computer software that an enterprise expects to use in production for more than one year is treated as an intangible fixed asset. Such software may be purchased on the market or produced for own use. Acquisitions of such software are therefore treated as gross fixed capital formation” (paragraph 10.92). Despite this clear recommendation, we differed, particularly in the area of own-account development expenditures. Our reason was the data problem. Since our implementation of SNA93 in 1997, many other countries have now implemented it as well and all software, useable for more than one year, is capitalised. This made our accounts less comparable with other countries, including USA and thus a problem for many users, both in the government and private sectors.

For pre-packaged and custom made software bought on the market, there are at least market transactions which, in principle, can be traced. This was not the case with respect to own-account software, since there was no information readily available in the enterprise records. Thus one needed to develop a methodology to estimate such expenditures. The United States was the first country to develop such a methodology and now most countries have replicated it. Own-account software investment in the USA and most other countries is built up from estimates of the labour cost of persons engaged in software development. Information on these persons comes from population censuses or special surveys of occupations of persons. Our Income and Expenditure Accounts Division has developed a detailed research paper on this issue, called Capitalisation of Software in the National Accounts: Sources and Methods. This paper is available at Statistics Canada’s web site as a methodology paper.

Reclassifying expenditures on software as capital expenditures from current expenditures in the enterprise books results in an increase in the value of GDP in two ways. First, we delete intermediate expenditures of enterprises on such software and allocate it to their investment expenditures, thus adding to GDP. In government enterprises, such expenditures are removed from current account and added to capital account: thus there is no net change. However, we

Statistics Canada - 369 - Collected Articles of Kishori Lal include the value of consumption of this capital in the current account, thus adding to GDP. Revision of total capitalisation in 2000 was of the order of about $11billion.

Decision: Capitalise all software development expenditures in the CSNA (income and expenditure accounts, IO tables and monthly real GDP by industry, and national balance sheet accounts) with effect from the beginning of 1981.

8. Electro-Magnetic Spectrum (Mobile Phone) Licenses

Issue: How should we treat the receipts from the sale of mobile phone licenses in the national accounts?

Pre-revision treatment: In the CSNA, payments by enterprises to the government for licenses to use electro-magnetic spectrum ( also called mobile phone licenses) were treated as taxes on factors of production.

Discussion notes:

SNA93 does not provide guidelines on the treatment of these licenses; hence, we need to develop our own general principles. The treatment of license payments for mobile phones in the national accounts has become increasingly important in recent years as mobile phone licenses have been auctioned for substantial amounts in several countries. Further, there is no uniform treatment of payments for such licenses. In Canada, as noted, such payments were treated as taxes on production. In USA, they are treated as sale of assets. In UK, license fees are prorated on an annual basis as rent. Policies of other countries also vary. In Finland, for example, the government has granted these licenses free of charge. Korea is the only country which treats such payments as payments for government services.

Consequently, the Inter-Secretariat Working Group on National Accounts (ISGWNA) has been deliberating on this important issue for more than a year. At the March 2001 meeting in Washington, the IMF distributed a paper, entitled ”IMF Working Paper, Treatment of Mobile Licenses in the National Accounts” , prepared by Robert Dippelsman and Nils Maehle, members of the IMF staff. The paper provides the pro and cons of various treatments. The United Nations Statistical Commission requested, at its meeting in March 2001, that the ISWGNA make decision as soon as possible on the national accounts treatment of mobile phone licenses. In response to this request, the ISWGNA organised a special meeting on this subject on April 5, 2001 in Washington and arrived at several conclusions, summarised as below.

ISWGNA conclusions: ISWGNA states that there are two assets: one is the electro-magnetic spectrum, the other is the license to use it for mobile phones. Electro-magnetic spectrum is a tangible non-produced asset of the government sector, similar to its other non-produced assets such as land and sub-soil mineral assets. Typically, governments auction licenses to use electro- magnetic spectrums for a determinate period of time; at the expiry of the licenses, the spectrums revert back to the governments. (The practice in the United States is different from that of most other countries; it sells the spectrum for an indefinite period, hence there is no need to issue a license). If the duration of the license is more than a year (though typically, it is for 10 to 30

Statistics Canada - 370 - Collected Articles of Kishori Lal years), the payment made by the licensee is to acquire an asset. There are many other conditions which also must be met for the licenses to be recognised as an asset. It is stated that the license is an intangible non-produced asset and the licensee’s payment for it is not a tax. There is no quid pro quo in taxes but, in this case, the payee is granted exclusive use of an asset. It is further stated that this payment is not for services provided, since non-produced assets do not, by definition, produce services.

The decisions arrived at the April 2001 meeting of ISWGNA were not available to us when we deliberated on the treatment of mobile phone licenses in our accounts, in late 2000, for implementation by the first quarter release of our accounts in May 2001. However, we were convinced that our pre-revision treatment of classifying payments for the license by enterprises as taxes on factors on production needed to be changed. In early 2001, the federal government auctioned mobile phone licenses for ten years for an advance payment of $1.5 billion, and the option chosen by the government was to accrue this revenue, roughly $150 million per year, over a 10 year period in the public accounts. We were already leaning towards this treatment. Our position is similar to that of the UK where mobile phone licenses were auctioned a year or so ago, for a very substantial amount of advance payment (about $60 billion) for a period of some 25 years and payments were prorated on an annual basis. We treat the advance (up-front) payment for mobile licenses as a prepayment for use of non-produced asset and classify it as a financial transaction. The payment is prorated on an annual basis and classified as rent or royalty. It is this position that we all agreed to implement in our 2001 SNA revisions, both in SNA and in government finance statistics. We intend to reexamine this treatment when the ISWGNA issues its final clarifications and they are accepted by the UN Statistical Commission. We are also looking with interest the treatment of mobile licenses in the forthcoming IMF’s Government Finance Statistics.

As noted, there was an auction by the federal government of the license for use of electro- magnetic spectrum in early 2001, for $1.5 billion over 10 years. At the end of this period, the license reverts to the government which will then auction it again. It was agreed to prorate this amount over ten years, approximately $150 million every year. There was no significant amount on this account before 1997.

Decision: Reclassify government receipts from the auction of electro-magnetic spectrum licenses as royalties on non-produced assets and prorate total receipts over the life of the licenses and bring these series back to 1997. Re-examine this treatment when the final clarifications issued on this subject by ISWGNA are adopted by the UN Statistical Commission.

9. Rents of agriculture

Issue: Should land rents paid by farmers remain part of their intermediate expenditures for services or be reclassified as payment of property income, and, hence part of their value added?

Pre-revision treatment: Land rents paid by farmers were treated as payments for services, contrary to the recommendation of SNA93.

Statistics Canada - 371 - Collected Articles of Kishori Lal Discussion notes:

At the time of the 1997 CSNA historical revision, there was an oversight in the classification of rents paid on land by farmers. This item is clearly delineated in agriculture surveys as an expenditure and was always assumed to be part of intermediate expenditures for all the many decades that we have calculated inputs and outputs of the agriculture industry. SNA93 clearly recommends a different treatment which somehow got missed in our deliberations. SNA93 notes: “Owners of land and sub-soil assets may put them at the disposal of other units by arranging contracts or leases under which the tenants, or users of the assets, agree to pay to owners property incomes in the form of rents” (paragraph 7.87). All payments of property income such as land rents or royalties on sub-soil assets are allocated in SNA93 to the industry paying them, not as part of their intermediate expenditures on goods and services but as part of their payments of property income. Land and sub-soil assets are non-produced assets, hence, by definition, they do not produce any services. There is no production involved in renting land. The treatment of land rent in the CSNA is similar to our treatment of royalties on sub-soil assets and, recently, that of royalties on the use of electro-magnetic spectrum, as discussed above under issue #8. .

Decision: Treat rent on land as payment of property income and move it from the intermediate expenditures of agriculture to its value added segment, effective 1961. At the same time, reduce farm rental output from the real estate industry and reduce its value added by the same amount. The result is that value added of agriculture increases by the same amount as the decline of value added in the real estate industry, leaving overall GDP unchanged.

10. Value of output of holding companies

Issue: Should the value of the output of holding companies be equated with their management fees or their operating expenses?

Pre-revision treatment: In the CSNA historical series of IO tables and industry measures, the value of the output of holding companies comprised only their non-financial operating revenues, such as management fees. These revenues were not sufficient to cover all their operating expenses, with the result that their value added became very small or even negative. This appeared counter-intuitive.

Discussion notes:

SNA93 does not provide specific guidelines for valuing the output of holding companies. Hence, we need to develop our own conventions that make economic sense. Holding companies have two main functions: a) holding securities or other equity interests in subsidiaries for the purpose of exercising control and b) influencing their management decisions. Holding companies generate two kinds of revenues: a) management fees charged to their subsidiaries and financial revenues, such as interest and dividends, by managing the financial assets of their subsidiaries. According to SNA conventions, financial revenues are not counted as output, hence the recorded output, limited to only management fees, is not sufficient to cover all operating expenses, resulting in small or even negative value added. As only the management fees, not the full

Statistics Canada - 372 - Collected Articles of Kishori Lal operating costs of holding companies, are charged to their subsidiaries as their intermediate expenditures, a bigger operating surplus in their accounts is recorded. Overall GDP for the combined holding companies and their subsidiaries remains unchanged. Our pre-revision practice produced two counter-balancing situations, much lower value added in holding companies and a corresponding higher value added in subsidiaries, because the value of output of holding companies did not reflect the full costs of their production.

After considerable discussion, it was agreed to value the output of holding companies by taking into account all their operating expenses. The value of this output would equal all intermediate expenditures plus labour costs and consumption of gross fixed capital formation. This higher revised output would be allocated to subsidiaries, thus their intermediate expenditures would be increased with a corresponding reduction of their operating surplus. Overall GDP for the combined sector of holding companies and subsidiaries would remain unchanged. This revised calculation would produce better economically meaningful results; with holding companies valuing their output at full costs and subsidiaries paying the full costs of services provided by their holding company.

Holding companies did not form a separate industry in IO industry classification in the back period. However, separate worksheets are available in the files for the period 1981 and onwards but not for pre-1981 period. Holding companies formed part of other finance, insurance and real estate industry during 1961 to 1980 and other financial intermediary industry during 1981 to 1997. Thus the conceptual change in the valuation of output can only be effected for 1981 and onwards.

Decision: Recalculate, for the period 1981 to 1997, the value of output of holding companies by adding all their operating expenses and allocate the revised output to their subsidiaries, with the result that value added of holding companies increases, counterbalanced by an identical decrease of value added of subsidiaries, keeping overall GDP unchanged.

11. Implementation of North American Industry Classification System (NAICS)

Issue: How should time series be maintained with the implementation of NAICS in the CSNA detailed industry statistics?

Pre-revision treatment: The CSNA detailed industry statistics from 1961 onwards are classified on the basis of three Standard Industrial Classification (SIC) Manuals, all issued by Statistics Canada or its predecessor Dominion Bureau of Statistics. The first one was issued in December 1960, called “Standard Industrial Classification Manual”, the second one was called “Standard Industrial Classification Manual, Revised 1970", and the third one was called “Standard Industrial Classification 1980". All three followed the same principles of classification, except that the latter ones had more details as new industries emerged and were comparable. There was not much difficulty in reporting the Canadian industrial structure over time.

Statistics Canada - 373 - Collected Articles of Kishori Lal Discussion notes:

NAICS is the joint product of three countries, Canada, Mexico and the United States, and it is scheduled to go into effect for reference year 1997 in Canada and the United States, and 1998 in Mexico. NAICS differs substantially from the SIC’s because it is based on a single organising principle, which is in contrast to the SIC’s which have no such single principle. NAICS is erected on a production-oriented or supply based conceptual framework, in which producing units that use identical or similar production processes would be grouped together. Some SIC based industries were grouped on production-oriented principles while other industries were based on demand based principles. This difference in orientation has created discontinuity in time-series comparability between SIC-based estimates and NAICS-based estimates. Even at major sector levels, such as retail trade and wholesale trade, the differences in values between the two classifications are substantial. This has created a dilemma both for compilers and users of detailed industry statistics.

It is very expensive, even if it were possible, to re-code on a NAICS basis, the establishments in the back period for all or even most sectors of the economy. Then, how far back can or should one go for this exercise? For example, in manufacturing, establishments have been re-coded to NAICS for 1992 onwards, but not much has been done in other sectors. Even in manufacturing, what should one do for the period before 1992? Given time and resource constraints, one must develop other options to serve users who require long time series of data.

We have developed data for 1997 on both a NAICS basis and an SIC basis at about 100 industry aggregation (also called link) levels. Note that the 100 industry link levels follow the hierarchical structure of NAICS. SIC based worksheet level industries have been allocated the new link level without regard to their own hierarchical structures. Very detailed reconciliations between the two series have been prepared and notes drafted. This information would be made available to enable users to reconcile the estimates for each of the100 industries based on the two classifications. In September 2001, we will publish estimates for all the 100 industries for 1981 to 1997 based on the existing worksheet level detail and on a NAICS basis for 1997 and 1998, at both current prices and constant prices. Users who only need constant price growth rate data can easily link 1998 to the 1981-1997 period because 1997 is calculated on the basis of both classifications. We will bring the 100 industries detail to 1961 in September 2002, as additional disaggregation is required for some industries in order to achieve better time series comparison.

We have aggregated approximately 700 NAICS based industries (5 and 6 digit codes) into 300 worksheet level detail for CSNA for the years 1997 and onwards. It is this detail which will be also concorded into the international ISIC3 for our presentation of national accounts, industry and productivity data to both OECD and other international bodies.

Decision: Implement NAICS into CSNA industry statistics starting with the reference year 1997. Produce both NAICS based and SIC based estimates at about 100 industry detail level for 1997 and make these estimates available to users. Produce the same 100 industry level detail based on SIC1980 for the period 1981 to 1996. Publish all these estimates in September 2001.

Statistics Canada - 374 - Collected Articles of Kishori Lal Constant Price Measurement

12. Constant price Macro Economic Measures

Issue: Should we switch to annual or even quarterly chain volume indices for constant price macro measures in CSNA from the Laspeyres volume indices based on 1992 weights?

Pre-revision treatment: All CSNA constant price series (both macro such as total GDP and its main components as well as detailed industry statistics such as GDP by industry) for the period 1992 and onwards used Laspeyres volume indices based on the fixed weights of 1992. CSNA constant price series for 1986 to 1992 were based on 1986 weights; constant price series for 1981 to 1986 were based on 1981 weights; constant price series for 1971 to 1981 were based on 1971 weights and constant price series for 1961 to 1971 were based on 1961 weights. When we rebased our series to a new base, we did not use the new weights for series of the earlier periods. Growth rates calculated for earlier series based on earlier weights were not changed unless there were statistical or conceptual revisions. Thus we have always chained the series at constant prices but our link was not annual but every 10 years in 1971 and 1981, 5 years in 1986 and 6 years in 1992. All the series were always expressed in the newest base year dollars.

In addition, we also published supplementary information for macro constant price series of quarterly accounts of the income and expenditure accounts using quarterly Fisher volume index. However, we highlighted in our constant price releases fixed weight Laspeyres volume series rather than series based on chain volume indices.

Discussion notes:

SNA93 clearly states its preference for the index number to be used for macro estimates such as GDP volume statistics as follows: “The preferred measure of year to year movements of GDP volume is a Fisher volume index; changes over longer periods being obtained by chaining: i.e., by cumulating the year to year movements” (paragraph 16.73a). However, SNA93 also notes: “Chain indices that use Laspeyres volume indices to measure year to year movements in the volume of GDP...provide acceptable alternatives to Fisher indices” (paragraph 16.73c).

A well known property of chain link series is that for the back period detailed components do not add to total, and with chain Fisher , even the current period detail does not add up to total. We observed this in our own chain linked series from 1961 onwards when detailed components for annual or quarterly series for 1961 to 1991 did not add to totals. Concerns about the lack of additivity for macro series need to be weighed against the usefulness of sound volume measures of their period to period growth rates.

SNA93's preference for the chain Fisher volume index is clearly stated. The USA has adopted this index for their constant price national accounts series since 1998. The alternative chain Laspeyres volume index may be adopted by the member countries of the European Union.

As noted, we have been publishing constant price series using the chain Fisher volume index in the income and expenditure accounts as supplementary information for some time. There is

Statistics Canada - 375 - Collected Articles of Kishori Lal agreement in the literature that the best measure of growth rate of GDP or its components is the one based on the most current weights. The Ideal Fisher index fully satisfies this requirement. The Fisher formula calculates each link in the chained volume index as the geometric mean of the Laspeyres (fixed-weighted) index and the corresponding Paasche (current-weighted) index. Thus, we decided to use the Fisher formula for calculating chain volume index for our macro accounts. Another issue that arose in this context was: should we use an annual or quarterly chain Fisher index? An annual chain, when benchmarked to the quarterly data, could create a problem for comparison between the fourth quarter of the last year and the first quarter of the new year. Comparative growth rates are exaggerated when benchmark revisions are large. To avoid this, certain quadratic minimisation techniques were applied. Alternatively, one might have used a quarterly chain index if data of sufficient quality were available. A judgement was made that quarterly weights could reliably be used. A detailed methodology paper on the use of the Chain Fisher Volume index for income and expenditure accounts was prepared and disseminated on Statistics Canada’s web site in May 2001.

In addition to endorsing the theoretical superiority of the Fisher volume index, there was also the issue of comparing Canadian data on GDP with our most important neighbour, the United States. The USA had already used the Chain Fisher index since 1998 and our numbers were no longer comparable with them, which was quite annoying to our major users, both in the government and private sectors. Thus it was agreed to switch to a quarterly chain Fisher volume index, with effect from the first quarter release of national income and expenditure accounts, and to go back to 1981. Given time and resources, the national data for the period 1961 to 1981 would be developed in 2002. The annual Provincial data will be developed later.

Decision: Produce constant price national quarterly income and expenditure accounts using a quarterly chain Fisher volume index with effect from the first quarter release in May, 2001 and bring back the series to 1981. At the same time, express all the volume series in chained 1997 dollars. Produce the series for the period 1961 to 1981 also using a chain Fisher volume index in 2002. Produce annual provincial income and expenditure accounts also using a chain Fisher volume index in 2002.

13. Constant price detailed annual industry measures

Issue: Should we switch to an annual chain volume index for detailed annual industry measures produced in the IO tables from the base weighted 1992 Laspeyres volume index?

Pre-revision treatment: We have constant price IO tables for 1961 to 1971 at 1961 prices, tables for 1971 to 1981 at 1971 prices, tables for 1981 to 1986 at 1981 prices, tables for 1986 to 1992 at 1986 prices and tables for 1992 onwards at 1992 prices. We have never produced a continuous time series of IO expressed at any single year dollars.

Discussion notes:

In view of time and resource constraints, as well as our primary objective of using identical or at least similar conventions for volume series, both at macro and detailed industry levels, such that the results report the same story, it was decided to proceed in this domain in two phases. In Phase

Statistics Canada - 376 - Collected Articles of Kishori Lal One ending in September 2001, try to reduce statistical differences as much as possible and in Phase Two, lasting over the next two years, to come as close as possible in our use of volume indices.

September 2001. As noted earlier, it was decided to adopt, effective May 31, 2001, the chain Fisher volume index for CSNA macro series. The 1992 fixed Laspeyres base for annual detailed industry volume measures was simply not acceptable. It was too far away from the current period and the results would have become totally incomparable. Thus, it was recommended to switch to the 1997 fixed Laspeyres base from the 1992 base for the annual industry series effective September 2001. All the current price changes listed in this report should be incorporated in the IO tables for the entire annual historical series back to 1961. Given time and resource constraints, it was also decided to make the volume changes from 1981 onwards only. Constant price IO tables for 1981 to 1986 will be recalculated at 1981 prices, IO tables for 1986 to 1992 at 1986 prices, IO tables for 1992 to 1997 at 1992 prices and IO tables for 1997 and 1998 at 1997 prices, all effective September 2001. We will then chain link the value added component only from the IO tables and produce continuous time series for value added by industry, all expressed at 1997 reference year prices.

Work during the next two years

In view of the adoption of a chain Fisher volume index for macro series such as GDP and its components and CSNA’s highly integrated system, it is necessary to examine the possibility of adopting an identical or at least similar volume measure for detailed industry accounts. The benchmarks for industry accounts in the CSNA are produced in the context of IO tables. There are certain unique features of IO tables which need to be taken into consideration in discussing the adoption of an appropriate index number formula for volume measures. In the current price IO tables, there are many identities (such as the identity that the supply of each product must be equal to its demand, and that the output of each industry must be equal to its intermediate expenditures on goods and services and its value added) which must be preserved. Indeed, balancing supply and use of all the detailed products is the most important activity in the compilation of IO tables at current prices. Once we have developed, at whatever detail level, consistent deflators for goods and services, their application to balanced current price IO tables automatically produces balanced supply and use tables at constant prices. Value added at constant prices is produced by using the SNA93 preferred double deflation method. Output at constant prices less intermediate expenditures at constant prices residually results in constant price value added. Value added and final expenditures at constant prices are automatically equal, because we remove from both the use and supply of products a common intermediate expenditure block.

There are more than 650 products and more than 250 industries ( for 1961 to 1997 tables, and even more for 1997 onwards at NAICS level), to each of which the IO identities must apply. Lack of additivity would be annoying to users of detailed industry statistics. In view of this, SNA93 suggests the following: “Although the preferred measure of real growth and inflation for GDP is a chain Fisher index, or alternatively a chain Laspeyres or Paasche index, it must be recognized that the lack of additive consistency can be a serious disadvantage for many types of analysis in which the interrelationships between various flows in the economy are the main focus

Statistics Canada - 377 - Collected Articles of Kishori Lal of interest. Most macroeconometric models fall into this category. It is therefore recommended that disaggregated constant price data should be compiled and published in addition to the chain indices for the main aggregates. The need to publish two sets of data that may appear to conflict with each other should be readily appreciated by analysts engaged in macroeconometric modelling and forecasting. Users whose interests are confined to a few global measures of real growth and inflation can be advised to utilize the chain indices and ignore the more detailed constant price estimates”( paragraph 16.75)

Though increasing in number, only a handful of countries produce a time series of annual IO tables at current prices and even fewer at constant prices. In Canada, we have a time series of annual current price IO tables going back to 1961. Constant price IO tables have used Laspeyres volume index, the same base as the pre-revision income and expenditure accounts. However, when the base year for calculating constant price IO tables was changed, the product detail (output of products and the intermediate use of products) in the IO tables for the back period was never chain linked. On the other hand, the value added component was chain linked for benchmarks to monthly constant price GDP by industry series (see more on this later). IO tables remained at original base year prices. Chaining IO tables would destroy the additivity for each of the hundreds of products and industries. Lack of additivity in the IO matrix format would negate IO’s most important property, (i.e., balances of supply and use of products and balances of output and the corresponding intermediate expenditures and the resulting value added).

It was decided not to chain link IO tables, and to the best of our knowledge, no other country does it. SNA93 clearly states: “It is therefore recommended that disaggregated constant price data should be published for as many of the flows of goods and services in the System as possible, with a change of base year about every five years. When the base is changed it is customary to link the data on the old base to the data on the new base rather than to carry the rebasing backwards” (paragraph 16.76). It further states: “Annual chaining is simply the limiting case in which rebasing is carried out each year instead of every five or ten years” (paragraph 16.77). Results from the use of annual chaining would be the closest to the ones derived from the use of chain Fisher volume index.

As noted above, value added by industry, produced through the preferred double deflation method, is chain linked and expressed at a single year dollars, the most recent one at 1992 dollars. These constant price value added series constitute the benchmarks for monthly detailed constant price industry measures which go back to January 1961. These monthly industry measures must be, as closely as possible, comparable to quarterly and annual macro income and expenditure accounts which now use a chain volume index. This has created a dilemma. We need to examine options which should eliminate or reduce any difference, between the volume growth rates for the total economy and its major components derived from macro series and detailed industry series. Having this in mind, it was decided to choose the SNA93's recommended annual chaining route, using annual Laspeyres volume index. Thus, we will deflate year t+1 table using prices of year t and year t+2 table using t+1 prices and so on. We will put together a series of tables, t+1 into t prices, t+2 into t+1 prices and so on. Calculate the time series of volume measure of value added by industry produced through IO tables by chaining, that is by cumulating the year over year movements, and then express these series at the same reference year 1997 dollars as the macro accounts. It would be instructive to compare the growth

Statistics Canada - 378 - Collected Articles of Kishori Lal rates, thus calculated, for the total economy resulting from industry value added series with the chain Fisher volume growth rates for macro series. If the differences of results from these two methods are statistically insignificant, we might recommend both methods, one for macro and the other for industry detail with confidence. This would be, as noted above, very much in tune with the SNA93 recommendation. One may go even further and also produce some well known aggregates of value added, such as for the total economy and its major sectors-business sector, manufacturing sector etc.,- by using a chain Fisher volume index.

Our objective is to produce chain linked value added series from IO tables for benchmarks to monthly GDP by industry measures for as current a period as possible. But at present, IO tables are produced with a lag of 32 months from the reference year. Thus, we need to project good quality balanced IO tables using supplementary and indicator information to reduce the lag from the present 32 months to, first, 20 months and, then, if possible, 8 months. If this works out, we will have quite current weights for the detailed industry statistics, not identical but close, to the weights used in the macro series.

Decision: Incorporate all current price changes listed in this report in the IO tables for the entire historical series back to 1961 and also produce IO tables for 1998. Recalculate constant price IO tables for 1981 to 1986 at 1981 prices, IO tables for 1986 to 1992 at 1986 prices, IO tables for 1992 to 1997 at 1992 prices and IO tables for 1997 and 1998 at 1997 prices, all effective September 2001. Also effective September 2001, chain link the value added at basic prices component only from the IO tables and produce continuous time series for value added by industry, all expressed at 1997 reference year prices. During the next 18 to 24 months, project current price IO tables with a lag of first 20 months and then, if possible, 8 months from the reference year. Effective September 2002, produce annual constant price IO tables for 1961 to 1999, if feasible, also project for 2000, using rolling annual Laspeyres volume index and chain link the value added by industry from these tables for the entire period. Effective September 2003, project, if feasible, IO tables for two extra years, 2001 and 2002. Produce, if statistically necessary, a chain Fisher volume index for the overall economy and also for major sectors such as business sector, manufacturing sector, etc.

14. Constant price detailed monthly industry measures

Issue: What should be the relationship between the detailed annual constant price value added by industry measures and the detailed monthly constant price value added by industry measures?

Pre-revision treatment: Monthly constant price value added (GDP) by industry is benchmarked to the annual measures produced through the IO tables. Conventions adopted in the annual series are, by and large, adopted in the monthly measures for the years the annual series are available. The calculated monthly movements are preserved as much as possible when benchmarked to the annual numbers using statistical minimisation techniques. As such, the change from December to January is not unduly exaggerated due to the incorporation of annual benchmarks. Rebasing is done the same way in both series, annual and monthly. Monthly series are projected for the latest 32 to 43 months as the benchmarks are not available.

Statistics Canada - 379 - Collected Articles of Kishori Lal Discussion notes:

In line with the above discussion and decision for the annual constant price series, monthly series should be produced in September 2001 using the Laspeyres volume index, re-weighted to 1997 from the present 1992 base. In September 2002, the monthly series will shift to an annual chain Laspeyres volume index. In September 2003, it will remain on annual chain Laspeyres volume index, except that the IO benchmarks will be available with lags of between 8 and of 20 months.

The issue whether one should use a chain Laspeyres or a chain Fisher volume index for detailed industry series was discussed above in relation to the annual series and the same notes apply to monthly series. If the differences of the results from using these methods are statistically insignificant, we recommend using chain annual Laspeyres method for detailed monthly industry series, as this would be in tune with the SNA93 recommendation. For the period without annual benchmarks, the monthly series will remain based on the latest benchmark numbers unless some new current price series or indicators are developed. A feasibility study on this question should also proceed during the next 18 to 24 months.

Again as in the case of the annual series, one must examine the possibility of also producing certain major aggregates such as total economy, business sector, manufacturing sector etc using a chain Fisher volume index for comparison with other macro series produced on the basis of this index. This would, of course, require developing current price data for the current periods.

Decision: Produce monthly real GDP by industry for 1997 to July 2001 in September 2001, rebasing the series to 1997 prices. Produce series for 1992 to 1997 at 1992 prices, for 1986 to 1992 at 1986 prices and for 1981 to 1986 at 1981 prices, all expressed at 1997 reference year prices. In September 2002, produce monthly real GDP by industry by incorporating IO annual benchmarks derived from an annual chain Laspeyres volume index, for the years 1961 to 1999, and, if feasible, also for 2000. In September 2003, add benchmarks for two extra years, 2001 and 2002, if feasible. Produce, if statistically necessary, a chain Fisher volume index for the overall economy and also for major sectors such as business sector, manufacturing sector, etc.

All the decisions listed in this report have been implemented except the chain volume index part for detailed annual and monthly GDP by industry statistics. Given that CSNA is highly integrated, it is suggested that a progress report be presented, in early spring 2002, for review by senior management. This review may include, if judged necessary, the conceptual underpinning of chain volume index formulation as well as the resource allocation for expediting the work.

Concluding Remarks

This is the third in the series of reports dealing with the implementation of SNA93 in the CSNA. The first, issued in 1998, recorded all hundred or so changes that were made in our accounts following the implementation of SNA93. The second report, also issued in 1998, recorded the remaining differences between the CSNA and SNA93 and their reasons. Since then, with the implementation of SNA93 in many other countries, particularly in the United States, issues of international comparability have assumed greater importance. This has led to reexaminations of our previous conventions and decisions. This report records these deliberations and the changes

Statistics Canada - 380 - Collected Articles of Kishori Lal made to our previous decisions that were mostly listed in our report on Remaining Differences. It lists 14 issues, on 13 of which we have changed our previous treatment, thus reducing the remaining differences and bringing us even closer to SNA93.

The capitalisation of all development expenditures on software is the only change which has affected the overall level of GDP at current prices. The shifting of farm rents paid in agriculture industry from its intermediate expenditures to payments for property income has increased value added in agriculture, counterbalanced by an identical decrease of value added in the real estate industry, thus leaving total GDP unchanged. The revaluation of output of holding companies has increased value added in holding companies, counterbalanced by an identical decrease of value added in their subsidiaries, thus leaving total GDP unchanged. Reclassification of payments by enterprises for electro-magnetic spectrum for mobile phones from taxes on factors of production to payments for property income on non-produced assets does not affect value added but it does reduce the value of taxes, counterbalanced by an identical increase of income from royalties in the government accounts. We have reexamined the treatment of stock options following the changes in the February 2000 budget of the Government of Canada, and have reaffirmed our earlier decision to continue adding such options to labour compensation, even if some of them are not taxed anymore. Materially, a very important conceptual change (worth more than $100 billion) in our treatment of government employees pension plans has increased the value of household assets, counter-balanced by an identical increase of government liabilities, and has also involved changes in sector income and outlay accounts. Again, this change had no effect on GDP. We have changed our valuation of value added from factor cost to the SNA93 recommended basic prices. This change makes our presentation comparable to most other countries which have implemented SNA93 and also makes it much more useful for productivity analysis. A very important change is in the area of deflating current price values of GDP and its components. We have now shifted from using a fixed weighted base (Laspeyres volume) index to SNA93's recommended current weighted (chain Fisher volume) index for macro series. This has brought our practice in line with that of the United States. Thus, when we now compare our GDP growth rates with those in the United States, we are comparing apples to apples, not apples to oranges. This report, like our earlier reports, documents the conceptual, statistical and methodological changes that we have made in our accounts and will be widely disseminated to our users. This helps them to fully appreciate the changes and their reasons, and thus provides them additional confidence in our statistics. ______

Note:

I am grateful to many colleagues working in the System of national Accounts branch at Statistics Canada for their help and advice and comments in the write-up of each of the issues listed in this paper, particularly, Abe Tarasofsky, Yusuf Siddiqi, Pat O’Hagan, Chris Jackson, Anna Macdonald, Nugent Miller, Philip Smith, Karen Wilson, Michel Girard and Hans Messinger.

This paper was presented at the OECD Meeting of National Accounts Experts, Paris, October 9– 12, 2001.

Statistics Canada - 381 - Collected Articles of Kishori Lal

Introduction

The 1993 SNA was implemented in Canada in November 1997 and all national accounts series- annual, quarterly and monthly, both at current and constant prices- were revised back to 1961. There were changes in classification of sectors and transactions, concepts and methodology. As well, we removed the statistical breaks in earlier series that arose due to our revision policy. In the spring of every year, we revise, if necessary, our national accounts series for the latest four years. Statistical breaks for earlier periods are removed only at the time of historical revisions, such as the one done in November 1997. This was the fifth and the most comprehensive historical revision of the Canadian SNA series since 1961, the earlier ones were done in the late 1960's, the late 1970's, in 1985 and in 1990. As our historical revisions have been done almost every decade, and more frequently since the 1980's, statistical breaks in the Canadian system have remained only for a short period.

There are three inter-related backcasting issues which compilers of national accounts must resolve when implementing the 1993 SNA in their own accounts. Should the new and the revised series that are consistent with the 1993 SNA be backcast? What methods should be used to backcast? How far back such series be backcast? These issues have to be resolved in the context of a country=s statistical development, availability of resources and the needs of users for long time series.

In Canada, we have an integrated system of national accounts going back to 1961. This system includes annual input-output tables, annual and quarterly income and expenditure accounts, monthly real GDP by industry, quarterly financial accounts and annual balance sheet accounts. Canadian compilers and users jointly decided that it would be appropriate to backcast the new and the revised series to 1961, rather than limiting the revision to a short back period, say to 1991. The availability of a well- developed integrated set of accounts back to 1961 also helped us determine the backcasting methods. We made every effort to recompile the series for the back period for the institutional units, reflecting the changes entailed by the implementation of the 1993 SNA, using, as much as possible, their own accounting or other records available in our earlier worksheets.

Classification of sectors and transactions

As we examined the 1993 SNA in the context of its implementation in Canada, we realised that the sector classification of several important institutions such as universities , special care facilities, several government business enterprises, and many more, would need to be changed to reflect the new guidelines. Hundreds of institutional units were reclassified, the reasons were fully documented, and a revised list was prepared and made available to users upon request. The effective date for change was the inception date of the institutional unit or 1961, whichever came later. For example, the universities were reclassified from the non-profit institutions serving households (NPISH) sector to the government sector. Similarly, many special care facilities were reclassified from the NPISH sector to the government sector. The Canadian Broadcasting Corporation was reclassified from the government business enterprise sector (corporate sector) to the government sector. To incorporate all these changes, every effort was made to use the accounting records or other records in our earlier worksheets of the institutional units for the back period. It was a time- consuming exercise but well worth in terms of proper historical time series analysis.

Statistics Canada - 385 - Collected Articles of Kishori Lal Reclassification of transactions in the government sector constituted the next group of changes made in implementing the 1993 SNA. We also took advantage of this opportunity to better realign the conventions used in the government finance statistics with those of the SNA. For example, should the profits of government monopolies, such as lotteries, be treated as investment income of government enterprises or as taxes on products? Should grants paid in lieu of taxes by senior levels of governments to local governments be treated as grants or as taxes? Should royalty revenues received by governments from business corporations for their mining and logging activities be treated as other taxes on production, or as investment income of government, or as sale of goods and services? Should seigniorage on coins be treated as government revenues from sale of goods and services or should it be ignored? The classifications of many of these significant transactions in the earlier series were different from those recommended in the 1993 SNA. A complete list of classification changes of transactions was prepared, the reasons for changes were documented, and the list was made available to users upon request. As in the case of reclassification of institutional units, our approach to backcasting in this domain was to recompile the revised series by incorporating accounting or other records related to these transactions for the entire back period.

Capital Account

There were several changes made in the capital account. In terms of value, the most significant were the reclassification of both purchases of non-military equipment by the Defence Department and expenditures on mineral exploration by mining units, from intermediate expenses to capital formation. In our capital stock series (though not in our SNA series), purchases of capital equipment by the Defence Department were always included in capital formation, and were also separately identified. In our pre-revision annual input-output tables for 1961 and onwards, the non-military equipment by the Defence Department was classified as intermediate expenditure and separately identified. These two sources permitted us to reclassify these expenditures from intermediate to fixed capital formation for the entire back period. In both the capital stock and the SNA series, on-site mineral exploration charges were always capitalised. The other exploration charges, called general or off-site charges, were classified as intermediate expenditures in the earlier accounts. Fortunately, total exploration charges were published for most of the back period by the Department of Energy, Mines and Resources. This permitted us to make an adjustment to our series back to 1961.

In addition, based on analysis in the 1950's, the assumption was made that a fixed proportion, namely 10 %, of operating expenditures for establishments in the manufacturing, trade and utilities sectors, should be classified to gross fixed capital formation. This required a specific adjustment to the reported series since the 1950's. We were not comfortable with this assumption, as it differed from the records that the establishments were using to develop prices for their commodities. For the 1997 SNA historical revision, it was decided to undo this adjustment for the entire period which was easy to do this as the adjustment was always separately identified.

Household expenditures

A major change was made in the classification of household expenditures of goods and services. We used the 1997 version of Classification of Individual Consumption by Purpose ( COICOP) as prepared by the OECD. This required us to recompile several underlying details in the household

Statistics Canada - 386 - Collected Articles of Kishori Lal expenditure series back to 1961. Commodities were reassigned from one function to another within the household expenditures categories This backcast exercise was facilitated by our input-output tables in Canada going back to 1961, with about 200 industries, 140 final demand categories and 600 commodities.

International Trade

The most important change for backcasting the international trade series involved the revaluation of exports and imports at the border of the exporting country. This change did not affect the value of total of goods and services but the level of value of services changed, counterbalanced by a change in the value of goods. In this case, the backcasting consisted in identifying the basic records on inland freight which were available in past publications under the transportation account and to reclassify its value to goods. This backcasting went as far back as 1926 in the balance of payments series.

Financial services

Several changes were made in the area of financial intermediation services indirectly measured (FISIM). By and large, the changes for the back period were made using basic records in our worksheets, not fixed proportion techniques. We had always allocated FISIM to users, both borrowers and lenders, even when the 1968 SNA did not recommend it. We continue to allocate FISIM to users as we have implemented the 1993 SNA but we have recalculated FISIM for own funds, for the period 1961 to date, such that only the borrowers= portion of service charges remains in FISIM. This recalculation was based on records in our worksheets for the back period. Note that we still differ from the 1993 SNA recommendation for own funds, as we believe that the 1993 SNA needs to reconsider its position on this issue. The calculation of the value of output of central bank and its allocation was changed for the entire back period. We now calculate the value of output of our central bank, the Bank of Canada, by its operating expenses and allocate the use of the entire output to the federal government. Previously, we used to allocate part of the output to the household sector. The value of the output of insurance industries has increased dramatically due to the inclusion of investment income of technical reserves. Again, as much as possible, we have gone back to basic records for this recalculation.

Concluding remarks

The 1993 SNA is a coherent, integrated and fully articulated document. Its implementation has entailed many changes from previous practices: reclassification of institutional units from one sector to another; reclassification of transactions from one class to another; reclassification of household expenditures from one function to another; expansion of several areas for capital formation; revision of guidelines for valuation of the output of several financial institutions, and many more. Such changes, particularly the ones which are materially important, made for the current period can meaningfully be backcast only by reference to the accounting and other records for the periods in question and by replicating the new methodologies for the back period. Fixed proration formulas for backcasting are not very helpful for the kind of changes required to implement the 1993 SNA. One must be very leery of adopting any proration techniques or other fixed proportion formulae relevant to a recent single year for backcasting purposes, of estimates of values of the entities due to changes

Statistics Canada - 387 - Collected Articles of Kishori Lal in sector classification or of estimates of transactions due to changes in their classification. Similarly, changes due to reclassification of functions of household expenditures for the back period made by using a fixed proration formula derived from a recent single year would also be hazardous, as they would carry back commodities which may not even have existed in the back periods.

Given many other compelling priorities, compilers in some countries might be tempted to limit the 1993 SNA series to the 1990's only, produce data for one overlapping year, and leave the back period on the old basis. As a result, users would need to compute their own long time series for economic and historical analysis. We strongly urge that, to the fullest extent possible, one should resist this temptation. The primary raison d=être of the SNA is to produce long time series of consistently defined and measured data, for analysis; outside users can never produce series as good as those produced by the official compilers, which are made available to everybody at the same time.

Note:

I am grateful to Lucie Laliberté, Yusuf Siddiqi and Abe Tarasofsky for their very useful comments and suggestions in the preparation of this paper. This paper was presented at the OECD Meeting of National Accounts Experts, Paris, 21-24 September 1999.

Statistics Canada - 388 - Collected Articles of Kishori Lal

Abstract

Under any degree of inflation, high or low, one observes that the value of changes in inventories (VPC) is generally different when it is calculated at the quarterly interval and the four quarters are aggregated into a year compared with its calculation done at the yearly interval. This is troubling, particularly because the VPC plays an exceptionally important role in the measurement of the value of output. It is argued in this paper that it is an inherent problem as one of the basic axioms of annual accounts is violated, namely, the assumption of price homogeneity over an accounting period. Given this inherent problem, some pragmatic alternatives need to be considered: a) the flows reflected in annual accounts supersede the flows reflected in the quarterly aggregations, the rationale being that for calculation of VPC, the annual data sources are always of better quality; b) the aggregation of the totals of the four quarters supersedes the totals of the annual accounts, the rationale being that in conditions of inflation, there is more price homogeneity during quarterly intervals than during annual intervals; c) allow the two aggregations, quarterly accounts adding to an annual accounts and annual accounts developed directly from annual data , remain different. The third alternative is the one that we prefer in Canada and is in practice.

Introduction

Among the many distortions of financial reporting caused by changes in the general price level (inflation or deflation) is the problem that the aggregate of data for the sub-periods that compose a given period may well be different from the corresponding data for the period as a whole. As is explained in the 1996 OECD publication entitled: @Inflation Accounting : A Manual on National Accounting Under Conditions Of High Inflation@ (in short Inflation Accounting), prepared by Peter Hill, factors that can produce this situation include uneven fluctuations throughout the period in either or both transaction flows (including inventory turnover) and general price levels. This note is concerned with a specific aggregation problem arising from changes in inventories.

When price changes occur while goods are held in inventory, one observes that the value of changes in inventories (VPC) is generally different when it is calculated at the monthly interval and the three months are aggregated into a quarter compared with its calculation done at the quarterly interval; similarly, the VPC is different when it is calculated at the quarterly interval and the four quarters are aggregated into a year compared with its calculation done at the yearly interval. This is troubling, particularly because the VPC plays an exceptionally important role in the measurement of the value of output. By and large, the current price value of output of any good produced equals the value of its sales and other uses plus the value of changes in its inventories. Once the VPC is different, the value of output is different, the value added is different and this difference is carried forward to all the sequence of accounts, up to the closing balance sheet when it is eliminated. Is this difference inherent or a result of surveys of varying quality for different intervals?

Statistics Canada - 391 - Collected Articles of Kishori Lal

Principles of measurement

Let us first look at the principles underlying its measurement in the national accounts. The 1993 SNA states: AThe basic principle underlying the measurement of changes in inventories is that output should be recorded at the time it is produced and valued at the same price whether it is immediately sold or otherwise used or entered into inventories for sale or use later. No output is recorded when goods produced previously are withdrawn from inventories and sold or otherwise used. It follows that entries into inventories must be valued at the basic prices prevailing at the time of entry, while withdrawals must be valued at the prices at which they are then sold.@ (paragraph 6.58). It further states: AWhen prices are stable, the measurement of changes in inventories is relatively simple. However, when there is inflation, significant price increases may occur while goods are held in inventory. Holding gains accruing on goods held in inventory after they have been produced must not be included in the value of output@ (paragraph 6.59).

The principle underlying the measurement of VPC is well stated in the 1993 SNA but it does not provide any operational guidelines for aggregation of VPC from sub-annual data into an annual series when the sub-period price changes are significant. When the sub-period price changes are significant, the economic principles underlying the process of aggregation by which annual accounts are constructed need to be reexamined. The above noted OECD document on Inflation Accounting states: ASimple addition of the values of flows occurring at different points of time throughout the accounting period as a whole may not necessarily produce the most meaningful and useful annual accounts under conditions of high inflation@ (page 31). A basic axiom of current price national accounts for any period, monthly, quarterly or annual, is that there is price homogeneity over that period. Though this is not explicitly stated in the 1993 SNA, it is implicitly assumed so. This axiom is explicitly stated in the IMF=s Textbook on Quarterly National Accounts Compilation, draft March 2000 (in short IMF=s Quarterly Textbook).

In paragraph 1.10 of the IMF=s Quarterly Textbook , it is stated that, at times of high inflation, annual national accounts are less useful because one of the basic axioms of annual accounts is violated, namely, the assumption of price homogeneity over time. Although this axiom never fully applies in practice (unless inflation is zero and relative prices are stable), with low inflation, the usefulness of annual accounts is not hampered too much. However, in a situation of high inflation, summing up the current price data from quarterly accounts over a year becomes meaningless. Under conditions of high inflation, the Inflation Accounting manual recommends to restate the sub-period accounts into constant price level (CPL) accounts such that the weights reflect the underlying economic reality. Note that in the CPL accounts for the sub-periods, the same price level is used throughout the year.

High inflation or moderate inflation or significant price changes for sub-periods within an accounting period result in similar conceptual difficulties for aggregation purposes, though their intensities are different. To emphasise this point, the OECD=s Inflation Accounting document states: A...It is worth noting that the procedures indicated above for the calculation of annual indices as weighted averages of sub-period indices using CPL values as weights should be applied even under conditions of low or zero inflation. Consider the limiting case of zero inflation in which the general index of inflation remains constant from sub-period to sub-period. Although there may be no general inflation, some degree of variation in relative prices is bound to be present in the accounts if only because of

Statistics Canada - 392 - Collected Articles of Kishori Lal seasonal variations in prices@ (page 52). It further states:@As soon as there is even moderate inflation it becomes imperative to do so using CPL accounts if precise and reliable measures of year to year quantity and price changes are required@ (page 53). Thus, in principle, annual accounts are meaningful only when the sub-period price changes are insignificant. Insignificant price changes produce, by definition, insignificant holding gains for changes in inventories. Thus, there is an inherent difference between the VPC for an annual total calculated by aggregating it from four quarters and calculated directly from annual changes in stocks when there are any holding gains in the sub-periods.

Statistical difficulties

In addition to the general reasons mentioned above, there are several specific statistical reasons why the results from the different intervals are generally not the same: the data sources are different; the data sources on the needed detail on composition of goods held in inventories, their turnover period and the prices charged for individual commodities are of poor quality at the annual interval but are rarely available at the sub-annual intervals. Additional conceptual and practical complications arise in valuing output and inventories in sectors, such as agriculture, whose output is subject to seasonal fluctuations in price due to changing demand or supply conditions over time. The 1993 SNA states: A Suppose an annual crop is harvested at one point of time, put in storage and then gradually sold off, or used over the remaining twelve months. Suppose further that the price gradually rises to reflect the increased scarcity of the good until the next harvest. In the absence of general price changes, the increase in the price of the good while being stored must be interpreted as measuring the value of an addition to work-in-progress.... The goods withdrawn from storage some months after the crop is harvested are economically different from those that entered because of the changing supply conditions over time.@ (Paragraph 6.108). Further to the conceptual issue just noted, there is the practical problem of not knowing what the actual price received is going to be when crops are sold through some type of marketing boards. Thus one must record an initial price for the quarterly accounts to be revised later when the actual price gets known. This approach, or one similar to it, is used in Canada for valuing output of wheat: the Canadian Wheat Board, considered as farmers= cooperative, receives a significant portion of the harvest and pays an initial price to farmers; its profits, known only when the wheat is sold many quarters later, are given back to farmers as final payment. It is the total payment, initial plus final that determines the net income of farmers and, hence, the value of their output, accruing for the period when the crop was produced.

The practical difficulties in measuring VPC are overwhelming for any interval, more so for sub- annual series, such that, in many OECD member countries, there is no independent estimate of VPC; instead, it is residually estimated, thus it includes all the errors of the other accounts. For example, of the 18 countries listed in a recent OECD document, Quarterly National Accounts Sources and Methods by OECD Member Countries (Paris 1996), 9 countries-Austria, Finland, France, Germany, Italy, Netherlands, Norway, Spain, Switzerland- calculate VPC as the difference between the production based GDP and all the components except VPC of the expenditure based GDP, thus it contains the errors from both sides of the accounts. The other 9 countries- Canada, United States, Japan, Australia, New Zealand, Denmark, Sweden, Turkey and United kingdom- make independent estimates, with data sources of varying quality, relatively good for those sectors for which physical quantities are available to mediocre and poor for others.

Statistics Canada - 393 - Collected Articles of Kishori Lal

As noted above, once the VPC is different, the valuation of output and of holding gains is also different and the difference arising from the sub-period (say quarters ) addition of values of flows to arrive at a higher interval (say a year) seems inherent as the holding gains occur when price of goods in the opening inventory is different from the price of goods in the closing inventory for a specific interval. Within an accounting period, be it monthly, quarterly or annual, VPC and holding gains are typically recorded by referring to the book values of opening inventory and closing inventory. If there is no opening inventory and there is no closing inventory within an accounting period, it is more than likely that the value of both VPC and holding gains would be set at zero even if there are price movements and there are physical changes of inventories during that period. This difference in current price valuation is demonstrated using an example, listed in the following table, of a producing unit A, producing only a single good B, using its quarterly and annual data.

Statistical example

Unit A produces in the first quarter 100 tons of good B, and the basic price per ton in the market is $10. Thus, the value of output is $1000. There is no opening inventory at the beginning of the first quarter. Nothing is sold from the first quarter=s output and the entire output is added to inventory. The value of output is $1000 and the value of the goods entering into inventory is also $1000, as recommended by the 1993 SNA. In the second quarter, Unit A does not produce any output and the price jumps to $15 per ton. The entire stock from the first quarter is now sold for $1500. Thus, goods withdrawn from inventory are valued at $1500. There is now a holding gain of $500 from goods held in inventory. This, again, is consistent with the recommendation of the 1993 SNA. Nothing is left in inventory at the end of the second quarter. Unit A produces 200 tons of good B in the third quarter and the price collapses to $8 per ton. 80 tons are sold and the rest is added to inventory. Output is valued in the third quarter at $1600 (200 tons at $8), sales are valued at $640 (80 tons at $8), and the goods entering into inventory are valued at $960 (120 tons at $8). Unit A does not produce any output in the fourth quarter as it has 120 tons of stock in inventory. Price goes up to $12 in the fourth quarter and Unit A sells its entire stock of 120 tons for $1440 (120 tons at $12). Withdrawals from inventory are valued at $1440 and Unit A makes a holding gain of $480, as the goods entering into inventory at the end of the third quarter are valued at $960.

Statistics Canada - 394 - Collected Articles of Kishori Lal

Current Price VPC , Output and Holding Gains

Y0 Q1 Q2 Q3 Q4 Q1..4 Y1 Output quantity 100 200 300 300 Sale quantity 0 100 80 120 300 300 Unit sale price 10 10 15 8 12 Output value 1000 1600 2600 3580 Sale value 0 1500 640 1440 3580 3580 Inventory-opening quantity 0 100 120 0 0 Inventory-closing quantity 100 120 0 0 Inventory-change in quantity 100 -100 120 -120 0 0 Inventory-opening book value 0 1000 0 960 0 0 Inventory-closing book value 1000 960 0 0 0 Inventory-change in book value 1000 -1000 960 -960 0 0 VPC 1000 -1500 960 -1440 -980 0 Inventory- holding gain 0 500 0 480 980 0

Let us now summarise the economic activity in each of the four quarters. In the first quarter, output is valued at $1000 and is reflected in an addition to inventory at $1000. In the second quarter, withdrawals from inventory are valued at $1500, the sales are $1500, and unit A makes a holding gain of $500. In the third quarter, output is valued at $1600, sales are valued at $640, and the addition to inventory is valued at $960; again, there is a balance between supply and use. In the fourth quarter, sales are valued at $1440, withdrawals from inventory are valued at $1440, and unit A makes a holding gain of $480. Adding all four quarters together, output equals $2600 ($1000+$1600), sales are valued at $3580 ($1500+$640+$1440), the value of changes in inventory is minus $ 980 ($1000-$1500+$960-$1440). The value of output for all four quarters at $2600 accounted for by sales at $3580 less VPC of $980. The holding gain of $980 is not included in the value of output or in VPC.

We now ask unit A in an annual survey, to provide information on output, sales and changes in inventory. Unit A reports that it produced 300 units in the year and sold all of them; there was no opening inventory and there is no closing inventory. The entire output of 300 tons is valued at the transaction price of $3580 in the market; thus, the supply is balanced with use. In this example, the construction of an annual account by adding four quarters generates value of output of $2600, while constructing the annual account from the annual survey generates value of output of $3580, a difference of $980. The same difference of $980 also emerges on the expenditure side: the value of

Statistics Canada - 395 - Collected Articles of Kishori Lal changes in inventory arrived at by adding four quarters produces a value of minus $980 whereas it is zero from the annual survey.

Three alternatives

We have already noted that the 1993 SNA does not provide any operational guidelines on aggregation over goods and services and over intervals of time within a given accounting period and that the OECD=s Inflation Accounting manual has specifically raised the issue of aggregation of quarterly accounts for annual accounts. There are two alternative ways to present annual accounts derived from annual surveys and annual accounts derived from the aggregation of four quarters for value of output, value of changes in inventory and values of holding gains: a. The flows reflected in annual accounts supersede the flows reflected in the quarterly aggregations, the rationale being that for calculation of VPC, the annual data sources are always of better quality. In our example, this means that the unit price used for valuing quarterly output is the actual unit sale price, whenever the sale occurs during the accounting period. Thus, the value of the first quarter=s output of 100 units is $1500 and the value of the third quarter output is $2080 (640+1440). As noted above, this approach is in the spirit of the 1993 SNA recommendation for valuing output and changes in inventory for goods subject to seasonal fluctuations, and which are stored for future sale, as in agriculture. b. The aggregation of the totals of the four quarters supersedes the totals of the annual accounts. The sales total remains the same at $3580. The value of annual output is $2600, and of VPC is minus $980. The value added of the annual accounts is lower by $980, as there are holding gains of $980 to be deducted. This approach does not require revision of the quarterly accounts and it follows the guidelines in the 1993 SNA for all sectors except those, like agriculture, where the higher values of the stored goods are not due to a price change but to an additional quality. The negative value of $980 of changes in inventory, when there is zero physical change of inventory remains counter- intuitive.

Note that both of the above alternative approaches produce identical closing balance sheets for the producing unit. The quarterly accounts aggregation approach compared with the annual accounts approach has lower financial assets in the amount of $980, but it has an additional $980 in holding gains in the revaluation account. The annual accounts approach, compared with the quarterly accounts approach, has higher financial assets of $980 but there are no holding gains in the revaluation account. c. The problem is inherent under any degree of inflation, high or low, as long as we produce current price quarterly and annual accounts. Hence, a third alternative could be considered namely, to allow the two aggregations, quarterly accounts adding to an annual accounts and annual accounts developed directly from annual data, remain different. The 1993 SNA, as noted above, does not provide guidelines for inter-temporal aggregation within a given period. However, the OECD Inflation Accounting manual does provide a solution to this inherent problem with its recommendation to restate the current price quarterly accounts into CPL accounts (under all conditions of inflation, more strongly under conditions of high inflation). In the

Statistics Canada - 396 - Collected Articles of Kishori Lal absence of developing quarterly CPL accounts or until CPL accounts become popular among the OECD member countries, some compromise convention needs to be developed. The issue that must be addressed is which alternative better reflects the underlying economic data during the period, in the most transparent way possible, both from the perspective of compilers and users.

Present Canadian Practice

In Canada, the quarterly and annual income and expenditure accounts are produced using both income approach and expenditure approach. Conceptually, the two approaches should yield identical value of total GDP but in practice the results are always different. An exhaustive probing is done to uncover any possible errors in either side of the accounts. Once this is done and the two totals are still not the same, the difference is divided by two. One half of the difference, called statistical discrepancy, is added to the income side and the other half with an opposite sign is added to the expenditure side of the accounts such that the two are now identical in value. This Canadian practice is unique amongst the OECD member countries. All other member countries add the total difference to one side of the accounts, either explicitly or implicitly. Our practice derives from a conviction that once all the possible errors have been corrected, it is not proper to assume that the remaining errors belong to only one side of the accounts. Note that in the input-output accounts, no statistical discrepancy is shown.

The Canadian quarterly income and expenditure accounts are produced with a lag of two months from the reference quarter. The VPC component of the quarterly expenditure on GDP (like other components) is independently estimated (unlike the residual estimation technique used in many European countries) by broad industrial sectors. At the end of February, the initial estimates of the fourth quarter of the preceding year, along with the annual data are released. During the year, the quarterly accounts continue being revised as new and revised data become available. At the end of May at the time of the first quarter release, the data for the latest four years and seventeen quarters are subject to revision. This is the time when the results from the annual input-output tables are incorporated in the accounts. Note that the annual national input-output tables are produced with a lag of 28 months from the reference year. Thus, for example in May 2000, the results of final input- output tables for 1996 and preliminary tables for 1997 are used as benchmarks in the national accounts for 1996 and 1997 and both the quarterly and annual data are revised consistently. However, the VPC is not benchmarked, thus we continue to have two independent estimates of VPC, one in the income and expenditure accounts and the other in the input-output tables. This practice needs some elaboration for proper understanding.

On the expenditure side of the income and expenditure accounts, all components (in aggregate and detail) -such as household expenditures, government expenditures, gross fixed capital formation, exports and imports of goods and services- are identical in the two sets (input-output and income and expenditure accounts) except VPC. The VPC of the input-output tables is identical in value only of the sum of VPC and published statistical discrepancy of the expenditure side of the income and expenditure accounts. Similarly, on the income side , wages and salaries, mixed income (net income of the unincorporated business), net taxes (taxes less subsidies) on production are identical in the two sets of accounts , except gross operating surplus. Gross operating surplus in the input-output accounts is identical in value of the sum of gross operating surplus and published statistical

Statistics Canada - 397 - Collected Articles of Kishori Lal discrepancy of the income side of the income and expenditure accounts. This convention, of course, assures that the aggregate GDP at market prices is identical in the two sets and it derives from several pragmatic considerations. There is a detailed reconciliation process in place before the input- output benchmarks are incorporated in the income and expenditure accounts. Initial input-output estimates in total and in detail are reviewed by the professionals of the income and expenditure accounts division. The sources of differences in each area are analysed and discussed for arriving at a consensus estimate for both sets of the accounts. This process lasts over several weeks of March and April and benchmark estimates are agreed upon by the staffs of both the accounts and the senior management for again all the elements of final expenditures except VPC. The VPC estimates in both sets are also challenged and major differences are discussed, analysed and are narrowed, where possible. Given the inherent conceptual and statistical difficulties for calculating VPC, we recognise that there is no perfect answer. Thus, VPC calculated at quarterly intervals and aggregated to annual totals is not forced to agree with the annual total calculated in the input-output framework. It is recognised that VPC in the input-output tables continues to contain some margin of error. Similarly, it is recognised that VPC in the quarterly accounts continues to have several weak spots. Thus our pragmatic practice is to publish annual VPC in the input-output tables which is equal in total to the VPC in the income and expenditure accounts (estimated using quarterly intervals and the quarterly values aggregated to annual totals) plus the statistical discrepancy. This Canadian compromise permits users to have VPC by broad industry groups in the income and expenditure accounts and the commodity composition of VPC in the input-output tables. The final differences are not huge but are not suppressed, thus users can make their own judgement on how far they should carry the analysis on VPC and draw conclusions.

Concluding remarks

Given the importance of this topic, it is incumbent upon the national accountants throughout the world to reexamine the issue of measurement of VPC at sub-annual and annual intervals under conditions of inflation and arrive at a consensus (or provide a majority view) on the following: a) recommend to restate the current price quarterly accounts into CPL accounts as very well articulated in the OECD Inflation Accounting Manual or b) annual calculations supersede the calculations made at quarterly intervals or c) the aggregation of the four quarters supersede the annual calculations or d) allow the two aggregations remain different. In the absence of CPL accounts, this paper has argued that given the inherent problem of aggregation under any condition of inflation, the preferred solution, both from the perspective of compilers and users of data, is to allow the two aggregations remain different.

Note:

I acknowledge with gratitude for their useful comments and suggestions in the preparation of this paper: many colleagues in the System of National Accounts Branch at Statistics Canada, particularly Abe Tarasofsky; and many colleagues at the Joint ECE/OECD/Eurostat Meeting on National Accounts, Geneva, April 26-28, 2000, where an earlier draft of this paper was discussed, particularly Esben Dalgaard of Statistics Denmark.

Statistics Canada - 398 - Collected Articles of Kishori Lal

Introduction

Both Canada and the United States have a very comprehensive set of statistics on national accounts and these are, by and large, consistent with the revised world-wide guidelines on national accounting, the System of National Accounts 1993 (1993 SNA). The 1993 SNA was produced under the joint responsibility of the United Nations, the IMF, the Commission of the European Communities, the OECD and the World Bank. The Commission of the European Communities and its statistical bureau, Eurostat, have produced a national accounts manual, the European system of accounts, ESA 1995, (Luxembourg, 1996) for use in the European Union. The 1995 ESA is broadly consistent with the 1993 SNA as regards the definitions, accounting rules and classifications. To ensure that the methodological provisions in the 1995 ESA are strictly applied, the Council of the European Union has adopted this manual in the form of a Council Regulation in 1996, thus giving it a solid legal basis. All member countries of the European Union follow, and must follow by regulation, the European system of accounts, ESA 1995. Due to this specific use of the manual and their measurement in mind, the concepts in the 1995 ESA have been expressed in operational terms to provide greater precision, common understanding and consistent application throughout the European Union. The OECD uses a single joint questionnaire, Questionnaire SNA 93/ ESA 95, to collect national accounts data from all member countries, countries of the European Union, and others, including Canada and the United States.

We fully recognise the importance of an internationally comparable set of national accounts statistics amongst all OECD member countries, particularly between Canada and the United States, so that economic analysts can use our statistics and draw conclusions with confidence. However, there exist some differences with the 1993 SNA in both countries. Further, the departures from the 1993 SNA are not necessarily identical in the two countries Thus, we will attempt to have a list, as complete as possible, of all such departures so that the users can make meaningful comparison of the published national accounts data by the two countries.

The 1993 SNA is a vast framework and is built around a sequence of inter-connected flow accounts linked to different types of economic activity taking place within a given period of time. It starts with the production account, links it to the income and outlay account, to the capital finance account and finishes with balance sheets that record the values of stocks of assets and liabilities held by institutional units or sectors at the beginning and end of the period.

In Canada, it is Statistics Canada which is solely responsible for all SNA components whereas in the United States. the Bureau of Economic Analysis (BEA) is responsible for most of the SNA components. Only a handful of SNA components are produced by organisations other than the BEA; these are: Federal Reserve Board produces financial accounts and balance sheet accounts as well as Index of Industrial Production and the Bureau of Labour Statistics compiles labour and productivity estimates at a detailed industrial level.

This paper is limited to those issues which affect primarily the production account, specifically the level of output, value added and GDP, both at the total economy level and by industry or sector, all at current prices. GDP per capita, distribution of equalisation payments to provinces within Canada, level of GDP for regular assessment to pay contributions to international organisations etc,

Statistics Canada - 401 - Collected Articles of Kishori Lal are some of the examples that illustrate why we need to pay attention to estimate proper levels. In any case, the starting position for calculating constant price growth rates, in most cases, is the current level of output and its industrial detail, GDP and its components. The methodologies used to develop constant price estimates are quite similar in the two countries, as both use Chain Fisher Volume indices, the indices also preferred by the 1993 SNA; however, undoubtedly, the two systems employ deflators, direct valuation, and quantity extrapolation to prepare volume estimates which are not always identical. Thus our similarities and differences in methodologies, conventions, assumptions, classifications etc which affect the current price values affect the constant price series, though not identically.

Statistics Canada - 402 - Collected Articles of Kishori Lal

Section A : 1993 SNA Production boundary, areas not yet implemented

In this section, we will first look at the evolution of the production boundary in the national accounts leading to the 1993 SNA. We will then expand on those areas which may significantly affect the level of output, value added or GDP and which have not yet been implemented in Canada and the United States.

1. Evolution of the production boundary to the 1993 SNA

Level of output and level of value added in any economy are immediately affected by economic production activities included in or excluded from the boundary of its national accounts. The boundary of economic production in the measurement of a nation=s income or wealth (or what we now call national accounts) has continued to change and has always expanded since the days of the Physiocrats in the 18th century. The Physiocrats believed that the only productive sector of an economy was agriculture and other sectors did not produce an output of any value. However, by the time the classic work of Adam Smith=s Wealth of Nations appeared in 1776, manufacturing was considered as another productive sector of the economy. It was commonly accepted up to the beginning of the 20th century that services were not productive. Services began to be accepted as productive in the market oriented economies in the 20th century; however, in the system of accounts in the centrally planned economies, called Material Product System (MPS), the only services accepted as productive were those related to the distribution and/or transportation of goods; all other services had no value. Since the 1950's, there have been three documents published by the United Nations and other international organisations for compiling national accounts for the market oriented economies, with economic production boundary expanding with each newer version. MPS was also published by the UN in the 1970's but, as noted above, it was applicable only to centrally planned economies. MPS became irrelevant after the collapse of central planning system in 1989.

As each newer version of national accounts included more activities than the previous one, it is, thus, important that we examine the production boundary outlined in the latest version, the 1993 SNA, and its implementation in the national accounts of Canada and the USA so that a proper comparison can be made of their published accounts vis-a-vis each other and internationally. At the international level, the 1993 SNA is the main document, though several countries have made some adjustments and modifications when implemented in their accounts.

As the SNA production boundary is defined with reference to the economic production activities, let us note the precise meaning attached to this term by the 1993 SNA. The 1993 SNA states: A Economic production may be defined as an activity carried out under the control and responsibility of an institutional unit that uses inputs of labour, capital, and goods and services to produce outputs of goods or services. There must be an institutional unit that assumes responsibility for the process and owns any goods produced as outputs or is entitled to be paid, or otherwise compensated, for the services provided. A purely natural process without any human involvement or direction is not production in an economic sense. For example, the un-managed growth of fish stocks in international waters is not production, whereas the activity of fish farming is production@ (paragraph 6.15). The 1993 SNA production boundary is the same as the boundary of economic production except that the SNA does not include the production of services by households for their own consumption.

Statistics Canada - 403 - Collected Articles of Kishori Lal

Examples of such services are preparation of meals in households, taking care of children and the elderly by other household members, etc. It may be noted that the production boundary of the 1993 SNA was augmented from the earlier United Nations SNA, called A System of National Accounts, published in 1968, (1968 SNA). In it, only production of primary goods, typically agricultural products, produced by households for their own consumption was included, whereas in the 1993 SNA, production of all goods produced by households for their own consumption is included. Imputation of the value of goods produced by households for their own consumption is limited, in both Canada and the United States, to farm products; however, goods other than agricultural, produced by households for their consumption, are insignificant in both countries but could be quite significant in many developing countries.

The effect of production boundary changes is not the same over time in a given country and may differ greatly amongst countries, depending upon their institutions and levels of economic development. With economic development, services previously performed in households may now be performed in the market and thus included in SNA. This would increase both level of output and GDP and their growth rates, yet no new production has occurred. Similarly, activities such as prostitution, narcotic drugs and certain economic activities performed by unauthorised producers are economic activities, included in some countries and excluded in others. They may also have been excluded in earlier periods and might now be included. Thus observed growth rates, both inter- temporal and inter-country, become problematic.

2. Production boundary, illegal production

Illegal production activities such as prostitution, narcotic drugs etc occur in all countries, more in some and less in others or that is perhaps what is perceived. Historically, it was assumed that such illegal activities should not be included in the value of production of a nation, as a nation=s production was thought to represent the welfare of its citizens. GDP was a moral GDP. National accountants have not lost any moral values in the recent period but have recognised that the national accounts must record all production in the market for them to be comprehensive, coherent and comparable, inter-temporally and internationally, and useful for policy purposes.

2.1 1993 SNA

It is explicitly stated in the 1993 SNA that illegal production forms part of the SNA boundary, whereas it was previously assumed that the1968 SNA excluded these activities as there was no explicit mention of them therein. The 1993 SNA states: ADespite the obvious practical difficulties in obtaining data on illegal production, it is included within the production boundary of the System. There are two kinds of illegal production : a) The production of goods or services whose sale, distribution or possession is forbidden by law; b) Production activities which are usually legal but which become illegal when carried out by unauthorised producers; e.g., unlicensed medical practitioners@ (paragraph 6.30). The 1993 SNA makes a cogent case for inclusion, stating: A Transactions in which illegal goods and services are bought and sold need to be recorded not simply to obtain comprehensive measures of production and consumption but also to prevent errors appearing elsewhere in the accounts if the funds exchanged in illegal transactions are presumed to be used for other purposes....The failure to record illegal transactions may lead to significant errors in

Statistics Canada - 404 - Collected Articles of Kishori Lal the financial account and also the external account of some countries@ (paragraph 6.31). Examples of activities which may be illegal but productive in an economic sense include the manufacture and distribution of narcotics, illegal transportation in the form of smuggling and services such as prostitution.

Note that the term illegal production as defined above in the 1993 SNA is not synonymous with the boundary of the so-called underground economy, also referred to by many other terms, such as informal, concealed, unmeasured, unrecorded, untaxed, etc. Most of the production activities in the underground economy are included and have always been included in the value of national accounts of many OECD countries, mostly due to the methods and conventions used in its measurement. Illegal production forms only a part of the underground economy, and is probably the most difficult to measure, hence ignored or used to be thought earlier to be excluded from the official measurement of production, as noted above.

2.2 Canadian practice

The production boundary in the Canadian System of National Accounts (CSNA) is very close to, though not identical with, the 1993 SNA; the main difference is the treatment of illegal production. Early in the 1990's, smuggling of cigarettes in Canada from the USA occurred on a fairly large scale: these were really Canadian cigarettes exported to USA and smuggled back to Canada. This activity was estimated in the amount of $700 million in 1991, $1.1 billion in 1992, $2 billion in 1993, $800 million in 1994 and only $200 million in 1995. The values of both imports and household consumption were changed in the published accounts. This activity dwindled by 1995, after taxes on cigarettes were lowered to levels so that smuggling was not very profitable anymore. Apart from this, we have not included any values for other illegal activities such as narcotics and prostitution. These activities do occur in Canada, so imputing zero value to them is problematic. How large are these activities, and have they grown faster or slower than the rest of the economy are questions whose answers impact on published estimates of production levels and growth rates. The CSNA examined this issue and published a manual in 1994, called The Size of the Underground Economy (Catalogue 13-603) wherein it stated that its best guess was that illegal production represented at the most 1% of GDP.

Amongst the OECD countries, to the best of our knowledge, only Finland has so far incorporated an estimate of output of prostitution in their accounts. In Finland, prostitution services make a tiny one- thirtieth of one percent of final consumption expenditures of households . Assuming that Canadians have attitudes and spending for prostitution services similar to those in Finland, we could add about $200 million to our household expenditures and to GDP, for the year 2000. A fair amount of narcotics is probably imported, thus we might be missing both imports and domestic consumption, without affecting GDP. However, the dealers in Canada do charge significant trade margins on the imported narcotics and we have no explicit estimate for this activity, thus it could be missing, but even here some of this missing activity might implicitly be included in our statistical discrepancy.

As noted above, the term illegal production is not synonymous with the production activities in the underground economy. Most of the underground economy is covered in Canada because of the way we use our methodologies, balancing procedures, conventions, specific imputations made to our

Statistics Canada - 405 - Collected Articles of Kishori Lal recorded estimates, etc. Again the readers may refer to our document, The Size of the Underground Economy, Studies in National Accounting, noted above. Statistics Canada compiles GDP through three independent methods, with annual IO tables providing benchmarks for the estimates, with the result that any significant activity missed in one method is usually picked up in the other and the commodity balancing done at very detailed level in the IO tables further assures the exhaustiveness of our measurement. There are very few countries having such a comprehensive set of procedures for calculating GDP. Some users ask us whether we are fully covering the activities of the First Nations in Canada. Most of the members of the First Nations live and work in the communities where other Canadians live and their economic activities are covered exactly the same way as those of other Canadians. About 250,000 members of the First Nations live on Indian Reserves. They may produce some products and sell them in the market off Reserves and they may buy products again from the market off Reserves and both of these transactions are covered in the normal routine of our surveys and are picked up in the commodity balances used in the IO tables. They may also produce goods which are entirely consumed within the Reserves and these may include hunting and fishing products for their own consumption and most likely, they are missed, as we do not have any specific imputation made for their value. We have no firm estimate of how significant is their value but we may hazard a guess from our imputation for farm products which are produced and consumed by the farming households. In the case of agriculture, we have made an imputation for the value of agricultural products consumed by the farmers and this value is less than $200 million in the current period. Most likely, the value of hunting and fishing for own consumption on the Reserves is much less than the imputation for farm output made in agriculture. Thus what we may be missing is quite small and statistically insignificant for the overall economy.

2.3 USA practice

Illegal activities, as defined in the 1993 SNA, are also excluded from the national accounts of the United States. Again, it must be noted that the production related to the illegal activities forms only a part of the underground economy and most of the underground economic activities are included in the published value of production and GDP in the United States. The BEA examined this issue in the mid-eighties and published several articles on this topic, in their Survey of Current Business (Carol Carson-May and June 1984, Robert Parker-June 1984, and Frank de Leeuw-March 1985), pointing out the material significance of such activities and how they were included in the official numbers. Further, the BEA has prepared an update of the work on underground economy for the Economic Commission of Europe in 2002. There is no reason to believe that the production of illegal activities in the USA is significantly different than in Canada wherein the best guess is about 1%.

Just to close the discussion on smuggling of cigarettes to Canada, there was no adjustment made in the United States. Exports of such cigarettes from Canada were recorded as imports from Canada in the USA but their further re-export was recorded as destined for some third country, not Canada. This required us to adjust our imports from USA to be consistent with our allocation of smuggling to the household consumption.

Statistics Canada - 406 - Collected Articles of Kishori Lal

2.4 Concluding remarks

It is reassuring to observe that both Canada and the USA have an almost identical SNA production boundary, both explicitly do not include the illegal production activities such as prostitution, narcotics etc., in their accounts and the value of their production in both countries is not very significant. It is quite possible that some of these activities might implicitly be included in the accounts in some miscellaneous series or in the statistical discrepancy between the two sides -income and expenditure- of the GDP. Developing explicit estimates of illegal production activities should not be ignored forever, as these activities do occur. There is an additional political reason why the statistically advanced countries, such as Canada and the United States, should make an effort to conform with the 1993 SNA recommendation on this issue so that some other countries where such activities are significant are encouraged to do the same.

3. Valuation of own-account construction by households

Households are engaged in the construction of their own dwellings or other structures for their own use, or on structural improvements or extensions to existing dwellings or structures. Typically the households buy the required materials from the market and provide their own labour to complete the job. This household activity falls within the production boundary of the SNA. Own-account construction by the households makes a significant proportion of total residential dwellings and other structures in most economies, the less developed an economy, the higher the proportion.

3.1 1993 SNA

The 1993 SNA recommends that the value of such output should be imputed on the basis of the prices of similar products sold on the market (paragraph 4.147). It further specifies: AIt will usually be necessary to value the output of own-account construction on the basis of costs as it is likely to be difficult to make a direct valuation of an individual and specific construction project that is not offered for sale. When the construction is undertaken for itself by a business enterprise, the requisite information on costs may be easily ascertained, but not in the case of the construction of dwellings by households or communal construction for the benefit of the community undertaken by informal associations or groups of households. Most of the inputs into communal construction projects, including labour inputs, are likely to be provided free so that even the valuation of the inputs may pose problems. As unpaid labour may account for a large part of the inputs, it is important to make some estimate of its value using wage rates paid for similar kinds of work on local labour markets. While it may be difficult to find an appropriate rate, it is likely to be less difficult than trying to make a direct valuation of a specific construction project itself @ (paragraph 6.86). The 1993 SNA emphasises the same valuation principle again (see paragraph 10.78) where it states that the value of finished structures will be seriously underestimated if the imputed value for unpaid household labour is not included. Guidelines in the 1995 ESA are similar to the ones in the 1993 SNA. It may be noted that an identical extension to two dwellings in the same locality, one produced by a construction contractor and the other produced by the household with its own labour must be valued similarly for capital investment as they provide identical services.

Statistics Canada - 407 - Collected Articles of Kishori Lal

3.2 Canadian practice

In Canada, and most likely in other OECD countries, new residential structures are, by and large, produced by the construction contractors whereas households are typically engaged in their alterations and improvements. Alterations and improvements make a very significant proportion, approximately 32% and ownership transfer costs account for approximately 17% of the total value of residential structures in the CSNA. Further, alterations and improvements by homeowners (owner- occupied dwellings) represent the largest proportion of alterations and improvements, approximately 75%. In estimating alterations and improvements, many sources are utilised such as : Survey of Household Spending, Homeowner Repair and Renovation Survey, Building Permit data, and the Survey of the Real Estate Rental and Leasing and Property Management industries. It is estimated that two-thirds of homeowner alterations and improvements are done by contractors and the rest done by households, using their own labour. In Canada, we have not imputed any value to the labour provided by the households. Our ballpark estimate for 1997, for example, of the value of homeowner alterations and improvements done by households is $2.4 billion of the total value of residential construction of $30 billion in the national accounts. This is built as follows: 32% of the total value of residential construction of $30 billion is accounted for by alterations and improvements, 75% of such alterations etc are made for the owner-occupied dwellings and a third of such activity is done by the homeowners with their own labour. Typically, the ratio of labour to materials in such construction activity is one to one, thus our accounts may be missing an imputed value of labour of $2.4 billion in 1997 or about 8% of total value of residential construction.

3.3 USA practice

Unlike in Canada, the BEA accounts include an imputation for own-account labour of owner- occupants but only for new single-family housing units. There is no imputation for labour by owner- occupants for improvements and renovations. Of course, direct costs of materials and supplies are always added in the own-account construction by the owner-occupants. The BEA practice of including an imputation of labour for owner-built new units , but not for improvements, reflects the methods used by the Census Bureau in preparing the source data for construction. The value of this imputed labour is US$8.2 billion for 2001, a value proportionately much smaller than the one suggested for the Canadian imputation because it does not include labour imputation for improvements which are very labour intensive.

3.4 Concluding remarks

In Canada, there is no imputation made for the value of labour provided by the homeowners to produce alterations and improvements to their own residential structures and in the United States, the value of imputation is limited to the new construction of single units only, though both the 1993 SNA and the 1995 ESA recommend to impute full value of own-account labour. Member countries of the European Union follow the 1995 ESA and are required to include an estimate of the value of free labour provided by the households for own account construction. It is our understanding that Australia also imputes such a value in the investment series for residential construction. Once we implement the 1993 SNA recommendation, the value added in the construction industry will increase

Statistics Canada - 408 - Collected Articles of Kishori Lal by the estimated amount of own account labour (classified as mixed income), and an identical increase will be shown in the value of residential construction. With this incorporation, the value of residential construction in Canada is estimated to increase by about 8% and the level of GDP will increase by about three-tenths of one percent and a bit less in the USA as there is already some imputation.

4. Investment in cultivated assets: livestock, plantations and orchards

In economic theory, national accounting and, by and large, also business accounting, there is general agreement that products used repeatedly or continuously over periods of time of (at least) more than one year to produce other goods and services should be classified as capital assets. In the national accounts, buildings, machinery and equipment have always been classified as capital assets because they are continuously used over long periods to produce other goods and services. In agriculture, investment in cultivated assets such as livestock, plantations and orchards is quite significant, yet, such investment was not recognised as a capital asset in the international manuals on national accounts, prior to the 1993 SNA, for reasons unknown.

4.1 1993 SNA

In the 1993 SNA, the asset boundary was extended to align it more closely with economic theory and reality, and it now includes investment in cultivated assets as capital assets. The 1993 SNA notes: ACultivated assets consist of livestock or trees that are used repeatedly or continuously over periods of time of more than one year to produce other goods and services. Thus, livestock that continue to be used in production year after year are fixed assets. They include, for example, breeding stock, dairy cattle, sheep reared for wool and draught animals. On the other hand, animals raised for slaughter, including poultry, are not fixed assets. Similarly, trees (including shrubs) that are cultivated in plantations for the products they yield year after year -such as fruit trees, vines, rubber trees, palm trees, etc.- are fixed assets. On the other hand, trees grown for timber that yield a finished product once only when they are ultimately felled are not fixed assets, just as cereals or vegetables that produce only a single crop when they are harvested cannot be fixed assets@ (paragraph 10.83). Capitalising such livestock and trees as cultivated assets rather than treating expenditure on them as intermediate consumption, adds to the value of output and GDP, both for agriculture and the total economy. Its economic significance, no doubt, differs greatly amongst countries as well as among various regions of the same country. For example, sheep reared for wool are very important in Australia and Scotland, vineyards are important in California, France, Italy, Ontario (Niagara belt) and British Columbia in Canada, and so on. Thus, it is necessary to be fully aware of its implementation amongst countries for proper analysis of their performance and their international comparability.

4.2 Canadian practice

We fully support the 1993 SNA recommendation in this regard. However, we have not yet implemented the 1993 SNA recommendation, primarily because traditionally, cultivated assets have never been included in the capital stock series, and this is not a strong reason. All expenditures to grow breeding and dairy cattle are, at present, intermingled with the expenditures to grow slaughter

Statistics Canada - 409 - Collected Articles of Kishori Lal cattle. The value of animals not yet slaughtered is added together with the value of dairy and breeding animals as part of inventory accumulation in final demand in the CSNA. Dairy and breeding cattle are not capitalised. Similarly, the output of the fruit bearing trees in the vineyards and other farms is not shown, and all the expenditures to grow these trees are intermingled with other intermediate consumption expenditures for the production of fruits. There is no capital formation recorded for fruit trees.

The following changes will be required to follow the 1993 SNA guidelines:

a) Expenditures to grow breeding stock (the term breeding stock is used to include both dairy cattle and breeding stock) will have to be separated from those for slaughter animals. b) In the inventory stock of cattle, the value of breeding livestock will have to be separated from slaughter animals and the breeding livestock will be added as capital formation. c) Expenditures to grow fruits will have to be separated from expenditures to grow fruit bearing trees. d) A new own account production of fruit bearing trees will be created, and it will be reported in gross fixed capital formation in the final demand. e) Rates of consumption of fixed capital (CFC) will have to be estimated for both breeding livestock and fruit bearing trees for use as expenditure on CFC in the value added of agriculture.

The result from these changes will be as follows:

a) Gross fixed capital formation for breeding livestock will increase by the value of their output, counterbalanced by an identical decrease in the value of inventory accumulation of cattle in the final demand. b) Gross fixed capital formation of fruit bearing trees will increase by the value of the new own account production. c) Gross value added in agriculture will not change for the change in the treatment of breeding livestock but net value added will decease by an identical increase in CFC for breeding livestock. d) Output of agriculture and also gross value added will increase in agriculture by the value of own account production of fruit trees. Additional CFC of fruit trees will be added in agriculture.

To repeat, expenditures on cultivated assets are not treated as investment in the Capital and repair expenditures surveys of Statistics Canada, nor are they treated as gross fixed capital formation in the CSNA. The importance of these cultivated assets is different across provinces and territories, thus adding another caveat to inter-provincial comparison. For example, dairy farming is more important in Quebec and orchards are more important in Ontario and BC, compared to other provinces.

We have looked anew at the information collected and published by our Agriculture Division in regard to the value of breeding livestock. As at December 2001, the value of breeding livestock was reported to be $9,368 million and inventory of slaughter animals $6,232 million and this information is collected annually. (see Agriculture Economic Statistics, Cat: 21-603, Table on Balance Sheet of

Statistics Canada - 410 - Collected Articles of Kishori Lal the Agriculture Sector, for more details). Typically, the dairy cattle produce milk for 6 to 7 years after they are two years old, so we have a good basis to calculate their depreciation rate. Thus, we can develop estimates for gross fixed capital formation for breeding livestock and their depreciation rates but separation of expenditures for producing breeding livestock from slaughter animals may require developing some methodology and conventions and some resources. As noted above, gross value added will not be affected by these changes.

There is no readily available information for the value of own account production of fruit trees and the expenditures for their production. However, let us make some ballpark estimate of the additional capital and its result on value added. In Canada, the value of production of fresh fruits is about $2.5 billion in recent years. Let us assume that the fruit trees yield fruits for about ten years if well maintained. Let us further assume that the cost of maintaining these trees is about $0.5 billion, thus the net yield of fruit trees is $2 billion. Assuming a rate of discount of some 8 to 10%, the market value of these trees may approximate $10 billion. Capital consumption will be approximately $1 billion, thus value added will increase by about $1 billion annually, if we conform to the 1993 SNA guideline in this case.

4.3 USA practice

As in Canada, cultivated assets are not capitalised in the national accounts of the United States, for reasons primarily statistical, thus their overall GDP and GDP for agriculture, are underestimated by the value of new investments in fruit trees. Again, the economic importance of these activities varies a great deal amongst the States.

4.4 Concluding remarks

In Canada, we have information for estimates of investment of breeding livestock but the expenditures to produce breeding stock are intermingled with those for slaughter animals; furthermore, the expenditures to grow orchards are not separable from expenditures to produce fruits. The situation in the USA is quite similar. In our opinion, it would be worthwhile to follow the 1993 SNA guidelines in this area. Our ballpark estimate is that we would add about $1 billion, to both value added and investment in agriculture in Canada, if cultivated assets are treated as investment. As the value of cultivated assets would be quite significant, particularly for agriculture and for regional accounts where such activities are important, it will be useful to have joint discussions with BEA to share expertise and develop a common methodology to make estimates.

5. Acquisition of entertainment, literary or artistic originals

Acquisition of entertainment, literary or artistic originals were not considered as gross fixed capital formation, and indeed, no intangible product was considered as a capital asset, in the international system of national accounts manuals prior to the 1993 SNA. In the 1968 SNA, the boundary of gross fixed capital formation was limited to tangible assets, such as buildings and machinery and equipment.

Statistics Canada - 411 - Collected Articles of Kishori Lal

5.1 1993 SNA

The 1993 SNA asset boundary has been extended to align it closer to economic theory. Several types of intangible assets were added: mineral exploration; computer software; entertainment, literary or artistic originals; and other intangible fixed assets (see paragraph 10.34 b). Expenditures on mineral exploration and computer software now form part of investment and capital stock in the national accounts of both the USA and Canada. Both output and value added increased by the amount of the new addition of investment recognised in the accounts and the amounts were substantial.

Classification of entertainment, literary or artistic originals as gross fixed capital formation is probably the newest of the new extensions of the asset boundary in the national accounts. The 1993 SNA states: AOriginals consist of the original films, sound recordings, manuscripts, tapes, models, etc., on which drama performances, radio and television programming, musical performances, sporting events, literary and artistic , etc.,are recorded or embodied@ (paragraph 10.94). It further states: AThe acquisition of an original constitutes gross fixed capital formation. The original is often retained by its producer, but it may also be sold after it has been produced in order to be exploited by another unit. When it is sold, the gross fixed capital formation is measured by the price paid by the purchaser to acquire the asset. If it is not sold, ...it may be necessary to value the original by its cost of production, as in the case of many other kinds of output produced for own gross fixed capital formation@ (paragraph 10.95).

Once a product is recognised as a capital asset, its service life needs to be determined in order to calculate its depreciation or consumption rate, for it to be used as part of gross value added in the using industry. Originals typically have copyright protection over a fairly long period. When a user buys the copyright for a given period of (at least) more than one year, it also acquires an asset which needs to be recognised. The debate is not settled yet on how to recognise and value licences and copyrights to use intangible assets like originals. This debate is similar to the one on how to value licences to use tangible assets like electro-magnetic spectrums for mobile phones. At the October 2001 meeting of the OECD National Accounts Experts, a Task Force on capitalisation of computer software was established. This task force has now recommended that the software original be recognised as a capital asset.

5.2 Canadian practice

Entertainment, literary or artistic originals are not recognised as capital assets in the CSNA. We do not have a big entertainment industry like the one in Hollywood but we have the National Film Board and the Canadian Film Development Corporation and some private sector enterprises involved in film production. We have many well known singers and song writers and authors. However, it is fair to assume that the economic importance of the entertainment originals of the film industry is far greater than that of literary or artistic originals in Canada. Annual expenditures of the National Film Board and the Canadian Film Development Corporation have been in the range of about $200 million recently. Most of their expenditures to help develop films in Canada are expensed in the entertainment and cultural industries and not capitalised. We have no other financial information at hand for the value of production of these originals by the private sector and by our singers, song writers and authors. Our ballpark guess is that the total annual expenditures probably do not exceed

Statistics Canada - 412 - Collected Articles of Kishori Lal

$500 million. Had we allocated these expenditures to produce capital assets as suggested by the 1993 SNA, the value added in the entertainment and cultural industries would have increased by the value of this new investment or about $500 million.

5.3 USA practice

As in Canada, the acquisition of entertainment, literary or artistic originals is not currently capitalised. The production of motion pictures, television, and sound recordings is important in the US economy, and implementing the 1993 SNA guidelines would increase the level of value added in the entertainment and cultural industries and the GDP for the overall economy by the value of this investment. It is our understanding that the BEA is planning to start research on this subject after its next comprehensive revision in 2003.

5.4 Concluding remarks

With the implementation of the 1993 SNA recommendation in this area, both the level of value added in the entertainment and cultural industries and overall GDP will increase in both countries, about $500 million in Canada and significantly more than ten times this value in the United States. This is one area where research needs to done to develop a reliable, acceptable and transparent methodology to estimate the value of such originals and share such research and findings with colleagues in other countries. Hence, it will be useful to have joint discussions with the BEA to share expertise and develop a common methodology to calculate these estimates. The importance of the activities associated with such originals is unevenly distributed amongst the Canadian provinces and, of course, amongst countries. Thus, its exclusion adds one more caveat to the reliability and accuracy of published GDP series at the international level.

6. Net acquisitions of existing (used) assets

Capital formation is a very important item in the national accounts: it is a significant component of GDP; it cumulates to capital stock and the consumption of fixed capital (CFC) derived from capital stock is again a significant component of gross value added. How existing used assets should be handled in the measurement of capital formation is a question which needs to be answered, as there are divergent practices on its treatment at the total economy level and by industry or sector.

6.1 1993 SNA

The 1993 SNA defines gross fixed capital formation as acquisitions, less disposals, of new or existing tangible and intangible fixed assets, major improvements to tangible non-produced assets and the costs associated with the transfers of ownership of non-produced assets (see paragraphs 10.33). There is no ambiguity on this issue in the SNA guidelines.

Statistics Canada - 413 - Collected Articles of Kishori Lal

6.2 Canadian practice

At the total economy level, our treatment of existing or used assets is consistent with the 1993 SNA. Acquisition of imported used assets by enterprises are added to their gross fixed capital formation, and the value of used assets exported by enterprises are removed from their gross capital formation, as recommended in the 1993 SNA. Similarly, at the sector level, sales of used vehicles by businesses and governments to the household sector are removed from their capital investment in the period in which such transactions occur, and are added to household consumption. This treatment is again consistent with the one recommended in the 1993 SNA. However, when one business enterprise sells its used capital assets to another business enterprise or acquires used capital assets from another enterprise, these transactions are ignored in the measurement of capital formation by industry. When an enterprise acquires only used assets to produce something, it is shown at present as if it has no capital. On the other side, when an enterprise sells its used capital and has no more capital, it is shown as if it still carries all the capital it had before. This convention in the measurement of capital investment and later capital stock is followed by the Investment and Capital Stock Division as well as in the CSNA. Thus, our treatment of used assets by industry is not consistent with the one recommended in the 1993 SNA, and neither is it consistent with business accounting.

6.3 USA practice

As in Canada, net acquisitions of existing (used) assets at the total economy level and by sectors are handled according to the 1993 SNA recommendations. It is reported by the BEA that the source data often do not allow it to separately identify transactions involving used assets at the industry level and this is same situation as in Canada.

6.4 Concluding remarks

Both in the USA and Canada, the present practice of ignoring the purchase/sale of existing assets by industry within the business sector is inconsistent with both the 1993 SNA and the business accounting principles. Thus both countries may carry incorrect capital stock by industry. Our present practice does not affect gross value added in total or by industry but it has two serious consequences: a) incorrect capital stock by industry leads to errors in the calculations of multi-factor productivity estimates by industry and b) it leads to incorrect CFC by industry, and thus incorrect net value added by industry.

7. Value of Consumption of fixed capital, macro series

Value of consumption of fixed capital (CFC) is an important item in GDP or in gross value added by sector or industry. This item is also known in the literature as depreciation or capital consumption allowance. It makes about 13% of GDP in Canada and as a single item, its value is higher than any of the following well known items listed in the GDP : profits, interest income, net income of unincorporated businesses, net taxes on factors of production or net taxes on products. The only item of higher value than CFC is the wages, salaries and supplementary labour income. In the business sector or in industries of the business sector, it is a charge against operating surplus, hence its calculated value helps to derive net value added from gross value added but the gross value added

Statistics Canada - 414 - Collected Articles of Kishori Lal does not change. However, in the non-market (government and NPISHs) sectors whose output and value added are measured by adding all the costs, its value affects both the output and the gross value added, hence total GDP of the economy.

One needs to have data on investment in assets, capital stock, service lives of assets to calculate CFC. In the business accounts, depreciation is always calculated for tax purposes but the rate of depreciation may be determined by tax authorities, not necessarily connected with economic service life. Further, the values used of fixed assets are typically the book values or the historic costs, not their current market prices. Thus, the profits and other incomes reported by the enterprises reflects these values of depreciation. In the national accounts, the rules for calculating CFC are different from those in the business accounts.

7.1 1993 SNA

The 1993 SNA recommends that "... consumption of fixed capital must be valued with reference to the same overall set of current prices as that used to value output and intermediate consumption... It should therefore be calculated using the actual or estimated prices and rentals of fixed assets prevailing at that time and not at the times the goods were originally acquired. The historic costs of fixed assets, i.e., the prices originally paid for them, may become quite irrelevant for the calculation of consumption of fixed capital if prices change sufficiently over time" (paragraph 6.180). Several methods to calculate CFC are noted in the 1993 SNA and in the OECD Manual, Measuring Capital, Paris 2001. The two most common methods are the straight-line depreciation (also called linear depreciation) and the geometric depreciation. With straight-line depreciation, the market value of an asset in constant prices is assumed to decline by the same amount each period. With geometric depreciation, the market value in constant prices is assumed to decline at a constant rate in each period. The 1993 SNA does not recommend one method over the other; it states: A Both the linear and the geometric, or declining balance, method are easy to apply. The choice between them depends upon knowledge, or assumptions, about the implied profiles of rentals which underlie them. It is not possible on a priori grounds to recommend the use of one in preference to the other in all circumstances. It is possible, for example, that linear depreciation may be realistic in the case of structures, while geometric depreciation is more realistic in the case of machinery and equipment@. (Paragraph 6.197). In the economic literature, particularly in the field of productivity analysis, there is a preference for geometric depreciation; further, the valuation of second hand goods in the market suggests a choice for geometric rates.

7.2 Canadian practice

Consumption of fixed capital is calculated by the Investment and Capital Stock Division (ICSD) of Statistics Canada using current market prices of fixed assets and estimates are published using both linear and geometric methods. However, this information is classified by industry, based on establishments rather than by sector based on institutional units such as corporations or companies. In the macro series of the CSNA, the value of consumption of fixed capital for the government sector, housing and agriculture is the one calculated by the ICSD, using current market prices of the capital stocks; however, for other sectors, the value of depreciation used is what enterprises report in their financial statements, typically using historic costs and tax determined depreciation rates.

Statistics Canada - 415 - Collected Articles of Kishori Lal

Our departure in the CSNA from the recommended treatment on CFC has been due to our statistical sources and partly also because we did not put high enough priority to develop estimates consistent with the 1993 SNA. What one would need is to develop a methodology and a database to redefine the value of the ICSD=s consumption of fixed capital classified by establishment-based industries to their institutional units, such as corporations which report profits. Only then, can the CSNA recalculate profits using the recommended CFC. During the last few years, two statistical developments have occurred at Statistics Canada which should help us to re-examine our existing approach: a) statistical information is now collected via a unified enterprise survey, thus making it feasible to integrate establishment data with sector/enterprise data, and b) the business register now includes both sector and industry classification for all economic entities. The new estimates of CFC, when applicable, will not affect GDP, but will change the value of consumption of fixed capital, counterbalanced by an equal change in the value of net operating surplus.

7.3 USA practice

The BEA values CFC for the macro SNA series for all sectors using assets at current cost and calculates depreciation using BEA-determined depreciation schedules (usually geometric) that are based on empirical research on used asset prices. Total profits are reported both using a) current cost and BEA-determined depreciation and b) tax-based historical-cost depreciation (referred to as Acapital consumption allowances@ or CCA). The difference between CFC and CCA is called the Acapital consumption adjustment.@ and is also published. 7.4 Concluding remarks

In the NIPA accounts, CFC as well as CCA and capital consumption adjustment are published; CFC equals CCA less capital consumption adjustment. This provides a rich database to connect with tax based calculations and to calculate the preferred national accounts method for CCF. It is our judgement that the CSNA will better serve its users if it also adopts the innovative and pragmatic approach used in the BEA NIPA accounts. Then the CSNA will be consistent both with the 1993 SNA and the NIPA accounts, a huge improvement over the existing situation.

8. Consumption of fixed capital, industries

Calculation of CFC is very important for a fuller analysis of the macro SNA series. It assumes an additional importance in the calculation of value added by industry estimates, specifically net value added, rather than gross value added. Net value added, compared to gross value added, is preferred, as it is more in tune with the theoretically correct Hicksian concept of income and sustainable development. The theoretically correct concept of value added, consistent with Hicksian income, is the maximum value added which an economy can produce without reducing capital.

It may be worth noting that, in the late 1960's and early 1970's, there was an interesting debate in the USA over the preference for net value added over gross value added or vice versa for productivity analysis between Professors Jorgenson and Griliches on one side and Denison on the other. A summary of this debate can be found in the article by Charles R. Hulten, ATotal Factor Productivity : A Short Biography,@ in New Developments in Productivity Analysis, edited by Charles Hulten, Edwin Dean and Michael Harper, 2001 , University of Chicago Press. Professor Hulten notes:

Statistics Canada - 416 - Collected Articles of Kishori Lal

A...Jorgenson and Griliches recognised that output must be measured gross of depreciation if it is to conform to the accounting system implied by the strict logic of production theory. This put them in conflict with Denison, who advocated a concept of output net of depreciation, and Solow, who used gross output in his empirical work but preferred net output on the theoretical grounds that it is a better measure of welfare improvement arising from technical progress@ (page 14).

8.1 1993 SNA

As noted above, the 1993 SNA recommends that CFC must be valued with reference to the same overall set of current prices as that used to value output and intermediate consumption, both for the macro accounts and the production accounts by industry. The balancing item in the production account is value added and it can be measured either gross or net: that is, before or after deducting consumption of fixed capital. The 1993 SNA states: A a) Gross value added is defined as the value of output less the value of intermediate consumption; b) Net value added is defined as the value of output less the values of both intermediate consumption and consumption of fixed capital A (paragraph 6.4). The 1993 SNA fully supports the theoretically preferred concept of net value added and it states: AAs value added is intended to measure the additional value created by a process of production, it ought to be measured net, since the consumption of fixed capital is a cost of production@ (paragraph 6.5). After strongly asserting its preference for net value added, the 1993 SNA adds a flexibility to produce gross value added. It states: A... consumption of fixed capital can be difficult to measure in practice and it may not always be possible to make a satisfactory estimate of its value and hence of net value added. Provision has therefore to be made for value added to be measured gross as well as net@ (paragraph 6.5). The result of this flexibility is that many countries produce only gross value added. From a conceptual point of view, it would have been appropriate, had the1993 SNA pushed hard for net value added as the primary featured measure. Further, given that the newer types of capital such as computers and computer software have short service life (three to five years, or even less), it would have been appropriate even from an operational point of view for the 1993 SNA to have pushed for the net value added concept.

8.2 Canadian practice

In the Canadian Input-Output tables, there has never been a separate estimate of consumption of fixed capital by industry, thus it has remained part of other operating surplus. However, as noted above, in the macro income and expenditure accounts, consumption of fixed capital is estimated for agriculture, own account housing, as well as for the government sector and it is calculated with reference to market prices. Furthermore, for the rest of the economy, depreciation is based on the historic cost of the value of capital stock and is determined by tax considerations. The ICSD calculates by industry capital stock and CFC at current market prices, the concept recommended by the 1993 SNA. Additional, though not very significant, resources will be required to make adjustments to the industry boundaries used in the ICSD calculations to bring them in line with the IO definition of industries. When such estimates of CFC by industry are incorporated in the Canadian IO tables, we will then have available both gross value added and net value added by industry. This additional detail, we believe, will substantially enhance the usefulness of our industry statistics .

Statistics Canada - 417 - Collected Articles of Kishori Lal

8.3 USA practice

Consumption of fixed capital (CFC) is published for National Income and Product Accounts (NIPA), as noted above in item # 7, only by legal form of organisation, such as corporate business, non- financial corporate business, farms, non-farm housing, government, etc., but not by industry. Both the NIPA=s and GDP by Industry accounts provide the capital consumption allowances (CCA) by industry, the former on a company basis (in table 6.22), and the latter on an establishment basis. Neither CFC nor CCA is available from the IO accounts.

BEA does publish CFC by industry on an establishment basis as part of BEA=s Fixed Assets accounts. These estimates of CFC have not been used for GDP by industry because of a) BEA=s concerns about the reliability of the industry distribution of fixed investment used in the perpetual inventory method, and b) potential inconsistencies between CCA from the capital stock and profits before tax converted from a company basis to an establishment basis in the GDP by Industry accounts.

8.4 Concluding remarks

The CSNA industry and IO accounts do not have CFC estimates and the BEA industry accounts have only CCA estimates, thus neither country produces the preferred CFC by industry. At Statistics Canada, CFC by industry is already calculated by the Investment and Capital Stock Division, but the industrial classification is not identical with the one used in the CSNA; this reconciliation is achievable with some additional, though not very substantial, resources. BEA does not believe that company-establishment conversion of CFC (if company estimates were available) is the best method to achieve the result of CFC estimates by industry. When source data become available that would allow BEA to estimate value added directly from establishment gross output and intermediate inputs, it may then consider using CFC from the Fixed Assets data as a measure of establishment-based CFC by industry.

We strongly believe that the usefulness of industry statistics will enhance substantially with the inclusion of CFC estimates, as this will permit to calculate both gross value added and net value added by industry.

Section B: Production account for institutional sectors, partially implemented

In the national accounts, the production account has always been given utmost importance. It is the first in the sequence of accounts compiled for any economy, thus, its compilation affects all the succeeding accounts in the system. The production accounts can be compiled for industries and then aggregated to the total economy or for institutional sectors and then aggregated to the total economy. The production account in the SNA contains three aggregate items apart from the balancing item. These are output, intermediate consumption, and consumption of fixed capital; the balancing item is value added, which can be measured either gross or net, that is before or after deducting consumption of fixed capital. Should such production accounts be prepared both by both industry and institutional sectors? Production accounts by industry have been compiled by most countries over a long period and only recently have such accounts been recommended and prepared by institutional sectors. In

Statistics Canada - 418 - Collected Articles of Kishori Lal this section, we will examine the sector boundaries used by Canada and the USA in this area vis-a- vis the international recommendations and their impact on inter-country and international comparability.

9. Business sector

It is quite natural to think of developing estimates of production of an economy distributed into two broad sectors, the market producers and the non-market producers, as their respective motives to produce goods and services are quite different. Market producers or producers in the business sector, must at least cover all their costs of production and make profits if possible, whereas the non-market producers may provide their output free or at prices which are not significant to cover their costs of production. Let us see the guidelines for aggregating institutional units for production accounts in the 1993 SNA.

9.1 1993 SNA

The 1993 SNA states: AThe production account is the first in the sequence of accounts compiled for institutional units, sectors and the total economy. The incomes generated by production are carried forward into subsequent accounts so that the way in which the production account is compiled can exert a considerable influence on the System@ (paragraph 6.1). It further states that production accounts are compiled for establishments and industries as well as for institutional units and sectors. Overall numerical consistency requires that the output of an institutional unit engaged in production should be equal to the sum of the outputs of the individual establishments of which it is composed (see paragraph 6.2)

The first UN document on National accounts, produced in the mid-1950's, did not specify a production account by industry or by sector, as it limited itself to the preparation of only macro estimates of national accounts. The UN 1968 SNA manual went much further and according to it, separate production accounts were required for industries but not for individual institutional sectors; only one consolidated production account was recommended for the economy as a whole. The 1993 SNA provides a framework for a full sequence of accounts and for this the only unit which cuts across the full set of accounts is an institutional unit or an institutional sector as only institutional sectors can be utilised to compile the income and outlay account, the capital and finance account and the balance sheet account. It is quite logical to have the same unit for the sequence of accounts and for a full appraisal of its performance. Compared to the 1993 SNA, the 1968 SNA did not provide a full set of accounts, hence could afford to ignore the compilation of a production account for institutional sectors.

The 1993 SNA requires the compilation of production accounts for five mutually exclusive sectors. It defines an institutional unit "as an economic entity that is capable, in its own right, of owning assets, incurring liabilities and engaging in economic activities and in transactions with other entities" (paragraph 4.2). The resident institutional units that make up the total economy are grouped into the following five mutually exclusive sectors (paragraph 4.6):

Statistics Canada - 419 - Collected Articles of Kishori Lal

9.2 Canadian practice

The CSNA produces production accounts for all years for which input-output tables are compiled, but the sector classification is different from that of the 1993 SNA. A business sector is created, which comprises all producing units of the non-financial corporations sector, the financial corporations sector and the household sector of the 1993 SNA. Two additional sectors, the general government and NPISHs (see additional details on their sector boundary below), produce goods and services primarily not for sale in the market but for their own consumption. All producing units of the Canadian economy are thus included in the production accounts of the business sector and the two non-market sectors. The business sector is not defined in the 1993 SNA but it is used both in the USA and Canada. Its boundary is close to, though not identical with, the market production of an economy, as the non-market sectors can and do produce minor secondary output for sale in the market. We have not been able to follow the recommendation of the 1993 SNA regarding an integrated set of production accounts for both institutional sectors and industries. Note that the government business enterprises are allocated to the business sector and they are classified by industry.

Our presentation is different mostly because of the way our production surveys have historically been conducted. Our production surveys have traditionally collected information on outputs, intermediate inputs etc., from establishments, without much regard to their relationship with institutional units (often called companies or enterprises in Canada) of which they formed part. Thus, it was not possible to reclassify information collected from establishments to their institutional units. However, the recent approach to collect all commodity outputs and inputs from unified enterprise units and the classification of these units linked with their establishments in our expanded Business Register may permit us to re-examine our approach to the development of production accounts for institutional sectors in addition to our regular production account by industry.

Though we have not yet implemented the 1993 SNA recommendation to produce full production accounts separately for each of the three sectors making up our business sector, it is possible to delineate, with some additional work, the financial corporations sector. The boundary of the finance and insurance sector #52 in the North American Industry Classification System (NAICS), is quite similar to the financial corporations sector of the 1993 SNA, except the NAICS industry 5242, Agencies, Brokerages and other Insurance Related Activities, which includes significant activity provided by the unincorporated enterprises. NAICS industry 5242 is separately identified in the CSNA worksheet level detail. NAICS Finance and Insurance less NAICS 5242 is almost identical with the financial corporations sector of the 1993 SNA.

The producing units in the household sector operate almost entirely in the non-financial part of the Canadian economy. It can be argued that these producing units, also called unincorporated business

Statistics Canada - 420 - Collected Articles of Kishori Lal enterprises, are more like quasi-corporations following the 1993 SNA terminology, and are thus classifiable with the corporations sectors in the production accounts. The 1993 SNA defines quasi- corporations as follows: AQuasi-corporations are unincorporated enterprises that function as if they were corporations...Such an enterprise must, of course, keep a complete set of accounts@ (paragraph 4.49). In Canada, all unincorporated enterprises submit the Canada Customs and Revenue Agency T- 1 Return, and thus have separate accounts. This fulfills a condition of the 1993 SNA for production accounts. However, they are not capable, in their own right, of owning assets and incurring liabilities, independent of households owning such enterprises: thus, they are not full-fledged corporations or quasi-corporations. In the subsequent sequence of accounts related to income and outlay, capital and finance and balance sheet, we recognise this characteristic of household-owned unincorporated enterprises and include them with the household sector, rather than with the corporations sector. Separating unincorporated enterprise units from the production account of the business sector will require a significant amount of work which will be possible only when we obtain data from unified enterprise surveys, with detail kept separately for corporate and unincorporated enterprise units.

9.3 USA practice

The BEA produces national income and product accounts (NIPA) which include tables providing information on production accounts for sectors. The overall definition and boundary of the sectors are similar to those in Canada but the details are different. These sectors are: (1) business, (2) households and nonprofit institutions, and (3) general government (see Table 1.7 Gross Domestic Product by Sector). The business sector measures production by all market producers (entities that produce goods and services for sale at a price intended at least to approximate the costs of production), including production by government business enterprises, as well as certain other types of production by nonprofit institutions serving business, Federal Reserve Banks, and services of owner-occupied housing and of buildings and equipment owned by nonprofit institutions. Note that the production measurement therein is limited to value added (also called Agross product@ or Agross domestic product@ in various BEA tables) only, and not the full production account which includes also output and intermediate consumption. It may be noted that the housing sub-sector under the Business sector in Table 1.7 includes government enterprises engaged in housing activities.

Government business enterprises are not classified by detailed industry in the GDP by industry accounts or IO tables; rather, they are aggregated under Special Industries in two blocks: Federal, and State and local. This presentation has an advantage that the users can very quickly assess the significance of government business enterprises in the economy but its disadvantage is that the NIPA=s industry or IO data are not comparable in detail with data for the same industry published by other departments which classify such enterprises by industry.

In addition to the information for the total business sector, the value added of all Corporate Business (in current prices only) and of Non-financial Corporate Business (in current and chained prices), are presented in NIPA table 1.16. Thus in the United States, the GDP of financial corporations can be residually calculated. Similarly, the value added of the non-corporate sector (mostly unincorporated business enterprises owned by the household sector) may also be residually calculated within the business sector. The macro aggregates by sector that are presented in NIPA tables 1.7 or 1.16 (or that

Statistics Canada - 421 - Collected Articles of Kishori Lal can be residually calculated from Table 1.16) are based on data by legal form of organisation. The data by legal form of organisation are benchmarked to estimates derived from samples of individual institutional units from the Internal Revenue Service for profits, net interest and CCA or from establishment data classified by legal form from the economic census for compensation of employees. These estimates are also available by industry. The IRS data are available annually, but the compensation of employees by legal form data are available only every five years and must be interpolated.

Output (called gross output in the US accounts) and intermediate consumption are not shown in NIPA tables, but can be inferred, at annual frequency, from the GDP-by-industry tables and input- output tables, for most of the sectors. Farm gross output (which includes farm housing ) can be obtained directly from GDP by industry and NIPA table 8.10, Farm sector output. Housing gross output can be obtained directly from NIPA 8.12, Housing sector output but the published GDP by industry estimate differs slightly because it excludes government enterprises. There is no issue for output of general government and private households because output equals value added by definition. The most difficult adjustment concerns NPISHs. In the GDP by industry accounts, the estimates of these organisations are embedded in the data for several industries and are not readily separable.

9.4 Concluding remarks

Both Canada and the USA compile production accounts for institutional sectors with similar, though not identical, boundaries: the business sector, the government sector and the NPISHs sector. It may be noted that neither country fully follows the guidelines of the 1993 SNA in this regard. In Canada, the production account for the three sectors includes output, intermediate consumption of goods and services and, residually, value added. In the United States, there is additional detail within these three sectors and value added is produced for each of these; however, gross output can easily be inferred for all the sectors from the annual GDP by industry accounts or IO tables, except NPISHs. Further, the USA can provide the sectoral breakdown of value added within the business sector whereas Canada does not currently have such estimates.

Both Canada and the USA need to develop production accounts for institutional sectors to conform with the 1993 SNA recommendation. This will help us to fully integrate with the income and outlay, capital finance and balance sheet accounts for institutional sectors which we already compile and which are consistent with the 1993 SNA guidelines. Production account by sector is one of the crucial recommendations of the 1993 SNA and should be given a high priority. In Canada, we may develop this database during the next few years as we obtain information from the unified production surveys and it is quite likely that the BEA, after the comprehensive revision in 2003, may start moving toward a sectoral presentation that more closely aligns with the 1993 SNA. Once done, we will have the same sectoral boundaries used for all of the accounts, (production, income and outlay, capital, and other accounts), and this will provide a rich database to analyse the performance of each sector of the economy in all its activities.

The business sector is widely used in productivity analysis in both Canada and the United States. In our judgement, it will be useful to retain it, as an alternative or supplementary sectoral presentation.

Statistics Canada - 422 - Collected Articles of Kishori Lal

10. Government Sector

The institutional units performing functions such as government administration, public order and security, defence, social protection and social security are unambiguously classified in all countries in the government sector. However, functions such as education, health, recreation and culture are performed in many countries by both government units and non-profit institutions, the latter may be controlled and financed or simply partly financed by the governments. The sector classification of such non-profit institutions varies amongst countries, depending upon conventions, traditions, and interpretation of international guidelines, thus making international comparisons problematic.

10.1 1993 SNA

In the 1993 SNA (paragraph 4.113), the government sector consists of the following institutional units:

The sector does not include government owned corporations (called government business enterprises in Canada), even when all the equity of such corporations is owned by government. These corporations form part of the corporations sector.

10.2 Canadian practice

The CSNA follows the same rules as the 1993 SNA for allocating units to the government sector with one modification relating to NPIs. We have not differentiated NPIs that are both controlled and mainly financed by government from NPIs that are only mainly financed by government units. We made a conscious decision to put the maximum weight on the mainly financed aspect as an operational guide to classify units as there is an excellent auditable information available to measure the share of finance. However, we have no readily available information to measure the degree of control in order to shift the classification of an NPI from one sector to the other when necessary. Thus, we have classified all NPIs mainly financed by government in the government sector, irrespective of level of control. Most of them are in the health and education sectors.

The value of output in the government sector equals its intermediate consumption, compensation of employees, consumption of fixed capital and taxes on factors of production (mostly on real estate property). As the methodology to estimate consumption of fixed capital for the government sector is different in the two countries and it affects the value of output, this issue is discussed under a separate heading below, issue # 11.

Statistics Canada - 423 - Collected Articles of Kishori Lal

10.3 USA practice

There are major differences between the Canadian practice and that of the United States. In the production account, the US government sector is pretty much limited to government administration and defence. In the USA as well as in many other OECD countries, the maximum weight has been put on the control aspect as a guide to classify NPIs. Government controlled hospitals and universities are in the government sector. However, NPI=s mainly financed but not controlled by the government are not allocated to the government sector.

There are certain conventions used in the estimation and presentation of industry tables which need to be noted for fuller understanding. The US accounts currently adopt the convention that purchases of goods and services by general government are final consumption expenditures, therefore, gross output of general government equals value added only and intermediate consumption is zero . In contrast, the 1993 SNA recommends that these purchases be classified as intermediate consumption and that final consumption consists of the collective services produced by these institutions.

It may be noted that in the upcoming comprehensive revision of the NIPA=s in 2003, BEA is preparing accounts that will permit the presentation of government as a producer of services. Plans (see more information in the January 2003 issue of the Survey of Current Business ) are to reflect government gross output and its components parts (value added and intermediate consumption), as well as its uses for final consumption expenditures, sales and own account capital formation, a presentation fully consistent with the 1993 SNA and the present Canadian practice. In the 1997 Benchmark IO tables, released in December 2002, the general government now includes force- account construction and own-account software inputs, in addition to labour compensation and consumption of fixed capital.

10.4 Concluding remarks

The government sector in both countries is quite large but in Canada it is even larger because many NPIs (which are mainly financed but not controlled by the government) are classified in the government sector. For example, the government sector in Canada makes about 16% of total value added, compared with about 10% in the United States. This huge difference primarily arises due to the different classification conventions used in the two countries. Canada includes, in addition to the activities noted for the USA, health services and the universities and these additions explain most of the difference in the published values for the government sector in the two countries. Thus, the values for activities limited to public administration and defence are quite comparable in the two countries if one goes into the detail but such detail is typically not presented or published in our regular aggregates. In any case, it is the published aggregate numbers that set the tone of economic, political and journalistic analysis.

As noted, the 1993 SNA recommends that the value of output of the government sector should include all their costs, intermediate consumption and all elements of value added; Canada follows this recommendation. The BEA is planning to follow the same in upcoming comprehensive revision of national accounts in 2003, and has already incorporated some of these recommendations in the 1997 Benchmark IO tables, released in December 2002.

Statistics Canada - 424 - Collected Articles of Kishori Lal

11. Consumption of fixed capital, government sector

The value of consumption of fixed capital included in the value added of the government sector in Canada is about 13%, an item of a fairly significant value and of similar importance vis-a-vis the rest of the economy. It needs to be stated that a change in the value of consumption of fixed capital for a given period in the business sector affects the level of profits or net income but it does not affect the level of gross value added, hence GDP. However, a change in the value of consumption of fixed capital in the non-business sectors (government and NPISHs) affects the level of gross value added as it is an additional imputed item for cost, hence affects the level of GDP of the sector as well as of the total economy. There are two methods widely used to calculate CFC, linear and geometric and the resulting value is different.

11.1 1993 SNA

As noted above, the 1993 SNA does not recommend one over the other method; it states: ABoth the linear and the geometric, or declining balance, method are easy to apply. The choice between them depends upon knowledge, or assumptions, about the implied profiles of rentals which underlie them. It is not possible on a priori grounds to recommend the use of one in preference to the other in all circumstances. It is possible, for example, that linear depreciation may be realistic in the case of structures, while geometric depreciation is more realistic in the case of machinery and equipment@. (paragraph 6.197). In the economic literature, particularly in the field of productivity analysis, there is a preference for geometric depreciation; further, the valuation of second hand goods in the market suggests a choice for geometric rates.

11.2 Canadian practice

The Investment and Capital Stock Division (ICSD) at Statistics Canada calculates estimates of CFC for all industries using three methods, two of them are the most popular: linear depreciation rates and geometric depreciation rates. ICSD has published series for the government sector back to 1961 using both methods. In every year, the series based on the geometric rate is about 75% of the estimate based on the linear rate. For example, in 2001, the geometric depreciation value was $16.6 billion, compared with $21 billion using linear depreciation, a difference of $4.6 billion. The CSNA has continued to use the linear depreciation rather than the preferred geometric depreciation for this sector, perhaps, for historical continuity reasons. Back a few decades ago, the CSNA was including CFC in the government sector and at that time, the only method applied was linear. The CSNA stayed with the linear estimates even when the preferred calculation became available by 1996 or so, before the 1997 CSNA historical revisions. This higher estimate of CFC by $4.6 billion in 2001 has resulted in an increase, by the same amount, of both value of output and value added of the government sector. The share of the government sector in the overall economy is thus higher by about one half of one percent, due entirely to the convention chosen for the CSNA, and this is not an insignificant difference.

Statistics Canada - 425 - Collected Articles of Kishori Lal

11.3 USA practice

The BEA uses geometric depreciation for most assets for all the sectors and industries , including the government sector; however, there are exceptions such as autos, computers, missiles, and nuclear fuel rods. For more information, see the most recent methodology for the Fixed assets estimates. If the geometric depreciation rate produces a lower estimate of CFC as in Canada, the published share of the government sector in the USA economy is lower compared with those countries using linear depreciation rates.

11.4 Concluding remarks

Here the issue is not of consistency with the 1993 SNA, as it does not recommend, on a priori grounds, the use of one method (say linear) in preference to the other (say geometric) method for calculating CFC but of the difference in methodology in the two countries (linear in Canada and mostly geometric in the United States) which affects the share of the value of output and of value added of the government sector in the economy. Many economists and analysts in North America prefer the estimates based on geometric depreciation. In principle, the life in the geometric method for depreciation is infinite but in practice, one truncates it and in Canada, it is truncated in the year which is five times the length of the average service life of the asset. Suppose the asset life of an asset is 10 years. In the linear methodology, the asset is completely removed from the stock at the 10th year whereas in the geometric methodology, it will be removed at the 50th year of the asset. Of course, the value of depreciation using the geometric methodology will be very small towards the end of the period.

There is a lot of sensitivity attached to the comparative role of government in the two economies. As noted in issue # 10, Government sector, the published values are proportionately much higher in Canada than in the USA because the CSNA includes most of the NPIs in the government sector whereas the same are classified to the NPISHs sector in the BEA national accounts. The difference in methodology for CFC in the government sector aggravates this situation further as the Canadian linear methodology generates a higher value of CFC compared with the geometric methodology.

It is strongly recommended that the CSNA should re-examine its present preference for the CFC estimates based on linear methodology for the government sector, as the alternative geometric methodology has many listed advantages: a) it is the preferred methodology by many analysts; b) it is used, for most assets, by the BEA, thus both countries will have similar methodology; and c) its application will reduce the apparent gap in the role of the government sector in the two economies,

12. Non-Profit institutions serving households (NPISHs) sector

Non-profit institutions (NPIs) serve corporations, and governments and they are classified to those sectors. There are also NPIs which exclusively serve households, and these are called NPISHs and are discussed here.

Statistics Canada - 426 - Collected Articles of Kishori Lal

12.1 1993 SNA

The 1993 SNA states: A Non-profit institutions are legal entities created for the purpose of producing goods and services whose status does not permit them to be a source of income, profit or other financial gain to the units that establish, control or finance them@ (paragraph 4.161). Some NPIs charge prices and fees that are economically significant. The 1993 SNA defines significant prices as A prices which have a significant influence both on the amounts the producers are willing to supply and on the amounts purchasers wish to buy@ (paragraph 4.161). NPIs which charge significant prices are typically part of the corporations sector and these, for example, include chambers of commerce, trade organisations, industry associations, etc. The majority of NPIs, however, are likely to be non- market producers that provide goods or services to other institutional units either free or at prices that are not economically significant. NPIs which are non-market producers and are controlled and mainly financed by government are allocated to the government sector; most of those not allocated to the government sector form the NPISHs sector. Thus, the NPISHs sector is a residual sector.

12.2 Canadian practice

The 1993 SNA emphasises the double criteria, controlled and mainly financed, for classification of NPIs in the government sector whereas, in the CSNA, we have used a single criterion of finance as a guide for allocation. The residual NPISHs sector in the CSNA includes only those NPIs which are not mainly financed by the government, whereas the strict adherence to the double criteria of the 1993 SNA would lead one to include also NPIs mainly financed by government so long as they are not controlled by government.

In the production account, the value of output of the NPISHs sector equals its intermediate consumption, compensation of employees, consumption of fixed capital and taxes on factors of production, as recommended by the 1993 SNA. Note that the NPISHs sector is separately identified in the production account but it is not separated in the CSNA from the household sector for the income and outlay account, the capital and financial account, and the balance sheet account.

12.3 USA practice

In the United States, all NPIs mainly not financed by government form part of the NPISHs sector, as in Canada but it also includes all those NPIs mainly financed by the government so long as they are not controlled by it. The phrase mainly financed needs to be elaborated. In Canada, the hospitals are public hospitals, entirely financed by the government whereas in the United States, the funding or financing provided by the government is less than 50%. Similarly, the universities in Canada are financed for some 70% of their expenditures by the government but the financing by the government is much lower in the United States, even lower than in the hospital sector there. Thus, the criteria used by the BEA for classifying NPIs to the NPISHS sector reflect the institutional structure and financing arrangements there which permit an inclusion of many more NPIs in NPISHs than in Canada.

In the BEA NIPA accounts (Table 1.7), value added of NPISHs is published but there is no estimate of their output and intermediate expenditures. Further, the value added of the nonprofit institutions

Statistics Canada - 427 - Collected Articles of Kishori Lal sub-sector includes only compensation paid to employees of these institutions but the value added related to their ownership of fixed assets, that is the consumption of fixed capital, is currently shown in the business sector. It is noted by the BEA that it is placed in the business sector by analogy to the treatment of owner-occupied housing. In both cases, the household or NPISH is thought of as running a separate rental business (in the real estate industry) in which the fixed assets are rented back to the owning household or NPISH. For the upcoming comprehensive revision in 2003, the BEA plans to change this treatment and start showing the rental value of NPISHs fixed assets in the NPISHs sector. Furthermore, the rental value of NPISH fixed assets will be allocated to the appropriate industry rather than shown in the real estate industry. These changes are described in an article, Preview of the Revised NIPA Estimates for 1997, Survey of Current Business, January 2003.

It is worth noting that the BEA differs from the 1993 SNA guidelines in the imputation of the rental value of fixed assets owned and used by NPISHs as it includes net interest as well as CFC, a convention which makes the value added of NPISHs higher than in other countries, including Canada, which follow the 1993 SNA. How important this addition is may be gauged from the following data: in 2000, gross value added of NPISHs (Table 1.7) was US$418 billion; imputation for CFC was US$40 billion and for net interest was US$17 billion (Table 8.21). These two imputations now form part the business sector but will be added to NPISHs in the upcoming comprehensive revisions, thus the value added for NPISHs would increase by 14%, of which 4% for net interest would be applicable only in the USA as no other country follows this convention so far. If we apply the same ratio in Canada, its overall GDP would increase by two-tenths of one percent.

In the GDP by industry accounts and the input-output tables, NPISHs are classified by industry along with other producers; thus, no separate output and their intermediate inputs are available. There are several non-profit hospitals as well as state and local government hospitals which now form part of NPISHs and they could be reclassified, but not done yet, as market producers, since their receipts generally approximate their expenses.

12.4 Concluding remarks

The concluding remarks concerning the government sector also apply here. In the United States, State and local government hospitals and universities are part of the government sector but all other hospitals and universities are in the NPISHs sector whereas all the NPIs in the hospital and education sub-sectors are included in Canada in the government sector. In Canada, NPISHs sector makes about 1% of total GDP whereas in the United States, it makes 4% and again a huge difference between the published data in the two countries. Apart from the classification differences, there is one additional reason for the difference in the share of value added of NPISHs sector in the two countries: In the United States, the consumption of fixed capital of NPISHs sector is currently allocated to the business sector whereas in Canada, it remains allocated to NPISHs but this will change, as noted above, in the upcoming comprehensive revisions. If we aggregate the two non-market sectors- Government and NPISHs- their share in the value added is 17% in Canada compared with 14% in the United States. Of the remaining 3% difference, about 0.5% is explained by the difference in the methodology of calculating CFC in the government sector and 0.4% is due to the current allocation of CFC of NPISHs sector to the business sector in the United States. The rest of the difference is

Statistics Canada - 428 - Collected Articles of Kishori Lal mostly due to the much heavier involvement of governments in Canada, compared with the USA, in the provision of public health services to the population at large.

The published numbers encourage the users to draw wrong conclusions that there is more government and less charity in Canada compared with the situation in the United States but in reality this is not true. One possible solution may be to have detailed discussion with our colleagues in the BEA and jointly agree on the classification rulings to be adopted by the two countries. In the meantime, let us put more emphasis on the results of the two non-business sectors together and publish, within the government sector, sub-sector detail for a) government administration and defence and b) other government activities including education, health etc.

As noted, the 1993 SNA recommends that the value the output of the NPISHs sector should include all their costs, intermediate consumption and all elements of value added; Canada follows this recommendation and the BEA plans to change its present practice in the upcoming comprehensive revisions such that it will be consistent with the 1993 SNA.

It is encouraging to note that the forthcoming (2003) UN Handbook on Nonprofit Institutions in the System of National Accounts (prepared in cooperation with the Centre for Civil Society Studies, Institute for Policy Studies, The Johns Hopkins University, Baltimore, USA) is proposing to produce a Satellite Account on Nonprofit Institutions whose boundary is quite similar to the aggregation of all the NPIs in the 1993 SNA. For example, The UN handbook on NPIs (see paragraph 2.14) defines the nonprofit sector as consisting of organisations that, (a) are not-for-profit and, by law or custom, do not distribute any surplus that may generate to those who own or control them; (b) are institutionally separate from government; (c) are self-governing; and (d) are non-compulsory. Such a satellite account will eliminate some of the differences in classification rulings which we have observed between the two countries.

13. Household Sector

One of the primary purposes of economic production is to satisfy the needs of households and in developing countries most of such production is carried out by the households themselves. Even in developed countries, households provide a significant proportion of economic output. Thus it is important to have a precise understanding of what should be included in its boundary in the national accounts.

13.1 1993 SNA

13.2 Canadian practice

In the production account of the CSNA, as noted above, all producing units of the household sector are merged with the two corporations sectors to form the business sector. Even when households hire

Statistics Canada - 429 - Collected Articles of Kishori Lal domestic workers, such as baby sitters and nannies, their activity is classified as part of Aother personal and household services@ industry in the 1980 Standard Industrial Classification or Aprivate households industry@ in the NAICS, but always as part of the business sector. Thus, in the Canadian production account, the household sector is not separately identified. Our departure from the 1993 SNA guideline is due to how our production surveys are conducted. The legal identification of the producing establishment as unincorporated or incorporated is of secondary importance for industrial statistics. Thus the detail on outputs and inputs by unincorporated sector, even when available for certain industries, is inadequate to produce the production account for the household sector. There are two statistical developments underway which should help us to separate the unincorporated enterprise units from the production account of the business sector. One is the enhancement of the Business Register at Statistics Canada which now includes the dual classification of enterprises and the establishments forming part of that enterprise, thus providing a sector classification of all establishments. The other development is the use of unified enterprise surveys, with detail available both for the enterprise and its constituent establishments. It is to be noted that the persons and unincorporated businesses sector of the CSNA relating to the income and outlay account, the capital and financial account, and the balance sheet account approximates the definition of the 1993 SNA with the one exception that the CSNA sector also includes NPISHs.

13.3 USA practice

In the United States, as in Canada, unincorporated enterprises owned by the household sector are merged with the corporate sector to form the business sector. In the 1992 Benchmark IO tables, there used to be one Household industry which related to the hiring by private households of domestic workers such as nannies and baby sitters and whose output consisted of compensation paid to such domestic workers. The value of this activity was insignificant, less than two-tenths of one percent of GDP. This activity, however, was treated as produced by the household sector, not by the business sector. In the 1997 Benchmark IO tables, this industry has been moved to Other services, following NAICS. It needs to be noted that in the upcoming comprehensive NIPA revision, the BEA is planning to add the owner-occupied housing industry to the household sector. For the income and outlay and capital finance accounts, the boundary of the household sector is closer to that in the 1993 SNA.

13.4 Concluding remarks

Both Canada and the USA differ from the procedure recommended by the 1993 SNA with respect to the boundary of the household sector for the production account. The share of value added by the household sector in the economy is zero in Canada and was less than two-tenths of one percent in the United States, produced by Household industry. The Household industry in the US IO tables has been moved to Other services, following NAICS, for the 1997 IO benchmarks. In all the OECD countries, own account housing services (owner occupied dwellings) are imputed to be performed by the household sector; most of agriculture, a significant part of the retail trade, repair services and professional services are performed by households. In the published detail, the share of household production is typically more than 20% of the total production in most of the OECD countries, compared with zero in Canada and an insignificant amount in the United States, thus making international comparisons by sector very problematic. It should be possible to sort out in a few years

Statistics Canada - 430 - Collected Articles of Kishori Lal the production boundary in the CSNA for the unincorporated enterprises owned by households in all the industries. The BEA is planning to allocate the owner-occupied housing to the household sector in 2003 and the CSNA may be encouraged to do the same, thus improving comparisons between the two countries as well as internationally.

Statistics Canada - 431 - Collected Articles of Kishori Lal

Section C: Valuation issues and industry detail, varied practices

In contrast to the production account by institutional sectors, the production account by industry has been in existence for a long time. Many countries produce production account by industry using unique valuation practices and conventions, some produce them through the input-output tables and all face the issue of maintaining consistent long time series. This section will cover most of these issues for Canada and the USA and compare their practices and point out any differences from both the 1993 SNA and each other.

14. Valuation of output

Output may be valued at the factory gate of the producing unit, including product taxes, excluding product taxes, including revenue received as subsidies, not including revenue received as subsidies, at the place of the purchasing unit, etc. Different countries may choose different valuation conventions for valid reasons such as their particular business accounting rules, or the particular principles used for the statistical surveys; however, this would make international comparisons very problematic. Hence, in the various vintages of the International manuals on SNA, guidelines are provided to value output in a way that makes it internationally comparable.

14.1 1993 SNA

The preferred basis in the 1993 SNA for the valuation of output of goods and services produced for the market is at basic prices, especially when a value added tax (VAT), or similar deductible tax, is imposed (paragraph 6.218). It is defined as follows: "The basic price is the amount receivable by the producer from the purchaser for a unit of a good or service produced as output minus any tax payable and plus any subsidy receivable, on that unit as a consequence of its production or sale. It excludes any transport charges invoiced separately by the producer" (paragraph 6.205 a). There are taxes on products and other taxes on production; similarly, there are subsidies on products and other subsidies on production. Since there may be some ambiguity regarding which taxes and subsidies are referred to in the 1993 SNA definition, the 1995 ESA clarifies and restates it as follows: AThe basic price is the price receivable by the producers from the purchaser for a unit of a good or service produced as output minus any tax payable on that unit as a consequence of its production or sale (i.e. taxes on products) plus any subsidy receivable on that unit as a consequence of its production or sale (i.e. subsidies on products). It excludes any transport charges invoiced separately by the producer. It includes any transport margins charged by the producer on the same invoice, even when they are included as a separate item on the invoice A (paragraph 3.48). Thus, in summary, the value of output of a product at basic prices represents the value of output at the gate of the producing unit excluding any taxes on product payable and including any subsidy on product receivable as a consequence of its production or sale.

14.2 Canadian practice

In the industry surveys conducted by Statistics Canada, producing units are instructed to value sales excluding any taxes payable on products and also excluding any subsidy receivable on products. Taxes on products are far more prevalent and significant in Canada compared to a handful of

Statistics Canada - 432 - Collected Articles of Kishori Lal products which receive subsidies. Thus, the value of output of most products in our industrial surveys is at basic prices, as defined in the 1993 SNA. We are different from the SNA guidelines only in those handful of products which receive subsidies as our value of output in those cases does not include the subsidies received. We have modified the 1993 SNA valuation guidelines in such cases and have called this valuation at modified basic prices, the modification being the exclusion of subsidies receivable on products. Our rationale for this modification is that we find it useful to record the value of output based on observed transaction prices as received by the producers and paid by the purchasers and listed on invoices, and hence verifiable. Our modified basic prices are always lower than the 1993 SNA basic prices by the amount of the subsidies on products receivable by any given producer and these modified prices are the ones used in the CSNA IO tables. The CSNA definition of modified basic price reads as follows:

The modified basic price used in the CSNA IO tables is equivalent to the price at the factory gate of the producing establishment. The advantage of using it is that the valuation of transactions between producer and purchaser is transparent and verifiable. These important accounting characteristics are not available in the event that the basic price, as recommended in the 1993 SNA, is used for valuing output. In such a case, the transactions will be reported, not at the prices prevailing in the market but at higher notional prices (market price plus subsidy per unit), which the purchaser does not pay and does not record: a feature that is not very appealing.

Another advantage, which is even more useful, is the facility and efficiency that our modified basic price approach provides for calculating constant price IO tables. We produce current price supply tables at modified basic prices and current price use tables at both purchaser prices and modified basic prices. We start with use tables at purchaser prices, as recommended by the 1993 SNA, and convert them to modified basic prices, the same prices as those used in supply tables. Once we have converted our purchaser price use tables into modified basic prices tables, both the supply and use tables are available at the same prices. This additional calculation is not discussed in the 1993 SNA. In Canada, we collect the sale prices of products excluding the taxes on products and without including the subsidies on products. This set of prices is immediately applicable to the values in the use tables which have already been converted at modified basic prices.

Deflating the basic price supply table would be difficult as basic prices are not observed in the market from the purchasers= point of view and would, therefore, require bold assumptions. Furthermore, deflating the use tables at purchaser prices will be very expensive as, in principle, each cell in the use table at purchaser price has its own unique price deflator. The 1993 SNA guidelines are problematic to implement cost-effectively for Statistics Canada which produces both current and constant price supply and use tables and constant price value added by industry, using the recommended double deflation approach, as benchmarks for many series in the CSNA. Given that modified basic prices are transparent and, more importantly, given their advantage for calculating constant price IO tables and value added by industry, the CSNA has not incorporated the 1993 SNA

Statistics Canada - 433 - Collected Articles of Kishori Lal recommendation to present value of output at basic prices for those products which receive product subsidies.

14.3 USA practice

In the industry and input-output accounts of the United States, output is valued at producer=s prices. These prices have the following features: a) they exclude wholesale and retail trade margins as well as transportation costs; b) they include all federal customs duties, as well as all federal, state, and local government excise and general sales taxes collected by the producers for later transmission to the respective governments; and c) they do not include government subsidies received by the producers in the valuation of their output. As the valuation of output of the producers affects the valuation of their value added and other items in all accounts, it is crucial that its impact and boundary be well understood for proper international or inter-country comparisons. The valuation of output in the industry and Input-output accounts of the USA is higher by the amount of taxes less subsidies on products when compared with the valuation of output at basic prices in the 1993 SNA, and is higher by the amount of taxes on products when compared with the valuation of output at modified basic prices in the CSNA industry statistics. These product taxes contribute significant amounts, about five percent of value added for the total economy, and, more importantly, they have very substantial impact in industries such as wholesale and retail trade and restaurants and hotels in which they are currently included. The different valuation conventions used in the valuation of output in the USA and Canada impose serious difficulties for inter-country comparison of their industry statistics. The national accountants at the BEA are aware of this controversy and may move towards the 1993 SNA recommendation on basic prices and convert their industry accounts from producers= prices. The 1997 benchmark IO accounts, released in December 2002, have the valuation of output at producers= prices, thus the subsequent annual IO accounts as well as the upcoming GDP by Industry comprehensive revision will stay also at producers= prices. Conversion at basic prices may be considered for the next benchmark revision in 2007. In the meantime, BEA=s GDP by Industry staff has been able to provide estimates at basic prices for some international submissions by excluding product taxes from both gross output and nominal value added , while including product subsidies.

14.4 Concluding remarks

We support the 1993 SNA preferred valuation of output at basic prices. In a handful of industries which receive product subsidies, we find it useful to modify the 1993 SNA definition by not including product subsidies in the valuation for two reasons: a) the 1993 SNA valuation is not transparent, hence not verifiable from the records of the purchasers, and b) our modification makes the calculations of constant price IO tables very efficient, a feature not discussed in the 1993 SNA. The 1993 SNA prefers valuation of output at basic prices but also notes that Aproducer=s prices may be used when valuation at basic prices is not feasible@ (paragraph 6.218). In countries where the industrial surveys can collect information on the value of output only at producer=s prices, there is no choice but to compile output at producer=s prices. However, international comparisons are difficult to make when the valuation practices are different amongst the same block of countries such as the OECD or two neighbours such as Canada and the United States. Though not elaborated in the 1993 SNA, an argument can reasonably be made that product taxes should not be recognised as a revenue

Statistics Canada - 434 - Collected Articles of Kishori Lal item as they do not add to the net worth of the producer collecting such taxes and then transmitting them to the governments. Business accounting principles do not record such tax collections on behalf of the government as revenue of producers. Producers= price valuation causes very significant additions to the output of trade industries, thus rendering international comparisons of industry statistics and resulting productivity calculations problematic.

Once the output values in the industry statistics of the USA are produced at basic prices, or preferably at modified basic prices, the industry series will lend themselves readily to inter-industry and international comparisons, particularly for multi-factor productivity analysis. See more comments on this topic in the next issue #15, Valuation of value added by industry.

15. Valuation of value added by industry

Value added is intended to measure the additional value created by a process of production. It is one of the most important constructs in the national accounts, hence its valuation must be carefully analysed. In economic theory, this concept has traditionally been known as value added at factor cost or GDP at factor cost. Some economists have preferred to use the concept value added at market prices which includes a slightly bigger boundary for valuation. As this is a very important construct, let us pursue its meaning as elaborated in the 1993 SNA.

15.1 1993 SNA

The 1993 SNA prefers that both output and the balancing item, value added, be valued at basic prices and it states: AGross value added at basic prices is defined as output valued at basic prices less intermediate consumption valued at purchasers= prices@ (paragraph 6.226). The preference for value added at basic prices rather than the traditional value added at factor cost derives from a proposition that one must take into account the full cost of factors of production in any analysis of production. Value added at basic prices is higher than the traditional value added at factor cost by the amount of other taxes on factors of production (such as property and payroll taxes) less other subsidies provided to factors of production (such as labour training). The value of output of any producing unit must be sufficient to pay the full costs of the intermediate consumption of goods and services and the full costs of the factors of production, labour and capital, used for its output, if the unit is to survive in the long run.

15.2 Canadian practice

The CSNA has always produced, for its industry statistics program, value added by industry (also called GDP by industry) at factor cost during the period from the 1950's to the year 2000. Our users were very familiar and comfortable with our concept of value added at factor cost, and this was also the concept used in economic textbooks. Despite this long tradition, we have now implemented the recommendation of the 1993 SNA for value added at basic prices. In the historical revision of the CSNA in 2001, we revised our GDP by industry series back to 1961 at basic prices. It may need to be repeated that our output by industry series are not at basic prices but at modified basic prices (as noted above in issue #14, valuation of output) but our value added by industry is at basic prices. The value added at basic prices of a producing unit in Canada is equal to its output valued at modified

Statistics Canada - 435 - Collected Articles of Kishori Lal basic prices less its intermediate consumption at purchasers= prices plus any subsidy receivable by that unit as a consequence of its production or sale (i.e. subsidies on products). Rather than adding subsidies on products to the value of output, we add them to our calculated value added from modified basic prices. Our measurement of value added at basic prices is identical to the one produced according to the 1993 SNA conventions.

In Canada, the most important taxes on factors of production are property taxes and payroll taxes. Payroll taxes are imposed by four provinces (Manitoba, Ontario, Quebec, and Newfoundland), paid by the employers, and the tax liability is calculated as a proportion of total wages and salaries, in every industry. Such taxes were treated as social insurance, hence part of supplementary labour income (SLI) in the pre-1997 CSNA historical revision. Should employer payroll taxes continue to be included as SLI or treated as taxes on production was the issue discussed in the CSNA in 1997. It was felt by the staff in the CSNA that any employer contribution not giving a specific economic benefit to the employees or his/her dependents should be considered as tax on production, not as SLI. Mandatory payments by the employers to the government, to cover pension benefits, employment insurance etc for the specific benefit of their employees, however, are part of SLI. Hence, it was decided to change the classification of payroll taxes from SLI to tax on production for the CSNA historical revision in 1997. The significance of our decision on payroll taxes may be gauged from the following data for 1998: total wages and salaries were $419 billion, total SLI was $56 billion and total payroll taxes (not included in SLI) were $7 billion. Had we added payroll taxes to SLI, total labour compensation would have increased by 1.5%. Unlike the payroll taxes which affect all industries, the property taxes mostly affect the real estate industry. These taxes are quite significant, amounting to more than 5% of total GDP, but in the real estate industry, they amount to about 20% of value added. Subsidies on the factors of production, mostly for labour training, are not materially important, amounting to just two-tenths of one percent of labour costs, and are prorated on the wages and salaries by industry.

15.3 USA practice

In the United States, the BEA=s GDP by Industry statistics and input-output tables use another concept of value added by industry: value added at market prices or producers= prices. As noted above, the value of output of industries in the USA is at producers= prices, which include all the federal, state and local government sales and excise taxes. Value added by industry is equal to its output at producers= prices less its intermediate consumption of goods and services at purchasers= prices. This calculation produces value added by industry which, in total, equals GDP at market prices. This presentation is quite appealing to users as it fits in with the notion that final expenditures on GDP at market prices must equal value added by industry at market prices. Many users in the USA are very familiar with this presentation and it is not the only country which adopts it. This valuation is higher than the one at basic prices, by the amount of product taxes collected by industries less subsidies on products received by them. Most of the product taxes, such as federal, state and local government sales and excise taxes in the USA (equivalent taxes in Canada are the goods and services taxes-GST- and the provincial sales taxes-PST) are collected primarily by trade establishments. Thus value added at market prices for the trade industry is much higher than its valuation at basic prices or at factor cost. On the other hand, value added at market prices for highly subsidised industries, such as agriculture, is lower than the one calculated at basic prices.

Statistics Canada - 436 - Collected Articles of Kishori Lal

There is a small statistical and conceptual difference in the calculation of value added in the two programs, GDP by Industry and IO tables. In the GDP by Industry program, inventory valuation adjustment (IVA) is distributed by industry; however, in the IO tables, IVA is shown only at the total economy level, via a special industry called Inventory Valuation Adjustment in the Use matrix, whose only input is IVA and is listed in the row of value added. This convention in the IO tables is unique, as no other country, to the best of our knowledge, uses it. If it does not require lot of resources, its present application may be re-examined and perhaps dropped in favour of its distribution by industry, thus helping international comparisons

The USA provides to the OECD gross value added at basic prices for the total economy and details for six main industrial groupings. (See Table 2, Gross domestic product: output approach, in the 2002 edition of the OECD publication, National Accounts of OECD Countries Main Aggregates). The six industrial groupings are: 1.Agriculture, hunting and forestry; fishing; 2. Industry including energy; 3. Construction; 4. Wholesale and retail trade, repairs, hotels, and restaurants, transport; 5. Financial intermediation, real estate, renting and business activities; and 6. Other service activities. The BEA may, in the near future, adopt the 1993 SNA recommendation on basic prices and convert their industry accounts from producers= prices.

15.4 Concluding remarks

As of 2001, 24 member countries of the OECD (all 15 European Union countries, Canada plus eight others) use the recommended basic price valuation for value added by industry series. The remaining six member countries (Japan, Korea, New Zealand, Switzerland, Turkey, and the United States) use producer price valuation for output and, thus, residually producer price/market price valuation for value added by industry.

Let us gauge the statistical importance of the various valuations, using the CSNA data. In 2001, GDP at factor cost was $964 billion; adding $53 billion for taxes less subsidies on factors of production resulted in GDP at basic prices of $1017 billion; further adding $75 billion for taxes less subsidies on products resulted in GDP at market prices of $1092 billion. These additions not only affect the level of total GDP by industry but the effects are unevenly distributed across industries. As noted, taxes less subsidies on the factors of production mostly affect the real estate industry. However, taxes less subsidies on products affect the trade and repair industries as well as restaurants and hotels. The different valuations make the inter-industry and inter-country comparisons of value added per labour unit or the labour productivity by industry measures very problematic.

Value added at market prices for trade industries is higher by a wide margin compared to its calculation at basic prices, because of a convention to include the taxes collected by these establishments on behalf of governments in their value added. Comments made above on the valuation of output equally apply here. The apparent presentational advantage of total value added by industry shown as equal to GDP at market prices needs to be assessed against its lack of international comparability and the conceptual difficulty of treating the collection of taxes by the business enterprises as their revenue item.

Statistics Canada - 437 - Collected Articles of Kishori Lal

As noted above, the USA already provides value added at basic prices for six broad groupings of industries to the OECD and may, in the near future, adopt the 1993 SNA recommendation on basic prices and convert their industry accounts from producers= prices. Once the United States, and hopefully the remaining five countries still using producers= price valuation, convert to the recommended basic price valuation, the analytical usefulness will increase of not only their accounts but those of other countries= accounts as well, by making them comparable.

There is an additional issue of a difference in the treatment of payroll taxes in Canada and the United States. As noted above, payroll taxes paid by employers in four provinces in Canada are classified as taxes on factors of production whereas the same type of taxes are classified as contributions to social insurance, hence a part of SLI and labour compensation in the United States. This difference in the treatment does not affect value added at basic prices but does affect the share of labour compensation which is an important concept in productivity analysis. Using the convention of the United States, the share of labour compensation in Canada would increase by about 1.5%. It will be worthwhile to have a joint discussion on this issue with our colleagues at the BEA.

16. Supply and use tables

Supply and Use tables (also called Input-Output tables or Make/Supply and Use matrices) have been produced for a long time by many countries, in most countries by their official statistical organisations and in others by some private research institutes. The most famous name in this subject is that of Professor Wassily Leontief of Harvard University who did the pioneering work in this area and was awarded the Nobel prize in 1973 for this work. The input-output tables were very popular in most of the centrally planned economies as they were used for economic planning purposes. In the market oriented economies, they did not form part of the first manual on the System of National Accounts (SNA) document promulgated by the United Nations in the mid-50's. Early in the 1960's, it was felt that the UN SNA manual was inadequate to deal with a growing need for analysis for industrial detail, and with issues relating to income and outlay, accumulation of capital formation and capital finance etc. To take these analytical needs into account, a revised SNA, called A System of National Accounts was published in 1968 by the United Nations. This document was prepared by an outstanding group of national accounts experts, with Professor Richard Stone of Cambridge University, chairing most of the sessions. IO tables became a very prominent part of this report, for some countries too prominent a part, as IO tables assumed a centre stage in the production accounts of the SNA. An excellent annex on mathematical discussion of IO framework was added.

16.1 1993 SNA

The 1993 SNA provides a full sequence of accounts. It starts with production, follows where the fruits of production (value added and final expenditures) go and records the impact they have on savings of each of the important players, called sectors, in the economy. It then joins these results with their capital acquisitions and notes their impact on their net lending. It then examines how the net lending is financed and finally strikes a balance sheet for each of the sectors of the economy. The Production account by both sectors and industries are recommended but these are very macro series, without much detail by industry and commodity. The detail is recommended in chapter 15, called

Statistics Canada - 438 - Collected Articles of Kishori Lal

Supply and Use tables and input-output, after all the sequence of accounts have already been noted and explained.

The 1993 SNA notes: AThe input-output tables and in particular the supply and use tables serve two purposes: statistical and analytical. They provide a framework for checking the consistency of statistics on flows of goods and services obtained from quite different kinds of statistical sources - industrial surveys, household expenditure enquiries, investment surveys, foreign trade statistics etc. The System, and the input-output tables in particular, serves as a coordinating framework for economic statistics, both conceptually for ensuring the consistency of the definitions and classifications used and as an accounting framework for ensuring the numerical consistency of data drawn from different sources. The input-output framework is also appropriate for calculating much of the economic data contained in the national accounts and detecting weaknesses@ (paragraph 15.3).

The 1993 SNA includes an integrated set of supply and use tables as well as symmetric IO tables. In the symmetric IO tables of the 1993 SNA, the number of rows and columns are identical and the same classifications or units are used in both rows and columns- such tables are industry by industry or commodity by commodity. The 1993 SNA states: AThe System recommends that the statistical supply and use tables should serve as the foundation from which the analytical input-output tables are constructed@ (paragraph 15.7).

16.2 Canadian practice

The CSNA has been producing annual supply and use tables, or input-output tables, starting with the reference year 1961. The dimensions of the Canadian IO tables are rectangular, meaning that the number of products is larger than the number of industries. Ever since the IO program was established at Statistics Canada in the early 1960's, it was decided to have rectangular IO tables because in the real world there are many more products than industries and that is how our industrial surveys have always been conducted. For IO tables to serve as an integration framework, it is essential to stick to the rectangular format. We were probably the first statistical institution who opted for the rectangular format. We could not have a better testimonial on the usefulness of rectangular format than from the father of input-output economics, Professor Leontief. Professor Leontief and Anne Carter mentioned two very important advantages of the rectangular format of the input-output tables over the traditional square input-output tables: a). It admits as much detail as is available in the basic census or survey records; and b). The meaning of each entry is straightforward because observed transactions are not combined with fictitious transfers, a feature of inter-industry square tables. ( See Anne Carter and Wassily Leontief, Goals for the input-Output Data System in the Seventies, published in BEA=s Survey of Current Business, July 1971, page 31).

In the CSNA, we have produced national annual IO tables, starting with 1961, and the latest one is for 1999. Previous to the current program with effect from 1961, the first IO table, The Inter-Industry Flow of Goods and Services, Canada, 1949, was published in 1958 by Statistics Canada (then Dominion Bureau of Statistics). This was a square 42 by 42 inter-industry IO table. It provided the basic data (factor cost and capital consumption allowances) and the weights for the industries detailed in the first publication, released in 1963, on Indexes of Real Domestic Product, 1946-61. The tradition for IO tables to be used as benchmarks for other sub-systems of the national accounts

Statistics Canada - 439 - Collected Articles of Kishori Lal started with our first IO table for 1949 and has flourished since then. The format of the IO tables from 1961 was drastically different from that of the 1949 table. The dimensions of the IO tables were not only vastly enlarged but the basic format changed from a square matrix to a rectangular one.

The dimensions of the annual IO tables at the detailed worksheet level for 1961-1980 were 204 industries and 650 commodities, for 1981-1996 were 243 industries and 650 commodities and from 1997 and onwards, the dimensions are 300 industries and 700 commodities. Since 1996, we have produced provincial (10 provinces and three territories) IO tables with the same dimensions as of national IO tables. Our IO tables serve precisely the same statistical roles as noted in the 1993 SNA. The current price annual national IO tables are produced with a lag of 29 months but are held back for another 4 to 5 months for the current price provincial IO tables to be completed as well as the national tables to be converted at constant prices and all are then simultaneously released to the public, with a lag of 34 months.

The Canadian statistical input-output tables have three broad sectors of the economy- business sector, government sector, and non-profit institutions serving households (NPISHs) sector. The business sector is coterminous with the aggregation of producing units of three 1993 SNA sectors, namely non-financial corporations sector, financial corporations sector, and the household sector. The business sector is disaggregated by industry. The NPISHs sector is disaggregated by industry except one miscellaneous industry; however, the government sector is disaggregated, not by industry, but by broad functions, such as education, health, recreation, administration, etc.

In the CSNA, we do not produce supply or output tables at basic prices as recommended in the 1993 SNA, but at modified basic prices. (See more on this in issue #14, Valuation of output). The Canadian modified basic price has the advantage that it is observed (and can be verified) in the transaction records of the producing units. The 1993 SNA basic price requires information which the purchasing unit does not have; hence it must be imputed for users of products. Our preference to connect our information with the accounting records of the institutional and producing units brings transparency to our statistical output.

16.3 USA practice

The history of producing IO tables in the USA starts with Professor Leontief=s original 1936 article, AQuantitative Input and Output Relations in the Economic System of the United States@, Vol 18, , No . 3. (August 1936), which contained a table for 1919. He later produced a table for 1929, and then for 1939. These tables were included in his most quoted book, The Structure of the American Economy, 1919-1939, published in 1951. Under the Eisenhower administration, official input- output work was closed down from 1954-59 as the technique was considered to be a tool of socialism. It is ironical that the same technique was considered a tool of capitalism in the People=s Republic of China. (See more details about it in an article by Karen Polenske, AHistorical and New International Perspectives on Input-Output Accounts@, in Frontiers of the Input-Output Analysis, Oxford University Press, 1989).

The BEA prepares detailed national benchmark input-output tables (make and use tables), with approximately 500 industries and 500 commodities, every five years; these tables incorporate

Statistics Canada - 440 - Collected Articles of Kishori Lal economic census data that are available for those years and are usually produced with a time lag of about five years from the reference year. The tables for other years are summary tables and are based on more limited sample surveys and analytical methods. The dimensions of summary tables were 85 by 85 before the 1997 IO tables; the summary version of the 1997 benchmark table contains 134 industries (see Appendix A. Industries in the 1997 Benchmark Input-Output accounts, pp 39-43, Survey of Current Business, December 2002), thus the dimensions of the future summary annual tables may 134 by 134.

With the completion of the 1992 benchmark IO Accounts, the BEA started producing alternative IO tables, which, with effect from the 1997 benchmark IO tables, are now referred to as standard NAICS based IO tables. The standard 1997 benchmark tables conform closely to the statistical data sources. The BEA notes: A In BEA=s standard make and use tables, all of the products -primary and secondary- that are produced by an industry are assigned to that industry. As a result, the data in these tables are consistent with the industry-based data that are collected and reported by other statistical agencies@ (see an article by Ann Lawson and others, Benchmark input-Output Accounts of the United States, 1997, published in Survey of Current Business, December 2002, page 27). These tables are more consistent with the GDP by Industry accounts and the gross state product accounts and with other industry data that are based on information collected using NAICS. In the 1992 standard tables, own- account construction done by non-construction industries was reassigned to the construction industry but in the 1997 standard tables, it is shown where work is done. The industrial boundary vis-a-vis secondary products in the Canadian IO tables is very similar to the standard IO tables in the United States, except for construction.

The 1992 benchmark IO tables were called traditional IO tables and their format was closer to the analytical or symmetric input-output tables of the 1993 SNA. The BEA has changed the terminology for the 1997 benchmark IO tables. The old traditional tables are now referred to as supplementary tables. The BEA notes: A In the supplementary make and use tables, some of the secondary products are reassigned to the industries in which these products are primary products. The data in these tables and in the total requirements tables that are derived from them are valuable for performing economic structural analysis, impact analysis, and other types of economic modelling@ (see the article by Ann Lawson and others, referred above, Survey of Current Business December 2002, page 27). The supplementary make and use tables are based on the standard make and use tables, except that some of the secondary products are reassigned.

16.4 Concluding remarks

Canada has a very extensive IO program, with national detailed tables produced annually since 1961 and provincial detailed tables produced annually since 1996. These tables are used as annual benchmarks for all the national accounts series throughout the CSNA. Provincial IO tables in earlier periods were produced only occasionally, four times during 1971-1992. Canada has always produced rectangular supply and use tables, with the number of products much larger than the number of industries. It can always produce, very quickly, a square industry by industry table, as it only needs a multiplication of two matrices, industry by product make matrix and product by industry use matrix. Canada does not produce a symmetric commodity by commodity square input-output table (see more

Statistics Canada - 441 - Collected Articles of Kishori Lal on this in issue #17, Symmetric input-output tables). In the United States, detailed benchmark IO tables are prepared every five years based on economic census data and summary tables are prepared for other years, based on more limited sample surveys and analytical methods. There are no five-year economic censuses in Canada except for agriculture, thus all the annual Canadian IO tables are based on the same set of information every year. The standard benchmark tables for 1997 in the USA follow the same industry boundary as in NAICS, a feature very similar to the one followed in Canada. Additionally, the BEA produces supplementary IO tables where some industries are redefined such that they retain, by and large, the principal activity, and, thus, go towards the format of symmetric or analytical table, as noted in the 1993 SNA. In Canada, only the construction activity from all other industries is reassigned to the construction industry.

17. Symmetric Input-Output tables

In the rectangular input-output tables or the supply/make and use tables, produced in Canada and other countries, the number of commodities is much larger than the number of industries. These rectangular tables must be converted into square tables for them to be inverted and used for input- output analysis such as developing total output requirements from any given change in final demand. The square tables are called symmetric IO tables, when the number of rows and columns are identical and the same classifications or units are used in both rows and columns: such tables can be industry by industry or commodity by commodity. It has been argued in the literature that for the input-output analysis, the IO tables must be such that they represent stable technological relationships. In the regular supply and use tables produced by all or almost all countries, industries produce their primary products but also some secondary products which, of course, are primary to other industries. If one converts these rectangular supply and use tables into square symmetric industry by industry tables, the technological relationships represented by such tables must refer to the production of all products by that industry. It is generally asserted that the stable technological relationships are better represented by pure product technology. To do this, one must remove from each industry its secondary output and all its inputs and add both of these to the industry where such outputs are primary. As the only source of information of the input structure for the secondary products is the industry where such products are primary, one must use such information to remove the inputs of secondary output. This operation sometimes generates negative inputs in those industries from which inputs are being removed. As negative inputs are counter-intuitive and non- explainable, mathematical proration techniques and assumptions are used to get rid of these anomalies. The result is a symmetric product by product matrix, developed through many assumptions. Whether a statistical organisation should develop a symmetric industry by industry table or a symmetric product by product table for input-output analysis is an issue which continues to remain unsettled. Let us first see what the 1993 SNA recommends in this area.

17.1 1993 SNA

Statistical units, in particular establishments grouped in industries serve as a common basis for the production accounts and the supply and use tables in the 1993 SNA. Industries always produce primary products but they sometimes also produce secondary products which are primary to other industries. The 1993 SNA recommends a different analytical unit, however, for input-output analysis. The 1993 SNA states: "For purposes of input-output analysis, the optimal situation would

Statistics Canada - 442 - Collected Articles of Kishori Lal be one in which each producer unit were engaged in only a single productive activity so that an industry could be formed by grouping together all the units engaged in a particular type of productive activity without the intrusion of any secondary activities. The appropriate analytical unit for purposes of input-output analysis is, therefore, a unit of homogeneous production, which may be defined as a producer unit in which only a single (non-ancillary) productive activity is carried out. If a producer unit carries out a principle activity and also one or more secondary activities, it will be partitioned into the same number of units of homogeneous production " (paragraph 5.46). It further states: "Although the unit of homogeneous production may be the optimal unit, ... it may not always be feasible to partition establishments... into a series of mutually exclusive units of homogeneous production. In situations of this kind, it will not be possible to collect directly from the enterprise or establishment the accounting data corresponding to units of homogeneous production. Such data may have to be estimated subsequently by transforming the data supplied by enterprises on the basis of various assumptions or hypotheses" (paragraph 5.47). The 1993 SNA further states: AHowever, the unit of homogeneous production is not normally observable and is more an abstract or conceptual unit underlying the symmetric (product by product) input-output tables@ (paragraph 15.14). Despite these recognised statistical difficulties, the 1993 SNA encourages the development of such analytical input-output tables based on a homogeneous production unit, or as close as feasible to such a unit.

17.2 Canadian practice

In the CSNA, we produce (or can easily produce) symmetric industry by industry input-output tables, but not product by product input-output tables. The symmetric product by product tables would have required significant resources and many artificial assumptions. In our judgement, their analytical usefulness is of dubious quality, hence their compilation continues to be considered not cost- effective, and particularly when we already have developed detailed symmetric industry by industry tables.

In the Canadian input-output tables, we do not subdivide establishments to create units of homogeneous production except in the case of own account construction. Both in industry surveys and capital expenditure surveys, a fair amount of information is provided by the respondents to help us separate both the value of own account construction and the corresponding materials and labour used for this activity from all industries, which then is added to the construction industry. Thus for one major activity in our economy, we follow the same direction as noted for the analytical IO tables in the 1993 SNA, also because we get the basic information to do this partitioning. Such information is not collected for other secondary activities in the producing industries. Thus, in the Canadian IO tables, each of the 300 industries has its principal product (s) and its secondary products. There are about 700 products listed in the IO tables but they are aggregated from several thousand products reported by respondents. In fact, the three hundred IO industries have subsumed several thousand separate technologies, simply because of how our data are collected. We have an input or use matrix with 300 industries and 700 commodities and we have an output or make matrix of the same dimensions. Assuming constant domestic commodity shares and constant industry technology, a 300 by 300 industry by industry input table can be quickly produced by multiplying the two matrices, the make matrix with 300 industries by 700 commodities and the use matrix with 700 commodities and 300 industries. Homogeneous production would require a 700 by 700 product matrix, each column showing a unique technology to produce that product but we would have to invoke 700 input vectors

Statistics Canada - 443 - Collected Articles of Kishori Lal from the 300 input vectors that we collect from industry surveys. This scale of artificiality would make any result of very dubious quality.

17.3 USA practice

In the supplementary (previously called traditional) IO tables produced by the BEA, secondary output including own-account construction and their inputs are removed from many industries in the standard tables and are added to the outputs and inputs of the industries where they are primary. The purpose of these reassignments is to assure that each resulting IO industry has a unique output and production process, represented by the mix of inputs, compared with other industries. These changes involve only the outputs and related inputs of some secondary products produced by an industry where the secondary product has a different input mix and production process from the industry=s primary product. For example, hotel and lodging places typically provide eating and drinking services as a secondary product to their primary product of hotel and lodging services. The inputs and production processes for these two activities, however, are very different and need to be separated in the supplementary IO tables for the purpose of preparing total requirements table. Consequently, the outputs and inputs associated with eating and drinking services provided by the hotels and lodging places industry are redefined to the eating and drinking industry for the supplementary IO tables. Note that the changes to industry outputs and inputs for secondary products are made not for all industries but many of them, though large adjustments are made only to a few, compared to only one of construction in Canada. The purpose of these redefinitions in the supplementary IO tables is to attain a greater degree of homogeneity in the inputs required by an IO industry to produce its commodities. Thus, we can say that the approach followed for reassignments of secondary products for several industries, in the supplementary IO tables of the United States, goes towards the conceptual underpinnings of the analytical or symmetric IO tables of the 1993 SNA.

The differences between the supplementary (previously called traditional) IO tables and the standard (previously called alternative) IO tables arise entirely from the importance of secondary products. Standard tables keep the secondary products whereas such products are removed for several industries in the supplementary tables. The difference between the standard and supplementary tables was larger for some industries in the 1992 benchmark IO tables when SIC was the basis of classification, compared with the 1997 tables which are NAICS based. For example, redefined auto repair output from the retail trade industry made up almost 40 percent of IO repair industry output in 1992 but now with NAICS as the basis of classification, this difference is much smaller for the 1997 IO tables, as auto repair output is classified to repair industry, not retail trade.

Until 1992, BEA produced only product by product total requirements matrices, as well as industry by product total requirements matrices. For the 1997 benchmark IO tables which were released in December 2002, an industry by industry total requirements matrix was added to other existing tables.

Statistics Canada - 444 - Collected Articles of Kishori Lal

17.4 Concluding remarks

The 1993 SNA encourages the development of symmetric product by product IO tables, based on a homogeneous production unit, or as close as feasible to such a unit. The Eurostat is in favour of transforming the supply and use tables into symmetric product by product IO tables (see Eurostat draft Input-Output manual, August 2002). The recommendation to produce symmetric product by product tables by both the 1993 SNA and the draft manual of the Eurostat is now challenged by some experts (see a paper by Bent Thage, Symmetric Input-Output Tables and Quality Standards for Official Statistics, presented at the 14th International Conference on Input-Output techniques, October 2002, Montreal).

The symmetric product by product IO tables recommended at the international level, particularly in the European Union countries, are of no more than 100 by 100 dimensions. Even 300 or 500 commodities are aggregations of some 50,000 or so commodities identified in the market. The meaning of commodity technology is questionable when 50,000 technologies are aggregated into a manageable set of 100 products or even 300 or 500 or 1000 products. Such an exercise can hardly be called generating homogeneous production units. One cannot derive any meaningful interpretation of product technology from such an aggregated matrix. Hence, there is no reason to spend vast resources to do such an exercise of very dubious quality.

One alternative is to produce, from the rectangular supply and use tables, symmetric industry by industry IO tables, with dimensions equal to the number of industries in the supply and use tables. The transformation of such tables is efficient, quite inexpensive, transparent, verifiable from the records and the resulting quality is the same as that of the basic tables. These are important ingredients to satisfy the quality dimensions increasingly required by the international organisations. One may also contemplate, as done in the USA and to a limited extent in Canada, redefining some important industries by removing their secondary products and their inputs and adding the same to the industries where they are primary. It is important to fully document these changes so that they are transparent to the users of both IO tables and statistics from regular industry surveys. We owe to our users a high quality and verifiable statistics, in as demystified a way as possible, so that they can use them with confidence.

The supplementary square (with 500 by 500 dimensions) benchmark IO tables in the United States, as noted above, redefine many important industries by removing their secondary products and their inputs and adding the same to industries where they are primary, thus going towards the analytical IO tables of the 1993 SNA, yet maintaining transparency. The construction industry, in the Canadian IO tables, is redefined as in the United States, with all the changes fully documented. The Canadian IO tables are rectangular but can readily be transformed into square (with 300 by 300 dimensions) industry by industry tables. Canada produces industry by industry total requirements matrix and the USA produces industry by industry (as well as industry by product and product by product) total requirements matrices to satisfy user needs for IO analysis. We have no hesitation to say that the technology reading available from the Canadian 300 by 300 industry IO tables is at least as good as, and most likely far superior to, the one available from the 100 by 100 product IO tables produced by some countries, with many assumptions and a lot of resources.

Statistics Canada - 445 - Collected Articles of Kishori Lal

18. Production account by industry

The production account by industry or some important segment of this account, such as the well- known Index of Industrial Production, has continued to be compiled for a long time, and its compilation predates the modern system of national accounts which came into existence immediately after the second world war. Even countries which do not yet compile national accounts collect at least the output of industries of national importance, for both administrative and policy reasons, and thus have some rudimentary information for production account by industries. The production account by industry is prepared in many countries through the input-output tables but there is no inherent reason that it must be so. Input-output tables provide, of course, a useful framework for such statistics; however, those countries which do not regularly produce such tables, may still compile a production account by industry. This item may repeat several arguments already made above in the Supply and Use tables so that the discussion flows smoothly.

18.1 1993 SNA

It needs to be re-emphasised that the 1993 SNA recommends that the production accounts (output, intermediate inputs, value added) be compiled for establishments and industries as well as for institutional units and sectors. It states: AOverall numerical consistency requires that the output of an institutional unit engaged in production -that is an enterprise- should be equal to the sum of the outputs of the individual establishments of which it is composed. As these outputs include deliveries of goods and services to other establishments belonging to the same enterprise, such inter- establishment deliveries are counted as part of the output of the enterprise as a whole even though they do not leave the enterprise@. (paragraph 6.2). Like the institutional sectors, the establishments that make up the total economy are grouped into industries, following international or country- specific standard industrial classification (SIC) manuals. The 1993 SNA recommendation that inter- establishment deliveries always must remain counted, no matter what level of industrial aggregation, is very important from a statistical point of view. Some countries used to produce, or perhaps still produce, industry statistics in which every aggregation of the same sector, say manufacturing, would have a different total value of output, depending on the level of aggregation of the detailed industries.

18.2 Canadian practice

The CSNA compiles annual production accounts for industries through the input-output tables (also called supply and use tables) at both current and constant prices, and the monthly GDP by industry only at constant prices. The annual input-output tables provide benchmarks for the monthly GDP by industry calculations. The value of output is at modified basic prices for the annual series and the value of value added is at basic prices for both annual and the monthly industry series. Note that output by industry is not calculated for the monthly GDP by industry statistics. Inter-establishment deliveries within the same enterprise are counted as output both in Canada and the United States. The number of industries at the most detailed worksheet level was 204 for 1961-1980, 243 for 1981-96 whereas from 1997 and onwards, the number has increased to 300.

In the input-output tables, three broad aggregates are specified: industries based on business sector establishments, industries in the NPISHs sector, and sub-sectors (not industries) belonging to the

Statistics Canada - 446 - Collected Articles of Kishori Lal government sector. The business sector consists of more than 250 industries but all establishments belonging to either of the two non-business sectors - NPISHs and the government- are excluded as they appear under their own sectors. The NPISHs sector has five separate categories: religious organisations, welfare organisations, sports and recreational clubs, educational institutions, and other organisations. Excepting other organisations, the other four categories in NPISHs follow the industry classification. This category is not really an aggregation of institutional units but of many establishment-based industries. The government sector has eight sub-sectors: hospitals, residential care facilities, university education, other education services, defence services, other municipal government, other provincial and territorial governments, and other federal government. Thus, the CSNA presentation of industries in the production account is different from the industrial groupings published by the Divisions conducting industry surveys at Statistics Canada. Each industry in the business sector or NPISHs and each sub-sector in the government sector contains, in detail, the output of goods and services, intermediate consumption of goods and services, and most of the elements of value added.

Though detailed outputs and inputs of goods and services in the government sector are not classified by industry, value added and its components are. This permits us to produce GDP by industry where each industry includes all establishments, business and non-business; however, the corresponding gross output and their intermediate inputs of goods and services by industry are not articulated by industry in the government sector. Similarly, the labour compensation of the establishments forming part of >other organisations= in the NPISHs sector is allocated to industries for GDP by industry estimates, but the other operating expenditures are kept together primarily because the establishments are small and numerous and their data are not of the highest quality. However, given that labour compensation forms a very significant part of total expenditures in the category other organisations, other operating expenditures might as well be prorated on the basis of labour costs.

18.3 USA practice

The production account by industry in the USA is compiled through two programs: Benchmark and annual IO accounts, and GDP by Industry. Benchmark IO tables for approximately 500 industries are prepared every 5 years, based on economic census data that cover most industries. Annual IO tables for the other four years are prepared for approximately 85 industries, (now with effect from 1997 benchmark tables 134 industries) based on benchmark relationships extrapolated using less comprehensive survey data. Detailed estimates of the intermediate consumption of goods and services, as well as value added, required by each industry (except general government, government enterprises) for the production of its output are prepared. Value added is shown for three componentsBcompensation of employees, indirect business tax and nontax liabilities, and other value addedBfor benchmark tables only. The term non-tax liability is not used in the 1993 SNA or in the national accounts of other countries. In the NIPA accounts, nontax liability includes other business liabilities to general government such as regulatory and inspection fees, special assessment , fines and forfeitures, rents and royalties, and donations. In Canada, these items are not shown separately and are part of other operating surplus by industry.

All Federal Government enterprises are listed together under three headings, rather than classified by industry: Postal service, federal electric utilities, and other federal government enterprises. Similarly,

Statistics Canada - 447 - Collected Articles of Kishori Lal all State and local government enterprises are listed together under three headings, rather than classified by industry: State and local government passenger transit, State and local government electric utilities, and Other State and local government enterprises. There is also one specific General government industry. Previously, the value of output of general government industries was limited to value added only and zero was reported as intermediate consumption. In the 1997 benchmark IO tables, the general government industry includes the intermediate inputs for force-account (called own-account in Canada) construction and own-account software in the standard tables. However, other intermediate consumption for conducting the general government activities are not covered in the Use tables but are shown in the final consumption expenditures. It should be noted that BEA plans to revise the presentation of government production to show intermediate inputs and production of services as part of the upcoming NIPA comprehensive revision in 2003. This change in the NIPA=s will be reflected in future IO and GDP by Industry accounts.

It may be noted that since June 2000, the GDP by Industry accounts provide an annual time series of nominal and real production accounts by industry going back to 1987 for all industries and back to 1977 for most industries. It should be further noted that there is a significant difference in methodology between the IO accounts and GDP by industry accounts and that only the latter program provides real estimates for industries .

18.4 Concluding remarks

Both Canada and the USA have extensive industry statistics programs, and their respective input- output tables play a crucial role. Both countries have classified their industries by sector but the sector boundaries are different as noted earlier in Section B, production account for institutional sectors.

In the United States, all establishments belonging to the business sector and the NPISHs sector are classified by industry but government business enterprises are not allocated by industry. All government business enterprises are aggregated in two groups, Federal Government enterprises and State and local government enterprises, and their individual establishments do not form part of their own industrial group. In Canada, government enterprises form part of the business sector and are classified with similar producing units. For example, the Federal, State and local government electric utilities are allocated to the same industry as the private sector industries in Canada but not in the USA. In the USA, government universities form part of General government industry and other universities are allocated to NPISHs but all universities are allocated to the government sector in Canada. In neither country, are the establishments in the government sector allocated to industries. In Canada, however, the value added for the two non-business sectors is classified by industry and it then is added to the value added of industries forming the business sector, thus producing comprehensive GDP by industry estimates.

The value of output for general government in the USA includes value added and (with effect from the 1997 benchmark IO accounts) force-account construction and own-account software and their intermediate inputs but the intermediate inputs for other general government activities are shown directly as final expenditures. In Canada and other countries, the value of such output includes both value added and corresponding intermediate consumption, thus the existing presentation in the USA

Statistics Canada - 448 - Collected Articles of Kishori Lal is different from most other countries. It should be noted, however, that BEA plans to revise the presentation of government production to show intermediate inputs and production of services as part of the upcoming NIPA comprehensive revision in 2003. This change in the NIPA=s will be reflected in future IO and GDP by Industry accounts. Thus in the GDP expenditures approach of the United States, government final expenditures have all the detailed purchases of goods and services plus the value of government services, which are equivalent to their value added. On the other hand, in Canada and other countries, government final expenditures have one item only, which is equal to the value of output of government services which are equal to their intermediate consumption plus value added. Total value added and total final expenditures in both countries= presentation are identical, but the details are different. This may be confusing for users.

In Canada, it will be useful (and it should not be very costly) to articulate by industry both the outputs and intermediate inputs in the government sector and other organisations in the NPISHs sector. Industries from the three sectors can then be re-aggregated following the same industry classification as in the Monthly GDP by industry program. This will permit us to have a full production account by industry for all three sectors as well as for the total economy. This development, when completed, will enhance the use of our industry statistics both within Statistics Canada and by outside analysts.

19. Long time series of industry Statistics

In both Canada and the United States, there have been many vintages of Standard Industry Classification (SIC) during the last 50 years and the newest one, the North American Industry Classification System 1997 ( NAICS), is based on quite different organising principles compared with the earlier SICs. How the two statistical organisations responsible for producing industry statistics should maintain a long time series of these statistics and at the same time implement NAICS is an important issue which needs to be carefully deliberated. Implementation of revised industrial classifications inevitably produces statistical breaks, yet the demand from users for continuous time series remains unabated.

19.1 Canadian practice

The CSNA detailed industry statistics from 1961 onwards are classified on the basis of three Standard Industrial Classification (SIC) Manuals, all issued by Statistics Canada or its predecessor the Dominion Bureau of Statistics. The first one was issued in December 1960, called AStandard Industrial Classification Manual@, the second one was called AStandard Industrial Classification Manual, Revised 1970", and the third one was called AStandard Industrial Classification 1980". All three followed the same principles of classification, except that the latter ones had more details as new industries emerged and were comparable. There was not much difficulty in reporting the Canadian industrial structure over time. In both the IO tables and the monthly GDP by industry measures of the CSNA, time series of industrial statistics are produced for some 150 industries (link level), consistently defined for the entire period, 1961 to 1997. At the worksheet level, there is even more detail available: statistics are produced for some 200 industries for the period 1961-1980, and for some 240 industries for the period 1981-1997.

Statistics Canada - 449 - Collected Articles of Kishori Lal

NAICS is the joint product of three countries, Canada, Mexico and the United States. It is implemented starting with the reference year 1997 in Canada and the United States, and 1998 in Mexico. NAICS differs substantially from the SICs because it is based on a single organising principle, which is in contrast to the SICs which have no such single principle. NAICS is erected on a production-oriented or supply based conceptual framework, in which producing units that use identical or similar production processes are grouped together. Some SIC based industries were grouped on production-oriented principles while other industries were based on demand based principles. This difference in orientation created discontinuity in time-series comparability between SIC-based estimates and NAICS-based estimates. Even at major sector levels, such as retail trade and wholesale trade, the differences in values between the two classifications are substantial. This created a dilemma both for compilers and users of detailed industry statistics.

It is very expensive, even if it were possible, to re-code the establishments on a NAICS basis, in the back period for all or even most sectors of the economy. Then, how far back can or should one go for this exercise? For example, in manufacturing, establishments have been re-coded to NAICS for 1992 onwards, but not much has been done in other sectors. Even in manufacturing, what should one do for the period before 1992? Given time and resource constraints, one must develop other options to serve users who require long time series of data.

With the NAICS implementation, starting with the reference year 1997, the link level detail for the entire period from 1961 onwards could only be produced at the 119 industry detail level. The 119 industry link level follows the hierarchical structure of NAICS. SIC based worksheet level industries have been allocated to the new link level without regard to their own hierarchical structures. Very detailed reconciliations between the two series have been prepared and notes drafted. This information is available to enable users to reconcile the estimates for each of the119 industries based on the two classifications for the year 1997. We have aggregated approximately 700 NAICS based industries (5 and 6 digit codes) into 300 worksheet level detail for CSNA for the years 1997 and onwards. It is this detail which will also be concorded into the international ISIC3 for our presentation of national accounts, industry and productivity data to both OECD and other international bodies.

There is no easy solution to resolve this problem of statistical breaks, yet the demand from users for continuous time series remains unabated. A joint working group of compilers and major users has been established to devise some methodological technique (based on judgement and statistics) which would satisfy both users and compilers. Our methodology has involved, fundamentally, an examination of the commodity composition of production of each link level industry. We have restricted our examination to the commodity composition because there is a continuous time series for commodities for the entire period. Further, we have selected only those industries for adjustments which have some significant statistical difference (mostly more than 2%) of their commodity composition between the SIC and NAICS basis. We have removed a commodity, part or whole, from one industry to be consistent with NAICS commodity production and the same value is added to another industry again making it as close as possible to the NAICS definition.

This cut and paste method, both for output and their inputs (always equal in value) has kept the balances intact and also kept our calculations in control. The cut and paste method worked as

Statistics Canada - 450 - Collected Articles of Kishori Lal follows: for each paired industry, we determined which additional commodity was produced and how much was added for the NAICS basis from the SIC basis. We also determined from the USE matrix the input structure (both intermediate consumption and value added components in full detail) of this commodity production. It is this detail which was cut and pasted.

In terms of operations, what this really involved is to determine which of the 243 worksheet level industries, in the 1981-97 period, would require further splits such that each split could be allocated to one of the 119 link level NAICS industries, the one which has similar or close to similar (difference no more than 2%) commodity production. Our examination led us to the conclusion that 204 of the 243 worksheet level industries could be allocated to one of the 119 link level industries for the entire period 1981 to date and the other 39 industries (38 in the business sector and one in the government sector) should be split such that each split is allocable to one of the 119 link level industries. The three digit code listed with each of the 39 industries below is the SIC-based IO industry code:

002. Field crop farms 003. Service industries incidental to agriculture 004. Fishing, and trapping industries 006. Forestry services industry 008. Other metal mines 013. Other non-metal mines (except coal) 016. Quarry and sand pit industries 172. Air transport and related service industries 173. Railway transport and related service industries 174. Water transport and related service industries 175. Truck transport industries 178. Taxicab and other transport industries 179. School and other bus operations industries 184. Other storage and warehousing industries 185. Radio and television broadcasting industries 188. Postal and courier service industries 191. Water systems and other utility industries n.e.c. 192. Wholesale trade industries 193. Retail trade industries 196. Credit unions and caisses populaires 197. Other financial intermediary industries 198. Real estate operator industries 199. Insurance and real estate agent industries 202. Computer and related services 204. Architectural , eng., & other scientific & tech services 205. Advertising services 206. Misc. business service industries 207. Educational service industries 208. Other health and social service industries 209. Health practitioners and medical laboratories ind.

Statistics Canada - 451 - Collected Articles of Kishori Lal

210. Accommodation service industries 212. Motion picture, audio and video prod. and distribution 214. Other amusement and recreational service ind. 216. Laundries and cleaners 217. Other personal service industries 219. Mach. & equipment., auto & truck rent., & leasing services 222. Other services n.e.c. 223. Other repair services and services to buildings & dwellings 239. Non-business- Government, Other educational services

These changes were fully documented with notes and comments and are available to users. What we really provide is the following: SIC based Make and Use matrices at 119 link level industries; a new make matrix of the 39 selected industries, values with negative signs and all their splits (which are consistent with the NAICS classification) as positive values such that this matrix adds to zero in total and in detail; a similar use matrix for the same 39 selected industries, and again this matrix adds to zero in total and in detail. Summing these two make matrices gives a new NAICS based make matrix; similarly summing the two use matrices gives a new NAICS based use matrix.

19.2 USA practice

The statistical system in the USA provides different industry time-series prepared by several different agencies. Mostly, these industry statistics are classified on an establishment basis, but some are classified on a company or an enterprise basis. Within BEA, different industry time series are prepared from source data classified on both an establishment and a company basis. BEA=s GDP by Industry program provides annual current-dollar estimates starting with 1947, and annual real estimates starting with 1977. Real estimates are not provided before 1977 due to a lack of source data needed for the double-deflation method. Estimates from 1947-87 are classified according to the 1972 Standard Industrial Classification (SIC) system (1977 revision) and estimates from 1987 forward are classified according to the 1987 SIC system. The estimates before 1987 were not converted to the 1987 SIC due to a lack of adequate source data. Instead, estimates for 1987 are shown on the basis of both the 1972 and the 1987 SIC.

The two classification systems are very similar in structure and content, despite the lack of a single organizing principle. The 1987 SIC revision resulted in no net change in the number of detailed four- digit industries; however, a minor restructuring resulted in 34 new industries that were balanced by 34 deleted industries. The most significant change in the 1987 SIC system was the introduction of a new major group for Engineering, Accounting, Research, Management, and Related Services (major group 87). This change was balanced by a significant reduction in the size of Business Services (major group 73) and Miscellaneous Services (major group 89). Other significant changes included the transfer of certain types of equipment from Electrical and Electronic Machinery (major group 36) to Instruments and Related Products (major group 38) and to Industrial Machinery and Equipment (major group 35), and the movement of savings and loan associations and credit unions from Credit Agencies Other Than Banks (major group 61) to Depository Institutions (major group 60).

Statistics Canada - 452 - Collected Articles of Kishori Lal

At their most detailed level, the GDP by Industry estimates are published for 66 industries at approximately the two digit-SIC level, with additional detail in some industry divisions and less detail in others. More aggregated results than two-digit detail are provided for farms, agricultural services, forestry, and fishing, construction, wholesale trade, retail trade, and other services. More detailed industry estimates are provided for transportation equipment, communications, and real estate. BEA=s decision to not implement the 1987 SIC revision before 1987 caused some difficulty for users of the industry data who were primarily concerned with long-term economic trends for detailed industries. BEA helped to bridge this gap not only by providing estimates according to both SIC systems for 1987, but also by providing estimates for combinations of industries with consistent definitions over time. For example, both current-dollar GDP by Industry and chain-type quantity indexes were provided for the combination of depository and nondepository institutions back to 1947 for current-dollars and back to 1977 for quantity indexes. BEA has not provided estimates of chained 1996 dollars prior to 1987 due to the non-additivity of chained dollars.

Conversion of industry data to the new North American Industry Classification System (NAICS) is in the early stages in the U.S. statistical system. Data from the 1997 economic census were classified according to NAICS 97, and BEA=s 1997 benchmark Input-Output accounts were released in December 2002 on a NAICS 97 basis. Several industry-based estimates in the NIPA, including sales and inventories for manufacturing and trade industries, are now provided on a NAICS basis. Most of the monthly and annual surveys conducted by the Bureau of the Census are now on a NAICS basis. BEA is scheduled to complete the conversion of its industry-based estimates to NAICS as part of the upcoming comprehensive revision in 2003. By that time, most if not all of the source data used for the NIPA and for GDP by Industry will be classified according to NAICS, with the producer price index prepared by the Bureau of Labor Statistics (BLS) the one major exception. See more details in regard to the implementation of NAICS in the USA in an article by John Kort, The North American Industry Classification System in BEA=s Economic Accounts, Survey of Current Business May 2001.

The current methodology for GDP by Industry imposes some fairly serious limitations on the ability to carry NAICS conversion back more than several years. Current-dollar estimates are based primarily on BLS administrative data for wages and salaries and on Internal Revenue Service (IRS) tax return data for corporate profits and other components of property-type income. Industry distributions for these income components are obtained from the NIPA and, with some adjustments, are used directly for GDP by Industry. Estimates of real GDP by Industry using the double-deflation method rely primarily on estimates of nominal gross output and price indexes for gross output and intermediate inputs.

Lack of consistent historical source data limits the feasibility of providing time-series data for GDP by Industry, based on the current estimating methodology, for the years prior to 2000. Estimates for the year 2000 will be provided on both a NAICS and SIC basis; however, because of the major differences between the two classification systems, maintaining consistency is difficult even at very high levels of aggregation. The NIPA components of gross domestic income will be available only from 2000 forward. Gross output by industry and the commodity composition of intermediate inputs will be available on a NAICS basis, starting in 1997, from the input-output accounts. It would be extremely difficult and costly to replicate existing procedures on a NAICS classification basis, primarily due to the lack of NAICS-based source data prior to 1997. BEA is exploring its options for

Statistics Canada - 453 - Collected Articles of Kishori Lal the conversion to NAICS for earlier years, however, the costs associated with any such conversion that is not highly mechanized is estimated as very significant, given BEA's current program commitments.

19.3 Concluding remarks

In the CSNA, a project was established to produce a continuous time series at the 119 link level industry detail level, approximately consistent with NAICS, back to 1961, by the end of March 2003. The BEA is considering making conversion to NAICS for earlier years but has not yet reached decisions on the level of detail or time that would be covered. Our cut and paste methodology primarily relates to the changes in commodity output in industries following different vintages of classifications. We would like to share our approach with our colleagues in the USA and in other OECD countries and hope that they find it cost-effective. In our judgement, our users require long time series of industry statistics, defined, as closely as feasible, on a consistent basis. If statistical organisations do not provide these series, most major users will develop their own independent and most likely different estimates. This will be very confusing.

It is important to remain fully transparent, and document every paired change (note that a change from one industry may go to more than one industry) with notes and comments and make such information available to all users for all those industries selected for splitting to yield a consistent series. This will permit users to have a full understanding for all the changes and they may select, if they so desire, to skip some paired information from their analysis and still have all the industries and commodities remain balanced.

20. Statistical discrepancy

Macro series in the national accounts are built from a myriad of sources. Expenditure based GDP is built from expenditure surveys of households, sales of retail establishments, records of governments on their detailed expenditures, investment surveys of enterprises, exports, imports, etc. Income based GDP is built from surveys of labour compensation, profits, net income, capital consumption, etc. GDP is also calculated by summing the value added of all the producing units and this information is built from industry surveys on production, their intermediate expenditures and value added. These three approaches to estimate GDP must, in principle, produce the same results but in practice the results are not identical, hence a statistical discrepancy appears amongst these three estimates.

20.1 1993 SNA

The 1993 SNA is a conceptual document and in it all three approaches produce the same results, thus there is no provision for a statistical discrepancy, hence no guideline on how to handle it in the real statistical world. Its handling is left to the conventions, judgements, and imagination of the national accountants of each country. Some countries add all the three different levels of GDP from the three approaches, divide by three and that becomes the official GDP. Other countries only use two approaches, GDP expenditure based and GDP income based; they may add the two different levels of GDP and divide the total by two and that becomes the official GDP. Some other countries may

Statistics Canada - 454 - Collected Articles of Kishori Lal accept one as more correct than the other, and then add the statistical discrepancy to the less correct side to make it equal to the one assumed to be more correct.

20.2 Canadian practice

In Canada, quarterly macro GDP is produced using both the expenditure approach and the income approach. The statistical discrepancy between the two series is divided by 2, one half is added to one side and the other half with reverse sign is added to the other side such that the two sides become equal. It is this equalled GDP which is the official level of GDP in Canada. At the annual frequency, IO tables are produced and they become benchmarks for the macro GDP estimates. All the elements in the macro series except the operating surplus on the income side and value of physical change in inventories (VPC) on the expenditure side are identical to those derived from the IO tables. In the macro series, the value of the statistical discrepancy plus VPC is made equal to the value of VPC in the IO tables; similarly, the value of the statistical discrepancy plus the operating surplus is made equal to the operating surplus in the IO tables. There is no provision for showing a statistical discrepancy in the IO tables.

As the accounts are struck, initially the statistical discrepancy may be large and it prompts us to reexamine our estimates, our methodology, our assumptions, our judgements etc. We go back to the records and make corrections where warranted. Once we have exhausted our probes, we are still left with a discrepancy, most likely smaller than the earlier one. At this point, we simply assume that the residual discrepancy is unbiased, one half is added on one side and the other half with reverse sign on the other side of our accounts, thus making the two sides identical in value.

20.3 USA practice

In the NIPA accounts of the United States, the statistical discrepancy between the expenditure based and the income based GDP is calculated and the entire amount is added to the income side as a separate item, thus making the two sides identical in value.

20.4 Concluding remarks

Canada and the USA follow different conventions to handle the statistical discrepancy between the two sides of the accounts which, in principle, must be identical. No matter what the national accountants do to handle the statistical discrepancy, it becomes problematic in times when the economy is growing or shrinking at very low rates, that is within the narrow range of zero. In such cases, the reported GDP growth rates might easily be due to how the discrepancy is handled rather than how the real economy is working. If there is no bias, the different handling of the statistical discrepancy has not much effect on the long term evolution of the economy but in the short term, its handling does affect the published growth rates and probably the economic policy issues emerging from such rates.

Statistics Canada - 455 - Collected Articles of Kishori Lal

Section D: 1993 SNA production boundary, ready to expand

In this section, we will select, for elaboration, some of the important guidelines in the 1993 SNA which we find quite problematic for implementation. In some areas, we have already expanded the 1993 SNA boundary and in others, we would like the existing boundary to be expanded.

21. Valuation of financial intermediation services indirectly measured (FISIM)

Banks and other financial institutions provide a variety of services. Those specifically charged for include currency exchange, handling of cheques etc; and the corresponding revenues form part of the institutions' output. An additional, and very significant, part of their income comes from charging higher interest rates to borrowers and paying lower rates to depositors than they would need to if they charged explicitly for all their services. This "hidden" charge (known as imputed banking service in the 1968 UN SNA) is called financial intermediation services indirectly measured (FISIM) in the 1993 SNA.

21.1 1993 SNA

The 1993 SNA defines FISIM as follows: "The total value of FISIM is measured in the System as the total property income receivable by financial intermediaries minus their total interest payable, excluding the value of any property income receivable from the investment of their own funds, as such income does not arise from financial intermediation. Whenever, the production of output is recorded in the System the use of that must be explicitly accounted for elsewhere in the System. Hence, FISIM must be recorded as being disposed of in one or more of the following ways- as intermediate consumption by enterprises, as final consumption by households or as exports to non- residents" (paragraph 6.125). It may be noted that the 1993 SNA does not include holding gains or losses in the valuation of output of any enterprise, financial or non-financial. In the valuation of output of certain financial institutions such as Insurance, it has specifically noted this restriction (see paragraph 6.138) while in other cases such as FISIM, it is not noted but always assumed. To further confirm this assumption, one may refer to the 1993 SNA definition of property income (paragraph 7.89) which provides no place for holding gains. This restriction is now challenged and we will come back to this later.

The debate on how to measure the value of FISIM and how to allocate it to all the users, is, at least, as old as the system of national accounts. In social and religious history, this debate goes back to the Middle Ages when charging interest was considered usurious; hence a sin. Even today, charging interest is not permitted in some countries. Staying with the limited national accounts history, an imputed banking output was recommended in the 1968 SNA but its allocation to users was not recommended. An artificial industry was created with zero output but it would use as an intermediate expenditure all the imputed banking services, thus generating an equivalent negative value added. This recommended arrangement did produce the value added for banks, quite similar in value to what they would report in their financial statements; however, there was no allocation to intermediate consumption of any industries using banking services, thus their value added remained exaggerated, counterbalanced by an identical negative value added in the above-noted artificial industry. Further, there was no allocation for use in final consumption or exports, thus use of banking

Statistics Canada - 456 - Collected Articles of Kishori Lal services was unrecorded, and total expenditures on GDP remained under-estimated. In the discussions leading to the development of the 1993 SNA, the 1968 SNA recommendation was considered deficient by many national accountants, particularly those of both the USA and Canada. Thus it was recommended (see the 1993 SNA paragraph 6.125 as quoted above) that the output of FISIM be allocated to all the users.

At the time of the presentation of the draft of the 1993 SNA to the UN Statistical Commission in March 1993 for its approval, some representatives again raised the issue of the difficulties of allocating FISIM to users. Not to further delay its approval, the following flexibility was added in the 1993 SNA: "In principle, the total output should, therefore, be allocated among the various recipients or users of the services for which no explicit charges are made. In practice, however, it may be difficult to find a method of allocating the total output among different users in a way which is conceptually satisfactory from an economic viewpoint and for which the requisite data are also available. Some flexibility has therefore to be accepted in the way in which the output is allocated. Some countries may prefer to continue to use the convention proposed in the 1968 version of the SNA whereby the whole of the output is recorded as the intermediate consumption of a nominal industry@ (paragraph 6.126). This flexibility has created problems for our submission of national accounts data to the OECD, as the European Union countries still use the 1968 SNA guideline, whereas the United States, Canada, Australia and New Zealand allocate FISIM to users.

21.2 Canadian practice

As noted above, the 1993 SNA recommends that the total value of FISIM be measured as total property income receivable by financial intermediaries minus their total interest payable, excluding the value of property income receivable from investment of their own funds. In the CSNA, we have not accepted this restriction of own funds in the calculation of FISIM both conceptually and methodologically (see issue #22, FISIM on own funds). In the case of financial institutions, the availability of own funds for loans is not very significant. In Canada, as in most other countries, own funds are used for fixed assets and other investments and only what is left is available for loans, etc. To put this in perspective, let us look at the following balance sheet information for banks for the year 1993 in Canada: Total assets were $548 billion, of which securities other than shares, and loans were $527 billion, and fixed assets and investments were $21 billion; total liabilities for intermediated funds were $ 507 billion and own funds were $41 billion, with total liabilities equalling $548 billion. Thus, own funds contributed only $20 billion ($41 billion less $21 billion) or 4% for total loans, the other 96% came from depositors. In our calculation of property income receivable from the investment of the available own funds ($20 billion in our example), we developed a methodology which, in our judgement, clarifies the definition of FISIM in the SNA such that the indirect measure of service output allocated to depositors is still made equal to zero (as there are no depositors) but the service output allocated to borrowers remains positive. Thus the Canadian definition of FISIM adds the bold bracketed insertion in the 1993 SNA definition and reads as follows: AThe total value of FISIM is measured in the System as the total property income receivable by financial intermediaries minus their total interest payable, excluding the value (calculated at the pure rate of interest) of any property income, receivable from the investment of their own funds.@ The meaning of the pure rate of interest used in our definition is not much different from the 1993 SNA definition of reference rate: A The reference rate to be used represents the pure

Statistics Canada - 457 - Collected Articles of Kishori Lal cost of borrowing funds -that is, a rate from which the risk premium has been eliminated to the greatest extent possible and which does not include any intermediate services@ (paragraph 6.128). We have operationally used the middle of the lending and borrowing rate for similar maturities, as a proxy for the pure rate of interest for our calculations.

In Canada, we allocate FISIM to all users, borrowers and depositors, and enterprises and final demand expenditures (persons, government and the rest of the world sector), all based on their respective assets and liabilities (distributed by detailed financial instruments). The total amount of loans should be similar in value to the total amount of deposits except when there are own funds available which are quite small. Given this, the distribution of total FISIM between borrowers and depositors should also be of similar value. In 1998, the value of FISIM output of deposit accepting financial intermediaries (banks and credit unions) was $23 billion and consumer loan companies (working primarily with own funds) was $3 billion, both totalling to $26 billion. About $13 billion of FISIM was allocated to enterprises and a similar amount to final expenditures, mostly to household consumption. In 1998, the distribution happened to be fifty-fifty, though it is a bit different for each set of financial enterprises; for example, for banks, FISIM allocated to enterprises was 54% and to final expenditures 46%. This distribution has been stable over the years. The money borrowed for house mortgages is allocated to real estate or own account housing enterprises, not to final demand expenditures, as this activity is deemed part of the business sector. In our submission to the OECD SNA data, we remove the FISIM allocation to final expenditures in order to remain consistent with the SNA series provided by other OECD member countries who have not yet allocated FISIM to users, including final consumption. Thus, our GDP total in the OECD publication on national accounts, say for 1998, is lower by $13 billion, or about 1.4%, than what we publish in Canada; similar deductions are made for all years. It is to the credit of the CSNA that it has always allocated this output to users, including final users, even when the 1968 SNA recommended otherwise

21.3 USA practice

In the United States, the calculation of the value of FISIM output is similar to its calculation in Canada. It is allocated to users, both industries and final demand expenditures (persons, governments and the rest of the world sector). However, the methodology of allocation in the USA is substantially different from the one in Canada. In the BEA accounts, all FISIM services are currently accrued to depositors only, whereas in Canada they accrue to both depositors and borrowers, almost fifty-fifty. The details are available from NIPA Table 8.21, Imputations in the National Income and Product Accounts. For example, in 2000, FISIM output was US$373.3 billion and was allocated as follows: Persons $265.4 billion; government $9.7 billion; business $77.1 billion and the rest of the world $21.2 billion. In that year, total GDP was $9810 billion. This allocation is different from the one in Canada: about 50% of FISIM is allocated to business compared to only 20% in the United States. This is almost entirely due to US allocation formula which assumes services to borrowers equal to zero. Most significant borrowing is typically done by the business sector (including home mortgages) and most significant depositing is done by households. The effect on FISIM allocation on total GDP is equal to its allocation only in the final demand which was $ 296.2 billion or equal to 3% , a ratio significantly higher than the one in Canada of 1.3%.

Statistics Canada - 458 - Collected Articles of Kishori Lal

In the submission of national accounts data to the OECD, the USA makes no adjustment to its published GDP. Thus, compared with both those countries which do not allocate FISIM to users and those which allocate but make adjustments for the OECD accounts, the GDP of the USA was higher, for example by US$296 billion in 2000 or by about 3% and a similar ratio applies to all other years too.

21.4 Concluding remarks

Both Canada and the USA allocate FISIM to users and have been doing so long before the 1993 SNA. Total FISIM output, as a ratio of GDP, is very similar, about 3%, in both countries. Both Canada and the USA regret the last minute insertion of flexibility in the 1993 SNA (paragraph 6.126) regarding the allocation of FISIM that countries may not allocate FISIM to users, industries and final demand users. As of now, all EU countries follow the 1968 SNA flexibility recommendation in this regard and do not allocate FISIM to users, both industries and final demand. The result of this flexibility is that the SNA/ESA questionnaire requires the member countries to remove any FISIM allocation to final expenditures, thus reducing the value of GDP, even if the country has such an allocation in its own official accounts. GDP provided to the OECD by Canada is lower than what it publishes, by the value of FISIM allocation to final demand. This unsatisfactory situation may be resolved in the next few years when, as it looks likely, EU member countries may start allocating FISIM to all users.

The BEA is re-examining the present allocation formula that assumes zero allocation to borrowers. It is studying the use of a reference rate to allocate some services to borrowers, and may decide to change the allocation in the upcoming comprehensive revision in 2003. When this is done, it will remove one more difference between the methodology of allocation and the substantial value difference by sectors between the two countries.

As noted above, four countries - Australia, Canada, New Zealand and the United States- allocate FISIM to all users (industries and final demand). In their submissions to the OECD, Australia and the USA do not make any adjustment on this regard in their published GDP, whereas Canada and New Zealand do. This needs to be looked at by the OECD, in the context of USA-Canada comparison.

In the measurement of output of FISIM, the 1993 SNA recommends removing any capital gains or losses from the property income received by the banks. However, financial institutions receive income from investing their financial property in bonds, loans, shares, securities and equity and it is the collective income from all these sources which they need to pay interest to depositors and the operating costs to run their institutions. It is the collective income which determines the rate of interest which they pay to their depositors. As national accountants, we cannot accept the deposit rate without simultaneously accepting all the elements of the income of financial institutions which they use to set such rates. Some, and increasingly more and more, financial intermediaries earn a substantial part of their income through capital gains associated with buying and selling securities and other financial assets. Following the 1993 guideline, their output is seriously understated Thus, the 1993 recommendation to exclude capital gains made by the financial institutions needs to be reconsidered, as it is producing results which are counter-intuitive for some and increasingly more and more financial intermediaries.

Statistics Canada - 459 - Collected Articles of Kishori Lal

A further anomaly may arise when financial intermediaries manage asset portfolios on behalf of customers or on own account. When banks undertake management of asset portfolios for customers, they receive commission income, hence production occurs. However, when done on own account, no production is recorded under the 1993 SNA standards as the income is from capital gains and there is no intermediation involved. The same activity produces two different numbers, depending upon whose behalf the activity is undertaken and this is problematic.

Both these problems and some others are being examined by an OECD Task Force on the measurement of the banking output. Both the USA and Australia have strongly argued to add capital gains in the valuation of FISIM. See their papers, Beyond 1993: The System of National Accounts and the New Economy by Rob Edward and others from the Australian Bureau of Statistics and the System of National Accounts for the New Economy: What Should Change by Brent Moulton from the BEA, submitted at the International Association for Official Statistics, London August 2002. The positions taken by both Australia and the USA are very well articulated and we fully support them. Many of us now urge that the 1993 SNA restriction on capital gains for calculating the value of output of financial institutions be re-examined.

22. FISIM on Own Funds

In the developing countries the banking system is usually not as well-established as in the developed countries, hence the borrowers seek loans from money lenders or some household finance companies which have only their own funds to lend. Even in developed countries, there are financial enterprises, typically in the consumer finance area, which have their own funds (they are not allowed to accept deposits by law) and they lend such funds to households for mortgages and to meet other personal financial needs. They are not financial intermediaries, as they do not intermediate between the depositors and the borrowers, there being no depositors. They exist in the market, have intermediate expenditures, may hire employees, make profits and meet all expenses by lending money. How should their output be valued?

22.1 1993 SNA

The 1993 SNA recommends that property income receivable from investment of own funds be excluded from the calculation of FISIM, as such income does not arise from financial intermediation (see paragraph 6.125). The 1993 SNA has no provision for their output but expenses cannot be ignored in any balanced system: hence there is a dilemma. The significance of this industry is different in countries, high in developing countries and far less in developed countries with well functioning banking sectors.

22.2 Canadian practice

We have noted above in our discussion of FISIM the modification that we have inserted in the 1993 SNA definition of FISIM. Staying with the same definition, the output of consumer finance companies, which have only their own funds, is calculated as follows: total property income receivable from borrowers less the imputed cost of property income of own funds (equal to own funds multiplied by the pure rate of interest). The entire output is allocated to borrowers, there being

Statistics Canada - 460 - Collected Articles of Kishori Lal no depositors. The value of this output in 1998 was about $3 billion. We have classified this output as FISIM purely for convenience, and it represented about 11% of total FISIM of $26 billion as noted above for 1998.

In a formal sense, we have deliberately gone beyond the 1993 SNA recommendation and expanded the 1993 SNA definition in a fashion that avoids the dilemma noted above, which recognises only expenses but no output. When a borrower pays interest to a bank, it has two parts: the pure interest and the service fee. Similarly when the same borrower borrows from a finance loan company and pays interest, this, too, has two parts: the pure interest and the service fee. Borrowers pay fees, hidden in higher interest rate, no matter where they go. We believe that borrowers of these funds receive a service from institutions using their own funds. We call this service FISIM but we could easily have called it FSIM, financial services indirectly measured, rather than financial intermediation services indirectly measured. We have added FSIM and FISIM together.

22.3 USA practice

As noted above in our discussion on FISIM, the BEA allocates the entire FISIM output to depositors. As there are no depositors in the case of finance companies using only their own funds, there is simply no FISIM in the NIPA on this account. The BEA is considering allocating FISIM also to borrowers in the next comprehensive revision of the national accounts. Once that is done, BEA will also have an output of FISIM for all those companies using own funds to lend to others. 22.4 Concluding remarks

Canada estimates FISIM for all finance companies who lend money to others but use only their own funds and their entire output is allocated to borrowers. The borrowers pay the implicit service fee (hidden in the interest rate) whether the funds they receive come from depositors or from the owners themselves. When the BEA starts allocating FISIM to borrowers, it may pursue a methodology similar to the one in place in the CSNA for developing FISIM on own funds.

The recommendation in the 1993 SNA, which does not recognise the output of finance companies but must recognise their expenditures has created a dilemma. To escape this dilemma, the 1993 SNA may choose the modification inserted by Canada, which is repeated below: AThe total value of FISIM is measured in the System as the total property income receivable by financial intermediaries minus their total interest payable, excluding the value (calculated at the pure rate of interest) of any property income, receivable from the investment of their own funds.@ Further, the 1993 SNA may call the new calculation FSIM, not FISIM. We are looking forward to the deliberations of the OECD Task Force on Banking Output, which must examine both the issue of holding gains and the recording of income from own funds.

23. Output of Central Banks

Financial intermediation is only one, and not necessarily the most significant one, of the many functions that central monetary authorities perform in most countries. Their other functions include formulating and implementing monetary policy, issuing and replacing bank notes, managing the public debt, etc. Delineating the cost structure of any of the activities, not necessarily the most

Statistics Canada - 461 - Collected Articles of Kishori Lal significant one, is always problematic. What is the most appropriate way to value the output of central banks?

23.1 1993 SNA

The 1993 SNA states: AThe services of financial intermediation provided by central banks should be measured in the same way as all other financial intermediaries@ (paragraph 6.132).

Canada requested the Inter-Secretariat Working Group on National Accounts (ISWGNA) in 1995 to re-examine the 1993 SNA recommendation on this subject, as we were finding it difficult to implement. Our difficulties arose from the fact that a very significant portion of liabilities of a cental bank are bank notes in circulation (more than 90% of the liabilities in the case of the Bank of Canada) and its property income derives from the assets in the form of the treasury bills and bonds provided by the central government to have access to these bank notes. Central banks are not like commercial banks as they perform, in addition to central banking services, many more functions, such as monetary policy, debt management, etc,. ISWGNA deliberated and issued a clarification on the valuation of central bank output. In its January 1996 issue of SNA News and Notes, the following is noted: A ISWGNA also discussed the 1993 SNA method of measuring output of central banks which has caused a number of concerns owing to the large positive or negative numbers for gross output and possibly even a volatility in output....ISWGNA agreed that the SNA treatment should continue to be recommended as the first approach, but, where this approach leads consistently to inappropriate results, output could, as a second best approach, be measured at cost as for other non-market producers. However, under no circumstances can it be construed that the central bank is part of the central government, regardless of how its output is measured@.

23.2 Canadian practice

The central bank in Canada, the Bank of Canada, serves four broad functions: formulate and implement monetary policy; issue and replace bank notes; provide central banking services; and manage public debt. Only activities related to the central banking services can generate FISIM. The bank does hold deposits and make advances as part of its banking activities but its other liabilities and assets are far more significant. For example, as at 31 December 1998, the assets of the Bank of Canada amounted to $35.3 billion and its liabilities of $35.3 billion had the following detail: Bank notes in circulation $32.6 billion and other liabilities of $2.7 billion consisted of $900 million deposits from banks and $1,800 million other liabilities, mostly securities sold under repurchase agreements. The Bank of Canada received revenue from investments amounting to $1,799 million, incurred operating expenses of $194 million, and remitted $1,679 million to the Receiver General for Canada. The Bank also reported its operating expenses by function: monetary policy $41 million; currency $57 million; central banking services $31 million and retail debt service $65 million. The CSNA decided to implement the modified recommendation (as issued in the SNA News and Notes in January 1996) for valuing its output, rather than the original formulation in paragraph 6.132 of the 1993 SNA. Its value in 1998 equalled all its operating costs of $194 million and was allocated entirely to the Federal Government.

Statistics Canada - 462 - Collected Articles of Kishori Lal

How can we assure ourselves and our users that our calculation of the value of output of the Bank of Canada of $194 million in 1998 would not be much different had we used the complicated route of the 1993 SNA definition of FISIM but with our modification? We can convincingly argue that the liability of $32.6 billion for banks notes in circulation is more like own funds (created from printing bank notes); further, its liability owing to banks did not incur much interest liability. We noted above that the assets of $35.3 billion produced revenue of $1,799 million or a rate of interest of 5.1%. Given that the assets of the Bank of Canada overwhelmingly consist of risk-free Government of Canada treasury bills and bonds, it is fair to assume that the service fee portion of the interest charged on government bills and bonds would be no more than the operating cost incurred by the Bank. This assumption provides a straight-forward calculation of the pure rate of interest of 4.55% ( total interest of $1,799 million less $194 million of operating expenses or a total of pure interest of $1,605 million divided by the total assets of $35.3 billion, giving a pure rate of interest of 4.55%). Interest cost on other liabilities of $1,800 million would approximate $92 million, using the same average rate of 5.1% for government borrowing. FISIM would be equal to investment income received of $1,799 million less the pure interest calculated for own funds of $1,483 million less investment income paid on securities of $ 92 million or a total of $224 million. This is higher by $30 million from our cost calculated output of $194 million, the additional income may cover some interest cost on the $900 million liability to banks. The interest rate on this liability would be much lower as there are legal requirements for the commercial banks to deposit some minimum required money with the Central Bank. We have gone through this exercise just to assure ourselves that the simpler method of calculating the value of output of the Bank of Canada produces results quite close to the complicated FISIM approach. We have no hesitation to assert that our calculation based on cost is robust and also easy to explain. Further, our allocation of this output to the Federal Government is intuitively appealing as it reflects closely the many activities, more significant activities than financial intermediation, regularly performed by our central bank.

23.3 USA practice

The BEA calculates the value of output of the central monetary authorities (Federal Reserve Board) as recommended in the 1993 SNA, thus FISIM is recognised. FISIM counts for most of their value, though a small amount, less than 20%, is due to some direct charges. Value of total output is less than $3 billion, (a ratio, similar to Canada, of less than 1% of total output of financial institutions). FISIM is allocated only to depositors.

23.4 Concluding remarks

The method of calculating the value of output of monetary authorities is different in the two countries but the resulting value is similar. Further, its allocation in the two countries is quite different. In the United States, it is allocated almost entirely to enterprises, hence leaving overall GDP unchanged, whereas in Canada, it is entirely allocated to the federal government , hence adding to GDP. The amount is quite small, hence it has no significant influence on our results, but it is useful to clarify the underlying concept.

Statistics Canada - 463 - Collected Articles of Kishori Lal

24. Valuation of Insurance services

The activity of insurance is intended to provide individual institutional units exposed to certain risks with financial protection against the consequences of the occurrence of specified events. It is also a form of financial intermediation in which funds are collected from policyholders and invested in financial or other assets which are held as technical reserves to meet future claims arising from the occurrence of the events specified in the insurance policies. It is generally recognized that the service output includes the transfer of risk, financial intermediation and administrative services such as the handling of claims. It is also recognised that insurers maximise profits by setting premiums based on their expectations or probability regarding future claims and investment returns.

24.1 1993 SNA

The 1993 SNA defines the output of the insurance activity, both life and non-life, in its Annex IV, The treatment of insurance, social insurance and pensions, (page 537, paragraph 18) as follows: a) Total premiums or contributions earned; b) Plus total premium or contribution supplements ; c) Less claims or benefits due; d) less increase (plus decrease) in actual reserves and reserves for with-profit insurance.

The ESA 1995 clarifies and restates the definition of the output of insurance activity, both life and non-life, in its Annex III, Insurance (page 271, paragraph 27) as follows: a) Total premiums earned; b) Plus premium supplements; c) Less claims due; d) Less increase (plus decrease) in technical provisions against outstanding risks and technical provisions for with-profit insurance.

The definitions in the two documents are identical except that the ESA 1995 clarifies the meaning of actual reserves as provisions against outstanding risks.

Items a), b) and c) typically define the value of non-life insurance whereas all the four items, a) to d) define the value of life insurance, except when the non-life claim is paid as annuity. The 1993 SNA states: AMost of these reserves relate to life insurance but they may be needed in the case of non-life insurance when claims are paid out as annuities instead of lump sums (see paragraph 6.138 d).

The 1993 SNA definition of the insurance output is different from the earlier 1968 SNA definition and includes an important item called investment income from technical reserves (also called premium or contribution supplement ). As the premium supplement is a new item and quite significant in value, let us note how it is handled in the 1993 SNA. The 1993 SNA states: A Although the reserves are held and managed by the insurance enterprises, they are treated in the System as assets of the policyholders. The income earned on the investment of the reserves is, therefore, attributed to the policyholders for whose benefits the reserves are held. The income is recorded as

Statistics Canada - 464 - Collected Articles of Kishori Lal receivable by the policyholders who pay it all back again to the insurance enterprises as premium supplements. These premium supplements must therefore always be equal in value to the corresponding income from the investment of the technical reserves@ (paragraph 6.138 b). The revenues which the insurance enterprises have at their disposal to pay the claims include both the actual premiums earned and the premium supplements.

As in other financial intermediaries, the 1993 SNA does not include income from own funds. It states: A The income concerned comes from the investment of the technical reserves of the insurance corporations, ...and does not include any income from the investment of the insurance corporations= own funds@ (Annex IV , paragraph 16). The 1993 SNA also notes that AAll changes in insurance technical reserves ... are measured excluding any nominal holding gains or losses@ (paragraph 6,138). This limitation has raised both practical and conceptual difficulties, to which we will refer later.

The 1993 SNA definition works quite well in normal circumstances when claims due occur pretty close to probability expectations used by insurance corporations in setting the premium rates. However, when exceptionally big events such as hurricanes, earthquakes, floods etc., occur, claims due become very large, resulting in a reduction, sometimes a big reduction of measured output. The resulting reduced measure of output is counter-intuitive when in fact, the actual activity in the insurance corporations goes up. This is the issue which needs to be resolved.

The 1993 SNA provides some guidelines for recording the destruction of property resulting from acts of war or exceptional events such as natural disasters when it deals with the issue of consumption of fixed capital. It notes: A Consumption of fixed capital is a cost of production. It may be defined in general terms as the decline, during the course of the accounting period, in the current value of the stock of the fixed assets owned and used by a producer as a result of physical deterioration, normal obsolescence or normal accidental damage. It excludes the value of fixed assets destroyed by acts of war or exceptional events such as major natural disasters which occur very infrequently. Such losses are recorded in the System in the account for other changes in the volume of assets@ (paragraph 6.179). The1993 SNA defines exceptional events also in paragraph 12.7.

The 1993 SNA guidelines on recording the losses due to exceptional events in the Aother changes in the volume of assets account@ are quite clear for the valuation of consumption of fixed capital. The same principles may apply to the destruction of property due to exceptional events. However, the issue of how to record the payment of exceptional claims still needs to be decided. This issue was discussed at the September 1999 OECD National Accounts Experts Meeting. It was suggested by several participants that this problem could be resolved if claims due are separated into two parts, regular claims due and exceptional claims due. Regular claims due should be recorded, as suggested in the 1993 SNA, in the production account but the exceptional claims due should be recorded in the capital account as capital transfers from the insurance corporations to the insured sectors. This solution would eliminate the volatility in the value of output of insurance corporations and would be more easy to explain. The recording of exceptional claims in the capital account seems quite sensible as they are more of capital nature. This solution, however, has not yet been implemented in any country.

Statistics Canada - 465 - Collected Articles of Kishori Lal

There are two outstanding problems in the valuation of insurance activities which continue to bother national accountants at the international level; these are a) how to treat capital gains or losses on in the investment income and b) how and where to record exceptional losses. Both these issues are now being examined by an OECD Task Force on Non-Life Insurance.

24.2 Canadian practice

In the pre-1997 CSNA Historical Revision, the output of non-life (property and casualty) insurance was deemed equal to premiums less claims. The output of life insurance companies was deemed equal to the operating expenses plus the dividends paid by the stock insurance companies. For both life and non-life insurance, total claims paid sometimes exceeded the premiums receivable, thus generating a counter-intuitive result of negative output. The most important change in the 1993 SNA was the inclusion of the investment income from the technical reserves in the calculation of output. As noted above, this was a change from the 1968 SNA where such investment was not included. It was a welcome recommendation to add the income earned by insurance companies from prepaid insurance premiums and reserves against future claims to the premiums earned. Thus, at the time of the CSNA historical revision in 1997, the value of the output of insurance activities was calculated in a manner Aquite similar@ to what is suggested by the 1993 SNA and this calculation was carried in the CSNA time series back to 1961. The calculation for profits etc for the macro accounts did not include any capital gains or losses, as recommended by the 1993 SNA. However, in the calculation of output in the IO industry accounts, the entire investment income (including capital gains or losses) was and remains included, thus not fully conforming to the 1993 SNA guidelines. As the IO output and its distribution to all the users, including final demand, provides benchmarks for all the CSNA accounts, the value recorded for the household expenditures in the macro accounts also, thus, includes the effect of capital gains on output of the insurance industry.

Capital gains or losses: In the calculation of the investment income from technical reserves, the IO accounts in the CSNA did not exclude the capital gains or losses from other investment income and this was, mostly likely, unintentional . Our practice is not consistent with the 1993 SNA which recommends to exclude capital gains and losses from the investment income. However, we now believe, and we are not alone, that our practice not to exclude capital gains or losses is a correct one. An insurance enterprise, in a competitive world, must endeavour to make the maximum investment income from the reserves to set the rate of premium as low as possible. Whether such income is derived from investing in deposits, bonds, securities, or equity is immaterial, once risks are taken into account. Thus, from a national accounts point of view, it is not logical to include some income and exclude some other income when an insurance enterprise takes both into account to set the premium rates. It is inconsistent to work with the reported premium rates without simultaneously taking into account all the investment parameters which the insurance enterprise uses to set such rates. As in other financial intermediaries, we do not separate the investment income of own funds from investment income of funds belonging to others. As investment funds are fungible as well as difficult to precisely identify on the balance sheet of an insurance enterprise, we have ignored this recommendation of the 1993 SNA and most likely it is not materially important.

Exceptional losses: The CSNA has not made, so far, any special adjustment to the recording of claims due to exceptional losses. In Eastern Canada, there was an exceptionally big ice storm in

Statistics Canada - 466 - Collected Articles of Kishori Lal

1998 which destroyed many properties and hydro lines. The claims which the insurance corporations paid were quite large, thus resulting in a decline of output during that period whereas, in reality, the insurance activity had increased due to the massive handling of claims. We did not make any adjustment to our annual production account because the effect was not exceptionally large; however, we did adjust the output for that month as well as for that quarter, thus avoiding having negative output of insurance industry in our monthly and quarterly GDP by industry program when this event occurred. The value of output of insurance enterprises during that short period was projected on the basis of average premiums earned. We were not entirely consistent conceptually for an annual account and its sub-period accounts, but our results were intuitively right and this was important.

24.3 USA practice

This write-up on the USA practice is based on Dennis Fixler=s paper, Rethinking the NIPA Treatment of Insurance Services for the Comprehensive Revision, presented at the BEA Advisory Committee Meeting, November 15, 2002. The current measure of insurance output in the USA is based on the 1968 SNA convention which does not include the investment income from technical reserves, thus it is equal to premiums less claims. This value is lower than the one resulting from the 1993 SNA which includes such investment income. It is noted that the BEA would implement the 1993 SNA recommendation on investment income from technical reserves in its valuation of insurance in the upcoming comprehensive revision in 2003.

BEA has argued, both at the OECD Meeting of National Accounts, October 2001 and at the BEA Advisory Committee Meeting , November 2002 that capital gains should be included in the investment income. As well, Fixler argues to include the income from own funds in the calculation of investment from the technical reserves.

Recording of exceptional losses in the insurance output is found to be problematic by Fixler and others. As demonstrated by the insurance flows generated by the terrorist attacks on the World Trade Centre, large claims associated with catastrophes lead to unusual national accounts estimates. Using regular conventions, insurance output would decline which is counter-intuitive, as the actual insurance activity increases to handle vast claims

The BEA proposes to use an estimate of expected claims, rather than claims due as recommended in the 1993 SNA, for calculating the value of output of insurance enterprises. It is noted that the Australian Bureau of Statistics has already adopted the use of a measure of expected claims. Their measure is a 5-year centred moving average for routine or normal claims and a 19-year centred moving average for catastrophic claims. The use of expected claims in the production account will undoubtedly require adjustments for the difference between actual claims and expected claims in many other accounts, leading up to the closing of balance sheet. The balance sheet must include the full weight of actual claims paid.

Statistics Canada - 467 - Collected Articles of Kishori Lal

24.4 Concluding remarks

There seems to be a consensus developing that the investment income from technical reserves should include all property income, including capital gains and losses and that the investment income from own funds should form part of its total, for it to be used in the calculation of insurance output. In the CSNA, we already follow this practice and the BEA would follow it in its 2003 Comprehensive Revision. Thus, we will have an identical methodology by 2003 except for the issue of expected versus actual claims. As noted, there has been only one exceptional event, the ice storm in 1998 in Canada which resulted in a small decline in our estimate of insurance output for the annual total. For the monthly GDP estimate, the CSNA used regular premiums as a proxy for output for the month in which this storm occurred, thus the decline in output was properly avoided. Expected claims have never been used in the calculation of insurance in the CSNA and our hesitation remains. Perhaps we are lucky in Canada, as the catastrophic events are extremely rare in our history, thus mitigating the need to worry about a concept which may hardly be applied. However, as a member of the international community of national accountants, our general position is to remain very transparent about our calculations and to assure to our users that our estimates are verifiable from the records of the transactors. Rather than calculating expected claims every year, using moving averages, we would like to suggest that only in the catastrophic year, the actual claims be replaced by some other estimated claims, perhaps based on the growth of actual premiums and the difference between the actual and estimated claims be handled through a capital transfer in the capital account. These one- time adjustments should be fully documented and provided to the users.

25. Valuation of output of government services

Government services are not provided in the context of a market but are supported by taxes and consumed collectively by the society at large. Consequently, there are no market prices to determine their value. A long-standing convention to determine their value, in the system of national accounts, has been to deem it equal to their costs. The boundary of which costs should be included and which costs should be excluded has not been clear-cut, both conceptually and in practice. Further, the accounting conventions used for presenting the government accounts for budgetary purposes have been changing overtime. In the government budgets of many countries, all expenditures, operating and capital, are expensed in the period they are incurred. However, there is a strong thrust now to separate current operating expenditures from the capital expenditures for presentation of government accounts to international organisations (see more on this issue in the International Monetary Fund=s Government Finance Statistics Manual, 2001).

In the system of national accounts, it is now generally agreed that the valuation of output of government services must include, a) all operating costs and b) costs of consumption of fixed capital formation. This convention, and its near universal practice, is a vast improvement over the earlier heterogeneous practices in many countries. Recently, the national accountants of some countries, particularly of the United States, have questioned the rationale of limiting the inclusion of the capital costs to only the consumption of fixed capital, rather than the full cost of capital services (consumption of fixed capital as well as the financial costs of capital). We will come back to this issue later but let us first record what the 1993 SNA says on this topic.

Statistics Canada - 468 - Collected Articles of Kishori Lal

25.1 1993 SNA

The 1993 SNA states: AThere are no markets for collective services such as public administration and defence, but even in the case of non-market education, health or other services provided to individual households, suitable prices may not be available. It is not uncommon for similar kinds of services to be produced on a market basis and sold alongside the non-market services but there are usually important differences between the types and quality of services provided. In most cases, it is not possible to find enough market services that are sufficiently similar to the corresponding non-market services to enable their prices to be used to value the latter, especially when the non-market services are produced in very large quantities@ (paragraph 6.90). It further states: AFor these reasons, and also to ensure that the various non-market services produced by government units and NPISHs are valued consistently with each other, they are all valued in the System by the sum of the costs incurred in their production: that is, as the sum of: intermediate consumption, compensation of employees, consumption of fixed capital, other taxes, less subsidies, on production. The net operating surplus on the production of non-market goods or services produced by government units and NPISHs is assumed to be zero A (paragraph 6.91). It is important to underline that the 1993 SNA recommends that the net operating surplus in the non- market sectors should be zero which means that it does not recommend including the financial cost of capital, as the inclusion of financial cost of capital would generate operating surplus, as it always does in the production of market output. It is this recommendation which has led to objections by the national accountants in some countries. It is worth noting that the inclusion of financial cost of capital (which in terms of value will be quite significant) will raise the value of output of government services and the value added of the government sector by an equivalent amount. Also the share of the government sector will increase in the overall economy.

25.2 Canadian practice

In the CSNA, the value of output of government services is consistent with the formula recommended by the 1993 SNA. It is equal to their cost of production: the sum of: intermediate consumption, compensation of employees, consumption of fixed capital and other taxes on production. Subsidies on production are not allocated to the government sector in the CSNA. No allocation of financial (or the interest) cost of capital is provided, at present, in the cost structure of government output. Let us see what would happen if we add full capital services in our government account. In 2001, the value of government current expenditures on goods and services was $204 billion, of which consumption of fixed capital amounted to $21 billion. For the same year, the value of government non-residential structures was $323 billion, residential structures $8 billion and machinery and equipment were worth $30 billion, or a total of gross fixed capital formation of $361 billion. In addition to the produced fixed capital assets of $361 billion, governments also own significant natural resources. A 5 to 6% interest rate on $361 billion assets would generate a net return of some $20 billion, an amount very similar in magnitude to the value of CFC of $21 billion. Thus, if we add the full cost of capital services, the value of government current expenditures would increase by $20 billion, or in other words, government expenditures will rise by about 10%, a very significant amount.

Statistics Canada - 469 - Collected Articles of Kishori Lal

25.3 USA practice and their objections

In the NIPA accounts produced by the BEA, the value of output of government services, at present, is consistent with the formula recommended by the 1993 SNA. However, they have raised serious objections to the recommendation of the 1993 SNA. Robert Parker and Jack Triplett stated the BEA position in 1995 as follows: A Use of depreciation as a measure of the value of services of government fixed assets is a partial measure of the total value. In theory, the service value of an asset should equal the reduction in the value of the asset due to its use during the current period (depreciation) plus a return equal to the current value the asset could earn if invested elsewhere (net return) .A (See their article, Review of the Comprehensive Revision of the National Income and Product Accounts: Recognition of Government Investment and incorporation of a new methodology for calculating depreciation, Survey of Current Business , September 1995 ). A recent review of the government sector of the U.S. national accounts by the Committee on National Statistics of the National Research Council, in a section entitled Going Beyond the System of National Accounts, puts the case against the convention of the 1993 SNA:@ The assumption of zero net return is implausible. If net return were really zero, it would imply substantial over-investment in public capital. In fact, however, serious shortages of many types of public infrastructure, ranging from schools to transportation systems, are widely perceived to exist@ (see C.M.Slater and M.H.David- editors Measuring the Government Sector of the U.S. Accounts. Committee on National Statistics, National Research Council, 1998, Washington, DC.

It is quite obvious that the BEA would like to add the full value of capital services (depreciation and net return to capital) in the current price valuation of government services, thus raising the value of their output and their value added.

25.4 Concluding remarks

The national accountants in the USA are not alone in raising objections to the recommendations of the 1993 SNA in regard to the value of output of government services. Australia has raised similar objections and we fully support their positions. These objections and concerns are also shared by a recent OECD document, Measuring Capital: A Manual on the Measurement of Capital Stocks, Consumption of Fixed Capital and Capital Services (Paris 2001). It states: AThe production account in the 1993 SNA is not, in fact, a proper production account @ (page 118). It further states that inputs of capital services need to A... be recorded in the production account alongside compensation of employees. The services may be valued by the actual or estimated pure rentals payable; that is, by the sum of depreciation and the capital, or interest, costs. While it would not be easy to estimate the value of capital services, it is no more difficult than estimating depreciation, or consumption of fixed capital@ (page 118).

The value of capital services is measured by very few countries and nowhere are they included in the value added series in published national accounts. Some countries have developed some volume measures but to the best of our knowledge, no country has yet produced any current price measures. The United States, Australia, the 2001 OECD Manual on Measuring Capital, all point to the lack of inclusion and its articulation of full capital services as a major weakness in the 1993 SNA production account. Lack of full capital services in the valuation of government services is particularly

Statistics Canada - 470 - Collected Articles of Kishori Lal damaging, both theoretically and empirically, as the values are very significant. Our rough calculations for Canada suggest an increase of $20 billion for 2001 due to net return on capital, an amount as big as the CFC for the government account.

We believe that the 1993 SNA should be revised and expanded to reflect the full capital services, both CFC and net return, in the valuation of current price output of government services. The value of capital services will have to be produced. Thus it is necessary to examine the feasibility to produce both current price and consistent volume measures of capital services, all integrated in the national accounts, to better serve the users and to bring more transparency to these estimates. This is one more area for joint research work between the CSNA and the BEA, particularly given its importance and its significance.

This note equally applies to the valuation of output of services produced by other non-market producers, such as NPISHs.

26. Valuation of defence services

Defence services, like many government services, are not provided in the context of a market, are supported by taxes and consumed collectively by society at large. Consequently, there are no market prices to determine their value. A long-standing convention to determine their value, in the system of national accounts, has been to deem it equal to their costs. However, there is a fair amount of divergence in practice amongst countries regarding the boundary of the costs to be included or excluded in any given calendar period, for the valuation of defence services. There is a universal agreement that all operating costs plus part of the capital costs consumed in any given period must be included; however, the boundary of what is capital in defence is not unambiguous.

26.1 1993 SNA

An issue has arisen in the most recent international system of national accounts, the 1993 SNA, regarding which capital investment should be considered as capital and which as operating. The 1993 SNA states: A In order to be treated as capital, a good must not only be durable but used repeatedly or be continuously in production over a number of accounting periods. However, if military weapons such as rockets, missiles and their warheads, are used in combat, they are used to destroy and not to produce. Thus, the actual use of destructive weapons can scarcely be treated as an input into an economic process of production@ (paragraph 6.168). The 1993 SNA further argues: AThe provision of defence, however, can certainly be construed as a form of production from which people benefit and for which they are prepared to pay, individually or collectively. Moreover, the provision of defence, like any other productive activity, does require the repeated or continuous usage of certain durable goods over a number of accounting periods. Thus, a distinction can be drawn between those durable goods that are actually used in much the same way as in any other type of production, and those which either are never used or, if they are used, do not constitute inputs into a productive process. This suggests a distinction between ordinary producers= durable goods of a kind used throughout the economy and destructive military weapons designed for combat@ (paragraph 6.169). It thus recommends:@...expenditures by the military on weapons of destruction and the equipment needed to deliver them should be classified as intermediate consumption. Conversely, the

Statistics Canada - 471 - Collected Articles of Kishori Lal construction of buildings for use by military personnel, including hospitals and schools, and also of roads, bridges, airfields, docks, etc. for use by military establishments should be treated as gross fixed capital formation@ (paragraph 6.170). The 1993 SNA draws a clear boundary between typically military equipment, usable for military purposes only and the other capital equipment acquired by defence services but usable for civilian purposes. It should be noted that this distinction is not observed consistently across countries. As defence expenditures are usually quite large, this inconsistent treatment makes international comparison problematic.

26.2 Canadian practice

In Canada, we have implemented the recommendation of the 1993 SNA in regard to defence expenditures. For example, the Defence Department purchased equipment worth $2.5 billion in 1999, of which the expenditure on military equipment was of the order of about $1.9 billion. Following the 1993 SNA recommendation, this $1.9 billion was expensed in that year and only $ 0.6 billion was capitalised. Typically, purchases of military equipment represent more than 60% of total equipment purchased by the Defence Department. Had we capitalised all the military equipment, consumption of fixed capital, hence the level of our GDP would have increased by about $1.5 billion on an annual basis in the 1990's. It should be noted, by those not immersed in SNA conventions that the classification of purchases of equipment from the operating to the capital category does not affect the level of final expenditures, hence GDP of a country. What affects the level of GDP is the inclusion of consumption of capital as an additional imputed current expenditure by the government sector.

26.3 USA practice

The BEA disagrees with the 1993 SNA recommendation in regard to defence expenditures and its position is well articulated in a recent paper, The System of National Accounts for the New Economy: What Should Change (prepared by Brent Moulton, June 2002 and presented at the International Association for Official Statistics, London August 2002). It notes: AThe SNA recognises the provision of defence as a productive service, and the labour and non-weapons equipment and structures that are used by the military are considered as productive inputs. Technologically sophisticated aircraft, tanks, and warships, however, are increasingly used as substitutes for personnel in defence activities. By not counting these critical inputs as providing capital services to the military forces, the SNA>s treatment seriously impairs the accounts in describing the actual production process of defence services@ (paragraph 35). It further argues: AThe failure to recognise most defence equipment as capital also makes the accounts less useful in measuring saving and wealth. Military equipment are valuable assets that are sometimes sold and that should be reflected in national balance sheets@ (paragraph 36). In BEA=s view, military equipment is used continuously in the production of defence services and is thus a fixed asset of the federal government. The BEA believes that investing in the development and acquiring of missiles and rockets and other weapons such as atomic bombs provides continuous defence services to the community at large, hence it is capital.

This additional (compared with the 1993 SNA) capitalisation has added about US$60 billion of value added and final consumption of the government sector in the USA in the recent years. All other

Statistics Canada - 472 - Collected Articles of Kishori Lal

OECD countries follow the 1993 SNA guidelines. The OECD needs to present, to the extent possible, comparable inter-country valuation conventions in their reports. The OECD receives from the BEA national accounts data on the comparable 1993 SNA basis wherein the BEA removes the capitalisation of military equipment, causing GDP on an SNA basis published by the OECD to be lower than GDP on a NIPA basis by about US$60 billion (representing the consumption of fixed capital for military equipment, which in the NIPA=s is included in general government value added).

26.4 Concluding remarks

The issue raised by the BEA that defence equipment provides continuous services has significant statistical implications and we fully support that its treatment in the 1993 SNA must be re-examined by the Inter-Secretariat Working Group on National Accounts with some urgency. Further, the 1993 SNA treatment requires an unattractive sequence of transactions whenever military equipment is traded. The sale of battleships, aircraft etc. between nations, in particular, is quite common and potentially significant. Under the 1993 SNA, their sale to a non-resident presumably requires the goods export entry be matched by a reduction in defence intermediate consumption. A payment in the form of reduced intermediate consumption lacks intuitive appeal as there cannot be negative intermediate consumption in reality.

Until the issues raised by the BEA and also Australia (and we are sympathetic to their well articulated concerns) are resolved at the international level, it is important that the OECD continue to publish internationally comparable series on defence expenditures such that the users can use these series with full awareness and confidence.

In addition to the debate whether equipment should be considered capital for defence services, the issue discussed above on including the full cost of capital services equally applies here.

27. Head Office activities

When an enterprise=s production takes place in two or more different establishments, certain ancillary activities may be carried out centrally for the collective benefit of all the establishments. For example, the purchasing, sales, accounts, computing, maintenance or other departments of an enterprise may all be the responsibility of a head office which is located separately from the establishments in which the activities of the enterprise are carried out. There are two choices for handling the activities of the head office: a) the head office is recognised as a separate establishment; it has output and value added and it sells its services to the establishments it serves; or b) it is not recognised as a separate establishment, it has no output or value added and its costs are distributed to the establishments it serves, in proportion, for example, to the latter's outputs or costs, and added to the latter's own costs.

Total value added in the economy does not change but the distribution by industry (if the establishments are in more than one industry) and value added by regions (if the establishments are in more than one region) will change, depending on which choice is made.

Statistics Canada - 473 - Collected Articles of Kishori Lal

27.1 1993 SNA

The 1993 SNA recommends choice (b) above as it does not recognise the provider of ancillary activities, such as a head office, as an establishment: thus it has no output. It states (paragraph 5.13):

A(a) The output of an ancillary activity is not explicitly recognised and recorded separately in the System. It follows that the use of this output is also not recorded;

(b) All the inputs consumed by an ancillary activity-materials, labour, consumption of fixed capital, etc.- are treated as inputs into the principal or secondary activity which it supports;

(c) It is not possible to identify the value added of an ancillary activity because the value added is combined with the value added of the principal or secondary activity.@

27.2 Canadian practice

At Statistics Canada, head offices are identified as separate units, with geographical locations to which employment and capital expenditures are assigned, but this practice is limited to establishment-based surveys for manufacturing industries. For all other industries, head offices are not separately identified. For purposes of industrial classification, the entire unit is assigned to a single industry, the one in which the bulk of the value added of the establishments it serves is generated. In the Canadian input-output tables, there is no separate head office industry in the CSNA industry statistics as each head office is classified to the same industry as the most significant establishment (s) it serves. This classification convention in the CSNA is different from the one recommended in the North American Industry Classification System (NAICS) where a separate head office industry has been created which includes all head offices, no matter which establishments and industry they serve. With the present practice, the value added, say, for manufacturing, is the same in the CSNA as in the 1993 SNA but its industrial distribution is different depending upon whether the CSNA or the 1993 SNA convention is used. Further, the value of output as well as an identical value of intermediate consumption is higher in the CSNA than in the 1993 SNA.

As there are no outside market transactions for the services provided by the head office to establishments, one needs to develop some imputation convention to value its output. The value of head office output is equated, in the CSNA, to its operating costs. The entire output of the head office is completely used up as intermediate consumption by its serving establishments, thus reducing the value added of each establishment by the amount of use of the head office service.

In the Canadian provincial input-output tables, the head office is recognised as a separate establishment and, for purposes of industrial classification, the entire unit is assigned to a single industry. That industry is the one in which the bulk of the value added of the establishments it serves is generated, as we have done for national industrial statistics. The head office produces output which is completely used up as intermediate consumption by its serving establishments, thus reducing the value added of each establishment by the amount of use of head office service. Value added for the country as a whole does not change but its provincial or regional distribution does. It is reduced in some regions counterbalanced by an identical increase in the region of the head office.

Statistics Canada - 474 - Collected Articles of Kishori Lal

The regional distribution of both output and value added are different using the CSNA conventions compared with the 1993 SNA conventions

27.3 USA practice

For the 1992 input-output accounts of the BEA, head office activities were dealt with in a manner that is consistent with the 1993 SNA; that is, no output was imputed for them, and the expenses related to their activities were shown as inputs to the industry the head office served.

Head office activities are shown differently in the 1997 input-output accounts, released in December 2002. There is a NAICS industry for the management of companies and enterprises. The output of this industry is measured by its total costs; these include some expenses paid by the head office on behalf of other establishments of the company, as well as expenses related directly to head office activities. The services of head offices are then sold to the industries they serve. The result does not change total value added in the economy, but it changes its distribution because compensation, depreciation, and taxes paid by head offices are included in the value added of the new industry, rather than in the value added of the industry being served.

27.4 Concluding remarks

In the CSNA industry statistics, particularly its input-output tables, head offices are separately identified only for manufacturing and therein each head office is assigned to a single industry, the one in which the bulk of the value added of the establishments it serves is generated. The value of its output, (limited to the manufacturing sector), is equated to its operating expenses and is fully used by its serving establishments. In the United States, for the 1992 input-output tables, the head office was not separately identified, and the tables followed the convention suggested in the 1993 SNA but in its 1997 input-output tables, head offices form a separate industry, and have their output and value added.

In the CSNA, we equated the value of the output of the head office, in the manufacturing sector, to its operating costs. The issue has arisen in regard to provincial accounts whether the present CSNA practice properly reflects the economic reality. Should the value of the output of a head office include, in addition to its operating costs, also a share of profits of its serving establishments. If it should, the share of profits allocated to the head office may be equated to its share of total wages paid by the enterprise multiplied by its total profits. One could devise some other convention to distribute profits. In any case, our present leanings are that in future the value of the output of the head office should be equated to its costs plus shared profits. This issue may be discussed with our colleagues in the BEA to arrive at a common joint decision.

The other issue of implementing NAICS head office industry in the Canadian statistics also needs to be settled as the USA has already decided to implement it in their industrial series effective 1997. In this context, it may be noted that it was not feasible to create a separate head office industry in the Canadian 1997 IO tables, even though they were based on NAICS, as there was no comprehensive data available. However, Statistics Canada has now implemented a survey program to collect head office data for all industries for the reference year 2002. Thus, it is planned to include a separate head

Statistics Canada - 475 - Collected Articles of Kishori Lal office industry in the IO tables for 2002 and onwards, which will make the detail in IO tables in the two countries fully comparable.

The 1993 SNA recommendation becomes particularly problematic when input-output tables or industrial statistics are produced at the provincial or regional level. At the regional level, when the head office is situated in a region different from that of the producer units it serves, the application of the 1993 SNA recommendation would imply no contribution of the head office to the value added of its region. The present recommendation regarding the recording of head office activities in the 1993 SNA is counter-intuitive and must be changed to bring it in line with economic reality, for both industrial and regional statistics.

Overall conclusions

This report provides, in a summary fashion, similarities and differences in the production accounts of Canada and the United States. The discussion is limited to those issues which affect the level of output, value added and GDP, both at the total economy level and by industry or sector, all at current prices. We have noted 27 issues, distributed under four broad headings: A). An examination of the production boundary recommended by the 1993 SNA and the effect of lack of its full implementation on the level of production in the two countries. B). A review of present practices in the two countries in compiling production account for institutional sectors vis-a-vis the recommendations of the 1993 SNA and their effect on both inter-country and international comparisons. C). A review of the conventions used for valuation of output and value added in the two countries vis-a-vis the recommendations of the 1993 SNA and their impact on their inter- industry comparisons as well as international comparisons. D). A review of the present conventions in the 1993 SNA and differentiate those which have become un-aligned with the economic reality in the world, thus may be changed.

Effect on Total GDP, due to lack of consistency with the 1993 SNA

Both Canada and USA have an identical production boundary which is consistent with that of the 1993 SNA boundary, with only a handful of exceptions. Our ballpark estimate is that what we are missing could add to no more than one percent of overall GDP. In the case of Canada, in the current period, we could add about $2.5 billion for labour for own account construction, about $1 billion for capitalising orchards, about $0.5 billion for capitalising entertainment originals, and $0.2 billion for prostitution; add a few more billion dollars for narcotics and still we have not reached a total of $10 billion or one percent of our current GDP. The situation in the USA is probably the same. Having said that, we would still suggest that these activities should not be ignored for ever. Their overall impact is very small but their effect on regions and individual industries may be significant. Such industries include agriculture, construction, cultural and personal services industries. Further, their impact varies greatly amongst countries, particularly developing countries and this makes international comparisons problematic. As statistically advanced countries, we need to set an example to support the full implementation of the 1993 SNA throughout the world.

Statistics Canada - 476 - Collected Articles of Kishori Lal

Effect on Total GDP, due to methodological differences with each other

There are several areas where the two countries use different methodologies and conventions which result in differences in the value of production. The two most noteworthy are the banking services (FISIM) and defence. The value of FISIM is calculated quite similarly in the two countries but its allocation to the users is very different. In the United States, the entire output is allocated to depositors but in Canada it is allocated to both depositors and borrowers as recommended in the 1993 SNA. As most of the depositors belong to the household sector and most of the borrowers belong to the business sector, the use of the US convention results in a much bigger allocation to final consumption by the households, hence GDP, compared with Canada. Our estimate is that this increases the total GDP in the USA by more than 1.5%, compared with Canada.

Another issue which the BEA has raised concerns the inclusion of military equipment in the capital for defence services, which the 1993 SNA does not recommend. This issue has been discussed in the BEA and other organisations in the USA and, in our judgement, they have made a convincing case for its inclusion, hence for revising the 1993 SNA in ths regard. The BEA already includes this in its published GDP and its value is about US$60 billion in recent years, or more than 0.5% of GDP. Both the USA and Australia are challenging the present 1993 SNA guidelines on capitalisation in the defence services and we find their articulation very convincing. We all would like that the 1993 SNA restriction on capitalisation of military equipment be removed.

Per our present practices in the allocation of FISIM and capitalisation of defence capital, the GDP in the USA is more than 2% higher compared to Canada. The BEA is re-examining the present allocation convention of FISIM and may adopt the recommendation to allocate FISIM to all users, barrowers and depositors and in the next few years, the present restriction on capitalisation of military equipment may also be removed, thus, our present differences may disappear in the near future.

Non-business sectors

Both Canada and the USA compile production accounts for the two non-business (non-market) sectors, government and NPISHs. Their individual values are not comparable between the two countries. For example, the government sector in Canada makes about 16% of total value added compared with about 10% in the USA. Most of this difference is due to the different classification conventions used: most non-profit institutions are allocated in the government sector in Canada while they are allocated to NPISHs in the United States. Thus the NPISHs sector makes about 4% of total GDP in the USA and only 1% in Canada. Additionally, in Canada, we have used a methodology to calculate CFC in the government sector which gives a higher estimate compared with the methodology used in the United States. The published numbers encourage one drawing wrong inferences that there is more government and less charity in Canada compared with the United States. The reality, however, is that the values for activities limited to public administration and defence are quite similar in the two countries and the major reason for the higher value of the government sector in Canada is the much heavier involvement of its government, compared with the USA, in the provision of public health services to the population at large. One possible solution may be for the CSNA to have detailed discussions with the colleagues at the BEA and jointly agree on the

Statistics Canada - 477 - Collected Articles of Kishori Lal classification rulings to be adopted by the two countries. In the meantime, the CSNA may put more emphasis on the results of the two non-business sectors together and publish, within the government sector, sub-sector detail for a) government administration and defence and b) other government activities including education, health, etc.

Valuation of output and value added by industry

This is one more area where the difference in the practices of the two countries is vast. Canada uses the 1993 SNA recommended basic price for valuing value added by industry. Canada also uses the recommended basic price valuation for output for all products which do not attract subsidies. For a handful of products which get subsidies in Canada, we have modified the 1993 SNA recommendation by not including subsidies in the valuation of output. In the United States, both the output and the value added are valued at producers= prices which include all the sales and excise taxes levied by the various levels of governments. This convention adds, both in the output and value added, the sales and excise taxes collected by the enterprises for transmission to the governments. Gauging from the Canadian sources, these taxes added $75 billion in 2001 in the total GDP of $1,092 billion, and a very significant amount of these taxes were collected by the trade industries. Thus the value added of trade industries, for example, in the USA is higher by the amount of sales and excise taxes compared with the value added in Canada. The results of value added by industry in the two countries are not comparable. This has a major impact on comparison of productivity estimates by industry for the two countries, unless adjustments are made to put the value added series on a comparable basis, which are not easy.

Valuation of financial services

The valuation of financial services in the 1993 SNA was a major improvement over that in the 1968 SNA. However, some guidelines of the 1993 SNA have been found problematic for implementation and some others are being felt as deficient as we examine the rapid changes in the financial sector since the write-up of the 1993 SNA. In the calculation of output of financial services (banks, insurance companies, other financial enterprises), there are two major issues which need an urgent attention: a) should capital gains/losses be included and b) should income received from own funds be included? The 1993 SNA recommends excluding both. The OECD has established two task forces, one on banking and the other on insurance to deliberate on these and other issues. The BEA has argued, with well-articulated position, to include both capital gains and income from own funds in the output. We support the BEA position that the 1993 SNA needs to be revised in this regard. It may be noted that we have already included income from own funds in the output of financial services in the CSNA but in the USA, it is ignored because FISIM is not yet allocated to barrowers, as noted above.

How to treat the destruction of property from exceptional events such as big hurricanes and terrorist attacks such as September 11, 2001, is another important issue for the calculation of insurance services. The 1993 SNA is considered deficient both by the BEA and Australian Bureau of Statistics (and we agree with them), hence it should be re-examined. At present, the practice in this area in the two countries is different. The above noted OECD task force is looking at this issue and

Statistics Canada - 478 - Collected Articles of Kishori Lal professionals from both the CSNA and the BEA are members of this task force. We are looking forward to their deliberations and conclusions.

Valuation of government services and defence

In the 1993 SNA, the valuation of government services, including defence is equated to their costs- intermediate consumption, compensation of employees and consumption of fixed capital. The BEA has been advocating to include the full cost of capital services -consumption of fixed capital plus financial cost or interest cost of capital- not just CFC in the valuation of government services. The BEA position is well articulated. We fully support that this issue be re-examined. Note that its inclusion will add a very significant amount to the value of government services and total GDP, for example in Canada, about $20 billion in the current period, a value of similar magnitude as CFC.

Head office activities

In addition to the valuation of financial services and government and defence services, the 1993 SNA is deficient in its handling of head office activities, particularly when accounts are prepared by regions. In the 1993 SNA, the head office is not recognised as a separate establishment, it has no output or value added and its costs are distributed to the establishments it serves. This convention produces counter-intuitive results when the head office is in one region and the establishments it serves are in other regions. In the NAICS classification, head offices are classified as separate establishments and form a separate industry. Head offices are separately identified in the 1997 benchmark IO accounts of the United States but in the Canadian national and provincial IO tables, head offices are recognised as separate establishments only in the manufacturing sector. This creates another difference in the practices on industrial classification in the two countries. This difference will disappear for the reference year 2002 and onwards when a separate head office industry is created in th Canadian IO tables.

Net value added

Net value added, compared to gross value added, is preferred, as it is more in tune with the theoretically correct Hicksian concept of income and sustainable development. Its calculation requires an estimate of consumption of fixed capital (CFC) by sectors and by industry. In the CSNA, there is no estimate of CFC by industry and even in the macro series, it is not the CFC which is used but the tax determined depreciation. In the BEA industry accounts, such CFC estimates are made and in the macro NIPA series, both tax determined depreciation and CFC are used. The US methodology to distribute enterprise- based CCA and capital consumption adjustment to industry-based CFC for the GDP by industry calculation is very innovative and pragmatic and, in our judgement, its adoption in the CSNA will greatly enhance the usefulness of its industry statistics.

Other issues

There are also similarities and differences in several other areas but they are not as significant as the ones noted above. Both Canada and the USA ignore the purchase/sale of existing capital assets by

Statistics Canada - 479 - Collected Articles of Kishori Lal one industry to another within the business sector, a practice inconsistent with both the 1993 SNA and business accounting principles that could lead to incorrect multi-factor productivity estimates by industry. Both Canada and the USA do not compile production account by sectors, thus deviate from the 1993 SNA recommendation. The 1993 SNA has wisely recommended production account for institutional sectors as only they (not industries) are and can be used for the full sequence of accounts such as production, income and outlay, capital finance and balance sheet. Both Canada and the USA produce standard IO tables based on NAICS but the industrial aggregations are a bit different, thus they are not straightforwardly comparable in industrial detail. Both Canada and the USA have extensive industry statistics programs, and their respective input-output tables play a crucial role. Both countries have classified their industries by sector but the sector boundaries are not identical and the presentation of industrial detail for the non-market sectors is quite different. All these issues are discussed in detail in this document and the reader may refer to the relevant sections for additional information.

Final remark

Both Canada and USA have a very comprehensive set of statistics on national accounts and, by and large, these are consistent with the revised world-wide guidelines on national accounting, the 1993 SNA and with each other. The 1993 SNA is already 10 years old and some issues have emerged since then which require changes to its present guidelines. Our differences, in a few cases, from the 1993 SNA guidelines are due to the fact that as we started implementing them in our accounts, we found some of them quite problematic, thus either modified them or simply did not implement them and these differences were documented providing our reasoning for differences, and widely disseminated. The ongoing review of the SNA may resolve some of these differences. Some of the differences in the practices by the two countries may disappear in the near future as the professionals in both the BEA and the CSNA have shown a very high level and friendly cooperation in the deliberation of this report and are increasingly looking forward to further cooperation on harmonisation and improvement of our national accounts and hopefully, through this, also the international standards. This paper is the first concrete step on this path. The most important purpose of this document is to provide, to the compilers and users in both countries, a fuller understanding of our present practices, our similarities and differences, so that they can make meaningful comparisons of the published national accounts data. This paper is limited to those issues which affect primarily the production account, specifically the level of output, value added and GDP, both at the total economy level and by industry or sector, all at current prices. The methodologies used to develop constant price estimates are quite similar in the two countries, as both use Chain Fisher Volume indices, the indices also preferred by the 1993 SNA; however, undoubtedly, the two systems employ deflators, direct valuation, and quantity extrapolation to prepare volume estimates which are not always identical. Thus our similarities and differences in methodologies, conventions, assumptions, classifications etc., which affect the current price values affect also the constant price series, though not identically. Thus, it will be useful to prepare a detailed document on the similarities and differences in regard to deflators used in the two countries.

One would hope that, in future, similar reports are prepared on other topics in the national accounts, which affect such important constructs as savings rate, net lending, the capital finance accounts and the balance sheet for the detailed institutional sectors and the economy as a whole. A detailed

Statistics Canada - 480 - Collected Articles of Kishori Lal comparative knowledge and understanding of these topics will be very helpful to study the performance of the financial markets in our two economies which are increasingly getting bound together.

Note:

I am deeply grateful for the very generous help and cooperation and detailed comments and clarifications provided by many colleagues at the BEA, particularly Brent Moulton (who organised most of the comments), Sumiye Okubo, Bob Yuskavage, Brooks Robinson, Ann Lawson, Mark Planting, and Karen Moses and at Statistics Canada, particularly Abe Tarasofsky, Karen Wilson, Yusuf Siddiqi and Trish Horricks, in the preparation of this report. However, the views expressed here are entirely mine and may not be attributed to any of the colleagues who helped me or to Statistics Canada or BEA.

Statistics Canada - 481 - Collected Articles of Kishori Lal