THE WORLDBNK iTDP-0OZI -

Public Disclosure Authorized Internal Discussion Paper

LATIN AmERICA AND THE CARIBBEAN REGION

Report No. IDP-078 Public Disclosure Authorized Measuring Price Distortionsfrom Commodity Taxation in

A. Estache

Public Disclosure Authorized and A. Fernandez and I. Roy (consultants) July 1990

Infrastructure Operations Country Department I Public Disclosure Authorized

Discussion Papers are not formal publications of the World Bank. They present preliminary and unpolished results of country analysis or research that is circulated to encourage discussion and comment; citation and the use of such a paper should take account of its provisional character. The flndings, trpretations, and canclusions eressed in this paper are entily those of fth autr(s) and shoul ro be attut In any man to theWorld Bank, to its affiliated organizations, or to members of its Board of Executive Directors or the countries they represent LAC DISCUSSION PAPER SERIES

Report No. T1.i Author and Date

IDP-3 "An Analyss of the Sources of Earnings Vari Among BrazIlan Mals" by Ma~clo Dabo and Geoge Psacharpouloe, December 1987 ID?-4 "The Ef am Ek of Credit and Export Crealt Insurance Programs" by Bruce Fitzgerald and Tmny Monson (Cosufant, Dce~ 1987 113.9 "Export Procesing Zone: The Economics of Offthore Manufactudng* by Peter 0. Warr (Consu~tant), August 1987 IDP-10 "Dumping, Ant-dumping and Effolency" by Bruce Yandle and ElIzabeth M. Young (Consultant), August 1987 IDP-1 1 "The Regulaton of the Qualty of Traded Commodie and Services- by Simon Rotenberg and Bruce Yandle (Consultants), June 1987 IDP-12 "Argentina: Proble for Achieving Macro Stabity" by F. Desmond MoCarthy and Alfrdo E. Thome, January 1988 IDP-13 "Argentna: Toward tO Year 2000" by F. Deomond McCarthy, June 1987 IDP-14 " Uberlzation: The Lessons of Expedence", Papers presented In the conference *Toward a New Trade Policy for Braz9, Sao Paulo. April 11 and 12, 1988 IDP-18 *Aspect of Privatization: The Case of Argentna 1976-81w by R. Luders (ponsultant), Apr 1988 IDP-17 "Aspects of Privatization: The Case of Chile 1974-85, by D. Hachette (consultant), April 1988 IDP-18 "Privatization In Argentina and Chile: Leesons from a Conpadon" D. Hachefe and R, Luders (consultants), Aprl 1988 IDP-19 "Principles of Water Supply Pricing In Developing Countes* by Mohan Munainghe, June 1988 IDP-20 "The Status of Energy Economic: Theory anJ Applkcation" by Mohan Munadinghe, June 1988 IDP-21 *What are the Prospect for Land Reform?" by Hans Binswanger and Miranda Elgin (consultant), August 1988 IDP-24 *Managlng Argentina's External Debt The Contribution of Debt Swap", Cad N~ Afrdo Rodrguez (oonsultmnt), January 1989 IDP-29 *Managing Mexc's Extemal Debt: The Contributon of Debt Reduc~on Schemes", Allen SanginUs (con~ulant), January 1989 IDP-30 *Debt Reduction Schems and the Management of Chflean Deb, Fepe Laraifn (consultant), March 1989 DP-31 *Managiril~ Exterral Dobt The Contributfon of Debt Reducion Schemes", Dionisio D. Camelro and Rogedo LF. Wemeck cnulrt), Juluafy 1969 DP-32 "Leading Econoi* Indicators for BraH: At Atempt at Forecasng Turning Ponts", Antonio Estache, February 1989 (DP-33 *Recoverig Growth wh Equity, World Bank Povery Alevlation ActMis In LatIn Amedlca", George Pacharopoulos, Aprl 1989 IDP-34 Brazil- External Dobt Development and Prospect", SivIna Va~nick, December 1988 IDP-40 'Cash Debt Buy Back and the Insurance Vake of Reserves", Sweder van WIjnbergen, June 1989 IDP-41 «Growth, Extemal De and the Real Exchange Rate In Meidco", Sweder van Wijnbergen, May 1989 IDP-43 "The Macroeconomics of Social Secudty In Braz: Flecal and Financial ~ese*, Douglas Puet (Consultant) and Emmanual Y. j~enez, June 1989 IDP-40 "Feedlng Latin Amedca's Chdren", Human Resources DMslon, Techncal Department October 1989 IDP-50 *Marginal Effetive Rate* on Capital Income In Argentina", Luca Babone and Michael McKee (Consuant), August 1989 DP-53 "A Bblography on Poverty and income DistrIbution In Latn Ameca", Human Resources DMslon, Techncal Depadment, November 1989 IDP-S4 'Ecuador Development lsues and 0pone for the Amazon Region", Country Department IV, December, 1989 DP-55 "Fiscal DefIcis, Ination and intere t Ratos", Rul Coutlnho, June 1989 DP-56 "Regulating Industdal Poflutfon In Developing Coun~des: Some Po~sible Framewor*s", Paul R. Poney (Consultant), December 1989 IDP-0 "Strategy for the Promoton of Non-Tradional Agricultural Expod~", Martin F. Raine, December 1989 IDP-61 "A Portfollo Model for Braz", Rul Counho, June 1989 IDP-64 "Why Tax Incenbies Do Not Promote ineetment In Brazil, A. Estache and V. Gaspar (consultant), February 1990 IDP-68 "Publc Sector "Debt Distrss" In Argenina's Recent Stablzation Eforis, Paul Beckemn, July 1990 IDP-69 "Affocatng Credit: How to D1et, Prie or Aucion", 8. Ramachandran, June 1990 DP-70 *LAC Ecoromic and Seator Work: A Cddial Review of Selected ~ssues", L Auernhelmer (consulant), June 1990 IDP-71 -Agricultural Trade in th Caribbean Community , John Nash, July 1990 IDP-72 *Poverty Allevlation in Braz, 1970-7, M. Louis Fox, July 1990 IDP-73 "F~scal Defcits, Exchange Rate Crisis and Infaton", Sweder van Wijnbergen, Januay 1990 July 1990

MEASURING PRICE DISTORTIONS FROM COMMODITY TAXATION IN BRAZIL

A. Estache The World Bank

A. Fernandez Instituto Tecnologico Autonomo de Mexico

I. Roy Universite de Montreal

* The findings, interpretations and conclusions are ours. They should not be attributed to either of the Institutions with which we are affiliated. Also, we are grateful to J. Mintz, J. Seade and J. Consentino Tavares for comments on an earlier version of this paper. The other usual disclaimers apply. ABSTRACT

The measurement of tax driven price distortions is based on the computation of effective tax rates imposed by the domestic indirect taxis on each sector. The accumulated effect of indirect on prices across sectors is tracked through a system of simultaneous equations provided by the 1980 input- output matrix. This accumulated effect of taxes on prices is the effective . It has two components. First, there is the effect on prices of statutory tax rates. Second, there is the effect on prices stemming from taxes levied on inputs by cascading sales taxes or by the use of exemptions rather than zero- rating to provide relief from VAT. This second effect is also referred to as the implicit tax effect. It illustrates the unexpected price distortions due to the reliance on indirect taxes other than VATs or on imperfectly designed VATs.

The methodology is then used to assess the effects of the 1989 changes imposed on the system by the New Constitution. The component of the new Brazilian Constitution improved the overall design of the indirect tax system by widening the base of one of its value added taxes (VATs) and suppressing some of the . According to simple computations based on the "Harberger triangle" formula, the reform reduces the welfare loss imposed by the indirect tax system by about 2.3%. However, the improvement in the design is insufficient as many intersectoral price distortions remain. The paper shows that the effective tax rates increase a little as a results of increase in statutory tax rates but that the implicit effect due to the unplanned taxation of inputs is somewhat lower but not significantly. The explanation for this weak effect of the reform is found in the failure to change the fundamental design of the two Braziliar VATs. In most cases, activities benefiting from tax incentives are excfpted from the VATs rather than zero rated and hence end up paying an implicit tax on their purchases of inputs. Table of Contents

Page No.

A. Introduction ...... 1

B. The Taxes ...... 2

1. Value-Added Taxes ...... 2 2. Other Significant Indirect Taxes ...... 3

C. The Importance of Tax Exemptions for Effective Tax Rates . . . . . 3

D. A Method to Measure Distortions Due to Indirect Taxation . . . . . 6

1. The Framework ...... 7 2. A Formalization of the Framework ...... 8

E. Intersectoral Tax Driven Distortions Before the New Constitution 14

1. The Value Added Taxes ...... 15 2. The Other Indirect Taxes ...... 18 3. Summing up the Individual Effects of Indirect Taxes . . . 20

F. Did the 1989 Reform Reduce Distortions? ...... 20

G. Concluding Remarks ...... 25

Bibliography ...... 27

Tables:

Table ls TOTAL IMPACT OF TAXES ON PRICES PER ACTIVITY IN 1988 . . . 15 Table 2: COMPARING TAX EFFECTS BEFORE AND AFTER THE NEW CONSTITUTION 22 Table 3: SECTORAL IMPACT OF INDIRECT TAXES ON PRICES ...... 25

Graphs

Graph 1: BRAZIL - EFFECTIVE AND STATUTORY DOMESTIC INDIRECT TAX RATES 16 Graph 2: BRAZIL - TAX EFFECTS BEFORE AND AFTER THE NEW CONSTITUTION . 23 Graph 3: BRAZIL - IMPACT OF INDIRECT TAXES ON PRICES - PRE- vs. POST-CONSTITUTION TAX SYSTEMS ...... 24 A. Introduction

Brazil's new Constitution, voted in 1988, includes significant changes in indirect taxes aiming at increasing the tax independence of subnational governments. The resulting redistribution of tax jurisdiction increases the proportion of economic activities subject to value added taxes (VATs) and reduces the number of taxes. This paper assesses the impact of these changes on intersectoral price distortions.

The quantification of the tax-driven distortions is based on a calculation of effective tax rates for final goods.11 The rates are defined as the tax element in the price of that good, accounting for taxes on inputs which arise through the production process. The effective rates are calculated using the 1980 input-output matrix for the current tax system and for the system prevailing after a full implementation of the major reform elements of the New Constitution. Even though the current production structure is not fully comparable to the 1980 structure, the results provide a useful quantitative approximation of the tax-driven distortions and raises some conceptual issues regarding the design of the Brazilian domestic indirect taxes.

The paper shows that the Brazilian indirect taxes were and remain, in spite of the changes brought by the new Constitution, a major source of intersectoral price distortions. The Brazilian VATs are not levied on the value added of the sectors subject to the tax only but also on many of the inputs of these sectors. This taxation of inputs transforms the VATs into cascading sales taxes, and implies that the effective tax rates on many goods and services are higher than the statutory VAT rates intended by the lawmakers. The widening of the state VAT base, resulting from the New Constitution, restores some of the credit mechanisms required to avoid the undesired cascading effects of the taxes and reduces somewhat the intersectoral distortions imposed by the Brazilian tax system. The reduction is however not very significant because the reform does not address some undesirable properties of the design of the Brazilian VATs.

The paper is organized as follows. Section 2 provides a description of the major indirect taxes. Section 3 provides an intuitive discussion of the price distortions due to the current design of the Brazilian indirect taxes. Section 4 presents the framework use to quantify the distortions. Section 5 assesses the Brazilian domestic indirect taxes both before and after the implementation of the tax component of the new Constitution.

1/ This approach is based on a methodology presented in Ahmad,E. and N. Stern, Effective Taxes and Tax Reform in India, Development Economics Research Center, University of Warwick, Discussion Paper No 25, 1983. For a recent detailed presentation of the approach see, Ahmad, E. and N. Stern, Taxation for Developing Countries, The Development Research Programme, LSE, DP 1, December 1986. - 2 -

B. The Taxes

Domestic indirect taxes can be divided into two groups: the VATs, which accounted for more than 65% of all indirect in 1988, and other taxes, essentially sector specific excise taxes and social contributions levied as sales taxes. Under the new Constitution, most excise taxes on specific industries are combined with the state VAT.

1. Value-Added Taxes

There are two primary VATs: the IPI (a tax on industrial products levied by the Federal Government), and the ICM (a tax on the circulation of goods levied by the states). Both VATs are gradually becoming consumption-based rather than production-based taxes, but remain hybrids.2/ Both are levied at multiple rates, and the IPI in particular has a highly complex rate structure.3/ The two taxes share some other important characteristics. The Brazilian VATs are rife with exemptions and special rules for the tax treatment of inputs. Almost all industries treated favorably in the tax code are exempted from VAT, except for industrial and production within the Zone (ZFM), which are zero-rated. Firms in industries which are exempted from VAT do not pay any direct VAT on their production; however, they are also not reimbursed for the taxes built into the cost of their inputs. Under zero-rating, in contrast, a firm receives credits for taxes paid on its inputs. This widespread use of exemptions rather than zero-rating results in liabilities imposed on inputs of the exempted sectors. In addition, the co-existence of the two taxes raises the question of of many activities.

The IPI is levied on industrial production only, so agriculture, minerals, and services are excluded from the tax base. The tax rates on different products and on different stages of production vary between 0% to over 3002. The highest tax rates are levied on tobacco and alcoholic beverages. Fiscal incentives offer important exemptions or tax reductions for some specific imports, especially high technology and capital goods; exports of all industrial products (zero-rated); the ZFM (zero-rated); and sectoral or regional fiscal incentives. As a result, the true size of the IPI tax base has shrunk, and receipts have declined from 4.4% of GDP in 1970 to 1.92 in 1988. These tax revenues are concentrated in several sectors, with tobacco, autos, and beverages providing about 60% of receipts.

The ICM is one of Brazil's most important sources of tax revenue, yielding around 5% of GDP. It is levied on sales of goods at all stages of prod- uction. Until 1989, its base only excluded products subject to specific taxes. Since 1989, these products have been integrated in the ICM base and only few

2/ A production-based VAT would tax factor incomes -- from wages, salaries, and capital incom.*. A consumption-based VAT taxes only the value added from the sale of consumer goods and services, while firms are allowed to deduct their purchases of capital goods.

3/ The IPI rate structure was somewhat modified on March 16, 1990 as part of an economic package presented by the New President. Many incentives have been suspended, rather than suppressed, and the rates have been increased. Most of the 4 -o rn4aii i n thin anar hnvpvar- rpmain. activities, mainly services, are left out of its base. Agriculture is treated uniquely, being generally unable to claim credit for taxes paid on inputs (with some exceptions). Generally, no credits are allowed for capital goods purchased. There are also exemptions, including some imports, fertilizers and pesticides and similar inputs for agriculture, and specific products--e.g., vegetables, eggs.

Until 1989, the standard rate for transactions within a state was 17%. The rate was somewhat lower for inter-state transactions (9% or 12Z depending on the destination). Following the new Constitution's reforms in 1989, intra-state rates are usually set at 12% for items included in the consumption basket of the poor, and 17% remains the standard rate or some luxury goods are now taxed at 25% in most states. ICM rates are actually levied on the price of the good including the tax--which implies a higher average effective tax rates than suggested by the statutory rate of 20.48Z.4/

2. Other Significant Indirect Taxes

Three other major sources of indirect taxation exist in Brazil: social contributions (FINSOCIAL and PIS/PASEP), a tax on services and specific taxes. FINSOCIAL is levied at 0.5% (in 1988) on monthly gross sales receipts of the sale of goods or services by all firms, to support social welfare programs. It is therefore in essence a cascading , and yields about 0.6% of GDP. PIS and PASEP are essentially forced savings programs for the private and public sectors respectively, producing around 0.5% of GDP. They are levied as both indirect and direct taxes. The indirect component is levied as a 0.75% tax on gross sales. The tax on services (ISS) is a tax imposed by municipalities on services not included in a federal tax base. While each municipality has its own list of exempt activities, tax coverage is in general quite complete. The tax rate is generally low, varying from 1% to 10%, averaging 5%. The tax yield is extremely low, approximately 0.3% of GDP.

Specific taxes were levied on transports, fuela, electricity, minerals and communications until the absorption of their base by the ICM in 1989. None of these taxes could be credited against other taxes and none ranted any significant exemptions. Tax rates differed across taxes and for any given taxes, rates could vary according to users or products.

C. The Importance of Tax Exemptions for Effective Tax Rates

This section demonstrates intuitively, through the use of a simple model, that the failure to maintain a smooth and full chain of credit transforms the VATs into cascading sales taxes. The model deals with a commodity i produced through the use of two inputs. One input is exempt from the VAT but may have

4/ The ICM is computed on the price inclusive of the ICM tax. Hence, for a commodity priced at 1 NCz and taxed at 17%, the actual producer price is (1-0.17), which is the true tax base. As tax is charged at 17% on 1 NCz, the real statutory tax rate is 20.48% (17Z(1-17Z)=20.48%). paid a VAT on its own inputs. The second input is subject to the VAT when acquired by its consumer. The model determines the effective tax rates on the output of the consuming sector and on each one of its inputs. For simplicity, the VAT is labelled t; it could be the IPI or the ICM. The model assumes competitive markets with constant returns to scale, so that prices are equal to average costs and taxes are fully shifted forward onto consumers.

The revenue (T from a VAT on a firm producing commodity i is calculated by subtractin the taxes paid on inputs (t*Ci) from the taxes paid on the gross sales revenue of the firm (t*Si)

Ti = to(Si) - t*(Ci) (1)

Now, assume that a firm i, subject to the VAT, purchases taxed inputs and exempted inputs, the value of its sales (Si is composed of the cost of its taxed inputs, including the values of the tax p id at the time of purchase of those inputs (Ci.* (1 + t)) plus the cost of its exempted inputs (Sm) minus the credits obtained on the taxes paid on its inputs:

S i= Ci*(1 + t) + Sm + VAi - t*Ci (2)

To identify the importance of breaking the credit chain for taxes paid on inputs, the composition of the sales value of the exempted inputs needs to be identified. The cost of the exempted input to the firm producing commodity i can be decomposed into: the costs of its own inputs which include the tax 5/ paid on those inputs (C* (1 + t)), and the value added of the sector (VA ):

S = Cmo(1 + t) + VAM (3)

Factoring out the tax rate in equation (1) and substituting equations (2) and (3) in equation (1) yields the following information:

Ti = t*(Si - Ci)

= t*(o *(1 + t) + S + VAj - toc - Ci)

= to(Sm + VAi)

= t*(Cmo(1 + t) + VA + VAi)

5/ This tax could be the same VAT levied on the final output, or any one of the excise taxes levied on some important inputs in Brazil. -5-

which can be rewritten ass

Ti = tO(VAm + VAj) + toCm + t2 CM (4)

Equation (4) illustrates the implications of the widespread use of exemptions of inputs in the Brazilian VATs. By levying a tax on the consuming industry i, the government is taxingt (i) the value added of that industry i (VAi); (ii) the value added of the exempted industry (VA ); (iii) the inputs of the exempted industry (Cm); and (iv) in the cases where the exempted industry itself uses taxed inputs, the tax levied on the inputs of the exempted industry as shown by the last term of equation (4). This last term would disappear if the inputs used in the production of the inputs of the final output were not taxed.

But the full importance of the price distortions built into the VAT emerges from a generalization of the previous result. Expression (4) can be generalized to the case where the exempted industry acquires a combination of exempted and taxed inputs, and allowance is made for differential rates in the taxation of inputs. In that case, expression (4) becomes:

Tf = (tf*Sm - Eti*Cfi) =

= (tfO(VAf + EVAe) + tf*( + ti)*(ECei) e i

+ (tf*(l + tj)O(ECfj) + ECfio(tf - ti) (5) j i where i : commodities taxed by the VAT f i final commodities taxed by the IPI e : exempted commodities j : commodities taxed by excise taxes Com : inputs m used in the production of good r

Equation (5) confirms and completes the results obtained from the two inputs simplification of equation (4). Equation (5) suggests that the effective tax rates imposed by VATs taxes are higher than the legal tax rates because:

(a) the VAT imposed on a final commodity is also imposed on the value added of the exempted inputs used by the industry, by virtue of the exemption from the VAT granted to that input;

(b) the VAT rate also applies on the inputs subject to excise taxes and to those excise taxes themselves; and -6-

(c) the inputg subject to a VAT rate lower than the VAT rate levied on the final commodity will be taxed at a rate equivalent to the difference between the two VATs.

These results also imply that the use of exemptions rather than zero rating leads to a bias against the use of inputs in the production of exempted inputs, and a bias against the use of exempted inputs themselves. In other words, exempting inputs hurts the sector producing them. Furthermore, the VATs are also levied on taxes paid at earlier stages, adding to the double taxation problem and to the resulting non-neutrality with respect to the use of inputs.

These conclusions are very significant for the IPI and the ICM since both taxes exclude sectors from their base through exemptions rather than zero rating. However, changes in the excise taxes levied on specific sectors or activities, such as the tax on services, can also have some important revenue and efficiency implication. Most of these taxes have or had their bases excluded from the bases of the IPI or the ICM. The mere exclusion of the bases of the specific taxes from the base of the IPI and the ICM is sufficient to transform the two VATs into sales turnover taxes.

An important corollary of these observations is that these distortions can have positive revenue effects for the government. By generating indirect revenue through the taxes built into exempted inputs, the distortions increase the tax revenue generated by the VATs. Hence, an additional impottant result follows from the above model: suppression of the distortions could lead to a decline in tax revenue. If tax rates remain the same, the intlusion of the base of the excise taxes in the VAT base and/or the suppression of the long list of exemptions for inputs would reduce distortions in relative prices at the cost of a lower revenue level for the government. For example, the replacement of exemptions by zero would lead to lower revenue by restoring the credit chain for taxes paid on inputs.

The results derived above provide a useful intuition regarding the sources of the problems in the design of the Brazilian indirect tax system. They do not allow, however, the quantification of the distortions in prices due the imperfect design of the tax system. In fact, the exact measure of the effects -- in terms of distortions and revenue -- of the taxation of all final commodities would require knowledge of the alternative uses of the exempted commodities and of the inputs taxed at lower rates. A full assessment would account for the general equilibrium effects tracked through the input-output matrix. This is the objective of the more general approach discussed below.

D. A Method to Measure Distortions Due to Indirect Taxation

This section presents a framework to quantify the impact on final prices of the interaction of the various taxes. Since the more exempt goods there are, the more likely it is that value added is taxed at different unintended rates, the calculation of the effect of taxation or. prices in a multi- commodity world requires general equilibrium modelling of the effects of taxes via inputs. The simplest intersectoral model for which data is available in - 7 -

Brazil is the input-output model. The interactions of the various taxes with the production structure are recognized explicitly.t/ The effects of the various taxes discussed in the previous chapter will be considered separately as well as jointly.

1. The Framework

The framework used is based on standard input-output analysis. The input-output model used here is a--fixed-coefficient--general equilibrium model reflecting the structure of the economy in 1980. It assumes, like the first model, competitive markets and constant returns to scale. Producer prices are equal to average costs, which depend on the returns to factors and on the prices on inputs. Input prices enter the price equation as a weighted average of the prices of all inputs used. The weights are the input-output technical coefficients which give the inputs required to produce a unit of the final output. If there were no indirect taxes, the producer prices would be equal to consumer prices. The introduction of indirect taxes in the model drives a wedge between the two prices for a given commodity. Because of the assumption of full forward shifting of the tax burden, the consumer price is larger then the producer price by precisely the amount of the indirect taxes paid by the sector producing the final commodity.

Essentially, the model calculates the effective tax rates for each sector in a way similar to that discussed in the previous section for a one input and output case, but it does of course approximate the interactions and complexities of the real world more fully. In particular, the model addresses both the cases where taxes on inputs are creditable and those where they are not because of their exclusion from the base or exemptions granted to their users. In both cases, the production equilibrium is reached when the producer price is equal to the average cost of production, but the difference lies in the determination of the average cost. When producers benefit from full credits, the average cost is a function of factor remuneration and a weighted average of input prices -- where the weights are the shares of the inputs in the total costs of the consuming sector -- without any tax element being built into the input price. On the other hand, when taxes on inputs are not fully creditable, the average cost depends on the same vector of factor remuneration but with a weighted average of input prices including a tax element.

Input price vectors are then derived as follows. To incorporate the Brazilian tax system explicitly into the model, a vector of tax rates for each tax is specified. The influence of those taxes on input prices is their weighted average, and the weights form a matrix summarizing the relevant features of the Brazilian tax system. For each sector, for each input required by that sector, and for each existing tax, the matrix indicates whether the sector gets a credit

6/ However, the role of the import taxes is ignored in this version of the model. This implies that the tax load computed for each sector is likely to be underestimated since most sectors use imported input subject to imports taxes, directly, through tariffs or indirectly through NTBs. In fact, the tax load computed is the domestic tax load. - 8 -

for taxes paid on inputs when it pays a tax on the purchase of an input. When the tax is creditable, the weight is shown to be equal to the share of the creditable input in the production cost of the consuming sector. Each tax vector required an adjustment discussed in Appendix I.

The total effect of each tax on the price for each sector will be decomposed into its two components. The first is the effect, determined by the legal or nominal tax rate for a given sector. The second effect is the indirect tax effect, representing the fully shifted forward implicit tax burden due to the exemptions, exclusions from the base, and all other price distortions introduced by the very complex regulation of the IPI. For each one of the taxes discussed below, the direct and indirect effects are measured in terms of both their effects on prices and their burden cn value- added. The burden on each sector is defined as the tax paid directly and indirectly by each sector as a proportion of its value-added.

Comparison of the total effect on prices with the direct tax effect is not only a measure of the distortion in prices: the results can also be used to assess the intersectoral differences in final prices due to any given tax. Even if a few sectors contribute most of the revenue for a tax, the effective tax rates and the implicit tax burden on all sectors could be more widely dispersed. A large spread of tax rates is not necessarily inefficient from an optimal taxation point of view, but the actual dispersion of tax rates across sectors does not seem to reflect efficiency considerations anymore than it reflects equity considerations. The revenue implications of the distortions can also be assessed from the results. Finally, the indirect effect is a good indicator of the incentive provided to industry to integrate vertically. Vertical integration, if complete, avoids the implicit tax burden imposed by those indirect effects.

2. A Formalization of the Framework

This section provides a mathematical presentation of the methodology. To track the effects of indirect taxes on relative sectoral prices through the input-output matrix, a system of equations is derived from the production equilibrium condition under perfect competition, assuming constant returns to factors:

Ppi = C (W , P p) (6)

where: Pi = producer price of good i

W* = vector of factor payments

P * = vector of input prices

Ci = production cost of good i - 9 -

In a world without indirect taxes, the producer price is equal to the consumer price:

Pci = Pp1 (7)

where: Pci = consumer prica of good i

When indirect taxes are accounted for, two additional elements must be factored in. First, the consumer price differs from the producer price by the amount of the tax:

Pci = PPi + Ti (8)

where: Ti = indirect taxes levied on good i

Second, for each tax analyzed, two cases need to be considered: (i) the sector taxed can credit the taxes paid on its purchases of inputs. In (ii) credits for taxes on inputs are not allowed. Case (ii) also corresponds to the situation of sectors exempted from the tax, rather than zero rated--or for activities not included in the base of the tax modelled. No tax is levied on their production but they cannot claim any credit for taxes paid on their purchases of inputs.

The previous considerations suggest that, for sectors entitled to credits, the production equilibrium can be expressed as

Ppi = C i (w *p)(9

which is the same as (6) while for the sectors not benefiting frGn credits for taxes paid on inputs, there is an implicit tax burden and the production equilibrium is expressed as:

Ppi = 1 ,P + Ti) (10) Differentiating the production equilibrium conditions, the following equations result:

PC = pi + Ti (11)

AA Ppi = Ej aij ppj (with credits allowed) (12)

PPi = E. aij (PPj + t.) (if no credits allowed) (13) .3 .3

where A denotes percentage changes

aij = share of input j in the total cost of good i; j, the number of activities varies from 1 to N

ti = tax levied on the consumption of good i - 10 -

In matrix notation, the previous system can be rewritten as

Pc PP + t (14) p

A P = AP + Al t (15) where:

all al2 ... aln a21 a22 ... a2n I I AI - I

I a a n1 n2 ... nn and

ja 11 al12 ... alln 1 1 1 a 21 al22 annj

1 1 1 aln1 aln2 ... ann

where:

al = 0 if sector i is entitled to a credit for taxes paid on the purchase of input j

alij = ai4 if sector i is not entitled to credits for taxes paid on the purchase of input j

The Brazilian tax structure was incorporated explicitly through the definition of a vector for each one of the taxes considered. This allowed the identification of the effects on prices of each one of the taxes individually. As a result, the price vector can be expressed as:

A A p = APp + A1 tIPI + A2 t1CM + A3 tIFFS + A4 tlSS + A5 tIE (16) - 15 -

Table 1: TOTAL IMPACT OF TAXES ON PRICES PER ACTIVITY IN 1988

SECTOR IPI ICM SOC.CON. ISS SPECIF. TOTAL Z 2 % Z % %

Agri./Livest./Fisheries 1.43 22.32 2.70 0.09 0.85 27.39 Minerals/Oil Extraction 1.43 17.88 2.34 0.19 4.55 26.41 Metals, Non-Metal & Steel 10.27 24.47 3.69 0.19 0.51 39.14 Machinery & Equipment 9.86 24.42 3.46 0.22 0.39 38.36 Automotive & Electronics 13.38 24.45 3.76 0.22 0.38 42.19 Timber & Paper 11.22 24.72 3.35 0.24 0.43 39.95 Chemistry & Derivatives 18.25 25.26 4.28 0.26 0.52 48.58 Textile, Clothing & Shoes 3.39 24.19 3.82 0.21 0.37 31.98 Food Supply 2.37 23.79 3.74 0.18 0.32 30.39 Other Industry 78.94 24.21 2.95 0.22 0.31 106.64 Electricity 1.24 0.82 0.47 0.15 49.74 52.41 Construction 2.96 1.51 0.94 6.94 0.45 12.82 Commercial Services 3.36 1.84 1.18 0.22 11.26 17.86 Transport & Communic. 2.21 2.16 1.20 0.17 7.12 12.86 Other Services 0.68 0.61 0.53 3.51 0.30 5.63

(1) Average Total 8.7 18.0 2.9 0.9 2.7 33.3 (2) Average Direct 5.5 15.0 1.0 0.4 1.9 23.8 (3) Average Indirect 3.2 3.0 1.9 0.5 0.8 9.5 (4) Weighted Average 4.26 9.54 1.74 2.16 2.41 20.11 (5) Standard Deviation 13.04 10.12 1.43 2.02 8.40 18.19

1. The Value Added Taxes

In 1988, the distortions in prices due the ICM are the largest in absolute terms with 182. The average direct impact on prices is 15% and the average indirect effect on prices is 3%. The activities most affected by the indirect effect are in industry. The less affected are in services. The average total impact of the IPI on prices was 8.7Z, the second most important in absolute terms. The average direct impact of the IPI on prices was 5.5%. The average indirect impact of the IPI on prices is 3.2Z--about 602 on the legal tax rate--with a standard deviation of 2.7. All sectors, without any exemption, are burdened by an indirect effect brought about by the imperfections of the tax crediting mechanisms. For no sector is the indirect impact 0%. It ranges between 0.07% (for communication equipment) and 5.96% for housing and food services. In terms of dispersion of effective tax rates, the IPI was ranked first however, reflecting mostly the extreme complexity of the . The high dispersion across sectors reflected the high tax rate levied on tobacco, perfumes, alcoholic beverages and automobiles--the standard deviation drops from 13 to 6.5 when these 4 sectors are not included in the average. Brazil - Effective and Statutory Domestic Indirect Tax Rates 1/

Agric./Liv./Fish. Minerais Metal/Non-Met./Steel Machin. & lquip. Auto & Electronic Timber/Paper Chemicals, deriv. Textiles & Shoes Food Other, Inc. alc/tab. Electricity Transp. & Communic. Construction Commercial Services Other Services r

Weighted Average

0 20 40 60 80 100 120 (percent)

ME Effective rate Statutory rate

1/ Effective minus statutory rate is the implicit tax rate due to imperfect design of domestic taxes. - 17 -

The same issues are raised by ICM and IPI exemptions, except that the intensity of the former is more limited. In short, thelexclusion from the ICM base of the sectors subject to the specific taxes hurts those sectors when they are used as inputs in the production of goods subject to the ICM. They impose an effective indirect tax burden unaccounted for by the ICM tax law, because specific taxes paid on inputs cannot be claimed as credits against the ICM tax liability. As a result and as discussed earlier, the effective tax rates on the products taxed by the specific taxes, such as services, transport, minerals, fuels and electricity differ widely by the user. The tax burden will depend on the relative share of taxed inputs and on whether the consumer is final or not, and will result in a fairly arbitrary and random set of effective tax rates.

Effective ICM tax rates are not dispersed across sectors as widely as IPI rates except for those sectors with a significant share of exempted inputs in total inputs. The exclusions from the base and exemptions are not as dramatic as in the case of the IPI, but there are still some significant exclusions, notably activities subject to the tax on services (ISS) or any of the specific taxes (on minerals, transport, fuels, electricity consumption, and communications) where the exclusion of some inputs from the ICM base breaks the chain of credits for taxes paid on inputs and makes the ICM more of a cascading sales tax than a VAT for some sectors.

The important sectors most affected by the price distortions are likely to be transport and construction, since they are subject to specific taxes while their inputs are subject to the ICM. In addition, the sectors which use the services of these sectors intensively are themselves affected by the heavy implicit burden imposed on those sectors which is probably shifted onto consumers to a large extent. From a revenue point of view, the relatively high tax burden imposed on those sectors may, however, not be excessive, given the fairly low demand elasticity with respect to prices of these sectors.

An important issue raised by the ICM and the IPI is the absence of credits for taxes built into the price of capital goods.7/ With a large number of exemptions--mostly for agricultural equipment--the ICM base approximates GNP, which is equivalent to the sum of wages, capital income (including rents), and the amortization of capital goods. From an economic efficiency point of view, this choice of base introduces an important distortion against capital, taxing both its income and its consumption during the production process. This favors the use of other inputs and discriminates against capital goods with shorter life spans. On the other hand, while a move from a production-based ICM to a consumption-based tax would restore efficiency, it would also produce a lower volume of revenue. The new base would be narrower.

The ICM treats agriculture uniquely. The problem comes from the fact that agriculture is considered the first link in the production chain. Producers pay taxes (ICM as well as others) as part of their inputs costs, but cannot claim

7/ However, this tax becomes a cost to be depreciated against the corporate during the economic life of the asset. - 18 -

any credits for them against their ICM liability. The magnitude of this distortion is however reduced by the zero-rating granted to the production of fertilizers, seeds, and insecticides.

In general, the tax treatment of agriculture is damaging. Its output is taxed at effective rates greater than those levied on the other economic activities, to the extent that it uces taxed inputs for which it receives no fiscal credit. The treatment of industrial machinery and equipment is especially discriminating against agricultural producers in relation to other producers. An additional distortion is created by the differential treatment of agricultural producers by region: the absence of credits for taxes paid on inputs is true only in parts of the country. Agricultural producers in the North, Northeast, and Center West regions benefit from true zero-rating, and thereby receive preferential tax treatment.

2. The Other Indirect Taxes

All the specific and excise taxes have been aggregated into a vector, so as to apply the methodology discussed earlier. This simplification does not alter the results. Nor should it hide any distortion, since most of the taxes are sector specific and do not cumulate on any given sector. The nominal tax rates can be tracked through the direct tax effects.

The average total price effect of the specific taxes was 2.7Z with a standard deviation of 8.3. The average indirect effect on prices of the specific and excise taxes is 0.8Z, with a standard deviation of 1.6. The direct effect is 1.8%, with a deviation of 6.8. These figures are quite low, but reflect the fact that many sectors are not directly affected by the specific and excise taxes. Only 5 sectors out of 42 are directly subject to any of the specific or excise taxes. Proportionally (looking at the ratio of indirect to direct effects), the distortions introduced by the specific taxes are much more significant than those identified for the IPI or the ICM. The highest indirect price effects were observed for communications, electricity and transports.

Most of the sectors subject to the specific taxes can be taxed twice or three times: once by the specific taxes, once by the ICM, and in some cases by the IPI also. Indeed, all the sectors subject to the specific taxes (electricity, minerals, fuels, transport, communications) are important inputs for many sectors subject to the ICM and the IPI. All these specific taxes are embodied in the price of these inputs. But most of these specific taxes are not creditable against the IPI or ICM liability. The only exemption is the IUM (tax on minerals), of which 90Z can be credited against the ICM and 1OZ against the IPI. Hence, the price distortion caused by the exclusion of minerals from the base of both VATs is less significant. For all the other excises and specific taxes, the formal exclusion of their base from the ICM and IPI bases does not mean that they are not also subject to them. On the contrary, it is precisely because of that exclusion that the output of the electricity, fuels, transport, and communications sectors is subject to both the specific tax and the IPI and/or the ICM, when the sector using them as an input is subject to either or both of the latter. - 19 -

An important corollary of this distortion is that all the Brazilian producers of these inputs lose competitiveness within the domestic market. The implicit tax burden imposes a price wedge between foreign and domestic prices missed when considering only the legal tax rates and the structure. The exact desincentive is even harder to assess, since the preferred protection instrument in Brazil is quantitative restriction, which also imposes an implicit tax burden. In sum, for comparable production costs, a domestic supplier of an input subject to the specific taxes and excluded from the VAT bases will be less competitive than any foreign supplier, a- long as the total level of protection (through tariff or quantitative restrictions) is not high enough to offset the price wedge imposed directly and indirectly by domestic taxes.

The conclusion drawn from the analysis of the specific taxes can be reiterated for the ISS. By design, the ISS is just one more tax on inputs for many activities. It also results in double taxation and loss of competitiveness for those activities included in its base when they are used as inputs by producers subject to the ICM and IPI, vis-&-vis substitutes not subject to the ISS. The tax is built into the price of these inputs, but it is not creditable against the IPI or ICM liability.

In practice, no sector is significantly affected by the indirect effect of the ISS on prices because of the ISS statutory rate is only 5Z. The average indirect price effect is 0.35%, with a standard deviation of 0.484. The total is 0.9% with a standard deviation of 2. The sectors most affected by the price distortions are the industrial sector in general and retail trade. The low average reflects the fact that only 5 of 42 categories of activities are subject to the 5% ISS tax.

Two of the social contributions are levied as indirect taxes. In concept, the social contributions are quite general -- everybody pays them, including agriculture. However, not all sectors pay them as an indirect tax. Essentially, service industries pay them as an income tax while other activities pay them as indirect taxes. When paid as an indirect tax, the legal rate is the same for all sectors, but the effective rates are bound to be different because the contributions are levied as cascading sales taxes on the receipts of a firm, with no allowance for taxes paid on any input. Hence, they lead to arbitrary relative price distortions, and provide an incentive for the vertical integration of economic activity.

The average direct price effect is 1.01% with a standard deviation of 0.6. The average indirect effect on prices is 1.8% with a same standard deviation of 0.9. The total effect is 2.9% with a deviation of 1.4. For this tax, the indirect effect is much larger than the direct tax effect. The indirect tax effect is mostly hurting the industrial sector. If the major users of agricultural output are indirectly taxed more highly than the average sector, the agricultural sector is only subject to low levels of indirect price effects--1.3. Services benefit from the same relatively favorable treatment, with the exception again of the construction and transport sectors. - 20 -

3. Summing up the Individual Effects of Indirect Taxes

The effects on prices of each taxes can be consolidated.8/ The averages are computed by weighing the tax rate levied on a sector--whether direct, indirect or total--by the value added of the sector. Hence the weights do not generally reflect the base since tax is levied on the whole value added of the economy. The results are also summarized in Table 1. It indicates that on average the total tax rate--accounting for all taxes--imposed on value added in Brazil is around 20%. Note that this does not really translate into equivalent revenue effect.

The dispersion around that mean is large--over 18. For every tax, that dispersion of tax rates increased as a result of the indirect effects. A similar computation for the legal tax rates suggests than the direct effect, is only 13.3Z on average. The difference of almost 7% is due to the implicit taxation of inputs due to imperfections in the design of Brazil's indirect tax system. In fact this indirect effect adds, on average, 50% to the legal tax rate. Finally, Table I highlights that the major sources of distortions are due to the design of the social contributions and the IPI. The ICM is not problem- free, but has an indirect effect relatively small in comparison to its direct effect.

To sum up, the quantitative analysis confirms the intuitive analysis of the distortions built into the 1988 Brazilian tax system provided by the simple analytical framework discussed above. The coexistence of two VATs (the IPI and the ICM) with specific input taxes ended up imposing a large unintended burden on the activities subject to those specific taxes. In addition to inducing inefficiencies in the allocation of resources among sectors, the distortions reduced the competitiveness of domestic inputs. Despite protection through tariff and non-tariff barriers, foreign inputs do not therefore necessarily carry as high a total tax burden as domestic substitutes. The importance of these effects varied significantly across sectors, according to production structure.

F. Did the 1989 Reform Reduce Distortions? 9/

The new Constitution merges the taxes on minerals, electricity, transport, and on communication services with the ICM. It maintains, at the municipal level alone, the tax on fuels with a reduced base. As a result of these changes, all the sectors presently taxed through specific taxes will no

8/ Recall that the results assume that all taxes are shifted forward through production costs to the last consumer of any given product, and reflect the 1987 legal tax structure.

91 These simulations were made before the recent decision by most states to maintain the standard ICM rate at 17% but to add a few rates for internal transactions. A fate of 25Z will be levied on luxury items. There is no agreement yet across states as to what should be considered a luxury item. Autos were originally included in the list, but the latest information suggests that some states have rejected the idea. In addition, a rate of 12% is likely to be levied on energy and minerals. - 21 -

longer be subject to double taxation, at least from the ICM. They will remain taxed indirectly by the IPI when they are used as inputs of exempted industries, as long as they are not included in the IPI base. The rate structure is very similar however. The states, which now can set their own rates, have decided to maintain the standard rate at 17% but allow a 12% rate for necessities and a 25% for luxuries. The major items included in the necessities are food and other agricultural items. For luxuries, the major items are those already highly taxed by the IPI. Those include tobacco, alcoholic beverages, weapons and ammunition for most states. Some states will also levy a 25% on cars and sports vehicles. Not all states will implement the rate differentiation to its full extent. Many will stick to the standard rate of 17%. Since the disaggregation level of the 1980 input-output matrix does not allow to differentiate adequately in the rate structure, all the results presented below for the effective tax rates after the implementation of the New Constitution will assume an ICM rate of 17%.

Some important cascading effects will remain. They will be due to the ISS and a tax on retail sales of fuels -- which represents only a part of the former IUCL. In addition, the new Constitution does not include any changes in the design of the IPI, FINSOCIAL and PIS/PASEP and these taxes were shown to carry a very high indirect tax burden. Table 2 summarizes the effects of the changes. Graph 2 allows a visual comparison of the changes.

The reform improves the overall design of the indirect tax system and hence should reduce the efficiency costs of the system fo a given tax rate structure.10/ However, its positive features only go part of the way in improving the system. On the one hand, it allows taxes paid on inputs currently subject to the specific taxes to be claimed as credits against ICM liability. This reduces a major source of distortion by restoring the chains of credits for taxes paid on inputs. As a result, the indirect effect is cut from 6.84% to 5.68% as reflected in Table 2. In addition, the overall dispersion of the tax rates declined as well. On the other hand, for many sectors formerly subject to a specific tax, the inclusion into the ICM also implies a higher legal tax rate and hence a higher direct tax effect, as refle,-ted in the increase in the direct effect for the ICM from 7.60% to 10.64% shown in Table 2. On average, the decline in the indirect tax effect will not be sufficient to offset the increase due to the direct effect and the total tax effect will end up beidg larger than under the 1988 tax system. The increases in effective tax rates are, however, concentrated in very few sectors. All the other sectors benefit from a due to the reduction in the indirect tax effects.

10/ The efficiency gains are not free: there will be a loss in revenue, since they imply lower effective tax rates on the sectors newly included in the ICM. This negative effect on revenue is likely to be offset by increases in nominal tax rates. - 22 -

Table 2: COMPARING TAX EFFECTS BEFORE AND AFTER THE NEW CONSTITUTION BEFORE AFTER WEIGHTED STANDARD WEIGHTED STANDARD AVERAGE DEVIATION AVERAGE DEVIATION

IPI Direct Effect 2.04 11.082 2.04 11.032 Total Effect 4.28 13.048 4.28 18.048 1CM Direct Effect 7.80 8.61 10.84 7.661 Total Effect 9.64 10.121 12.26 8.861 SOCIAL CONTRIBUTIONS Direct Effect 0.56 0.672 0.56 0.572 Total Effect 1.74 1.428 1.74 1.428 ISS Direct Effect 1.62 1.608 1.52 1.608 Total Effect 2.16 2.024 2.18 2.024 SPECIFIC TAXES Direct Effect 1.54 8.848 n.a. n.a. Total Effect 2.41 8.897 n.e. a.a. TOTAL (1) Direct Effect 18.27 14.988 14.78 14.168 2) Total Effect 20.11 18.190 20.44 17.287 (2) - (1) (Indirect Effect) 6.84 6.68

Table 3 and Graph 3 show the sectoral impact of the changes induced by the new Constitution. The effect is a decline in the average price for most sectors, even if the aggregate effect on prices is positive. The net increase is explained by a few sectors, especially transport and communications, and minerals. In most case, the reduction obtained by the others are minor, except for electricity.

Despite these important changes, significant distortions remain. First, as IPI statutory rates are not part of the reform. no significant decline is expected from current inter-sectoral differences in marginal IPI tax rates. The base remains narrow since the allocative distortions imposed by the ZFH have not been reduced. Also, the ICM treatment of agriculture is not improved. Neither VAT is closer to becoming a true consumption-based tax, and their credit mechanisms still discriminates against inputs not incorporated into the final output during the production process.

All in all, Brazil is better off with its reformed tax system, even if many design issues remainto be addressed. A ratio of the deadweight losses under the two systems, based on the "Harberger Triangle" formula, indicates that the tax reform reduced the deadweight lose by 2.3%. As revenue will improve.by about 1.6% as a result of the reform, the average deadweight burden of each unit of tax revenue collected is cut by 3.82. The modesty of the gain, however, confirms the conclusions drawn from the analysis of the changes in intersectoral price distortions. Brazil - Tax Effects Before The Brazil - Tax Effects After The New Constitution New Constitution

26 25

20 ------20---*-*-20*

16 - - 15 ------

01

10 - - - -.-.. 10-1

5 --.------5 - .

0 IPI 0 IOM sooea Is Speciflo TOTAL INI scm socil Contributions Taxes Contributions Ss TOTAL

Direct Effect Total Effect Direct Effect E Total Effect Impact of Indirect Taxes on Prices - Pre- vs. Post-Constitution Tax Systems

Agric./Liv./Fish. Minerals Metal/Non-Met./Steel Machin. & Equip. Auto & Electronic Timber/Paper Chemicals, deriv. Textiles & Shoes Food Other, inc. alc/tab. Electricity Transp. & Communic. Construction Commercial Services Other Services I 0 20 40 60 80 100 120

i 1988 System 2 Post-Constitution - 25 -

Table 3: SECTORAL IMPACT OF INDIRECT TAXES ON PRICES UNDER THE CURRENT AND POST-CONSTITUTION TAX SYSTEMS

SECTOR 1988 TOTAL DIFFERENCE SYSTEM POST-CONSTITUTION

Agr1culture/Livest./Fisheries 27.89 25.98 -1.41 Mi nerals/Ol Extraction 26.40 27.13 0.78 Metals, Non-Letal and Steel 89.14 87.83 -1.81 Machinery A Equipment 38.86 86.92 -1.44 Automotive A Electronics 42.19 40.80 -1.89 Timber & Paper 89.95 38.39 -1.56 Chemicals A Derivatives 48.88 46.84 -1.74 Textile, Clothing A Shoes 31.98 30.67 -1.81 Food Supply 30.89 29.14 -1.25 Other Industry 106.64 105.52 -1.12 Electricity 52.41 26.42 -26.99 Construction 12.86 28.34 18.48 Commercial Services 12.82 12.11 -0.71 Transport and communications 17.86 28.17 10.31 Other sevices 6.68 5.24 -0.39 (1) Average 3.26 82.09 -1.17 (2) Weighted Average 20.11 20.44 0.84 (3) Standard Deviation for (2) 18.19 17.24

G. Concluding Remarks

The results allowed to identify the major sources of unexpected price distortions and the major sectors affected by these distortions. The major sources are found in the overall design of the tax system. First, both the old and the ew systems allow for overlapping tax bases and hence double taxation through ndirect taxes. This overlapping is largely explained by the use of exemptions rather than zero-rating in the design of the VATs. Second, the IPI rate structure is unusually dispersed and as such is the major contributor to distortions in relative prices across sectors.

The distortions in relative prices introduced by taxes may be justified for some of the sectors--alcohol, tobacco, ...-- from an optimal taxation point of view. However, the major sectors affected by the price distortions also include many of the sectors producing capital goods. Indirect taxes are in fact the major source of distortion in the investment decision in Brazil and hence one of the factors contributing to slower growth.11/ This conclusion may seem counterintuitive as many of the sectors with larger than average price distortions are also major recipients of fiscal incentives. Once again the

11/ For additional results on this issue see: Estache, A. and V. Gaspar, Why Tax Incentives Don't Promote Investment in Brazil, Latin American and The Caribbean Region Internal Discussion Paper #64, February 1990. - 26 -

explanation is found in the design of the VATs. A common form of incentive in Brazil is to exempt from the IPI. This form of incentive can induce very significant additional implicit tax effects which hurt the sector instead of assisting it.

These concluding comments point the following direction for short term reforms:

- widen the IPI base to include all industrial activities currently exempted or excluded and grant true zero-rating to agricultural and services inputs into industrial production;

- allow full credits for taxes paid on all inputs into production, including capital goods to transform the VATs into true consumption based VATs

- simplify and reduce the spread of the IPI rate structure, maintain highly differentiated rates only for a few sectors--tobacco, alcohol, ...-- or replace the IPI on these sectors by an excise tax.

In the longer run, a single VAT covering all sectors of the economy would significantly improve the neutrality of the Brazilian indirect tax system, in particular with respect to the use of services inputs. - 27 -

BIBLIOGRAPHY

Atkinson, A. B. and Stiglitz, J.E., "The Structure of Indirect Taxation and Economic Efficiency', Journal of Public Economics, April 1972, pp. 97-119.

Ahmad, E. and N. Stern, Indirect Taxes and Prices in India: The Calculation of Effective Tax Rates using Input-Output Methods, Development Economics Research Centre, University of Warwick, Discussion Paper No. 14, June 1982

Ahmad,E. and N. Stern, Effective Taxes and Tax Reform in India, Development Economics Research Centre, University of Warwick, Discussion Paper No 25, 1983

Ahmad, E. and Stern, "Tax Reform in Pakistan: Overview and effective taxes for 1975/76, Pakistan Economic Review, Spring 1986

Ahmad, E. and N. Stern, Taxation for Developing Countries, The Development Research Programme, LSE, DP 1, December 1986

Ahmad, E. and Stern, "Alternative Sources of ; Illustration for India for 1979/80", in Newbery and Stern (eds), 1987

Estache, A. and V. Gaspar, Why Tax Incentives Don't Promote Investment in Brazil, Latin American and The Caribbean Region Internal Discussion Paper #64, February 1990

Fundagdo Instituto Brasileiro de Geografia e Estatistica, Diretoria de Pesquisas, Brasil - Novo Sistema de Contas Nacionais, Metodologia e Resultados Provis6rios, Ano-Base 1980, IBGE, December 1988

Harberger, A.C., Taxation and Welfare, The University of Chicago Press, 1974.

Mc Lure, C.E., The Value-Added Tax - Key to Deficit Reduction?, American Enterprise Institute, 1987

Newbery, D. and N. Stern (eds), The Theory of Taxation for Developing Countries, Oxford Universtity Press, 1987

Varsano, R. Estudos para a Reforma Tribut&ria - Tomo 3: TributagAo de Mercadorias e Servicos", Instituto de Planejamento Econ8mico e Social, Instituto de Pesquisa, Texto para Discussao Interna No. 106, February 1987

Varsano, R., DescrigAo do Sistema Brasileiro de Tributagio Indireta, mimeo, July 1988 - 28 -

Appendix 1 Page 1 of 3

CONSTRUCTION OF THE TAX VECTORS

This appendix discusses the modelling of the regulation of each tax in a way consistent with the framework used to assess them. For each tax, the major assumptions are discussed first. The description of the vector of tax rates applying to all the sectors is given next. When a given tax includes several types of tax bases, a matching number of tax vectors are defined. Each tax vector is multiplied by its proxy base, as given by the input-output matrix, tailored to reflect the true base. The matrix multiplying a given tax vector will only have non-zero coefficients for the commodities subject to the tax rates given in the tax vector by which it is multiplied. The tax yield of a given tax will be given by aweighted average of the tax vectors defined where the weights are given by the matrices approximating the true base.

The IPI. The dispersion of IPI rates and the disaggregation of the IPI base are is unusually large for a VAT. This is a source of difficulty when approximating the tax bases by the sectors defined in the input-output matrix. The matrix counts 43 sectors which is significantly lower than the number of commodities identified in the IPI regulation. Each sector defined in the matrix is subject to a large number of different IPI rates. It is thus necessary to calculate an average tax rate or a representative tax for each one of the sectors. When no commodity was clearly outstanding in the matrix sector, a simple average was calculated to obtain the legal rate to be used in the calculation of the effective tax rate for that sector. In the second case, the rate applying to the most representative commodity was selected. The final set of rates is given at the end of this appendix.

An additional adjustment was made to the IPI to account for the fact that the margin of commercialization of industrial products at the retail level is not subject to the IPI but is included in the proxy base used in the calculations. Using the commercialization margin and the final consumption value of production given in the input-output matrix, an average margin of commercialization of industrial production of 10.32Z was calculated. This allows to calculated the effective IPI rates as

t'ipi = tipi/(l+m) where: t'ipi is the effective tax rate on any commodity

tipi is the legal tax rate on the same commodity

m is the retail margin. - 29 -

Appendix 1 Page 2 of 3

The ICM. Since the ICM rates are inclusive, the relevant tax rates are somewhat higher than the rates quoted in the law. The law distinguishes between five rates: 0%, 9%, 12%, 13% and 17% according to rule explained in a previous chapter. When adjusting for the fact that they are inclusive of the base, the rates become: 0%, 9.89%, ... , 13.64% and 20.48% respectively. In most cases, the 20.48% was used. The exceptions were extraction of minerals where a rate of 15% was used and services which are exempted. The resulting vector is given at the end of this appendix. To limit the number of vectors built in the analysis the IOF was built in the ICM vector.

The IOF. For sectors producing goods, a rate of 1.35% was applied. Those same sectors and the financial services sector pay ai tax of 0.6%.

The ISS. A tax of 5% was used in almost all sectors providing services. Financial services (sector 116 are exempted). However, the taxation of services is subject to a very specific fiscal rule. That rule specifies that the providers of services that purchase themselves other services will only pay the tax of 5% on the net value of the services they are supplying to the market. The modelling of the rule required the calculation of an effective tax rate for the purchase of services by the sectors that also supply services. As a result, the model includes two vectors of taxes on services. The first one applies to those services not requiring the uses of other services as inputs. The second one applies to those sectors requiring the application of the special rule discussed above. The difference between the two vectors is that in the second one, the tax rates on services are weighted by the share of net value of production in total value of the services acquired as inputs by those providers of services using other services as inputs.

FINSOCIAL and PIS/PASEP. The social contributions are all lumped into one tax vectnr. The vector includes two parts. The first for one includes the 0.75% rate levied on the gross sales revenue of the firms selling the products (excluding any type of service). The second one is a rate of 0.6% levied on the total revenue of firms supplying goods and all services.

Specific Taxes. All the other taxes are lumped together in one tax vector building in their individual characteristics. The vector includes the tax on electricity consumption, the tax on transport, and the communications tax. The tax also includes a 13% export tax levied on non-industrial exports.

The Tax on Electricity. With a few minor exemptions business use of electricity is taxed at two different rates. There is a different rate for commercial and industrial use. Two tax vectors were used to distinguish between the two types of uses. The first vector gives a tax rate of 60%. The second one a tax rate of 16%. Private consumers are taxed at 50%. - 30 -

Appendix 1 Page 3 of 3

Other Taxes. For the various excise taxes not discussed so far, the procedure followed was the following. For the products derived from oil refinery, the tax rate of 29Z applying to gasoline was used. For communications, the rate of 20% applying to domestic phone services was used. For the transport sectors, the rate used was 5%. The same rate was also used for construction for those sectors which are not exempted. The exemption of the transport tax benefiting the transport of books, minerals and lubrifiers was modelled by using a zero coefficient in the matrix.