Poverty Monitoring, Measurement and Analysis (PMMA) Network

Incidence of Commodity Taxation on Income Distribution in

Pahan Prasada Sri Lanka

A paper presented during the 4th PEP Research Network General Meeting, June 13-17, 2005, Colombo, Sri Lanka. An Analysis of Incidence of Commodity Taxation on the Income Distribution in Sri Lanka

Pahan Prasada, Ranmuthumalie de Silva and Jeevika Weerahewa1

Abstract Commodity taxation constitutes an important component of the overall . In most developing countries where the formal base is not well established, taxation of goods and services via instruments play a significant role in sustaining the state revenue flow. The objective of this paper was to assess whether some commodity tax instruments were progressive or regressive on the income distribution in Sri Lanka, with reference to the year in which Sri Lanka Integrated Survey was conducted. Analysis was conducted using the food and non-food expenditure data extracted from the said household survey (1999/2000). The concept of welfare dominance was adopted and Lorenz and Tax Concentration curves were used as analytical tools. The empirical analysis was concentrated on total of 49 commodities where 23 commodities were used to analyze the incidence of GST, 34 commodities to analyze the incidence of NSL and 14 commodities to analyze the incidence of Import . Commodities were selected according to the frequency of usage. The results revealed that GST on 16 items which epitomize 24.02 percent total households expenditure was progressive while seven items which represent 4.43 percent total household expenditure was regressive, NSL on 11 items which denote 22.66 percent total expenditure was progressive whereas 22 items which account for 19.75 percent total expenditure was regressive and NSL on wheat displayed a special situation where the share of taxable consumption was exactly similar to the share of income. Import Tariff on six commodities which embody 10.69 percent total household expenditure was progressive and eight commodities which represent 14.43 percent total household expenditure was regressive. Wide provincial variation was visible among commodities when incidence of taxation was considered. These findings highlight the fact that targeting of indirect tax instruments with wide coverage could have a momentous impact on already significant inequality in Sri Lanka.

1 Paper to be presented at the Poverty and Economic Policy (PEP) meeting, Colombo, Sri Lanka, June 13-17, 2005. Pahan Prasada and Ranmuthumali de Silva are assistant lecturers and Jeevika Weerahewa is a senior lecturer at the Department of Agricultural Economics and Business Management, Faculty of Agriculture, University of Peradeniya Sri Lanka. Authors acknowledge the financial assistance provided by the International Development Research Center (IDRC) to carry out the study 1. Introduction plays a vital role in the budgets of most countries where it is the dominant source of revenue for financing public expenditure. This is a fact that needs no elaboration since the developing world is grappling continuously with the challenge of meeting increasing expenditure burdens, with budget deficits escalating each year and facing the grim picture of comparatively low tax revenue to GDP ratios. On average, the ratio is 0.18 for most developing countries while it is as high as 0.30 in the developed world (Gemmell and Morrissey, 2003). Indirect commodity taxation, without detailed reference to the consumption patterns and profiles of the respective income classes, is widely used to address this issue. Such tax instruments could prove detrimental in terms of equity considerations because inappropriate commodities and services are selected as potential sources of tax revenue.

A possible argument against indirect tax is that it contradicts the state’s own welfare aims. Governments of developing countries in general have numerous, elaborate, pro-poor transfer payment schemes that constitute a significant proportion of annual budgetary allocations, aiming to strengthen the livelihoods of the poor masses. Indirect tax covering items consumed by all income classes bear disproportionately on people with low income. In such an instance, the effect of indirect offsets the purchasing power offered through the transfer payments. This contradiction of its own interest by the state would, in fact, lead to a wastage of public money (by way of an unnecessary transaction cost) spent on implementing the transfer payments as well as tax collection.

Historically, Policy makers have not given their attention to the concept of redistribution when designing taxes. This may be because there are numerous subsidies and direct transfer payments which are conventionally identified as pro-poor instruments. Another issue that should be addressed is the prevalence of “implicit” forms of taxation in most developing countries. Price controls, punitive exchange rates and other consequences of economic liberalisation discriminate against the agricultural sector, (still the main source of employment for the rural poor) and tend to act as implicit forms of taxation, adding to the burden of indirect commodity taxes targeting daily consumption, irrespective of income class.

Impacts of fiscal interventions are usually examined quite critically in developed countries. Tax structure and in developing countries in general have received much less attention. Nevertheless, taxation is a theme that needs to be considered seriously since governments of developing countries always need to balance the need to raise revenue against the social burden of taxes on the poor. This balance is bound to be politically sensitive. The direct impacts of a tax manifest in terms of revenue, producer and consumer surpluses in a partial equilibrium setting while the indirect impacts could create pressures on the income distribution. The subject of income distribution is inextricably linked to that of taxation, because virtually every tax system changes or attempts to change the proportion of income enjoyed by different groups in society. It is common knowledge that when the demand for items subject to indirect tax is inelastic, retail prices tend to include the ultimate tax burden, which is passed on to the consumer. For essential goods and services, such price increases will be imposed on all consumers whatever their income, an apparent violation of the vertical equity principle in tax design.

Refaquat, 2003 attempts to determine the incidence of the said tax on the economy as a whole and on different individual commodities. The author points out that the individual commodity analysis has no bearing on the overall estimation and as such can be performed independently. According to the results obtained, the study concludes that the General (a form of indirect taxation) has been ‘progressive’ and ascribes the outcome to the particular selection of taxable items and exemptions. That is, the author believes under an alternative set of commodities the results could have been entirely different. Munoz and Cho, 2003 consider the case of replacing the sales tax with a value added tax (VAT) in Ethiopia and conclude that VAT is progressive contrary to the common presumption, reinforcing the findings by Sahn and Younger (2001) on eight other African countries using welfare dominance approach. Rajemison et.al, (2001) have made a novel attempt at determining the incidence of tax on intermediate goods. The technique used for the analysis is the input-output table and this study is unique in comparison to the general approach of focusing only on final goods and services. The case of Sri Lanka The contribution of tax revenue to the total government revenue in Sri Lanka was 86.32 percent in the year 2000. There were five major contributors to this revenue namely, , stamp duties, tax on Treasury bills held by Central Bank, taxes on goods and services and taxes on international . The year 2000 witnessed the implementation of several measures aimed at enhancing the revenue mobilization efforts. These efforts resulted in raising the revenue collected from domestic goods and services through Goods and Services Tax (GST), National Security Levy (NSL), Tax, License Fees and . The total contribution of indirect commodity taxes GST, NSL and import tariff to the total tax revenue was 55.59 percent, implying that Sri Lankan tax revenue has a higher dependency on commodity taxes. In accordance with the trend towards globalization and tax liberalization, a shift from import tariff to tax on domestic goods and services was visible and this was evident from the changes in the tax collection. The revenue from domestic goods and services as a percentage of total tax revenue rose from 36 percent in 1999 to 39 percent in 2000, while the share of international trade-oriented taxes as a percentage of total tax revenue declined to 41 percent in 2000 from 43 percent in 1999.

Tax structure of Sri Lanka is restructured yearly at the planning of the budget and even throughout the year via certain amendments. Therefore, it is important to investigate the impact of different tax instruments if we are to decide required amendments to the tax structure to be able to achieve a sustainable flow of income in an equitable manner, which is the ultimate objective of a tax instrument.

According to the recent literature, the distributional impact of tax instruments has been investigated in large number of developing countries in Asia and Africa but none of the researches have been done in the Sri Lankan context. Therefore, this paper focuses on the distributional impact of ‘National Security Levy’, ‘Goods and Services Tax’ and Import Tariff’ on fifty selected commodities. Though income tax contributed to 15.05% of total tax revenue in 2000 total registered taxpayers were only 0.02 % of the total population (Appendix 1). Therefore income tax does not have a significant effect on poor groups; rather the burden is borne by richer 0.02 % of the population.

The commodity taxes such as GST, NSL, and Import Tariff could have a significant effect on distributional impact since these are not specifically charged on a target groups as income tax. The basic principle behind investigating distributional impact is to find out whether a specific tax instrument on a target commodity is regressive or progressive. A tax is considered progressive if the share in the proceeds of the tax that emanates from poor households is smaller than their share in income or consumption of the total population. It is termed regressive if the share if the share of tax precedes coming from poor households is larger than their share in income or consumption and the incidence is considered proportional if the share of tax payments across all households is equivalent to their share of income or consumption.

The objective of this paper is to assess the incidence of different commodity tax instruments focusing on its progressivity and regressivity using concentration curves as the analytical tool.

The paper is organised in the following sequence. The introduction looks at the overview of the subject of commodity taxation with reference to the developing countries while the latter part of the introduction details the nature of the tax instruments implemented in the year 2000. A conceptual background behind the theme of analysis and few theoretical definitions of the analytical tools are given in the next two sections. The following section discusses the issues related to data and the tools used in the analysis. Fourth section details the outputs derived in the concentration curve analysis. Results and the discussion of the findings are included in the fifth section. The final section comprises the discussion on policy implications, some limitations faced and possibilities for further research as separate components. The annexure, which contain the figurative outputs of the ‘DAD’ analysis, are ordered in the following sequence. The first one lists the regressive and progressive instances of GST while the second annexure incorporates the regressive and progressive cases of NSL. Next, the provincial variations for NSL and GST are documented followed by the figures, which show regressive and progressive outcomes for Import Tariff. The fifth annexure incorporates the frequency of usage of items, which were used to select commodities for the analysis.

Overview of the tax structure in Sri Lanka Tax revenue is the major contributor of government earnings, which are the basis of different welfare related instruments, an important consideration for all developing countries. During the year 2000, 86.28 percent of government revenue of Sri Lanka was collected through taxes. As illustrated in the Table-1. Income tax, Goods and Services Tax (GST), Excise Tax, National Security Levy (NSL) and import tariff play significant roles in generating government tax revenue.

Table 1: Economic classification of government tax revenue by components Item Tax collection Rs. Percentage of total million tax revenue Income tax 27437 15.05% Stamp 8163 4.48% Tax on treasury bills held by central - - bank Tax on goods and services 122802 67.33% Turnover tax 1708 0.94% Goods and Services Tax(GST) 43893 24.07% Excise tax 42655 23.39% National Security Levy(NSL) 33539 18.39% Licence fee 1007 0.55%

Tax on international trade 23970 13.14% Import 23970 13.14% - - Total 182392 100% (Source: Central Bank Annual Report, 2000)

Economic classification of tax revenue

13% 1% 15% 4% 18% 1%

25% 23%

income tax turnover tax Goods and Services Tax Excise Tax National Security Levy Licence Fees tax on imports

Figure 1: Economic classification of tax revenue

Tax collection Department of Inland Revenue is responsible for majority of tax collection as clearly mentioned in their mission statement. ‘To collect the taxes under the law by encouraging voluntary compliance and by deterring and , and to enhance public confidence in the integrity and related legislation fairly, uniformly and courteously and thereby facilitate and foster a beneficial tax culture’ During the year 2000, 52.54% of total tax revenue was collected by the Department of Inland Revenue but whether they are moving towards the direction guided by the mission statement is questionable. Decision-making has been highly politicized and no research has been conducted to investigate the effectiveness of these decisions. Further many cases of tax evasion, tax avoidance, unfairness of tax collection decisions and ineffectiveness of tax collection have been reported. Therefore it is worthwhile to investigate the effectiveness of these tax instruments if they are to move to towards their mission by restructuring Sri Lankan tax structure by overcoming inefficiencies.

Commodity taxation in Sri Lanka Commodity taxation referred to taxes charged on supplies made in Sri Lanka. These are indirect taxes and tax collected by registered sellers and responsible for returning to the Inland Revenue. Goods and Services Tax (GST), National Security Levy (NSL) and Import tariff bear a higher share of tax revenue.

Tax instruments The sectoral composition of revenue collected from each tax instrument varies as illustrated in following tables. Goods and Services Tax (GST): GST is a tax on consumer expenditure. It is collected on business transactions and imports. Most business transactions involve supplies of goods and services. GST is payable if they are: Supplies made in Sri Lanka; By a taxable person; In the course or furtherance of business; and Are not specifically exempted, zero-rated or excluded wholesale or retail supplies of goods

A taxable person is and individual, firm, company etc. which is or is required to be registered for GST. A person makes taxable supplies above certain value limit are required to be registered. Supplies are outside the scope of the tax if they are; Made by someone who is not a taxable person Made outside Sri Lanka or Not in furtherance of business.

Taxable supplies and different GST rates All goods and services, which are GST- rated at either 12.5% or 0%, are known as taxable supplies. There are two rates of GST as mentioned above and the total value of supplies, which comes under these two categories, creates the taxable turnover. This excludes National Security Levy. Exempt supplies are business supplies, which have no GST, charged on them at either the standard or zero rate and the value of these supplies are not included into suppliers’ taxable turnover unless such supplies are exported.

12.5% - most of supplies of manufactured goods and services.

Zero rating Goods - The goods, which are exported to a customer outside Sri Lanka, are zero rated provided the appropriate conditions are met.

Services – Services are zero rated where the supply is directly connected with: The international transportation (including transshipment) of goods and passengers; Any moveable or immovable property outside Sri Lanka Any repair of any foreign ship or aircraft, refurbishment of marine cargo containers or any other goods imported for the purpose of re-export; A copyright, patent, license, trademark or similar intellectual property right to the extent that such rights is for use outside Sri Lanka.

Exemptions The exemption schedule to Goods and Services Act, No 34 of 1996 was reproduced in 1998. The meaning of the wordings in the exemption schedule is as follows.

Supply - means only the supply is exempted from GST. Import – means only the imports are exempted from GST. Supply or import – means either supply or import is exempted from GST. Import and supply - means both import and supply are exempted from GST

Exemption schedule i. The supply or import of coconuts, desiccated coconuts, coconut poonac, tea, green leaf, paddy, cardamoms, cinnamon, cloves, nutmeg, pepper or any other unprocessed produce

of any agricultural, horticultural, forestry, animal husbandry, poultry or fishing undertaking; ii. The supply or import of potatoes, onions and chilies or vegetable seeds; iii. The supply or import of rice, rice flour, wheat, wheat flour or any other grains; iv. The supply or import of bread of any description; v. The supply or import of milk or powdered or condensed milk excluding any article manufactured or produced from milk. vi. The supply or import of sugar, jaggery and sakkara vii. The supply or import of kerosene, diesel, and liquid petroleum gas and petrol viii. The supply or import of dried fish and maldive fish.

There are three main contributors of revenue collected from Goods and Services Tax namely, manufacturing, non-manufacturing and imports.

Table 2: Economic classification of revenue collected from GST Item Tax collection Percentage Rs. million Manufacturing 11378 25.93% Non-manufacturing 13643 31.08% Imports 18872 42.99% (Source: Central Bank Annual Report, 2000)

Economic classification of revenue collected from GST

Manufacturing Non-manufacturing Imports

Figure 2: Economic classification of revenue collected from GST

National Security Levy (NSL): National Security Levy (NSL), which was implemented in 1995 and was in force till 2001, to fund the civil strife in Sri Lanka2. The NSL was the single largest revenue provider for military expenses (contributing up to 55% on average), during the years of its implementation. Military expenditure was approximately 5.6% of the national GDP (Central Bank, 1991-2001). The rates at which NSL was levied were as follows: (i) two percent of the value of all imported capital goods (with exemptions for imports used in the manufacture of ; items imported for temporary use and repairs; goods imported for sale at duty-free shops), and (ii) four and a half percent of the value all other items imported and manufactured within the country (Sri Lanka Tax Guide and Inland Revenue Department reports, 1999/2000). The above rates of taxation were operational for all economic activities in the period. Almost all the items which came under NSL regulation increased retail prices. The exceptions were some domestic products, where the burden of the tax was fully borne by the producer, since a price increase would render their items uncompetitive in the open market.

Income Tax: This is classified as a Instrument since the income is directly taxed. However, the collection of revenue from this Tax instrument id not significant due to the lack of information and lapses in administrative structure in the country. Income tax consists of three main sectors namely, personal, corporate and Save the Nation Contribution. Table 3: Economic classification of revenue collected from Income Tax Item Tax collection Percentage Rs. million Personal 10820 39.44 Corporate 15757 57.43 Save the Nation Contributions 860 3.13 (Source: Central Bank Annual Report, 2000)

2 The conflict in the North and East in Sri Lanka began in 1983 and has been having devastating social and economic impacts on the economy ever since. The strife has developed significantly over the years, creating a potential threat to national integrity. The causes are largely socio-political but the conflict has given birth to a host of socioeconomic consequences; loss of human life, a huge financial burden on the government budget, destruction of public property, poverty and inequality impacts both in the areas directly affected by the fighting and in the rest of the country. Many national and international initiatives have been tried to bring lasting solutions to the on-going conflict albeit with little tangible success. As of now, the civil strife is temporarily under check owing to a foreign mediated ceasefire.

Import tariff: Import tariff is a tax charged on commodities imported to Sri Lanka. Despite a growth in imports by over 20 percent in 2000, revenue from import duties declined by 14 percent. This reduction was due to the removal of crude oil imports, the temporary duty waiver afforded to refined petroleum imports, the partial duty waiver on essential food items and rationalization of the tariff structure as pointed out in the Central Bank Annual Report. The Budget 2000 reduced the three band structure (10, 20 and 35) to two bands (10 and 25) with a special duty rate (protective tariff) of 35 percent provided for selected agricultural commodities.

The tax rates and the percentage imported for selected commodities used to analyze the tax incidence are illustrated in the table 4.

Table 4: Import tariff rate and the imported percentage of selected commodities Item Percentage imported Paddy (samba+kekulu) 35% 35 Sugar Rs 3.50/= per kg 89 Dried fish Sprats dried not salted Free Other 10% 69 Onion 35% 61 Wheat Free (with effect from 29 100 February 2000) Masoor dhal 5% 100 Potatoes 35% (in addition to import duty a 71 surcharge of 35% was applied for the period of 30th August to 8th December 2000) Dried fruit 25% 100 Tinned fish 10% 100 Milk powder 10% 81 Milk foods (tinned, 25% 5 condensed milk, other milk foods)

Fats 25% 47 Cigarettes 100% 100 (Source; Food Balance Sheet, 2000)

2. Conceptual Background of Tax Incidence Studies The distributional impacts of a tax can be analysed in several ways; taking the household as the unit of analysis (considering the impacts on income and expenditure) or else considering individual or entities (corporate/ non-corporate) as the unit of analysis. Also, the effects can be analysed in a temporal sense as pre-tax and post-tax on any given unit of analysis. Spatial impacts also matter in terms of empirical validity, since such analysis can provide useful evidence for regional policy planning. The spatial element has additional value when one clearly perceives geographical variations in the consumer response to fiscal interventions which affect commodity prices. The dynamic/static nature of tax impacts, (where the secondary adjustments in response to the tax instrument would differ from direct responses that are gauged in a static sense) could add another dimension to analysis of the effects of a tax.

The above discussion suggests that, depending on the context of the analysis, any tax study could be very complex. Nevertheless, even the simplest tax analysis (i.e. static effects at one point of time) is capable of providing valuable input to policy formulation. According to Morrissey and Gemmell, 2003 the most popular measure of incidence used in earlier studies is average tax progression. Later, there had been extensive use of other methods; namely, concentration curve and inequality measurement method, marginal social cost method, computable general equilibrium methods etc. Most analyses of indirect tax incidence are concerned with the share of taxes paid by different groups. The only data necessary are (i) a variable that defines the groups, and (ii) an estimate of the taxes paid by each group, where one considers “taxes paid” to be the loss in real income. The most common source of these data is a nationally representative household survey, or a household income and expenditure survey, as published by national statistical agencies. Usually, the groups are defined by welfare levels – poor vs. non-poor or each quintile of the welfare distribution – so one needs a variable that ranks people by welfare. Most studies choose to rank by household expenditures.

Estimates of the taxes paid by each group are more difficult, and it is helpful to explain first what to do once such an estimate is obtained. The simplest sort of comparison notes that group 1 pays so much of tax‘t’, group 2 so much, etc. For the most common case, where one wants to group people by welfare status, the groups might be poor and non-poor, or people in each decile of the welfare distribution. It is easy to make a much more useful comparison based on the theory of welfare dominance, which also offers the advantage of involving individual agents rather than groups of agents.

The theory of welfare dominance provides general criteria that allow us to conclude that one distribution of welfare is better than another for broad classes of social welfare functions (Saposnik, 1981; Shorrocks, 1983; Foster and Shorrocks, 1988; Slemrod and Yitzhaki, 1987; Lambert, 1993). One particular application of this theory is particularly useful for tax incidence analysis. Shorrocks (1983) shows that if the generalised Lorenz curve for one distribution of welfare is everywhere above the generalised Lorenz curve for another, then the first distribution is preferable to the second under any social welfare function that is (1) increasing in the welfare variable, (2) anonymous3, and (3) equality-preferring. Since the Lorenz curves tend to be quite close together, even for major tax changes, concentration curves are easier to see and work with.

Another key concept in tax incidence terminology is tax progressivity. This concept can be clearly defined, but measurement can be complicated. A tax is considered progressive if the average tax rate rises with increasing income (i.e. if the marginal tax rate is above the average tax rate); proportional when the average tax rate is constant; regressive when the average tax rate falls with increasing income (i.e. if the marginal tax rate is below the average tax rate)

To understand the context the following issues warrant attention: • Tax analysis often employs choice of expenditure as a proxy for income. Income data is rarely available in general household surveys; whenever they are available the reliability is questionable. Use of expenditure data has certain theoretical advantages since expenditure values are better indicators of lifetime/permanent income. Existing theories of household consumption behaviour –life cycle hypothesis and permanent income

3 Anonymity means that the welfare function does not take account of where each person in the ordering is, or whether a person changes position from one ordering to another. Equality preference means that if we generate a distribution by taking an existing distribution and transferring a small amount of welfare from a better-off individual and transfer it to a worse-off one, the new distribution is preferred to the old. The potential of this comparison to be applied generally makes it quite attractive. To use it, we need only construct two generalised Lorenz curves, one pre- tax and one post-tax, and check to see if one is clearly above another. hypothesis- argue that expenditure better represents permanent income, thus acts as a better proxy for long-term welfare. • Unit of Analysis concerns the choice between individual and household as the basic observational unit. Since expenditure is assessed for different households, the household becomes the unit of analysis. Per adult equivalent is another unit of analysis, which is preferred over the individual because of its ability to capture household size. (Demeri, 2000) • Average and marginal measures, as two parameters, have their own advantages but marginal measures (as structural tools of analysis) are more useful for forecasting purposes.

Even under highly regimes, it is likely that poor people would be faced with high implicit tax rates on their wages. This is another important behavioral consideration in tax incidence studies.

3. Tools of Analysis- Definitions

3.1 Concentration Curve

The concentration curve is an important normative and descriptive tool, used in evaluating the impact of tax and transfer policies. It can capture the horizontal and vertical equity concepts related to tax impacts on social welfare.

The concentration curve is defined as, 1 p C = T(q)d(q) T( p) ∫ ------(1) µT 0

1 µ = Q ( p)d(q) = µ − µ Where T ∫ t X N is average ‘taxes’ across the population. Since 0 population size is normalized to 1. CT(p) shows the proportion of total taxes paid by the p bottom proportion of the population, and X and N signify the gross and net incomes respectively.

In general use, concentration curves are usually estimated by ordering a finite number n of sample observations (X1; N1)…. (Xn; Nn) in increasing values of gross incomes, such that X1 < X2 ………< Xn, with percentiles pi = I / n; where i =1…..n.

3.2 Concentration Co-efficient

Concentration Co-efficient is calculated by aggregating the distance between p and the concentration curves C(p) . This is important to compute aggregate indices of progressivity and vertical equity.

3.3 Lorenz Curve The Lorenz curve is one of the most popular graphical tools for illustrating and comparing income inequality. It provides complete information on the whole distribution of income relative to the mean, and therefore gives a more comprehensive description of the relative standards of living than any one of the traditional summary statistics of dispersion pertaining to income distribution. The Lorenz curve has the advantage of being able to establish orderings of distributions in terms of inequality.

The Lorenz curve is defined as follows:

p 1 L = Q(q)dq / Q(q)dq ( p) ∫ ∫ ------(2) 0 0

1 p = ∫ Q ( q ) dq µ 0

p The numerator ∫ Q()q dq sums the incomes of the bottom p: proportion (the poorest 100p%) 0 of the population. The denominator sums the incomes of all. Since population size is normalized to 1, the denominator gives average income µ. L(p) thus indicates the cumulative percentage of total income held by a cumulative proportion p of the population, when individuals are ordered in increasing values of their income. For instance, if L(0:5) = 0:3, then we know that the 50% poorest individuals hold 30% of the total income in a population.

3.4 Indicator for measuring the degree of burden on poor groups An indicator is used to calculate the burden of each tax instrument on poorest 20 percent of the population and it is calculated as follow.

⎛ Shareof taxborneby poorest ⎞ ⎜ ⎟ ⎜ 20% as a percentageof total tax ⎟ ⎜ collection ⎟ Taxburdenon poorest 20 percent of the population =⎜ ⎟ *100 ⎜ Shareof incomereceived bypoorest ⎟ ⎜ ⎟ ⎜ 20% as a percentageof total income ⎟ ⎜ ⎟ ⎝ earned by the population ⎠

If the indicator is higher than 100 it revealed the fact that tax instrument on a particular commodity is regressive and poorest 20 percent of the population is worse off whereas if it is less than 100 the tax instrument is progressive and poorest 20 percent is better off. The value of it shows the degree of progressivity or regressivity.

3.5 Household expenditure share The household expenditure on each commodity as a share of total household expenditure was used to analyze the burden of taxes in depth by integrating household expenditure in to the picture.

⎛ Averageexpenditureon a commodity by a household ⎞ Household expenditure share = ⎜ ⎟ *100 ⎝ Total averageexpenditureby a household ⎠

4. Data and Methodology The sample frame was based on secondary data from the Sri Lanka Integrated Survey carried out in 1999/ 2000. The income and expenditure data were collected from 7,500 households from all the provinces in the country covering 30,581 individuals. Section 6 of the SLIS questionnaire collected expenditure (money value of purchases, home production and in-kind receipts) information on 62 items, and quantity information on 57 items of food. The data were classified on a spatial basis and the survey was formulated as a two-stage stratified design, with district and sector classification. The primary sampling unit was the Grama Niladhari (GN) division and the secondary sampling unit was the household.

The analysis was based on taxed consumption of different commodities by the sampled population. The aim was to assess the burden of a tax on a given commodity on the different income classes of the sample population as shown by the concentration curves.

The total survey sample frame of 7,500 households was used for the plotting of concentration curves and Lorenz curves and cardinal measures of inequality were made for the total data set. The distributive analysis has been conducted using ‘DAD’ software for evaluating the coefficients of concentration and plotting respective curves of concentration for different commodities. The Lorenz curve was used for total household expenditure, providing a foundation for inferences on the concentration curves.

From the information provided in the survey, expenditure data was specifically selected and total household expenditure was used, as an indicator of income with the aim of minimizing possible unreliability that could have occurred if income data had been used. The expenditure of each household had been bestowed for each commodity and there were 62 food items and 32 non- food items. Data related to the province in which the household live was also obtained from the above data set.

For the purpose of calculating taxable consumption, tax rates and bases were obtained from custom tariff guide, GST guide and NSL guide for the year 2000 and Food Balance sheet (2000) was used to stumble on percentage contribution of imports.

To analyze the tax incidence commodities were selected based on the frequency of usage and tax exemptions. Finally 34 commodities to calculate incidence of NSL, 23 commodities to calculate incidence of GST and 14 commodities to calculate incidence of import tariff was selected which represent both food items and non-food items. (Annex 5). Electricity and Liquid Petroleum (LP) were selected irrespective of its’ frequency of usage as these were considered as important commodities to be analyzed.

Guideline for making inferences The following figure provides a guideline for interpretation of the graphical results of the concentration curve analysis. The 45-degree line is the line of perfect equity, i.e. if every household had identical income or expenditure this line would represent the distribution of the variable across the population. Any disparity in the distribution of income/expenditure would result in the cumulative income/distribution curve caving downwards from the 45-degree line. The curve titled “cumulative expenditures” represents the cumulative expenditure curve as is the case in this particular study for the 7,500 households in the Sri Lanka Integrated Survey 1999/2000. This serves as the benchmark for comparing different concentration curves and determining the level of regressivity or progressivity of the tax. The cumulative expenditures curve is represented by the Lorenz curve for the total expenditures in this study.

Cumulative share of taxed consumption

Regressive 45 degree line tax

Cumulative expenditures

Progressive tax

Cumulative share of households, from poorest to richest

Figure 3- Progressivity and Regressivity Compared with the Lorenz Curve

If the concentration curve is above the Lorenz curve and below the 45-degree curve, the tax instrument is classified as regressive, i.e. the impact of taxed consumption of the particular item concerned is concentrated more on the lower income classes. Alternatively, if the concentration curve for the commodity falls below the Lorenz Curve, this indicates that a larger component of the taxed consumption falls on the higher income classes. If the concentration curve crosses the Lorenz curve, the share of the tax burden borne by both rich and poor groups is either higher than their share of income or lower. The foremost difference of a curve in this nature compared to curves describe above is that the incidence of a tax instrument is same for both groups.

Calculation of taxable consumption Taxable consumption was calculated from the data related to the total consumption of the selected commodities given in the SLIS data set. Post tax prices were calculated as follow with the aim of deriving taxable consumption.

Post tax price of commodities except sugar If pre-tax price is Rs100/=,

Pr iceafter charging import tariff = Im port tariff rate +100 = Y

(before NSL and GST)

Post − tax price = (Y * NSL) + (Y *GST) +Y

Post tax price of sugar If pre-tax price is Rs 100/=,

⎛ 100 ⎞ Pr ice after charging import tariff = ⎜ ⎟ *3.5+100 = Z ⎝ priceof sugar ⎠ (before NSL and GST )

Post − tax price = (Z *NSL) + (Z *GST ) + Z

Average price of sugar prevailed during the year 2000, was Rs 31.34/= per kg.

Commodities are classified into five categories as fresh and dried food commodities, fish and meat food commodities, processed food commodities, stimulants and non-food items..

Table 5: Post tax price of fresh and dried food commodities Commodity NSL GST Import Post tax price tariff (if pre-tax price=100) Beans 4.5% Exm 104.5 Brinjals 4.5% Exm 104.5 Coconut 4.5% Exm 104.5 Dhal (Mysoor) 4.5% 12.5% 5% 122.85 Dhal(Kadala) 12.5% 112.5 Dried fruit 12.5% 25% 140.625 Fresh chillies 4.5% Exm 104.5 Jak 4.5% Exm 104.5 Ladies fingers 4.5% Exm 104.5 Limes 4.5% Exm 104.5 Onion (Bombay) 4.5% Exm 35% 141.075 Onion (Red) 4.5% Exm 35% 141.075 Potatoes 4.5% Exm 35% 141.075 Rice (Kekulu) 4.5% Exm 35% 141.075 Rice (Samba) 4.5% Exm 35% 141.075 Tomatoes 4.5% Exm 104.5

Table 6: Post tax price of fish and meat food commodities Commodity NSL GST Import Post tax price tariff (if pre-tax price=100) Canned and 12.5% 10% 123.75 preserved fish Chicken 4.5% Exm 104.5 Dried fish (like 4.5% Exm 10% 114.95 katta) Dried fish (like 4.5% Exm 104.5 Sprats) Eggs 4.5% Exm 104.5 Fresh large fish 4.5% Exm 104.5 (like Mora) Fresh small fish 4.5% Exm 104.5 (like Salaya)

Table 7: Post tax price of processed food commodities NSL GST Import Post tax price tariff (if pre-tax Commodity price=100) Bakery products 12.5% 112.5 Biscuits, Cakes, 4.5% 12.5% 117 Other confectionary and Jam Bread 4.5% Exm 104.5 Cereal foods in 12.5% 112.5 tins and packets Cereal 12.5% 112.5 preparations Coconut oil 4.5% 12.5% 117 Fats (e.g. 12.5% 25% 140.625 Margarine etc) Foods and drinks 12.5% 112.5 purchased and consumed outside Hoppers 12.5% 112.5 Milk foods 12.5% 25% 140.625 Milk powder 4.5% Exm 104.5 Non alcoholic 12.5% 112.5 beverages except tea Salt 4.5% 12.5% 117 String hoppers 4.5% 12.5% 117 Sugar 4.5% Exm Rs 3.5/=per 111.17 kg Tea 4.5% Exm 104.5 Wheat flour 4.5% Exm Free 104.5

Table 8: Post tax price of stimulants Commodity NSL GST Import Post tax price tariff (if pre-tax price=100) Arrack 12.5% 112.5 Beedi * 12.5% 112.5 Beetle(for 12.5% 112.5 chewing) Cigarettes 4.5% 12.5% 100% 234

Table 9: Post tax price of non-food items Commodity NSL GST Import Post tax price tariff (if pre-tax price=100) Electricity 4.5% 12.5% 117 LP gas 4.5% Exm 104.5 Personal care 4.5% 12.5% 117 items Textile 4.5% 12.5% 117

* - Beedi is a local low cost substitute for cigarettes Exm - Exempted under GST

Taxable consumption for each commodity was calculated by using following equations. ⎛ household expenditure ⎞ ⎜ ⎟ ⎜ on a particular com modity ⎟ Tax charg ed on a com modity (Taxableconsumption) = *tax rate ⎜ Post tax price(if pre − tax ⎟ ⎜ ⎟ ⎜ ⎟ ⎝ price is Rs100 / =) ⎠

Im port tariff on sugar =consumption(kg) *3.5

When calculating monthly taxable consumption for each household following adjustments were done with the aim of minimizing the complex city. Though there was a surcharge on potatoes, since the duration on which this was enforced was very short and this was not considered for the analysis. To calculate incidence of import tariff both Bombay onion and Red onion were summed and same was applied for rice as well, where kekulu and samba were summed. The figure for textile expenditure was derived by summing all the items related to this in the SLIS data set.

5. Results and Discussion Concentration Curve based Analysis of Commodity Tax Instruments The progressivity and regressivity of three tax instrument NSL, GST and Import tariff on different commodities are given in following tables. When the type of the commodity is considered most of the commodities which shows a regressivity are commodities which are generally classified under basic commodities where as progressivity is visible among commodities which are consumed by higher income groups. Table 10: Incidence of tax instruments on fresh and dried food commodities Progressive(P)/Regressive(R) Coeffici Share of GST NSL Import ent of exp. Commodity tariff concent from

ration total exp. Beans R 137.5% 0.208 0.602 Brinjals R 112.5% 0.250 0.479 Coconut R 137.5% 0.2 1.389 Dhal (Mysoor) R 137.5% R 150% R 125% 0.209 0.875 Dhal(Kadala) P 37.5% 0.494 0.715 Dried fruit P 32.5% P 33.75% 0.586 0.922 Fresh chillies R 112.5% 0.229 0.354 Jak fruit R INC 0.274 0.153 Ladies fingers R 150% 0.192 0.380 Lime P 50% 0.392 0.271 Onion (Bombay) R 175% R 137.5% 0.137 0.563 Onion (Red) R 106.25% R 137.5% 0.225 0.569 Potatoes R 105% R 106.25% 0.264 0.821 Rice (kekulu) R 125% R 107.5% 0.217 4.410 Rice (Samba) P 62.5% R 107.5% 0.328 4.534 Tomatoes R 112.5% 0.253 0.488 Table 11: Incidence of tax instruments on stimulants Progressive(P)/Regressive(R) Coeffici Share of GST NSL Import ent of exp. Commodity tariff concent From

ration total exp. Arrack P 38% 0.494 4.818 Beedi * R 103.75% 0.154 0.932 Beetle (for chewing) R 150% 0.162 1.274 Cigarettes P 47.5% P 50% P 50% 0.378 2.727

Table 12: Incidence of tax instruments on processed food commodities Progressive(P)/Regressive(R) Coeffici Share of GST NSL Import ent of exp. Commodity tariff concent From

ration total exp. Bakery products P 31.25% 0.399 1.023 except bread Biscuits, Cakes, P 33.75% P 25% 0.247 1.529 Other confectionary and Jam Bread R 162.5% 0.166 1.255 Cereal foods in tins P 50% 0.459 1.188 and packets Cereal preparations P 62.5% 0.491 0.832 Coconut oil R 125% R 162.5% 0.192 0.981 Fats (e.g. Margarine P 25% P 27.5% 0.585 0.813 etc) Foods and drinks P 46.25% 0.466 2.305 purchased and consumed outside Hoppers P 50% 0.456 1.012 Milk foods P 60% P 64% 0.608 1.590 Milk powder P 75% P 67.5% 0.321 3.134 Non alcoholic P 12.5% 0.698 1.883 beverages except tea Salt R 187.5% R 200% 0.092 0.225 String hoppers P 32.5% P 43.75% 0.516 0.868 Sugar R 137.5% R 187.5% 0.204 1.424 Tea R 175% 0.135 0.832 Wheat flour S 100% 0.282 1.814

Table 13: Incidence of tax instruments on non-food items Progressive(P)/Regressive(R) Coeffici Share of GST NSL Import ent of exp. Commodity tariff concent from

ration total exp. Electricity P 100% P 100% 0.318 0.291 LP gas P 25% 0.467 0.335 Personal care items R 125% R 112.5% 0.254 0.134 Textile R 137.5% R 137.5% 0.209 0.010

Table 14: Incidence of tax instruments on fish and meat food commodities

Progressive(P)/Regressive(R) Coeffici Share of GST NSL Import ent of exp. Commodity tariff concent From

ration total exp. Canned and P 31.5% P 71.25% 0.365 1.502 preserved fish Chicken P 25% 0.260 4.104 Dried fish (like katta) R 137.5% R 137.5% 0.236 1.238 Dried fish (like R 156% 0.122 0.966 Sprats) Eggs P 62.5% 0.368 0.869 Fresh large fish (like P 43.75% 0.425 2.846 Mora) Fresh small fish (like R 100% 0.237 1.740 Salaya)

* Beedi is a local low cost substitute for cigarettes. Special – Lorenz curve overlaps with the concentration curve P - Progressive exp. - Expenditure R - Regressive S - Special INC - inconclusive When a tax instrument on a particular commodity is progressive it implies the fact that the burden of tax shared by rich groups in the population is higher than that of poor groups whereas if it is regressive share of tax burden borne by poor groups is higher than rich groups. The indicator given elucidates the degree of progressivity or regressivity. For example the indicator for NSL on beans is 137.5 percent and it divulges the fact that the share of tax borne by poorest 20 percent of the population is 137.5 percent of their share of income. Though NSL on Fresh small fish (like Salaya) is regressive and both NSL and GST on electricity is progressive the share of tax borne by the poorest 20 percent of the population is exactly similar to their share of income (indicator is 100%).

Though the above indicator specifically concentrated on poorest 20 percent of the population concentration coefficient articulates the degree of regressivity or progressivity by taking the population as a whole.

According to the results, it is clear that progressivity or regressivity of a particular tax instrument depends on the commodity, thus selection of commodities becomes a crucial issue in the design of taxes.

NSL on 86 percent fresh and dried food items which represent 72.58 percent total households expenditure on this group and processed food items such as coconut oil, sugar, salt, tea etc is regressive. These important commodities are essential and poor groups are highly affected by these. This passing of burden on poor groups obliterates the concept of equity and does not serve the purpose of taxation.

There is a clear distinction between incidence of taxation between NSL and GST because only 30.43 percent (7 out of 23) commodities considered under GST which includes only 4.43 percent total household expenditure are regressive whereas 67.647 percent (23 out of 34) items which denote 19.75 percent total household expenditure are regressive in the case of NSL. This discrepancy is due to the exemptions provided under GST. Of the commodities on which NSL is regressive 86 percent come under the exemption list of GST.

Both NSL and Import Tariff on Milk powder are progressive and under GST this commodity is exempted. Though Rice (samba), chicken, eggs, LP gas and lime are exempted in GST, NSL on these commodities is progressive. These findings divulged that, though the presence of an exemption list is important there is a room to introduce GST on certain commodities with out getting poor groups affected.

All the three tax instruments on Cigarettes display progressivity while both NSL and GST on Electricity appear progressive. These findings reveal that there is a room for continuation of taxes on commodities which are progressive.

A special observation is visible in the case of wheat flour. Lorenz curve of expenditure on wheat flour overlaps with the concentration curve of taxable consumption of wheat flour for NSL which shows that the share of NSL borne by different income groups are exactly similar to their share of income.

Of the import tariff on all commodities analyzed 43 percent (6 items out of 14 items) of the instances which constitute 14.43 percent of total household expenditure show regressivity. Import tariff on all the agricultural commodities demonstrates regressivity, which results in transferring a higher burden on poor groups thus the effectiveness of import tariff is questionable.

Provincial variation Though the above discussion illustrates the picture of incidence of tax instruments by considering Sri Lanka as a whole, provincial variations were discernible both according to the tax instrument and commodities.

When impact of NSL on milk powder is considered in the Western province, richer groups bear a higher burden of tax than poor groups but at the coordinate (0.95, 0.85) concentration curve crosses the Lorenz curve and that implies the fact that the share of tax borne by richest 5 % is lower than the share of their income though the term progressivity expects them to bear a higher share.

Same scenario is visible in other provinces other than North western province but the coordinates at which Lorenz curve crosses the concentration curve vary from one province to the other which shows that the tax borne by richest 20 % in Central province, 10 % in Southern province, 30 % in North east province, 55 % in North central province, 30 % in Uva province and 35 % in Sabaragamuwa province are lower than the share of their income. The situation at North western province is different to rest, where 13% to 85% of income groups bear a higher share of taxes compared to their share of income while other income groups basically bear a tax which is proportionate to their income.

Such provincial variations are visible when consider about the incidence of taxes. NSL on coconut is regressive for all the provinces but the degree of regressivity varies according to the province. The order of regressivity according to the degree of it is as North western, Uva, Southern, North central, Western, Central, North-east and Sabaragamuwa where Sabaragamuwa is having the lowest regressivity. Similar type of variation is visible in NSL on sugar and the order of regressivity according to the degree of it is as Western, Uva, North central, Southern, North east, Central, North western where Western is having the highest regressivity.

Though the concentration curve for NSL on wheat overlaps with the Lorenz curve for all inland data provincial variations are visible. NSL on wheat is regressive in the Western province whereas it is progressive in the Central province and rest of the provinces does not show a clear progressivity or regressivity. Instead the concentration curve crosses the Lorenz curve at different points as described NSL on milk powder.

GST on canned and preserved fish is progressive in Southern, North east and North western provinces and GST on cigarettes is progressive in North central, Uva and Sabaragamuva where as a clear progressivity or regressivity is not visible when consider about other provinces.

GST on salt is regressive in all the provinces where as GST on electricity is progressive in Western, Central, Uva and Sabaragamuwa provinces and regressive in Southern province.

The progressivity of a tax is not all that matters when one is attempting to evaluate the impact of it. Economic efficiency, administrative convenience, and the potential for raising a reasonable amount of revenue are important considerations. Progressivity is an issue that will remain uppermost in the minds of policymakers because of its socio-political ramifications.

An interesting feature associated with individual commodity analysis as against the overall incidence analysis is that such results help to determine the probable incidence of a particular good or service that could be brought under taxation in future reforms, and alternatively to distinguish items which should be exempted from taxation on equity grounds. Such disaggregated analysis would be especially useful when commodity consumption patterns are dependent on socioeconomic and other demographic variables that cannot be incorporated in a broad-based nationwide analytical framework. For example, if a hypothetical commodity A is being consumed by several communities in the estate sector, incorporation of the tax incidence on this commodity would not create any additional information for a national perspective. The incidence data on this particular commodity as a single item would inform policy advice on commodity taxation.

6. Conclusions, limitations and suggestions for further research Findings of the study could be summarized as follows, • Of the instruments on commodities 96 percent of them were imposed on basic essential goods where as progressivity was visible among commodities which were exceedingly consumed by higher income groups. • Irrespective of the tax instruments regressivity and progressivity was basically dependent upon the type of commodity on which a particular tax was imposed. In this regard list of exemptions matters on determining progressivity or regressivity nature of a tax instrument. • NSL on wheat flour was the only situation where the share of tax borne by different income groups was exactly proportionate to the share of income received by them.

From the above findings, following conclusive remarks can be made.

According to results it is clear that GST has a better scope than NSL. This has been achieved by them by including an exemption list where it is lacking in NSL. Therefore it can be concluded that exemption schedule is important when considered about commodity taxes and this can be used to eliminate taxes on commodities, which are regressive while introducing taxes on commodities, which demonstrate a progressivity.

Though the idea of import tariff is to protect local produces in addition to obtaining revenue findings revealed that this results in reducing the welfare of poor groups. Therefore it is worthwhile to investigate whether the gain acquired through this tariff by the government and the local produces can compensate the loss faced by poor consumer groups.

The whole country situation is not a common scenario of all the provinces and provincial variations are visible. Welfare dominance based on the partial equilibrium framework provides only partial answers to the problem of estimating the burden of taxes and the imposition of such a burden on different welfare groups. Some of the conspicuous inadequacies in the approach can be listed as follows: • The inability to account for behavioural responses to tax implementations over a long period. This is critical from a practical point of view. It is in the nature of the average consumer to shift to viable available substitutes away from commodities whose prices have increased because of taxes. These issues can only be incorporated if time series data on consumer expenditure patterns are available as panel data sets. • The second issue is the case of intermediate goods, when certain items could be used as inputs for other final outputs. The best example is gasoline, which could be taxed both as a final good and an intermediate input for public transport, another significant expenditure item in the daily budget of the lower income classes. • The third limitation is the restrictions implicitly imposed by existing market conditions, since the assumptions of perfectly competitive conditions are rarely fulfilled in actuality. In non-competitive market environments taxes imposed on commodities do not abide by the simple shifting assumptions made in the partial equilibrium framework.

The study suffers from its dependence on partial equilibrium assumptions. These provide realistic foundations for import-based items. They are not adequate when a certain item has domestic origins with significant local value addition. In such a situation a price increase to incorporate a tax burden will make the item uncompetitive compared to imported substitutes.

The choice of the appropriate time duration, during which the income or the consumption should be gauged in order to estimate the incidence of tax burden poses an additional challenge to the tax analyst. In other words, it is about whether one should rank the households based on their annual income or based on their lifetime income. The lifetime income will obviously show lesser variability compared to that of the annual or monthly income (or expenditure). This has an important implication given the fact that the annual household consumption is more related to the lifetime income than the income in a particular year. Furthermore, lifetime income incorporates two considerations which the annual income concept overlooks; the inclusion of common patterns of cash inflow, asset accumulation and consumption and the reduction of variability in income due to unemployment or changes in family status. This fact also reinforces the propriety of selecting expenditure as a proxy for income in tax analysis.

In any empirical study the sample frame itself is an important factor. In the Sri Lanka integrated survey, the sample was meticulously designed to represent the total population of the country. Therefore, the inferences, statistical or otherwise, would be descriptive of the total population. Fiscal implications (like tax incidence) need not always be carried out at national level. Identification of different populations based on different objectives e.g. communities, ethnicities, geographical areas could be quite revealing, especially when the rest of the country (areas other than the north east) is not uniform in its socioeconomic parameters.

Another area for potential research is spending incidence analysis, a separate discipline by itself, where one could look at who will ultimately benefit economically from military expenditure on account of civil war. Going a little further, the two fiscal aspects, that of tax incidence and that of spending incidence could be integrated to provide what is referred to as budgetary incidence analysis, where all stakeholders involved in the issue are captured in an economic perspective.

Expenditure incidence could follow the same line of methodology as that adopted here. There are studies of this nature available in the tax incidence literature. Concentration curves could be used as a tool to produce the graphical outputs, helping to compare different expenditure instruments. Another potential addition to this kind of welfare dominance studies is the statistical tests conducted to validate the inferences obtained by concentration curves. This is especially important when the graphical outputs are inconclusive.

Results based on the welfare dominance approach utilised in this study are by no means conclusive. Such results could be qualified further and their reliability could be augmented by the judicious use of alternative methodologies. There are other ways of pursuing empirical and theoretical evaluation of tax reforms, with reference to the differential tax incidence; the situation when one tax instrument is substituted with another so as to maintain the revenue at its original. References Central Bank of Sri Lanka, 1999-2000. Annual Report. Colombo: Central Bank

Demeri, L., 2000, “Benefit Incidence: A Practitioner’s Guide”, Poverty and Social Development Group, Africa Region, The World Bank.

Department of Inland Revenue of Sri Lanka, 2000 .Administration Report of the Commissioner General of Inland Revenue. Colombo. Department of Inland Revenue

Department of Inland Revenue of Sri Lanka. Goods and Services Tax (GST): Registration Guide. Colombo. Department of Inland Revenue.

Department of Inland Revenue of Sri Lanka, 1998. GST Guide. Colombo. . Department of Inland Revenue

Food and Agricultural Organization, 2000. Food Balance Sheet. On line available at www. FAO.org.

Foster, J.E., Shorrocks, A.F., 1988, “Poverty Orderings”, Econometrica 56, pp.173-77.

Gemmell, N. and Morrissey, O., 2003, “Tax Structure and Incidence on the Poor in Developing Countries”, Nottingham: Nottingham University Press.

Lambert, P., 1993, “The Distribution and Redistribution of Income: A Mathematical Analysis”, 2nd Edition, Manchester: Manchester University Press.

Ministry of Finance and Planning of Sri Lanka. The Gazette of the Democratic Socialist Republic of Sri Lanka: Notifications. Colombo. Ministry of Finance and Planning

Munoz, S. and Cho, S., 2003, “Social Impact of Tax Reform- The Case of Ethiopia”, Washington: International Monetary Fund.

Refaquat, S., 2003, “Social Incidence of the General Sales Tax in Pakistan”, IMF Working Paper, 2003-16

Sahn, D.E. and Younger, S.D., 2001, “Estimating the Incidence of Indirect Taxes in Developing Countries” , World Bank -Poverty and Social Impact Analysis(PSIA) Division

Saposnik, R., 1981, “Rank-Dominance in Income Distributions”, Public Choice, 36, pp.147- 151.

Shorrocks, A.F., 1983, “Ranking Income Distributions”, Economica, 50, pp.3-17.

Slemrod, J. and Yitzhaki, S., 1987, “Welfare Dominance- An Application To Commodity Taxation”, National Bureau of Economic Research, working paper