Journal Issue
Share
Article

Income Distribution and Tax and Government Social Spending Policies in Developing Countries

Author(s):
Ke-young Chu, Hamid Davoodi, and Sanjeev Gupta
Published Date:
March 2000
Share
  • ShareShare
Show Summary Details

I. Introduction

For many developing countries, widespread poverty and the small income share of the poor have been a source of particular concern. Increasing international economic interdependence, uncertainties arising from this interdependence, and some evidence of widening income disparities in recent decades have further heightened the economic profession’s interest in income distribution, its changes, and the underlying factors.

The evidence of widening income disparities also has heightened the economic profession’s interest in the role of fiscal policy as a redistributive instrument in the short run and in the long run, as well as in the progressivity (or the lack) of tax and transfer policies. At the same time, questions have been raised on the effectiveness of tax and transfer policies as a redistributive tool.

Assigning a more activist role to tax and transfer policies in developing (and transition) countries often gives rise to many challenges. The interaction between the progressivity of tax and transfer policies and income distribution is complicated by a set of factors that are unique to developing (and transition) countries, which, in general, have a number of weak fiscal features: a low tax-to-GDP ratio, reflecting poor governance, weak tax administration, and widespread tax evasion; the predominance of indirect taxes and a limited menu of capital and wealth taxes; and a limited role of formal cash transfer and social protection policies.2 These features cast doubt on the ability of tax policy to redistribute income.3

Noting that the poor rarely pay income taxes in many developing countries and that education and health spending and other in-kind transfers account for a large share of the budget, many have argued that the expenditure side of the budget should be a primary redistributive tool (Tanzi, 1974 and 1998; Harberger, 1998). The major contribution of tax policy as a redistributive instrument should be to raise the revenues needed to finance efficient pro-poor and other essential government expenditures, and to avoid generating horizontal inequities. To this end, tax systems should have broad bases, limited exemptions, and low rates. This view has formed a basis for an agreement among many researchers and policymakers on the relative role of tax and expenditure policies in income redistribution in developing (and transition) countries.

As regards income distribution and the role of fiscal and other economic policies in developing (and transition) countries, a number of questions arise: How do these countries differ in income distribution—both among themselves and from industrial countries? Is income distribution in these countries becoming more unequal? What is the redistributive role of tax, transfer, and other expenditure policies?

Unfortunately, the limited availability of high-quality data for developing countries prevents a rigorous analysis of these questions. While numerous studies have focused on income distribution in individual countries and, to some extent, on international comparison of income distribution, scarce high-quality data have limited studies of long-term changes in, and international comparison of, income distribution in developing countries.

This paper provides an overview of the changes in income distribution in developing (and transition) countries in recent decades, and assesses the incidence of taxes and government expenditures in these countries. For the overview of income distribution, this paper relies on a set of newly available “high-quality” income distribution data.4 For the assessment of tax and government expenditure incidence, it relies on existing incidence studies on individual countries.

This paper is organized as follows. Section II provides an overview of the changes in income distribution in developing countries in recent decades. The sample countries include several transition countries. Section III provides a survey of the studies on the incidence of taxes and expenditures, paying a particular attention to the incidence of government spending on education and health, reviewing the available evidence for a large number of developing countries. Section III also offers an analysis of the changes in income distribution in developing countries from the 1970s through the 1990s and discusses possible underlying factors, including the role of tax policy. Section IV concludes the paper.

II. Overview of the Changes in income distribution

A. Income Distribution in the 1990s

As indicated later in this paper, the analysis of income distribution and distributional implications of taxes and government spending is subject to many conceptual and practical difficulties. Should income or consumption be used? How should the benefit of government spending be valued?

The difficulties are particularly severe for developing (and transition) countries. Income distribution data for these countries are expanding, but still are not adequate. Internationally comparable data on before-tax and after-tax household incomes for same countries are virtually nonexistent. The available data do not allow an assessment of the long-term evolution of income distribution for a large number of developing countries.

Table 1 is based on the high-quality data compiled by Deininger and Squire (1996). Estimates of before-tax and after-tax Gini coefficients and quintile income shares in the 1990s are available only for 20 developing (and transition) countries. In general, these estimates are based on household incomes including government cash transfers. For a considerable number of countries, however, the estimates are based on consumption data. Unlike the income data used for tax or expenditure incidence studies in individual countries, the Deininger and Squire data have been compiled on the basis of a common methodology and are intertemporally and internationally comparable. For the 9 “b” countries, estimates are based on before-tax (but after-transfer) incomes; for the 13 “a” countries, estimates are based on after-tax (and after transfer) incomes (or consumption). However, even these data do not include both before-tax and after-tax household incomes, ruling out the possibility of using the data for explicitly assessing the role of tax and transfer policies in the changes in income distribution; for only two countries (Poland and Romania) are estimates available on the basis of both before-tax and after-tax incomes, but for different years.

Table 1.Twenty Developing (and Transition) Countries: Income Distribution, 1990s 1/
GiniAdj.

Gini
SharesRatio

Q5/Q1
Per Capita

GNP$

1992
Income

Level
Q1Q5
(In percent)
Before tax (“b”) countries
Median343064261,585
Mean383474483,228
Czechoslovakia *252111363eq2,450
Romania*29259374eq1,130
Yugoslavia *32287395eq
Poland*33296396eq1,910
Bulgaria*343074261,330
China*38346427470li
Hong Kong, China454154910ueq15,360hi
Colombia514745415ueq1,330
Thailand524845916ueq1,840
After tax (“a”) countries
Median3274151,035
Mean3484261,198
Romania *259354eq1,130
Poland*298384eq1,910
Sri Lanka309394540li
Taiwan317395hi
Pakistan318405420li
India329415310li
Hungary *3273962,970
Indonesia339425670li
Mauritius3774362,700
Jamaica3874571,340
Jordan416487ueq1,120
Nigeria417487ueq320li
Peru4555010ueq950
Sources: Deininger and Squire (1996); International Monetary Fund, International Financial Statistics Government Finance Statistics, and World Bank’s World Development Report, 1994.

denotes transition countries.

eq = relatively equal income distribution (after-tax Gini smaller than 30); Adj. Gini = “b” Gini - 4.ueq = relatively unequal income distribution (after-tax Gini greater than 40).hi = high-income country; li = low-income country (per capita GNP less than $670).Q1 = poorest quintile; Q5 = richest quintile

Each entry in the table corresponds to the latest available observation in the 1990s for each country; Discrepancies are due to rounding errors.

Sources: Deininger and Squire (1996); International Monetary Fund, International Financial Statistics Government Finance Statistics, and World Bank’s World Development Report, 1994.

denotes transition countries.

eq = relatively equal income distribution (after-tax Gini smaller than 30); Adj. Gini = “b” Gini - 4.ueq = relatively unequal income distribution (after-tax Gini greater than 40).hi = high-income country; li = low-income country (per capita GNP less than $670).Q1 = poorest quintile; Q5 = richest quintile

Each entry in the table corresponds to the latest available observation in the 1990s for each country; Discrepancies are due to rounding errors.

Subject to the above caveats, after-tax Gini coefficients, on average, are lower than before-tax Gini coefficients, and have a smaller range as well. For the 9 “b” countries, for which inequality estimates are based on before-tax incomes, Gini coefficients range widely, between 25 percent and 52 percent, averaging 38 percent; by contrast, for the 13 “a” countries for which inequality estimates are based on after-tax incomes, Gini coefficients range between 25 and 45 percent, averaging 34 percent. The difference of 4 percentage points between the two averages does not necessarily suggest the redistributive impact of taxes. First of all, the two samples comprise two different sets of countries. Moreover, the before-tax incomes are likely to reflect the impact of tax policy. It is interesting, however, that the small difference between the averages is similar in its magnitude to the differences of 3 and 4 percentage points between before-tax and after-tax Gini coefficients for Poland and Romania—the only two countries for which both before-tax and after-tax income data are available.5

By comparison with industrial countries, before-tax (after-transfer) Gini coefficients for developing countries are smaller, on average, but more widely dispersed. The before-tax Gini coefficients for the 9 developing countries in Table 1 average 38 percent, compared with the average of 44 percent for the before-tax and before-transfer “market-income” Gini coefficients for 11 industrial countries in Table 2.6 The before-tax Gini coefficients for developing countries range between 25 percent and 52 percent, whereas the market-income Gini coefficients for industrial countries range between 34 percent and 54 percent. However, tax (and transfer) policies in industrial countries reduce their Gini coefficients much more than their counterparts in developing countries. For the 11 industrial countries, the average disposable income Gini coefficient is lower than the market income Gini coefficient by 16 percentage points and lower than after-tax Gini coefficient for developing countries by 6 percentage points.

Table 2.Income Distribution in Industrial Countries, 1970s/80s–1990s(In percent)
CountriesGiniPeriodPercent Change in Gini
PeriodMarket

Income
Disposable

Income 1/
Market

Income
Disposable

Income
Average Gini

(1990s)
44.327.715.75.0
Average Gini

 (1970s/80s)
40.426.4
Percentage-Point Change in Gini

 (1970s/80s-1990s)
3.91.3
Australia1993/9446.330.61975-9436.65.2
Belgium199554.529.61983-952.3
Canada199428.41975-940.2
Denmark199442.021.71983-9411.2-4.9
Finland199539.223.11986-9511.49.1
France199029.11979-90-1.7
Germany199443.628.21984-941.26.4
Italy199351.034.51984-9320.812.7
Japan199434.026.51984-944.9
Netherlands199442.125.31977-9414.211.8
Norway199539.925.61986-959.4
Sweden199548.723.01975-9517.3-1.0
United States199545.534.41974-9513.110.0
Sources: OECD, 1997, OECD Economic Outlook, No. 62 (December): 49-59; Oxley, Burniaux, Dang and d’Ercole (1999, Tables 1, and A1).

Disposable income is defined as sum of 1) earnings, 2) self employment and capital income, 3) transfers received from general government, and 4) direct taxes and social security contributions paid by individuals.

Sources: OECD, 1997, OECD Economic Outlook, No. 62 (December): 49-59; Oxley, Burniaux, Dang and d’Ercole (1999, Tables 1, and A1).

Disposable income is defined as sum of 1) earnings, 2) self employment and capital income, 3) transfers received from general government, and 4) direct taxes and social security contributions paid by individuals.

B. Changes in Income Distribution, 1970s–1990s

Table 3 provides an overview of the changes in income (or consumption) distribution for groups of countries for which data on (1) either before-tax or after-tax Gini coefficients and (2) tax structures are available for recent decades (1970s–1990s). For changes in Gini coefficients from the 1970s to the 1980s, consistent data are available for only 19 countries. For changes from the 1980s to the 1990s, data are available for 10 countries.

Table 3.Developing Countries: Overview of Income Distribution and Tax Burden/Structure, 1960s-1990s(In percent unless otherwise indicated)
1970s1980s1990s
Gini

(percent)
Q5/Q1

Ratio
Gini

(percent)
Q5/Q1

Ratio
Gini

(percent)
Q5/Q1

Ratio
Income distribution

1970s-80s
All countries44124411
15 “b” countries46134612
4 “a” countries367367
Tax burden/structure (percent of GDP)TotalDirectIndirectTotalDirectIndirect
1970s-80sTaxTaxesTaxesTaxTaxesTaxes
All countries14671678
15 “b” countries15671678
4 “a” countries14581669
Income distribution 1/

1980s-90s
All countries327348
5 “b” countries318339
5 “a” countries356377
Tax burden/structure (percent of GDP) 1/TotalDirectIndirectTotalDirectIndirect
1980s-90sTaxTaxesTaxesTaxTaxesTaxes
All countries2251221511
5 “b” countries2471121611
5 “a” countries2041320412
Sources: Tables 6, 7, 9, 10, and International Monetary Fund, International Financial Statistics and Government Finance Statistics data base.b = Gini coefficient and Q1 and Q5 estimates based on before-tax incomes.a = Gini coefficient and Q1 and Q5 estimates based on after-tax incomes.
Sources: Tables 6, 7, 9, 10, and International Monetary Fund, International Financial Statistics and Government Finance Statistics data base.b = Gini coefficient and Q1 and Q5 estimates based on before-tax incomes.a = Gini coefficient and Q1 and Q5 estimates based on after-tax incomes.

The average Gini coefficient for the sample countries was stable during the 1970s–1980s. The average remained at 44 percent for the 19 countries. The average Gini coefficient for the 10 sample countries increased only slightly, from 32 percent to 34 percent, during the 1980s-1990s.

The stable average Gini coefficient during the 1970s and the 1980s, however, masks considerable changes in the coefficients for some countries. For example, during the 1970s-1980s, the Gini coefficient for Thailand increased by four percentage points, while the Gini coefficient for Turkey declined by seven percentage points. During the 1980s–1990s, Thailand and two transition countries (Bulgaria and Hungary) experienced a large increase in the Gini coefficient (see Section III for further discussion).

It is well recognized that transition countries have experienced an increase in the degree of inequality in the size distribution of income. It is not immediately clear, however, why Gini coefficients for a large number of developing countries increased precipitously during the 1980s–1990s.

On the basis of the averages, data are not clear about the distributive role of taxes. From the 1970s to the 1980s, the tax burden, on average, increased slightly, with no notable change in the composition of direct and indirect taxes. From the 1980s to the 1990s, neither the average tax burden, nor the tax composition changed significantly.

III. Role of Taxes and Social Spending

Income distribution has many aspects. The distribution of market incomes is an important aspect of income distribution. The distribution of disposable incomes is another. One could also look at the distribution of disposable incomes together with government in-kind transfers, such as education and health services. More broadly, one could analyze the distribution of disposable incomes and all government services, including defense, justice, and infrastructure services. In a country with a large-scale provision of free public education and health care, this last measure of income distribution could be substantially different from that of disposable incomes. Even in a country where markets play a dominant role, the distribution of income based on the first and last measures could be substantially different.

There are many factors that affect the distributions of market and disposable incomes. These factors include the distribution of physical, financial, and human capital; rates of returns to these forms of capital; and formal and informal institutions. Taxes and transfers affect the different between market and disposable incomes in the short run, but they can also affect the distribution of market incomes over time. Some taxes can affect individuals’ work efforts. Excessively high tax rates can drive economic activities out of the formal sector or out of a country. Government social spending policies have distributional implications not only because social spending can offer immediate benefits (e.g., health and education services), but they also affect the distribution of earning capacities of individuals and households, and thus help shape the distribution of market incomes over time. Some social expenditures (e.g., expenditures on primary schooling) can affect income distribution with a long time lag.

This section traces the distributional implications of tax and social spending policies through some of these diverse channels. Section A provides an overview of the literature on how taxes and social expenditures are distributed among different income groups. Section B discusses how income distribution, measured on the basis of before-tax or after-tax incomes, has changed in recent decades. Section C uses econometric techniques to detect any long-term relationship between the distribution of either before-tax or after-tax incomes and certain aspects of tax and social spending policies (i.e., tax structure and secondary enrolment rate). Section D discusses the relationship between income distribution and taxes and government social spending policies in selected countries.

A. Selective Literature Survey

Existing surveys of tax incidence studies

This section discusses five surveys of tax incidence studies of developing countries: Bird and de Wulf (1973); De Wulf (1975); and subsequent surveys by McLure (1977) and Shah and Whalley (1990 and 1991).7

Bird and de Wulf (1973), covering 29 studies on 17 Latin American countries, conclude that tax systems in these countries were often ineffective as redistributive tools. Bird and De Wulf note that only 4 of 29 studies suggest a mild redistributive impact of the tax system, whereas the remaining studies suggest “rough proportionality or even regressivity over most income classes” (Bird and De Wulf, 1973, p. 671).8 Although de Wulf (1975), in a subsequent survey of 66 studies covering 23 countries, concludes that the tax system in developing countries tends to be progressive (and the degree of progressivity varies from steep to moderate), he reverses himself in a subsequent survey by noting that it is “difficult…not to conclude that tax systems in LDCs must be regressive” (de Wulf, 1981, p. 20). Underlying his reasoning is the relative dominance of consumption taxes, which he assumes to be mostly regressive.

McLure (1977), in a survey of seven studies on the tax burden faced by the urban poor in developing countries, notes that tax rates for urban households are progressive but not smooth. For example, although the effective tax rates for the urban rich (top 2–5 percent of urban households) are at least twice those of the urban poor (bottom 40 percent of urban households), the fourth urban quintile and the urban poor face similar effective tax rates. He concludes that, because policy decisions are made at the margin, a tax incidence analysis would be more useful to policymakers if it focused on the distributional implications of changes in tax systems rather than on the incidence of an existing tax system.

Shah and Whalley (1990 and 1991), in a brief survey of seven tax incidence studies, conclude that, with some exceptions, the overall tax system is broadly progressive. Regarding individual taxes, they find excises, personal income taxes, and urban property taxes to be progressive; sales and import duties to be regressive; corporate taxes to vary, depending on assumptions used; and general indirect taxes to have a U-shaped incidence pattern.9

Recent tax incidence studies

Some methodological improvements in measuring the incidence of taxes have been made since the surveys by Bird and de Wulf (1973) and de Wulf (1975). Fully-specified computable general equilibrium models have gradually replaced various (forward or backward) shifting assumptions featured in a large number of studies surveyed by Bird and de Wulf (1973) and Shah and Whalley (1990 and l991), although in some cases the results of judgmental studies were close to general equilibrium simulations (Devarajan, Fullerton, and Musgrave, 1980).

General equilibrium analysis seems to add an interesting dimension that was absent in partial equilibrium studies. For example, personal and company taxes in Kenya are found to be progressive on the basis of a general equilibrium analysis (Mwega, 1986), but to be regressive on the basis of a partial equilibrium analysis (Westlake, 1973). Similarly, indirect taxes in the Philippines are found to be broadly neutral, but only after taking into account their general equilibrium effects (Devarajan and Hossein, 1998).

Mostly, however, tax incidence studies still visually inspect the effective tax rates by income group and present a judgment of the degree of progressivity. Certain methodological difficulties, dating back to the surveys of Bird and De Wulf (1973), continue to persist. For example, income concepts used in incidence studies continue to vary widely, from taxable income (Chowdhury, 1988) and gross income (Bolkowiak and others, 1996), to permanent income (Gil-Diaz, 1982). In addition, there are significant differences in taxes studied, units of analysis, and underlying assumptions.

For a better understanding of the state of knowledge on the incidence of taxes in developing countries, this paper offers the results of a systematic survey of all tax incidence studies conducted since the surveys of Bird and De Wulf (1973) and De Wulf (1975). The intention of the survey is to record the following crucial features for each study: the country and period covered, concept of income used, taxes included, unit of analysis (e.g., individuals or households), coverage (e.g., all individuals, wage earners, or pensioners), measure of tax progressivity, and the study’s conclusion about the tax incidence.

Table 4 suggests the following points:

  • As regards overall tax systems or both direct and indirect taxes in 19 countries studied, only 13 of 36 cases are progressive, 7 are proportional, 7 are regressive, and the rest have mixed findings or insignificant effects.
  • As regards income tax systems in 8 countries, 12 of the 14 cases are progressive, one is regressive, and one has mixed findings. In the 5 cases of payroll taxes in 3 countries, 3 cases are regressive, and 2 cases have insignificant effects.
  • Eight studies report a decline of the progressivity of direct taxes over time.

Table 4.Developing Countries: Tax Incidence
Periods

Studied
Countries

Studied
Taxes Studied
All

Taxes
Direct

Taxes
Indirect

Taxes
IncomePayroll
Frequency3631455
PP1
P121122
Prop711
Insig12
Mixed8112
R73
RR1
1971-75ColombiaPP
MexicoP
KoreaProp
JamaicaMixed
IndiaP-
1976-80MexicoP
PakistanP
KoreaProp
KoreaR
IndiaP-
IndiaP
1981-85KoreaProp
JamaicaMixed
KoreaR
IndiaP-
BangladeshMixed
JamaicaRR
IndiaP, Prop
JamaicaProp
1986-90JamaicaP
PhilippinesProp
PhilippinesMixed
KoreaR
HungaryP-
HungaryR
PolandR
PeruMixed ()
1991-95GuatemalaP
JamaicaInsig
PolandMixed
BulgariaP
HungaryP-
IndiaP-
BulgariaInsig
PolandMixed
Sources: Calculated from the data in AnnexTable 1.PP = Strongly progressive; P = Progressive; P- = Decline in progressivity.Prop - Proportional; Mixed = Mixed.R = Regressive; RR = Strongly regressive; Insig. = Insignificant effect.() = Selective taxes studied.
Sources: Calculated from the data in AnnexTable 1.PP = Strongly progressive; P = Progressive; P- = Decline in progressivity.Prop - Proportional; Mixed = Mixed.R = Regressive; RR = Strongly regressive; Insig. = Insignificant effect.() = Selective taxes studied.

Some studies, not reported in the table, suggest that indirect taxes may not be as regressive as assumed in the surveys of Bird and De Wulf (1973) and De Wulf (1975).10

Benefit incidence of government spending11

In an exhaustive survey of some 25 years ago, McLure (1974) defines expenditure incidence as “how government spending affects private incomes” and benefit incidence as “who receives benefit of government services.” 12 Benefit incidence analysis is meaningful only if they can be interpreted properly, given its limitations.13 The concepts of targeting and progressivity, commonly used to interpret such data and used in this paper, need to be defined as follows: 14

  • Government spending is considered to be well (poorly) targeted if the poorest quintile’s share of benefits from such spending is larger (smaller) than the richest quintile.15 This means that the poorest 20 percent benefit more than the richest 20 percent, in absolute terms.
  • Government spending is considered to be progressive (regressive) if the benefits to the poorest quintile are larger (smaller) than the benefits to the richest quintile, relative to their income or expenditure. With progressive (regressive) spending, benefits represent a smaller (larger) fraction of income or expenditure at higher income or expenditure quintiles.16

These definitions imply that if spending is well targeted, it will be progressive, but progressive spending may not be well targeted. It also implies that if spending is poorly targeted, it may be progressive or regressive. Targeting an expenditure well is a much more demanding objective than making it progressive.

Benefit incidence of government education spending. Thirty-one of the 55 studies for 25 developing countries for which central government spending data are available on “all education” (primary, secondary, and tertiary education) find government educational spending to be progressive (Table 5).17

Table 5.Developing Countries: Incidence of Selective Social Expenditures
Frequency: Number of Countries
EducationTargetingProgressivity
AllPri-

mary
Secon-

dary
Ter-

tiary
AllPri-

mary
Secon-

Dary
Ter-

tiary
Targeting55545452
Good2242230
Poor33123152
Incidence31372611
Progressive3137266
Regressive0005
HealthTargetingProgressivity
Targeting38
Good21
Poor17
Incidence30
Progressive30
Regressive0
TransfersTargetingProgressivity
Targeting14
Good4
Poor9
Inconclusive1
Incidence15
Progressive14
Regressive0
Inconclusive1
Sources: Davoodi and Sachjapinan (forthcoming).Central government data.
Sources: Davoodi and Sachjapinan (forthcoming).Central government data.

Spending on education, on average, is poorly targeted in 33 studies; there are regional differences, however. Education spending is well targeted in Asia and Latin America, but poorly targeted in sub-Saharan Africa, the Middle East, and transition economies.18 Subsaharan Africa’s record stands out among the latter economies: the poorest quintile receives the least (13 percent of benefits), and the richest quintile receives the most (32 percent of benefits).

  • In all regions, government spending on primary education is well targeted, but there are notable differences across the regions. The poorest quintile in sub-Saharan Africa receives from government primary education spending slightly more than the richest quintile, whereas the poorest quintile in Latin America receives more than four times the richest quintile.
  • Government secondary education spending, on average, is well targeted in Asia and Latin America but poorly targeted in sub-Saharan Africa, the Middle East, and transition countries.
  • Government tertiary education spending mostly benefits the richest quintile in all regions. Sub-Saharan Africa stands out in this respect. The poorest quintile receives 4.5 percent of benefits from government tertiary education spending, whereas the richest quintile receives as much as 59 percent.

The data compiled by Davoodi and Sachjapinan (1999) also allow an analysis of the changes in benefit incidence over time. The available data indicate that the poor’s share of the benefits from education spending has increased relative to the richest quintile. Of the 11 countries for which benefit incidence data for education are available, 8 countries show improvements for the poorest quintile while 3 countries show a deterioration.19

Benefit incidence of government health spending.Table 5 indicates that 21 of 38 studies find government health spending to be well targeted, and all 30 available studies find government health spending to be progressive.

Government spending on “all health” 20 (comprising expenditures on health centers, hospitals, and hospital inpatients and outpatients) for a sample of 14 developing countries are progressive over the period 1974–1995 for which data are available.21 Much like the benefit incidence of government education spending, there is a striking diversity in the degree of progressivity among countries.

  • Government spending on health, on average, is well targeted in a sample of 29 developing and transition economies over the period 1978–1995. Sub-Saharan Africa and transition economies are the only areas in which government health spending is poorly targeted. In contrast, the poorest quintiles in Asia and Latin America receive, respectively, 1.5 and 3 times as much in benefits as the richest quintiles.22
  • Benefit incidence of health spending has changed over time. Of the 10 countries for which data at two points in time are available for health spending, the poorest quintile’s share of benefits has increased relative to the rich in five countries but has decreased in another five. Similarly, changes in incidence for subcategories of health spending do not follow a uniform pattern.

Benefit incidence: other government expenditures.Table 5 offers a summary of the results for selected transfers. As in the case of “other government expenditures,” transfers are not well targeted in many countries.

Not surprisingly, the well-targeted programs include a number of expenditure items that involve some targeting mechanism, such as food stamps (Jamaica) or self-targeted food subsidies (Tunisia). In Sri Lanka’s (Edirisinghe, 1987), although both universal and targeted programs are found to be progressive, the food stamp program in fact resulted in a greater degree of progressivity and a better targeting. Universal food subsidies in the sample are poorly targeted, with upper-income quintiles obtaining more benefits in absolute terms (e.g., Tunisia).

A number of housing expenditures are poorly targeted (four of six cases), as well as pension and social security benefits (Chile, Costa Rica, and Uruguay). However, a better targeting is sometimes achieved when pensions are combined with pro-poor family allowances and other benefits (see e.g., Milanovic, 1995). Meanwhile, residential utilities are often poorly targeted, except for new investments in Colombia from 1970–74.

A limited number of studies allow for meaningful comparisons over time. They generally show that there has been some improvement in progressivity and targeting over time. For example, in Chile (Aninat, Bauer, and Cowan, 1999), the share of cash transfers received by the poorest quintile increased from 33 percent in 1992 to 40 percent in 1996. The redistributive impact of cash transfers, along with social spending, was seen to have compensated for some deterioration in the distribution of income during this period. In Colombia, pro-poor investments from 197074 resulted in a larger percentage of households in the poorest quintile with access to public utility services (electricity, water and sewage, and street lighting). In Costa Rica, the share of pension benefits received by the poorest quintile increased from 6 percent in 1986 to 9 percent in 1992, although this is still somewhat lower than their 1983 share (10 percent).

B. Changes in Income Distribution, 1970s–1990s

Changes from the 1970s to the 1980s

For the period 1970s–1980s, consistent income distribution data are available only for 19 developing countries: before-tax (but after-transfer) income data for 15 countries (“b” countries), and after-tax (and after-transfer) income data for only 4 countries (“a” countries). During the 1970s, Gini coefficients for “b” countries ranged between 35 percent and 54 percent; those for “a” countries between 31 percent and 47 percent (Table 6).

Table 6.Nineteen Developing Countries: Income Distribution, 1970s–1980s(In percent unless otherwise indicated)
1970s (1)1980s(2)Change (3)=(2)-(1)
SharesRatioSharesRatioSharesRatio
GiniQ1Q5Q5/Q1GiniQ1Q5Q5/Q1GiniQ1Q5Q5/Q1
AverageAll countries44550124454911000-1
“b” countries46551134655112000-1
“a” countries367447367447-10-10
Indiaa3194053194151010
Pakistana3194053284151000
Bangladeshb3574363774562021
Korea, R.b367437367437-1000
Indonesiaa378445338425-30-20
Singaporeb3774264174774-151
Sri Lankab3974674564786011
Thailandb425481046453134-143
Venezuelab435471044549101020
Jamaicab4545012435499-11-1-3
Costa Ricaa47453154645113-20-2-2
Trinidadb49251234234513-71-7-10
Philippinesb49454154655210-32-2-5
Guatemalab50654959263279-3917
Turkeyb51457164455010-72-7-7
Malaysiab51456154845312-31-3-3
Colombiab525581351456150-1-22
Panamab53357245235621-10-1-3
Mexicob54360225345816-11-3-6
Source: Deininger and Squire (1996).“b” = Based on before-tax income“a” = Based on after-tax incomeDiscrepancies are due to rounding errors.
Source: Deininger and Squire (1996).“b” = Based on before-tax income“a” = Based on after-tax incomeDiscrepancies are due to rounding errors.
Table 7.Ten Developing Countries: Changes in Income Distribution, 1980s–1990s 1/(In percent unless otherwise indicated)
1980s(1)1990s(2)Change (3) = (2)-(1)
SharesRatio

Q5/Q1
SharesRatio

Q5/Q1
SharesRatio

Q5/Q1
GiniQ1Q5GiniQ1Q5GiniQ1Q5
AverageAll3584373874583-111
“b” countries3574383874593-121
“a” countries3584363774472-111
Hungarya23113433273969-453
Bulgariab23103332893745-151
Polandb25103542883754-121
Indiaa3194153194150000
Pakistanb328415318405-10-10
Jordana3874774164872000
Mauritiusa406468377436-31-2-1
Perua43650845550102-112
Thailandb464531350457154042
Colombiab5145615514541500-20
Source: Deininger and Squire (1996).“b” = Based on before-tax income“a” = Based on after-tax income

Discrepancies are due to rounding errors.

Source: Deininger and Squire (1996).“b” = Based on before-tax income“a” = Based on after-tax income

Discrepancies are due to rounding errors.

During the period 1970s–1980s, although neither the averages nor the ranges of the Gini coefficients changed significantly for the 19 countries, some of these countries experienced considerable changes in their Gini coefficients. For example, the “before-tax” Gini coefficients for Trinidad and Tobago and Turkey declined by seven percentage points (income distribution became less unequal), while those for Sri Lanka and Guatemala increased, respectively, by six and nine percentage points (income distribution became more unequal). The Philippines, Indonesia, and Malaysia experienced a smaller, but still considerable, improvement in their income distribution, while Bangladesh, Singapore, and Thailand experienced a considerable deterioration.23

Based on the economic growth and income distribution performances of the 19 sample countries, it is not evident that a prolonged economic growth, in and of itself, while having a powerful impact on poverty reduction, necessarily leads to an improvement in income distribution. Among the high-growth countries, Malaysia and Indonesia achieved a considerable improvement in income distribution, but both Thailand and Singapore experienced a significant deterioration. Such low-growth countries as Trinidad and Tobago and the Philippines significantly improved their income distribution; Guatemala achieved a low growth and a deterioration in income distribution.24

Changes from the 1980s to the 1990s

For the group of 10 sample countries for which comparable income distribution data are available, the average Gini coefficient increased by three percentage points from the 1980s to the 1990s. As from the 1970s to the 1980s, the modest increase in the average Gini coefficient masks a considerable variation in the changes in Gini coefficients across countries.

From the 1980s to the 1990s, however, unlike from the 1970s to the 1980s, a much larger number of countries experienced a considerable increase in their Gini coefficients.

The 10-country sample for which comparable income distribution data are available for the 1980s and the 1990s include several transition countries, which experienced a substantial deterioration in income distribution during the period 1980s–1990s. Thus, the Gini coefficients for Poland, Bulgaria, and Hungary increased by between 4 and 9 percentage points. Bulgaria, having maintained a Gini coefficient of between 18 percent and 26 percent in the 1980s, experienced an increase in the Gini coefficient to 34 percent in 1993, substantially higher than previously, but still low by market-economy standards. Similarly, Poland has also experienced an increase in the Gini from between 21–27 percent to 33 percent.

As from the 1970s to the 1980s, the countries for which income distribution deteriorated included high-growth countries (e.g., Thailand), as well as low-growth countries (Jordan and European transition countries). On the contrary, Mauritius improved their income distribution while experiencing a low economic growth.

C. Econometric Estimation of Gini Equations

Econometric estimation of Gini equations reveals possible factors underlying the changes in Gini coefficients. Equation (1) assumes a number of potential determinants of income distribution:

where

g = Gini coefficient (percent),

r = ratio of direct to indirect taxes (percent),

d = ratio of direct taxes to GDP (percent),

s = secondary school enrollment rate (percent),

u = urbanization (percent),

k = transition country dummy,

x = inflation dummy, and

i, t = country subscript, and time subscript for three decades (1970s, 1980s, and 1990s).

While the Deininger-Squire Gini estimates have been compiled on a reasonably comparable methodology across countries and through time, they are still not methodologically homogeneous, based on either consumption or before- or after-tax income that often include government transfers. The consumption Gini coefficients would reflect the effects of taxes.

The specification of Equation (1) is based on the recognition that, in view of these characteristics of the data, the explanatory variables included in the equation would affect the distribution of “market incomes.” The specification is aimed at testing the distributional implications of the nature of tax regime and the secondary school enrollment rate. Given the nature of data that include Gini coefficients based on consumption, as well as on pre-tax and after-tax incomes, taxes can affect consumption-based Gini coefficients. In many developing countries, an increase in secondary school enrollment is a critical means of improving the distribution of human capital and earnings capacity.

Table 8 offers the results of estimation. The results represent highly tentative evidence on the relationship between income distribution and its several potential determinants: the tax regime, secondary school enrollment, urbanization, and inflation. The hypotheses is that, other things being equal, the Gini coefficient is associated negatively (or income equality is associated positively) with (1) the ratio of direct to indirect taxes and (2) the secondary school enrollment rate. Since the distributional effects of the tax regime would depend not only on the tax structure, but also on the amount of tax revenue relative to GDP, the equation includes the product of the ratios of direct to indirect taxes and of direct taxes to GDP.

Table 8.Developing Countries: Estimation of Gini Equations 1/
Constant

(c)
Ratio of

direct tax

to indirect

tax 2/

(r)
(r) times

ratio of

direct tax

to GDP

(rd)
Secondary

school

enrolment

rate

(s)
Urbani-

zation

rate

(u)
Transition

country

dummy

(k)
Inflation

dummy

(x)
(1.1.A)
Coefficients41.722-0.0250.0008-0.2010.229
t-ratios15.19-2.722.21-3.624.80
Adjusted R-square0.200
Number of observations85
(1.13)
Coefficients37.069-0.0190.0004-0.0540.213
t-ratios13.75-2.071.08-0.994.33
Adjusted R-square0.230
Number of observations78
(1.2.A)
Coefficients33.779-0.020-0.0620.313-19.583-1.010
t-ratios11.64-2.50-0.936.76-6.25-0.38
Adjusted R-square0.620
Number of observations55
(1.2.B)
Coefficients34.176-0.021-0.0770.322-2.54
t-ratios11.56-2.47-1.116.96-0.61
Adjusted R-square0.430
Number of observations48
(A)Average818434547
Range16-6754-208004-958-100
(B)Average808364246
Range16-6754-122224-958-86
Sources: Based on the data from Deininger and Squire (1996) and International Monetary Fund, International Financial Statistics and Government Finance Statistics.

Estimates marked by A and B, respectively, are for the sample of developing and transition countries and for the sample of developing countries.

For the sample of developing and transition countries, the average of Gini coefficients is 41; the average of the ratio of direct taxes to GDP is 5.4 percent (A). For the sample of developing countries, the average of Gini coefficients is 43; the average of the ratio of direct taxes to GDP is 5.1 percent (A).

Sources: Based on the data from Deininger and Squire (1996) and International Monetary Fund, International Financial Statistics and Government Finance Statistics.

Estimates marked by A and B, respectively, are for the sample of developing and transition countries and for the sample of developing countries.

For the sample of developing and transition countries, the average of Gini coefficients is 41; the average of the ratio of direct taxes to GDP is 5.4 percent (A). For the sample of developing countries, the average of Gini coefficients is 43; the average of the ratio of direct taxes to GDP is 5.1 percent (A).

Both an increase in the ratio of direct to indirect taxes and an expansion of the secondary school enrollment rate would tend to reduce the Gini. Urbanization may have positive, negative, or neutral effects. Economic growth that allows an expansion of a high-income urban sector could increase income inequality; a rural-to-urban migration of unskilled workers might not have any effect on income inequality. Some studies have suggested that high inflation tends to be associated with high income inequality, although, over time, income distribution depends more on real factors (e.g., skills distribution) than on inflation.

The equation is estimated with various constraints for coefficients for two samples: one including developing and transition countries (estimates indicated by A) and the other including only developing countries (estimates indicated by B). In the sample including both developing and transition countries, the transition country dummy would capture the low Gini coefficients that the transition countries inherited from the socialist era. The estimates suggest the following:

  • The effects of tax ratio and urbanization variables are statistically significant and the significance is fairly robust. In addition, the estimate of the coefficient for the variable representing the product of the ratios of direct to indirect tax and of direct tax to GDP is statistically significant for the (1.1.A) regression.
  • The role of secondary education is unclear. For the A sample, without the transition country dummy, the estimate of the coefficient for the secondary school enrollment rate is statistically significant. However, the estimate is not statistically significant either for the B sample or for the A sample with the transition country dummy. The statistical significance for the A sample without the transition country dummy might reflect the difference of secondary school enrollment rate between transition and developing countries.
  • The estimation suggests that inflation does not affect the long-term evolution of Gini coefficients.

The explanatory power of the equations is rather limited, indicating that there are many other factors that influence income distribution.25

D. Role of Taxes and Government Social Spending Policy

Overview

The estimated equations reported in Table 8 suggest some evidence of the effects of the tax regime and secondary school enrollment on income distribution. One of the equations, Equation (1.1.A), indicates statistically significant effects of the two variables. The interpretation of the magnitudes of the estimated coefficients require some caution: 26

  • Other things being equal and in the neighborhood of averages, a revenue-neutral increase of 40 percentage points (from 80 percent to 120 percent) in the ratio of direct to indirect tax revenues would reduce the Gini coefficient only by 0.75 of a percentage point. A one percentage-point increase in the ratio of direct tax revenue to GDP and an increase of a same rate in indirect taxes, would raise the Gini coefficient by 0.06 of a percentage point.
  • Other things being equal, a 10 percentage-point increase in the secondary school enrollment rate would reduce the Gini coefficient by 2 percentage points. This is a relatively large improvement in income distribution.

The Gini coefficients explained by the equation consist of those based on before-tax incomes, after-tax incomes, and consumption. The sign and statistical significance of the coefficient for the ratio of the direct to indirect taxes are consistent with the conclusions of the literature survey, which indicate the progressivity of direct taxes and of education expenditure, including secondary education expenditure. While the small magnitude of the effect of tax policy does not appear to suggest an active use of tax policy for redistribution, country-specific analyses might suggest different conclusions.

From the 1970s to the 1980s, five countries significantly reduced Gini coefficients (one of them on the basis of after-tax income), but five other countries significantly increased Gini coefficients (all on the basis of before-tax incomes) (Table 6).

The countries that significantly reduced Gini coefficients are Trinidad and Tobago, Turkey, the Philippines, Indonesia, and Malaysia. Indonesia and Malaysia achieved sustained high economic growth, pursued with strong poverty reduction and equity objectives. Malaysia and the Philippines had broadly progressive and well-targeted educational and health programs. Indonesia had progressive and well-targeted (albeit increasingly less effective) education programs, although its health programs were poorly targeted. Moreover, Indonesia and Malaysia raised their tax-to-GDP ratio, by increasing direct taxes but reducing indirect taxes as a ratio to GDP. In particular, both countries reduced international taxes significantly (Table 9).

Table 9.Nineteen Developing Countries: Changes in Tax Structure, 1970s–80s(Percentage points of GDP)
Changes: 1980s less 1970sChanges

in Income

Distribution
AllDirectIndirect
AllDom.

Goods
Inter.

Goods
AverageAll1.61.00.50.6-0.1
b1.50.90.60.8-0.2
a2.11.40.40.10.3
Trinidad and Tobagob5.96.4-0.4-0.3-0.1iid
Turkeyb-1.90.4-2.1-0.7-1.4iid
Philippinesb0.00.0-0.10.7-0.7iid
Malaysiaa1.92.3-0.40.6-0.9iid
Indonesiab2.83.0-0.50.2-0.7
Costa Ricab3.4-0.21.7-0.11.8
Jamaicab3.80.05.05.5-0.5
Mexicoa4.03.4-0.4-0.70.3
Panamab0.10.60.10.6-0.5
Korea, R.b1.70.41.10.70.4
Colombiab0.0-0.70.91.2-0.3
Indiab2.51.01.80.21.5
Pakistana2.10.61.60.90.6
Venezuelaa0.1-0.50.6-0.41.0
Bangladeshb1.10.21.00.10.9did
Singaporeb0.60.9-0.70.4-1.1did
Thailandb1.60.90.61.0-0.4did
Sri Lankab1.80.31.31.9-0.6did
Guatemalab-1.30.0-0.90.3-1.1did
Sources: International Monetary Fund, International Financial Statistics and Government Finance Statistics.iid = improvement in income distributiondid = deterioration in income distribution
Sources: International Monetary Fund, International Financial Statistics and Government Finance Statistics.iid = improvement in income distributiondid = deterioration in income distribution

The countries that significantly increased Gini coefficients are Bangladesh, Singapore, Thailand, Sri Lanka, and Guatemala. Of these, government education and health spending incidence studies are available only for Bangladesh, which had poorly targeted primary, secondary, and tertiary education programs, although its health program was considered progressive and well targeted. Singapore achieved a high economic growth, but Guatemala’s growth was low. The deterioration in Thailand’s Gini coefficient is notable: that country’s high economic growth reduced poverty, but was regionally unbalanced, and the returns to education expenditures differed significantly among schooling levels and occupations (Ahuja, et al, 1997).

From the 1980s to the 1990s, only two countries significantly reduced their Gini coefficients (Jamaica and Mauritius), but nine countries significantly increased theirs, sharing the widespread experiences of both industrial, developing, and transition countries (Table 7). No public spending incidence analysis is available for Mauritius, but available studies suggest that Jamaica’s foodstamp program was progressive and well targeted, although its tax system was considered to have either a highly regressive (income tax) or mixed or insignificant (overall tax system) incidence. Moreover, Jamaica’s agricultural production responded strongly to the introduction of market-oriented reforms of the 1980s (Handa and Kin, 1997).

Of the nine countries for which income distribution deteriorated, China, Poland, Bulgaria, Romania, and Hungary were in the transition and in general achieved a low economic growth. Poland, Bulgaria, and Hungary experienced a large decline in their tax-to-GDP ratio (Table 10). Of the other four, Hong Kong and Thailand achieved a high economic growth, but Jordan and Nigeria experienced a low economic growth. No public spending incidence study is available for either Jordan, Nigeria, Hong Kong, SAR, China, or Thailand.

Table 10.Ten Developing Countries: Changes in Tax Burden/Structure, 1980s–90s(In percentage points of GDP)
Changes: 1990s less 1980sChanges

in Income

Distribution
AllDirectIndirect
AllDom.

Goods
Inter.

Goods
AverageAll-1.0-0.3-0.3-0.1-0.2
b-2.4-1.0-0.3-0.30.0
a0.40.3-0.30.2-0.5
Mauritiusa-0.30.0-1.21.2-2.4iid
Pakistanb0.50.40.10.3-0.3
Indiaa-0.10.6-0.7-0.70.0
Colombiab1.82.41.41.8-0.4
Perua1.1-0.1-0.50.9-1.4
Jordana6.51.14.93.91.0did
Poland*b-4.4-0.3-1.1-1.0-0.1did
Thailand*b2.82.30.10.3-0.2did
Bulgaria*b-12.6-9.6-1.8-3.01.2did
Hungary*a-5.10.0-4.2-4.40.2did
Sources: International Monetary Fund, International Financial Statistics and Government Finance Statistics.iid = Improvement in income distributiondid = Deterioration in income distribution
Sources: International Monetary Fund, International Financial Statistics and Government Finance Statistics.iid = Improvement in income distributiondid = Deterioration in income distribution

The increase in income inequality observed for the sample countries reflects, to a considerable extent, the increase in inequality in several transition countries included in the sample. However, even if the transition countries were removed from the sample, the upward trend in Gini coefficients would be notable, with many market-oriented developing countries sharing a similar experience with industrial countries.27

The role of changes in the tax system in this deterioration is not obvious. On average, these countries did not change their tax structure significantly, although many transition countries, which experienced an increase in their Gini coefficients, also suffered from a substantial reduction in their tax revenue.

None of the eight developing countries whose tax reforms were studied in a recent World Bank report has a pre-reform schedule of progressive tax rates successfully generating a progressive distribution of tax burdens (Thirsk, 1997). In general, tax reforms in developing countries have been aimed at increasing revenues with a degree of efficiency and with some attention to equity; rarely have they been aimed specifically at improving equity. While the literature survey suggests that income taxes tend to have a progressive effect, the regressions indicate that the magnitude of the effect might not be large.

Developing countries in general have been expanding primary and secondary school enrollment rates through public education programs. Transfer programs often have taken more of the form of untargeted food, fuel, fertilizer, and other subsidies than of cash transfers, such as unemployment benefits and pensions, which are largely limited to public sector employees. Transition economies started the transition to the market in the early 1990s with an expansive social programs and generally high levels of social indicators. The subsequent compression of output has forced many of them to reduce social spending. As a result, many of the transition economies experienced a deterioration in social indicators.

A number of reasons have been offered to explain increases in before-tax income inequality in the context of industrial countries.28 Two of them might be applicable to developing countries:

  • The opening up of low-income developing countries (e.g., China and India) to foreign direct investment and trade tends to reduce the wages of unskilled workers in middle-income developing countries (e.g., Thailand), which are competing with low-income countries.
  • With globalization, developing countries gradually adapt their traditional income-equalizing social norms to those tolerating a higher degree of income inequality.

Among the countries that defied this broad trend were those that had progressive and well-targeted education and health programs that increased the earnings potential of low-income workers and some countries that achieved a sustained high economic growth, although not all high-growth countries achieved a reduction in inequality.

It is worth noting the difficulties in comparing before-tax and after-tax income inequality in developing countries. Unlike for industrial countries, for which separate Gini coefficient estimates are available for before-tax, before-transfer “market incomes,” and after-tax and after-transfer “disposable incomes,” Gini coefficients for developing countries are available only for either before-tax (but after-transfer) or after-tax (and after-transfer) incomes. Both types of Gini coefficients are available only for two countries on a common methodology.

Data are not available for before-tax and before-transfer Gini coefficient estimates and after-tax and after-transfer Gini coefficient estimates. However, the relatively small cash transfer programs in developing countries suggest that the two types of Gini coefficients for developing countries would not differ, on average, as substantially as in industrial countries.

Selected country experiences

The remainder of this section discusses the nature of tax reforms and social expenditure policy and their distributional implications in Hungary, Indonesia, and Thailand, three of the countries that experienced a large change in the Gini coefficient. All of these countries increased the ratio of direct and indirect taxes and expanded secondary school enrollment. However, their achievements in income distribution diverged. While Indonesia reduced their Gini coefficients, Thailand and Hungary experienced an increase in their Gini coefficients (Table 11).

Table 11.Selected Developing Countries: Changes in Gini Coefficient, Tax Structure, and Secondary School Enrolment
Before tax

(b) or after tax

(a)Gini
Percentage point

change in Gini
Percent change in

DT/IT
Percent change in

secondary school

enrolment
70S–80S80s–90s70S–80S80s–90s70s–80s80s–90s
Large decline in Gini
Turkeyb-7151128
Trinidad and Tobagob-7136149
Philippinesb-3100116
Malaysiab-3153118
Indonesiaa-3147190
Large increase in Gini
Singaporeb4125115
Thailandb4138112
Sri Lankab690131
Guatemalab9119146
Large decline in Gini
Mauritiusa-3105116
Large increase in Gini
Polandb4103117
Thailandb4172148
Bulgariab55991
Hungarya9124122
Sources: Tables 6, 7, 9, and 10.DT = direct taxes; IT = indirect taxes.
Sources: Tables 6, 7, 9, and 10.DT = direct taxes; IT = indirect taxes.

Hungary

During the 1980s–1990s, Hungary’s transition to a market-oriented society resulted in a large structural change in its economy, which suffered a long transitional recession. While modestly increasing both the ratio of direct to indirect taxes (by 24 percent) and the secondary school (by 22 percent) enrollment rate, Hungary suffered a 9-percentage-point increase in the after-tax Gini. The changes in the tax system and social protection programs discussed below had direct implications for the after-tax (and after-transfer) Gini. The changes in education programs probably influenced the before-tax Gini, which is not reported in this paper, and, thereby, the after-tax Gini. It is worth noting that the Gini increased, in spite of the increases in the ratio of direct to indirect taxes and the secondary school enrollment rate.

The reform of the Hungarian tax system began with the establishment of personal income and value-added taxes in 1988, followed by the introduction of a modern corporate tax in 1992. Since then, the attempt has been to improve the efficiency of the tax system by

adjusting marginal tax rates and brackets for personal and corporate taxes as well as by reducing tax exemptions. A broadening of the VAT base through rate rationalization and extension of coverage has accompanied this. The Hungarian tax system now relies more on direct and broad-based consumption taxes than in 1988. Some 40 percent of the revenue in 1998 accrued from personal, corporate, and value-added taxes compared with 30 percent in 1990.

Adjustments to the tax system have been made almost every year in the past decade. In 1988, the personal income tax consisted of 11 brackets, and the tax rates ranged from 20 to 60 percent; by 1999, there were three brackets ranging between 20 and 40 percent. The general profit tax was 50 percent in 1988, but was reduced to 20 percent of the corporate income in 1997. In 1988, the VAT comprised three rates: 0, 15, and 25 percent. The zero rate applied to most food products, fuels, pharmaceuticals, household heating, and exports; the 15 percent rate was levied on most services; and the 25 percent rate applied to about 40 percent of consumer expenditures. A major change was effected to the VAT rate structure in 1993; 80 percent of food products and household heating were subjected to the lower rate of 10 percent, and all public services began to be taxed at 25 percent. In 1998, the 10 percent rate was increased to 12 percent and zero-rating confined to pharmaceuticals and exports. Financial, education, health, and postal services are still tax-exempt.

Like other transition economies, Hungary started the transition to the market with expansive social welfare programs, including state-provided health and education services. The education system, which produced a secondary education enrollment rate of some 80 percent, enjoyed a pupil-teacher ratio that was even lower than in industrial countries. From the mid-1980s to the mid-1990s, the large reduction of untargeted price subsidies was compensated for by an increase in cash benefits to vulnerable groups, including old-age pensioners, families with children, and the unemployed.

Indonesia

During the 1970s–1980s, Indonesia achieved sustained economic growth, improved income distribution, and reduced poverty. Indonesia also made progress in improving the tax system and in expanding education programs. Between the 1970s and the 1980s, the after-tax Gini coefficient declined by three percentage points; the ratio of direct to indirect taxes increased by 50 percent, although this ratio declined subsequently, and the secondary school enrollment rate almost doubled. It appears that both tax and social spending policies had reinforced each other in reducing the after-tax Gini.

Indonesia’s tax reform during the 1980s was aimed at increasing revenues and enhancing the efficiency and simplicity of the tax system but without causing an adverse impact on the poor. The 1983 tax reform package aimed at broadening the base of the income taxes, but reducing their rates. It also included a VAT and a luxury sales tax, which were implemented in 1985. Subsequent reforms broadened the property tax. The VAT, from which unprocessed foodstuffs, certain farm products, and services were exempted, was extended later to wholesale trade and other services. The result was a broader-based, more equitable tax system.

Indonesia expanded social programs largely on the strength of its sustained economic growth. Without a considerable increase in social expenditures as percent of GDP, social indicators improved. From the 1970s to the 1980s, the secondary school enrollment rate almost doubled. Significant improvements in school enrollments among the lower-income groups took place in the late 1970s. Indonesia had achieved nearly universal primary education by the late 1980s, although there were concerns that the improvements were achieved at the expense of the quality of education.

Thailand

Thailand is a high-growth country that reduced poverty substantially. The proportion of people below the poverty line, the head count index has fallen from about 23 percent in 1981 to about 13 percent in 1992 (World Bank, 1996). More recently, in response to the financial crisis, Thailand expanded social safety net programs. During the 1970s–1990s, however, Thailand experienced a large increase in the before-tax Gini. It is unclear why the increase in the Gini occurred, in spite of a large expansion of secondary school enrollment. Between the 1980s and the 1990s, the Gini coefficient increased by four percentage points while the secondary school enrollment rate increased by 48 percent.29 It is also unclear how the changes in the tax system have changed the after-tax Gini, which is not reported. The ratio of direct to indirect taxes increased by 72 percent. This expansion of direct taxes perhaps tended to reduce the after-tax Gini. As discussed below, however, the tax system has been changed frequently, in response to fiscal needs.

Thailand’s changing direction on tax reforms has often been triggered by macroeconomic imbalances. During the 1973–76 period, for example, indirect tax rates were reduced to alleviate the burden of higher inflation and lower real GDP growth that followed the oil shock of 1973–74. The ensuing deficit spending in the next two years, which was intended to counteract slower real income growth, necessitated higher revenues. As a result, a series of tax rate increases were instituted during the 1977–79 period, followed by income tax rate reductions between 1980 and 1986, in the aftermath of the oil shock of 1979–80.

The burden of higher taxes in the 1977–79 period seems to have fallen mostly on indirect taxes that tend to be regressive; in the 1970s, indirect taxes, for example, made up almost 80 percent of all taxes. Reduction in tax rates in the 1980–86 period led to a decline in the buoyancy of personal income taxes that may have contributed to the increase in after-tax income inequality, since for a given income increase, the average effective tax rate did not increase in the 1980s as much as it did in the 1970s.30 The tax structure is complex and has been characterized by “base erosion resulting from special allowances, high standard deductions, and by failure to tax fringe benefits and non-neutrality in the tax treatment of different income sources on different transactions” (Tanzi and Shome, 1992). In 1989, Thailand carried out further tax reforms aimed at simplification, neutrality and revenue generation. The number of personal income tax (PIT) brackets were reduced from eleven to six and a greater number of low income tax payers were left out of the income tax net. No attempt was made to reduce expense deductions and allowances for business incomes.

In the early 1990s, Thailand introduced another series of tax reforms. VAT was introduced in 1992 at a rate of 7 percent which replaced an inefficient and complex business tax with 21 rates ranging from 0.10 percent to 50 percent; the number of PIT brackets was further reduced from six to four; the top PIT rate was reduced from 50 percent to 37 percent, and more incomes became subject to the lower tax rates. These tax simplifications, other things being equal, would not necessarily lead to a deterioration in income inequality. While the ratio of direct to indirect taxes rose by 70 percent from the 1980s to the 1990s, given the observed increase in income inequality in the 1990s, these reforms have not reversed the rise in income inequality, 31 nor the legacy of tax reforms of the earlier decades.

Social spending does not seem to have had much effect on income inequality, although the secondary school enrolment rate increased substantially. Government spending on transfer programs (in kind and cash) as well as employment generation programs constitute a small portion of government expenditure (about 1.6 percent during the 1990–95 period), and various programs are not well targeted toward regions with high incidence of poverty and inequality. Despite the increase in income inequality, government spending on education and health has remained steady in the last two decades at about 19 percent and 7 percent of total expenditures, respectively. There is no benefit incidence study of government transfer programs or social spending that can be used for analyzing the incidence structure of the past spending.

IV. Summary and Conclusions

The available data indicate that, before the effects of redistributive tax and transfer programs, income inequality in developing countries, on average, is lower than in industrial countries. However, while industrial countries improve income distribution effectively through taxes and transfers, developing countries do not have adequate redistributive programs to achieve a post-tax, post-transfer income equality comparative to those of industrial countries.

  • Based on the data, compiled by Deininger and Squire (1996), that are broadly comparable intertemporally and internationally, before-tax (but after-transfer) Gini coefficients for developing (and transition) countries, on average, are lower than “market-income” Gini coefficients compiled by the OECD for industrial countries. For the 1990s, the former average 38 percent; the latter average 44 percent.
  • Adequate data are not available for developing countries to assess the impact of the tax and transfer programs. There are indications that the tax and transfer programs in developing (and transition) countries are not as effective as in industrial countries. The before-tax and after-tax Gini measures for two separate groups of developing countries average, respectively, 38 percent and 34 percent. The difference does not exceed four percentage points between the two Ginis for two transition countries for which comparable data are available.
  • Existing studies on tax and transfer incidence support the argument that their redistributive effects are not as large as in industrial countries (see below).

Studies of tax incidence suggest that their redistributive effects are minor in developing (and transition) countries.

  • Developing countries have a tax structure dominated by indirect taxes and with a limited menu of capital and wealth taxes. In general, their weak tax administration gives rise to tax evasion, a marked difference between de jure and de facto tax regimes, and a low tax-to-GDP ratio. They have only limited formal cash transfer and social protection programs. These features cast doubt on the ability of tax (and transfer) policies in developing countries to redistribute income effectively. Corruption and poor governance also limit the effectiveness of taxes and transfers as redistributive instruments.
  • Only 13 of the 36 overall tax systems studies were found to be progressive, the rest were either proportional or regressive. Over time, the progressivity has declined in several developing countries.
  • It should be noted, however, that the survey found 13 tax systems to be progressive. In particular, most of the income taxes were progressive.

Education, health, and transfer programs in developing countries, in general, had a progressive incidence, but many of them were not well targeted.

  • All primary and secondary education programs, but only half of the tertiary education programs assessed, had a progressive incidence. Their targeting, however, was less effective. While all health programs were progressive, only half of them were well targeted.
  • While 14 of the 15 transfer programs were progressive, 9 of them were not well targeted.
  • The incidence of government spending programs is not easy to assess because their effectiveness matters crucially. Government expenditure on an ineffective primary education program might be more a cash transfer to teachers than a benefit to schoolchildren.

In recent decades, many developing countries have experienced an increase in income inequality based on both before-tax and after-tax measures.

  • The experiences during the 1970s–1980s appear somewhat different from those during the 1980s–1990s. From the 1970s to the 1980s, based on the data for a limited sample of developing countries, the average Gini coefficient remained unchanged, but the Gini coefficients for individual countries increased or declined, significantly in some cases.
  • From the 1980s to the 1990s, based on the data for a limited sample of developing and transition countries, the Gini coefficients for most countries increased, substantially in some cases. The tax and transfer policies in developing and transition economies were not sufficiently effective in limiting the increase in after-tax and after-transfer Gini coefficients.
  • Industrial countries also experienced a large increase in market-income Gini coefficients. Relative to developing and transition countries, however, industrial countries successfully used their taxes and transfers to limit the increase in disposable income Gini coefficients.32
  • While many global and country-specific factors might be contributing to the widespread increases in income inequality, it appears that sound economic and social policies help either limit a deterioration of income distribution or achieve its improvement. High economic growth, alone, does not appear to ensure an improvement in income distribution.
  • Possible factors underlying these changes are the impact of opening up low-income developing countries on middle-income developing countries and changes in social norms. The opening of low-income countries might have an adverse effect on the wages of unskilled workers in middle-income developing countries.
  • Countries that pursued sound macroeconomic and structural policies, including sound social policies, improved their income distribution, in spite of the limited equity objectives of their tax reforms. For example, some countries (e.g., Jamaica, Indonesia) improved income distribution, in spite of the limited equity orientation of their tax reform efforts. The experiences of 12 OECD countries in the late 1980s and the early 1990s also have shown a limited redistributive effect of certain taxes. For example, studies of industrial countries have shown no evidence for a positive association between the progressivity of the personal income tax on one hand and the difference between before-tax and after-tax Gini coefficient on the other.33
  • The use of tax instruments for redistribution remains an interesting issue that needs to be further explored. The survey finds income taxes to have progressive incidence. Should developing countries actively promote an expansion of these taxes to promote their expansion? Alternatively, should the developing countries promote an increase in tax revenue and use the revenue to expand well-targeted social programs? This question would be difficult to answer without considering individual countries’ specific circumstances, such as the nature of their tax regime and tax administration. Countries that have capacity to increase tax revenue with a degree of progressivity without causing disincentive effects on work efforts would enrich their redistributive policy instruments. The econometric estimation of the Gini equations, however, suggests that the magnitude the effect might be small. If the progressivity of the tax system were achieved at the cost of revenue, relative to the case of a neutral tax regime, the gains in redistribution on the tax side could be more than offset by lost opportunities to use progressive expenditure policy instruments.

ANNEX I
Annex Table 12Tax Incidence Studies, 1975-1998: Summary of Characteristics
Author/s, YearCountry/iesPeriodIncome ConceptHouseholds and/or

Individuals Covered
Taxes IncludedMeasures of Progressivity 1/Author’s Conclusions 2/
Country studies
Chowdhury, 1988Bangladesh1980 to 1984Taxable incomeCA and NCA 3/Income taxNot indicatedRegressive in CAs, some degree of progression in

NCAs
Hassan and Bogetic, 1996Bulgaria1992Household income and

expenditure
AllIncome taxNot indicatedModest progression
Milanovic, 1994Bulgaria1989Household incomeAllPayroll taxGini coefficientsSmall impact Payroll taxes increase income

concentration by 1.1
Foxley, 1979Chile1969Household income and

expenditure
AllDirect and indirectNot indicatedSemi-proportional
Engel, Galetovic, Raddatz, 1998Chile1996Household income and

expenditure
AllDirect and indirectGini coefficientsSlightly regressive
Gillis and McLure, 1978Colombia1974National incomeHigh-incomeDirect and indirectNot indicatedProgressive
Berry and Soligo, 1980Colombia1970Household income and

expenditure
AllDirect and indirectNot indicatedSubstantial progressivity
Chia, and others, 1992Cote d’Ivoire1986Gross household incomeAllDirect, indirect plus

subsidy
Welfare impactLarge redistributive effect but criss-crosses rich-poor

spectrum
Holthus and Shams, 1994Cote d’IvoireEarly 1990sNot indicatedNot indicatedOverall tax systemNot indicatedSlight improvement in redistributive effect
Sahn and Younger, 1998Cote d’Ivoire1985Household income and

expenditure
AllDirect and indirectConcentration curvesKerosene, excises export and import duties are

regressive, rest are progressive 4/
Milanovic, 1994Czechoslovakia1988Household incomeAllPayroll taxGini coefficientIncome concentration slightly reduced
Santana and Rathe, 1993Dominican Republic1989Household incomeAllDirect and indirectNot indicatedA degree of progressivity
El-Edel. 1979Egypt1958-59,1964-65,

1974-75
Household income and

expenditure
AllTotal net taxGini coefficientsProgressive
Younger, 1993Ghana1977-78,1981-82,

1987, 1990
Household income and

expenditure
AllDirect and indirectProgressivity coefficientRegressive in 1977-78 and 1981-82 but less so in

1987, then proportional in 1990
Galper and Ramos, 1992Guatemala1992Household incomeAllOverall tax systemNot indicatedProgressive
Bahl, and others, 1996Guatemala1994 (estimate)Household incomeAllOverall tax systemNot indicatedProgressive
Hicks and Lee, 1997Guatemala1992Household income and

expenditure
AllDirect and indirectNot indicatedMildly progressive
Sahn and Younger, 1998Guinea1993-94Household income and

expenditure
AllDirect and indirectConcentration curvesTaxes on exports and kerosene are regressive, rest

are progressive 4/
Dublin, 1979Guyana1969-70Household income and

expenditure
AllCentral government taxesNot indicatedLargely progressive
Milanovic 1994Hungary1989Household incomeAllPayroll and direct taxesGini coefficientsPayroll taxes increase income concentration by 1.2;

direct taxes reduce income concentration by 1.5
Jarvis and Pudney, 1995Hungary1988 to 1994Taxable incomeAllPersonal income taxNot indicatedProgressive, but progressivity reduced over time
Toth and Abraham, 1996Hungary1994Total declared incomeAllPersonal income taxNot indicatedProgressive
Kattuman and Redmond, 1997Hungary1987 to 1993Household incomeAllPersonal income taxDecomposition analysisProgressive and redistributive, but declining in

importance
Newberry and Revesz, 1997Hungary1991Household income and

expenditure
AllDirect and indirectNot indicatedProgressive, but reduced when analysis is restricted

to active households
Jha and Srinivasan, 1989India1983-84Personal expenditureAllIndirect taxesModified Kakwani indexMostly progressive or proportional
Murty, 1989India1979-80Household income and

expenditure
AllCommodity tax and

personal income
Atkinson indexProgressive
Aggarwal, 1990India1961-62 to 1983-84Personal incomeAllPersonal income taxIndex of redistributive

impact
Redistributive, despite decreasing progressivity
Aggarwal, 1991India1961-62 to 1983-84Personal incomeAllPersonal income tax“Global” index of

progressivity
Redistributive, despite decreasing progressivity
Aggarwal, 1994India1961-62 to 1983-84Personal incomeAllPersonal income taxRSTP and RISP IndicesRedistributive, despite decreasing progressivity
Achdut, 1996Israel1979-93Household incomeAllDirect taxesDecomposition analysisFrom 1979-84 taxes reduced inequality, however,

redistributive effect weakened from 1985 onwards
Wasykenko, 1986Jamaica1975Household incomeAllDirect and indirectNot indicatedBoth pro-poor and pro-rich
Bird and Miller, 1989Jamaica1983-84Household expenditureLow-incomeIndirect taxesNot indicatedProportional
Wasylenko, 1990Jamaica1983 and 1986Comprehensive

household income
AllDirect and indirectNot indicatedIn 1983, first 8 deciles progressive, then regressive;

estimated more progressivity after reform
Alm, Bahl, and Murray, 1991Jamaica1983Comprehensive

individual income
Self-employed and employeesIncome taxNot indicatedHighly regressive
Sjoquist and Green, 1992Jamaica1990-91Not indicatedAllNot indicatedNot indicatedSlightly more progressivity after 1986 reform
Holthus and Shams, 1994JamaicaEarly 1990sNot indicatedNot indicatedOverall tax systemNot indicatedNo improvement in distributional effects
Mwega, 1986Kenya1976Household incomeAllDirect and indirectNot indicatedMixed but broadly progressive
Heller, 1981Korea1976Household incomeAllDirect, indirect, and

property
Not indicatedRoughly proportional: progressive for top 10 percent
Leipziger, and others, 1992Korea1970 to 1986Household and personal

income
AllOverall tax systemReview of 4 studiesNeutral with respect to income distribution
Choi. 1997Korea1976 to 1991Household and personal

income
AllNational and localReview of 6 studiesMore or less regressive
Sahn and Younger, 1998Madagascar1993Household income and

expenditure
AllDirect and indirectConcentration curvesTaxes on exports and kerosene are regressive, rest

are progressive 4/
Andic, 1974Malaysia1970Cash and in-kind

household income
Low-incomeRubber export taxes,

import duties, excises, land

and religious taxes
Not indicatedRoughly proportional
Gil-Diaz, 1982Mexico1972,1980Permanent household

income
AllDirect, indirect and

inflation
Not indicatedProgressive but uneven in 1972, smoother in 1990
Malik and Saqib, 1989Pakistan1978-79Household incomeAllAll federal taxes, except

export duties
Suits indexSlightly progressive
Escobal, and others, 1993Peru1985-90Household expenditureAllFuel taxesNot indicatedGasoline tax is progressive, kerosene tax is

regressive.
Yoingco and Guevara, 1993Philippines1990Gross household incomeAllDirect and indirectNot indicatedGenerally progressive, though some regressivity in

first quintile noted
Devarajan and Hossain, 1998Philippines1988, 1989Household incomeAllDirect and indirectNot indicatedBasically neutral
Milanovic, 1994Poland1989Household incomeAllPayroll taxGini CoefficientsIncome concentration increased by 1
Bolkowiak, and others, 1996 Poland1994Gross household incomeEmployees, pensionersDirect taxesNot indicatedProgressive over first four quintiles, then regressive
Bolkowiak, and others, 1996Poland1994Gross household incomeEmployees, pensionersIndirect taxesNot indicatedProgressive among employees, U-shaped among

pensioners
Huang, 1976Tanzania1971Household incomeAllDirect and indirectNot indicatedProgressive
Sahn and Younger, 1998Tanzania1991,1995Household income and expenditureAllDirect and indirectConcentration curvesKerosene tax is regressive, rest are progressive 4/
Kibuka, 1977Uganda1961 to 70Household income and expenditureAllDirect and indirectNot indicatedProgressive, biased against rural households
Kakwani, 1996Ukraine1989,1991,1992Personal incomeAllPersonal income taxGini coefficient, progressivity indexIncreasing progressivity
Milanovic, 1994Yugoslavia1989Household incomeAllPayroll and direct taxesGini coefficientsDirect taxes no impact; payroll taxes reduced income

inequality by 0.6
Surveys
McLure, 1977Colombia1970Progressive, but not smooth
Brazil1962-63
Pakistan1966-67
India1963-64
Gujarat1967-68
Mysore1968-69
Lebanon1968
Jamaica1971-72
Shah and Whalley. 1990PakistanOverall tax system is broadly progressive
Colombia
Sri Lanka
Jamaica
Brazil
Jamaica
Thirsk, 19975/BoliviaNo case of tax reform has led to progressive
Colombiadistribution of tax burdens
Indonesia
Jamaica
Korea
Mexico
Morocco
Turkey
Sources: As indicated.

Unless otherwise indicated, these studies do not employ an explicit quantitative measure of progressivity or redistributive impact. Instead, effective tax rates by income groups are visually inspected.

When several types of taxes are included in the study, the conclusion refera to the overall tax burden (e.g., direct plus indirect taxes), unless otherwise indicated.

Company assessees (CAs) and non-company assessees (NCAs).

Progressive taxes include the VAT, all excises, taxes on wage, transport, and gasoline, and, except for Côte d’Ivoire, import duties.

Thirsk’s paper is an overview chapter from an anthology of articles on tax reform in developing countries and not, strictly speaking, a review of the literature.

Sources: As indicated.

Unless otherwise indicated, these studies do not employ an explicit quantitative measure of progressivity or redistributive impact. Instead, effective tax rates by income groups are visually inspected.

When several types of taxes are included in the study, the conclusion refera to the overall tax burden (e.g., direct plus indirect taxes), unless otherwise indicated.

Company assessees (CAs) and non-company assessees (NCAs).

Progressive taxes include the VAT, all excises, taxes on wage, transport, and gasoline, and, except for Côte d’Ivoire, import duties.

Thirsk’s paper is an overview chapter from an anthology of articles on tax reform in developing countries and not, strictly speaking, a review of the literature.

Bibliography

    AchdutLea1996Income Inequality, Income Composition, Macroeconomic Trends: Israel, 1979–93EconomicaVol. 63Supplement pp. S127.

    AggarwalPawan K.1990An Empirical Analysis of the Redistributive Impact of the Personal Income Tax: A Case Study of IndiaPublic FinanceVol. 45 pp. 17792.

    • Search Google Scholar
    • Export Citation

    AggarwalPawan K.1991A New Global Measure of ProgressivityNational Institute of Public Finance and Policy Working Paper No. 6/91 (New Delhi: NIPFP).

    • Search Google Scholar
    • Export Citation

    AggarwalPawan K.1994A Local Distribution Measure of Tax ProgressivityPublic FinanceVol. 49 pp. 111.

    AhujaVinodBenuBidaniFranciscoFerreira and MichaelWalton1997Everyone’s Miracle? Rising Poverty and Inequality in East Asia (Washington D.C.World Bank)

    • Crossref
    • Search Google Scholar
    • Export Citation

    AlmJamesRoyBahl and Matthew N.Murray1991Tax Base Erosion in Developing CountriesEconomic Development and Cultural ChangeVol. 39No. 4 (July) pp. 849872.

    • Crossref
    • Search Google Scholar
    • Export Citation

    AndicFuat M. and Aris BinOthman1974The Impact of the Tax System on the Poor: with Special Reference to Rural Poor: Draft Report” (mimeograph).

    • Search Google Scholar
    • Export Citation

    AninatEduardoAndreasBauer and KevinCowan1999Addressing Equity Issues in Policymaking: Lessons from the Chilean Experience” in Economic Policy and Equityed. by VitoTanziKe-youngChu and SanjeevGupta(Washington: International Monetary Fund).

    • Search Google Scholar
    • Export Citation

    AlesinaAlberto1999Too Large and Too Small Governments” in eds.VitoTanziKe-youngChu and SanjeevGuptaEconomic Policy and Equity (Washington: International Monetary Fund)

    • Search Google Scholar
    • Export Citation

    AtkinsonA.B.1999Equity Issues in a Globalizing World: The Experience of OECD Countries” in Economic Policy and Equityeds.VitoTanziKe-youngChu and SanjeevGupta(Washington: International Monetary Fund).

    • Search Google Scholar
    • Export Citation

    BahlRoyJorgeMartinez-Vazquez and SallyWallace1996The Guatemalan Tax Reform (Boulder: Westview Press).

    BerryR. Albert and RonaldSoligoeds.1980Economic Policy and Income Distribution in Colombia (Boulder: Westview Press).

    BirdRichard M. and Barbara DianeMiller1989The Incidence of Indirect Taxes on Low-Income Households in JamaicaEconomic Development and Cultural ChangeVol. 37 (January) pp. 393409.

    • Crossref
    • Search Google Scholar
    • Export Citation

    BirdRichard M. and L.De Wulf1973Taxation and Income Distribution in Latin America: A Critical Review of Empirical StudiesStaff PapersInternational Monetary FundVol. 20No. 3 pp. 639682.

    • Search Google Scholar
    • Export Citation

    BolkowiakIzabel and others 1996Tax Burden on Gross Incomes of HouseholdsInstitute of Finance Working Paper No. 49 (Warsaw: Institute of Finance).

    • Search Google Scholar
    • Export Citation

    BovenbergLans A.1987Indirect Taxation in Developing Countries: A General Equilibrium ApproachStaff PapersVol. 342 pp. 333373.

    ChiaNgee-ChoonSadekWahba and JohnWhalley1992Tax Incidence Analysis in Developing Countries: The Case of Côte d’Ivoire (Washington: World Bank).

    • Search Google Scholar
    • Export Citation

    ChoiKwang1997Tax Policy Reforms in Korea” in Tax Reform in Developing Countriesed. by WayneThirsk(Washington: World Bank).

    • Search Google Scholar
    • Export Citation

    ChowdhuryNuimuddin1988Income Tax Incidence in Bangladesh, 1980–84Bangladesh Development StudiesVol. 16 (September) pp. 8197.

    • Search Google Scholar
    • Export Citation

    ChowdhuryOsman H.1991Revenue Neutral Value-Added Tax in Babgladesh: Some General Equilibrium IllustrationsBangladesh Development Studiesvol. 19no. 4 pp. 4963.

    • Search Google Scholar
    • Export Citation

    ClareteRamon L.1991A General Equilibrium Analysis of the Tax Burden and Institutional Distortions in the Philippines” in Tax Policy in Developing Countries: A World Bank Symposiumed. by JavadKhalilzadeh-Shirazi and AnwarShah(Washington: World Bank).

    • Search Google Scholar
    • Export Citation

    DavoodiHamid R. and SwaitreeSachjapinan1999How Useful are Benefit Incidence StudiesforthcomingIMF Working Paper. (Washington: International Monetary Fund).

    • Search Google Scholar
    • Export Citation

    DevarajanShantayana and ShaikhHossain1998The Combined Incidence of Taxes and Public Expenditures in the PhilippinesWorld DevelopmentVol. 26 pp. 963977.

    • Crossref
    • Search Google Scholar
    • Export Citation

    DeiningerKlaus and LynSquire1996A New Data Set for Measuring Income InequalityWorld Bank Economic Review10 pp. 565592.

    De WulfLuc1975Fiscal Incidence in Developing Countries: Survey and CritiqueStaff PapersInternational Monetary FundVol. 22No. 1 pp. 61131

    • Search Google Scholar
    • Export Citation

    DublinKeith Gordon Newton1979An Estimation of Tax Incidence in Guyana” (Ph.D. dissertation; Urbana-Campaign: University of Illinois).

    • Search Google Scholar
    • Export Citation

    EdirisngheNeville1987The Food Stamp Scheme in Sri Lanka: Costs Benefits and Options for Modification Research Report No. 58 (Washington: International Food Policy Research Institute).

    • Search Google Scholar
    • Export Citation

    El-EdelM. Reda A.1979Impact of Taxation on Income Distribution: An Exploratory Attempt to Estimate Tax Incidence in EgyptPrinceton-Egypt Income Distribution Project Conference Paper.

    • Search Google Scholar
    • Export Citation

    EngelEduardo M.R.A.AlexanderGaletovic and Claudio E.Raddatz1998Taxation and Income Distribution in Chile: Some Unpleasant Redistributive ArithmeticNBER Working Paper No. 6828 (Cambridge, Massachusetts: National Bureau of Economic Research).

    • Search Google Scholar
    • Export Citation

    EscobalJavier and others 1993Public Administration and Income Distribution in Peru” in Government Spending and Income Distribution in Latin Americaed. by RicardoHausmann and RobertoRigobon(Washington: Inter-American Development Bank).

    • Search Google Scholar
    • Export Citation

    FoxleyAlejandro1979Redistributive Effects of Government Programs: The Chilean Case (Oxford: Pergamon Press).

    FreemanRichard B.1995Are Your Wages Set in BeijingJournal of Economic Perspectives (Summer1995).

    GalperHarvey and FernandoRamos1992The Incidence of Guatemalan TaxTechnical Memorandum No. 32 (Atlanta: Georgia State University) cited in The Guatemalan Tax Reform, 1996, ed. by Bahl Roy, Jorge Martinez-Vazquez, and Sally Wallace (Boulder: Westview Press).

    • Search Google Scholar
    • Export Citation

    GillespieIrwin W.1965Public Expenditures and Income Distribution” in Essays in Fiscal Federalismed. by Richard A.Musgrave(Washington: The Brookings Institute).

    • Search Google Scholar
    • Export Citation

    GillisMalcolm and Charles E.McLureJr.1978Taxation and Income Distribution: The Colombian Tax Reform of 1974Journal of Development EconomicsVol. 5 pp. 23358.

    • Crossref
    • Search Google Scholar
    • Export Citation

    Gil-DiazFrancisco1982The Incidence of Taxes in Mexico, A Before and After Comparison” (mimeograph).

    GroshMargaret1994Administering Targeted Social Programs in Latin America: From Platitudes to Practice (Washington: World Bank).

    HammerJeffrey S.IjazNabi and James A.Cercone1995Distributional Effects of Social Sector Expenditures in Malaysia, 1974–89” in Public Spending and The Poor: Theory and Evidenceed. by Dominiquevan de Walle and KimberlyNead(Baltimore: Johns Hopkins University Press).

    • Search Google Scholar
    • Export Citation

    HandaSudhanshu and DomienKing1997Structural Adjuistment Policies, Income Distribution a, d Poverty: A review of the Jamaican ExperienceWorld Developmentvol. 25no. 6 pp. 915930.

    • Crossref
    • Search Google Scholar
    • Export Citation

    HellerPeter S.1981Testing the Impact of Value-Added and Global Income Tax Reforms on Korean Tax Incidence in 1976: An Input-Output and Sensitivity AnalysisStaff PapersInternational Monetary FundVol. 28 (June) pp. 375410.

    • Crossref
    • Search Google Scholar
    • Export Citation

    HarbergerArnold C.1998Monetary and Fiscal Policy for Equitable Economic Growth” in eds.VitoTanzi and Ke-youngChuIncome Distribution and High-Quality GrowthThe MIT PressCambridge, England.

    • Search Google Scholar
    • Export Citation

    HassanFareed M.A. and ŽeljkoBogetic1996Effects of Personal Income Tax on Income Distribution: Example from BulgariaContemporary Economic PolicyVol. 14 (October) pp. 1728.

    • Crossref
    • Search Google Scholar
    • Export Citation

    HicksNorman L. and HaeduckLee1997The Impact of Fiscal Policies on the Poor in GuatemalaJournal of Public Budgeting Accountingand Financial ManagementVol. 9 (Fall) pp. 50115.

    • Search Google Scholar
    • Export Citation

    HolthusManfred and RasulShams1994Tax Reforms and Equity in ACP CountriesIntereconomics: Review of International Trade and DevelopmentVol. 29 (September/October) pp. 239243.

    • Crossref
    • Search Google Scholar
    • Export Citation

    HuangYukon1976Distribution of the Tax Burden in TanzaniaEconomic JournalVol. 86 (March) pp. 7386.

    HubbardGlen1995Distribution Tables and Tax Policy” in Distributional Analysis of Tax Policyed.David F.Bradford(Washington: AEI Press).

    • Search Google Scholar
    • Export Citation

    JarvisSarah and StephenPudney1995Redistributive Policy in a Transition Economy: The Case of HungaryCEPR Discussion Paper No. 1117 (London: Centre for Economic Policy Research).

    • Search Google Scholar
    • Export Citation

    JhaShikha and P.V.Srinivasan1989Indirect Taxes in India: An Incidence AnalysisEconomic and Political WeeklyVol. 24 (April) pp. 81130.

    • Search Google Scholar
    • Export Citation

    KakwaniNanak1996Income Inequality, Welfare and Poverty in UkraineUniversity of New South Wales School of Economics Discussion Paper No. 94 (Australia: University of New South Wales).

    • Search Google Scholar
    • Export Citation

    KattumanPaul and GerryRedmond1997Income Inequality in Hungary, 1987–1993University of Cambridge Department of Applied Economics DAE Working Paper No. 9726 (U.K.: University of Cambridge).

    • Search Google Scholar
    • Export Citation

    KibukaRobin D.1977Income Distribution and Fiscal Incidence in Uganda, 1961–1970: A Thesis” (Ph.D. dissertation; Cambridge, Massachusetts: Harvard University).

    • Search Google Scholar
    • Export Citation

    LeipzigerDanny M. and others 1992The Distribution of Income and Wealth in KoreaEDI Development Studies (Washington: World Bank).

    • Search Google Scholar
    • Export Citation

    McLureCharles E.1977Taxation and the Urban Poor in Developing CountriesWorld DevelopmentVol. 5No. 3 pp. 169188.

    McLureCharles E.1974On the Theory and Methodology of Estimating Benefit and Expenditure IncidenceProgram of Development Studies (William Marsh Rice University).

    • Search Google Scholar
    • Export Citation

    MalikMuhammad Hossain and Najam usSaqib1989Tax Incidence by Income Classes in PakistanThe Pakistan Development ReviewVol. 28 (Spring) pp. 1326.

    • Search Google Scholar
    • Export Citation

    MilanovicBranko1995Impact of Cash and In-Kind Transfers in Eastern Europe and Russia” in Public Spending and the Poored. by Dominiquevan de Walle and KimberlyNead(Baltimore: Johns Hopkins University Press).

    • Search Google Scholar
    • Export Citation

    MilanovicBranko1994Cash Social Transfers, Direct Taxes, and Income Distribution in Late SocialismJournal of Comparative EconomicsVol. 18 (April) pp. 175197.

    • Crossref
    • Search Google Scholar
    • Export Citation

    MurtyM.N.1989Redistributive Taxation and Public Sector Pricing: Some Theoretical Issues and The Indian Experience (New Delhi: Commonwealth Publishers).

    • Search Google Scholar
    • Export Citation

    MwegaF. M.1986Incidence of Taxes and Transfers in Kenya: A General Equilibrium AnalysisEastern African Economic ReviewVol 2 (June) pp. 613.

    • Search Google Scholar
    • Export Citation

    NewberryDavid M.G. and TamaszRevesz1997The Burden and Disincentive Effects of Hungarian Personal Taxes, 1988–96CEPR Discussion Paper No. 1570 (London: Centre for Economic Policy Research).

    • Search Google Scholar
    • Export Citation

    SahnDavid and StephenYounger1998Poverty and Fiscal Policy: Microeconomic Evidence” (mimeograph; Cornell University).

    SantanaIsidora and MagdalenaRathe1993The Distributive Impact of Fiscal Policy in the Dominican Republic” in Government Spending and Income Distribution in Latin Americaed. by RicardoHausmann and RobertoRigobon(Washington: Inter-American Development Bank).

    • Search Google Scholar
    • Export Citation

    SeldenThomas M. and Michael J.Wasylenko1992Benefit Incidence Analysis in Developing CountriesWorld Bank Policy Research Working Paper No. 1015 (Washington: World Bank).

    • Search Google Scholar
    • Export Citation

    ShahAnwar and JohnWhalley1990An Alternative View of Tax Incidence Analysis for Developing CountriesPRE Working Paper No. 462 (Washington: World Bank).

    • Search Google Scholar
    • Export Citation

    ShahAnwar and JohnWhalley1990An Alternative View of Tax Incidence Analysis for Developing CountriesPRE Working Paper No. 462 (Washington: World Bank).

    • Search Google Scholar
    • Export Citation

    ShomeParthasarathi1995Tax Policy Handbook. (Washington: International Monetary Fund)

    SjoquistDavid L. and DavidGreen1992Distribution of Tax BurdensJamaica Tax Review Project Report No. 3 (Atlanta: Georgia State University) cited in Bahl Roy, 1997, “The Jamaican Tax Reform: Its Design and Performance,” in Tax Reform in Developing Countries, ed. by WayneThirsk (Washington: World Bank).

    • Search Google Scholar
    • Export Citation

    TanziVito1974Redistributing Income Through the Budget in Latin AmericaBanco Nazionale del Lavoro Quarterly ReviewVol. 27(March) No. 108 pp. 6587.

    • Search Google Scholar
    • Export Citation

    TanziVito1998Fundamental Determinants of Inequality and the Role of GovernmentIMF Working Paper 98/178 (Washington: International Monetary Fund).

    • Crossref
    • Search Google Scholar
    • Export Citation

    TanziVitoKe-youngChu and SanjeevGupta1999Economic Policy and Equity. (Washington: International Monetary Fund).

    TanziVito and Ke-youngChu1998Income Distribution and High-Quality Growth. MIT Press.

    TanziVito and LudgerSchuknecht1997Reconsidering the Fiscal Role of Government: The International ExperienceAmerican Economic Review872 pp. 164168.

    • Search Google Scholar
    • Export Citation

    TanziVito and ParthasarathiShome1992The Role of Taxation in the Development of East Asian Economies” in TakatoshiIto and Anne O.KruegerThe Political Economy of Tax ReformThe University of Chicago Press (Chicago and London).

    • Search Google Scholar
    • Export Citation

    ThirskWayne1997Overview: The Substance and Process of Tax Reform in Eight Developing Countries” in Tax Reform in Developing Countriesed. by WayneThirsk(Washington: World Bank).

    • Crossref
    • Search Google Scholar
    • Export Citation

    ThirskWayne1997Overview: The Substance and Process of Tax Reform in Eight Developing Countries” in Tax Reform in Developing Countriesed. by WayneThirsk(Washington: World Bank).

    • Crossref
    • Search Google Scholar
    • Export Citation

    TothI.J. and A.Abraham1996Income Structure and Distribution of the Tax Burden (a study of the personal income tax returns of 1994)Acta Oeconomica: A Periodical of the Hungarian Academy of SciencesVol. 48Nos. 3–4 pp. 27195.

    • Search Google Scholar
    • Export Citation

    WagstaffAdamEddyVan Doorslaer and PierellaPaci1989Equity in the Finance and Delivery of Health Care: Some Tentative Cross-Country ComparisonsOxford Review of Economic PolicyVol. 5 (Spring).

    • Crossref
    • Search Google Scholar
    • Export Citation

    WarrPeter G. and BhanupongNidhiprabha1996Thailand’s Macroeconomic Miracle: Stable Adjustment and Sustained GrowthOxford University Press.

    • Crossref
    • Search Google Scholar
    • Export Citation

    WasylenkoMichael J.1986The Distribution of Tax Burden in Jamaica: Pre-1985 Reform (Syracuse, New York: Metropolitan Studies ProgramSyracuse University for Board of Revenue, Government of Jamaica).

    • Search Google Scholar
    • Export Citation

    WasylenkoMichael J.1990Tax Burden Before and After Reform” in The Jamaican Tax Reformed. by Roy W.Bahl(Cambridge, Massachusetts: Lincoln Institute of Land Policy).

    • Search Google Scholar
    • Export Citation

    WestlakeM.J.1973Tax Evasion, tax Incidence and the Distribution of Income in KenyaEastern Africa Economic ReviewVol. 5no. 1 pp. 127

    • Search Google Scholar
    • Export Citation

    WhalleyJohn1984Regression or Progression: the Taxing Question of Incidence AnalysisCanadian Journal of EconomicsVol. 17no. 4 pp. 654682.

    • Crossref
    • Search Google Scholar
    • Export Citation

    YoingcoAngel Q. and Milwida M.Guevara1993Study on the Incidence of the Philippine Fiscal System: Summary and RecommendationsAPTIRC BulletinVol. 11 (January) pp. 4044.

    • Search Google Scholar
    • Export Citation

    YoungerStephen D.1993Estimating Tax Incidence in Ghana: An Exercise Using Household DataCornell Food and Nutrition Policy Program Working Paper No. 48.

    • Search Google Scholar
    • Export Citation

    YitzhakiShlomoJeffrey D.Lewis1996Guidelines on Searching for a Dalton- Improving Tax Reform: An Illustration with Data from IndonesiaThe World Bank Economic ReviewVol. 10no. 3 pp. 54162.

    • Crossref
    • Search Google Scholar
    • Export Citation

    ZeeHowell1999Inequality and Optimal Redistributive Tax and Transfer PoliciesIMF Working Paper 99/60 (Washington D.C.: International Monetary Fund).

    • Crossref
    • Search Google Scholar
    • Export Citation
1The authors wish to thank Vito Tanzi, Michael Keen, John Norregaard, Gustavo Yamada, and Chi-Wa Yuen for helpful comments, and Randa Sab and Erwin Tiongson for efficient research assistance. An earlier version of this paper was presented at a World Institute for Development Economics Research conference on “Rising Income Inequality and Poverty Reduction: Are They Compatible?” held in Helsinki, Finland, July 16–18, 1999. The paper will be published in the conference proceedings. The authors are thankful to Giovanni Andrea Cornia for comments and suggestions on an earlier draft.
2Alesina (1999) discusses the role of poor governance and the vicious cycle of low tax-to-GDP ratio and tax evasion.
3See Tanzi (1998), who notes these and other practical problems involved in administering a progressive tax system in developing countries.
5The income shares of the poorest quintile range between 4 percent and 11 percent for the “b” countries, and between 5 percent and 9 percent for the “a” countries. The ratios between the income shares of the richest and poorest quintiles range between 3 and 16 for the former countries and 4 and 10 for the latter. Gini coefficients, in general, are low for transition countries (e.g., Czechoslovakia, Romania, Yugoslavia, Poland, and Bulgaria).
6The Gini measure of income inequality as reported in the OECD (1999) accounts for the economies of scale of household consumption, whereas the measure in Deininger and Squire (1996) does not. The OECD data might better capture capital incomes and the incomes of self-employed persons than the Deininger and Squire data on developing countries.
7Any study of tax incidence confronts problems, including the coverage of incomes and taxes, assumptions about me shifting of the tax burden and its measurement, estimation of counterfactual before-tax income distribution, behavioral responses of individuals, and the time horizon of the analysis. These problems make cross-country comparisons of tax incidence and even intertemporal within-country comparisons difficult, and more so when a country’s major tax reform has made the task of reaching a firm conclusion regarding tax incidence difficult.
8These conclusions are based primarily on the calculation (and visual inspection) of average effective tax rates by income class whenever reported in a study and some notion of departures from proportionality.
9These conclusions are based on studies that rely on a standard tax incidence analysis with shifting assumptions that in general determine the outcome of the incidence studies. Typically, personal income taxes and payroll taxes are assumed to be borne by the taxed income recipient, whereas indirect taxes are assumed to be shifted forward to consumers of taxed commodities. Incidence of corporate taxes is controversial, however. In an earlier study, Whalley (1984) shows how a tax system can be made to appear sharply progressive or sharply regressive by changing a number of shifting assumptions.
10For example, VATs have a progressive incidence in four African countries (Sahn and Younger, 1998). Sahn and Younger, 1998 provide no explanation for this nonconventional finding. See Bird and De Miller (1986); Srinivasan (1989); and Shah and Whalley (1990 and 1991).
11This section draws on Davoodi and Sachjapinan (forthcoming).
12There are three methods of measuring the incidence of public expenditure: the individual’s own valuation of public goods, expenditure incidence analysis, and benefit incidence analysis. The first approach involves eliciting the prices that individuals are willing to pay for public goods (Aaron and McGuire, 1970). This is a demanding task, because of the well-known problems associated with the provision of public goods (e.g., free-riding, non-rivalry). The other two methods attempt to circumvent these problems by making various assumptions about the “ultimate” beneficiaries of publicly financed goods and services. Identification of the “ultimate” beneficiaries of government spending under any of the three approaches is also a problem that is shared with the identification of the tax burden under tax incidence analysis, because benefits of government spending can be shifted just as much as tax burden can.
13For example, as Tanzi (1974) points out in the context of Latin America, “… those benefits are never allocated. What gets allocated is the cost. But costs may not be any indication of benefits and the fact that two children get the same cost imputed to them tells nothing about the benefits that they will derive from that experience… a child from the urban middle class is more likely to get a great benefit from the same educational spending than one from the rural subsistence sector.”
14These definitions are used, among others, by Selden and Wasylenko (1992); and Castro-Leal, Dayton, Demery and Mehra (1999).
15For analytical convenience, the poor are assumed to be in the bottom quintile and the rich in the top quintile.
16Note the asymmetry in the definition with that for tax progressivity.
17This pattern holds regardless of whether benefits are expressed as percent of income or expenditure.
18The coverage of the Middle East is too narrow as Tunisia is the only country with data in this region.
19Using the two latest incidence data for each country, changes in benefit incidence are measured as changes in the ratio of Q1 to Q5, where Q1 and Q5 have been defined previously. As for categories of education, there are more countries showing improvements than deterioration for government primary education and tertiary education spending, but there are as many cases of improvement as deterioration for government secondary education spending.
20Unlike education, which has three well-defined subcategories, health spending is disaggregated into many categories. The findings here are based on a pattern of health disaggregation that is frequently found among the countries surveyed.
21This pattern holds regardless of whether benefits are expressed as percent of income or expenditure.
22Spending on each subcategory of health spending is, on average, well targeted except for spending on hospital outpatients. Sub-Saharan Africa and transition countries are the only regions in which each subcategory of health spending is poorly targeted, thus mirroring the pattern observed at the level of total health expenditure.
23Although the changes in Gini coefficients in general are consistent with the changes in quintile shares, results for some countries need to be interpreted with caution. For example, Trinidad and Tobago and Turkey, for which data show an improvement in income distribution, had only three surveys during the period 1970s–1980s. The before-tax Gini coefficient for Trinidad and Tobago, at 51 percent, was unusually high in 1971, compared with 1959, 1976, and 1981, when it ranged between 42 percent and 46 percent. The before-tax Gini coefficient for Turkey, declined from 56 percent in 1968 to 51 percent in 1973 and to 44 percent in 1986, with corresponding increases in the income share for the poorest quintile from 3 percent in 1968 to 4 percent in 1973 and 5 percent in 1986.
24Trinidad and Tobago and Turkey, the two countries that each reduced their Gini coefficients by seven percentage points, achieved the improvement in income distribution by reducing both the income shares of the poorest and the richest quintiles, but reducing the latter more than the former. Among the countries that experienced a deterioration in income distribution, both Thailand (a high-growth country) and Guatemala (a low-growth country) achieved a large increase in the income share of the richest quintile accompanied by a reduction in the share for the poorest quintile.
25The estimation, based on data representing averages for decades, does not support the hypothesis that inflation has a strong effect on income distribution. These results differ from those of the regressions based on annual or quarterly data, which often suggest significant effects of inflation on income distribution. This difference might suggest that, while inflation affects short-run changes in income distribution, it may not affect significantly the changes in income distribution through decades. The inflation dummy used in Equation (1.2) assumes 1 if the rate of inflation exceeds 100 percent per annum; 0 otherwise. Alternative inflation thresholds do not change the results significantly.
26Assume that r (ratio of direct to indirect taxes) is raised from 80 percent to 120 percent, by increasing d (ratio of direct taxes to GDP) from 5.4 percent to 6.6 percent (and, to keep the ratio of direct and indirect taxes to GDP constant, by reducing the ratio of indirect tax to GDP from 6.7 percent to 5.5 percent). On the basis of Equation (1.1.A), the effect of these changes on the Gini may be estimated by -0.025(40) + 0.0008[40(5.4) + 1.2(80)] = -0.75.
27See Atkinson (1999), for a discussion of the increasing inequality in industrial countries.
28See Freeman (1995) for a discussion of the effect of the opening up of low-income countries on the wages of unskilled workers in industrial countries and Tanzi (1998) and Atkinson (1999) for a discussion of the effects of changes in social norms for income distribution.
29Studies of Thailand’s income inequality based on household surveys show that the increase in income inequality over the last three decades have been driven primarily by structural factors, such as levels of education (particularly secondary education), location of households (e.g., urban/rural regions) and occupational characteristics of households (e.g., entrepreneur, farm laborer). See Ahuja, Bidani, Ferreira and Walton (1997).
30The buoyancy of PIT declined to 2.2 in the 1980s from 2.7 in the 1970s (Warr and Nidhiprabha, 1996).
31Of course the counterfactual exercise always remains a credible explanation, i.e., the rise in income inequality would have been more in the 1990s, had the tax system in the 1990s remained the same as in the 1980s.
32The implications of this redistributive effort for an improvement of social indicators is a separate issue. See Tanzi and Schuknecht (1997) for a discussion of how the large increase in transfer programs industrial countries in recent decades had only limited effects on their social indicators.
33See Wagstaff and others (1999). The before-tax and after-tax Gini refer to the Gini based on personal income tax alone.

Other Resources Citing This Publication