Information about Asia and the Pacific Asia y el Pacífico
Journal Issue

Do public expenditures reduce inequality?: Concern over income distributional aspects of fiscal policies is growing.

International Monetary Fund. External Relations Dept.
Published Date:
September 1974
  • ShareShare
Information about Asia and the Pacific Asia y el Pacífico
Show Summary Details

Luc De Wulf

Until recently development policy has been primarily concerned with economic growth. In light of the widely accepted view that economic growth is a precondition for a more equal distribution of income, little attention was given to existing inequalities prevailing in many less developed countries. These income inequalities, however, are becoming less and less politically acceptable. Concern with the income distributional aspects of fiscal policies has thus acquired new respectability.

In a previous article in Finance and Development (March 1974), Richard M. Bird dealt with the impact of taxation on income distribution and concluded that available evidence did not support the pessimistic view that taxation was not used—and could not be used—to improve the distribution of income generated by the economic system. On the contrary, the article argued that the tax system has influenced primary (pre-tax) income distribution through policies affecting employment and the relative rewards received by labor and capital. In addition it was argued that those who would like to see fiscal policy used as a tool to improve the distribution of income should give closer attention to wealth taxation. But while taxes can, in certain circumstances, improve income distribution by making the rich relatively less rich, they cannot substantially improve the wretched living conditions of the poor. Government expenditures are obviously more relevant for this purpose.

Empirical research

Robert S. McNamara, President of the World Bank, expressed this feeling when he stated at the Annual Meeting of the Bank Group and the Fund in 1972 that a “shift in the patterns of public expenditures represents one of the most effective techniques a government possesses to improve the condition of the poor.” In order to guide governments in such policies he urged more empirical research into the distributional effects of government expenditures. A number of such studies have already been made. What were their conclusions and what are the main problems facing researchers in this field?

Interest in who benefits from government expenditures is not a recent phenomenon, nor one that is limited to less developed countries. In the mid-1960s, for example, elaborate studies were carried out into the benefit incidence of government expenditures in Canada and the United States. However, my concern is primarily with developing countries and I have reviewed 14 studies covering government expenditures in Brazil, Colombia, El Salvador, Guatemala, India, Malaysia, Panama, and Puerto Rico. (The authors were usually academics, some of whom had previously estimated tax incidence in the countries concerned.) Eleven of these studies suggest that the upper-income groups receive benefits from government expenditures that are a smaller share of their incomes than their lower-income counterparts. Only three studies indicate that government expenditures increase all incomes proportionately and are thus neutral from a distributional point of view. The impression left is clearly that government expenditures in most countries redistribute incomes in favor of the disadvantaged. But do these figures mean what they seem to say?

Are these results valid?

Goods and services sold in the market place fetch a price on which both producer and consumer agree. According to the propositions of welfare economics, the rational consumer, limited by his income and guided by his preferences and needs, decides, before purchasing any goods, whether the utility he expects to derive from the last unit of that product equals the sacrifice involved in parting from a certain amount of purchasing power. In doing so, the consumer provides the observer with some objective measure of his utility or benefit. As there is a price for all goods and services he can buy, it is theoretically possible to measure the total utility he derives from the consumption of a certain number of products. Admittedly, such a procedure is very simplistic. It values the products in terms of the utility derived from the last unit bought and ignores the fact that this depends crucially on the number of identical products already bought, which in turn tends to vary systematically with the income position of the buyer. It ignores, for instance, that, although the market price for the different buyers is identical, the utility derived from the acquisition of a radio is different for the person whose radio is the only source of information and musical entertainment and for the person who already owns a sophisticated stereo set but wants an additional radio for use near his swimming pool.

These considerations are to some extent taken into account when economists make value judgments in discussing the prevailing income distribution and the “ideal” degree of tax progression. However, such welfare propositions—enormously oversimplified as they are—are all too swiftly shunted aside when one turns to the benefits from government expenditures.

The problem in this case is far greater since not even such an approximate valuation process exists for the products and services provided by government. Decisions as to the types and quantity of products to be produced and on their distribution are reached through a political decision-making process. No equivalent to the market mechanism checks the quality, variety, or adequacy of their supply. But the government does incur costs in providing a package of goods that, given some political constraints, seems desirable to those groups that control the government. These costs are then, for want of any other number, taken as the measure of benefits generated. Under this procedure, a new presidential palace or a new irrigation system, each of which may cost $10 million, is assumed to produce $10 million of benefits (though perhaps differently distributed). This accounting method of valuing benefits is obviously an inadequate basis for any welfare judgment.

Systematic bias

To have any meaning, the benefits derived from government expenditures really need to be evaluated from the point of view of the beneficiaries. Unless this is done, the results of expenditure incidence studies are likely to be systematically biased; in particular, the accounting method of benefit valuation, which has been employed in all studies of less developed countries to date, tends to overestimate the benefits allocated to the families at the lower end of the income scale. Since such families rarely participate in decisions about the level and the composition of the public expenditures, their valuation of these expenditures may differ considerably from that of the government.

But governments do, presumably, spend money to provide certain services to the community, and not to maintain a certain number of civil servants on the payroll or to be kind to certain contractors. Since the pattern of payments is an intermediate, not a final, product, distributional analysis should concentrate on the distribution of services rendered and not on the flow of money generated—on educational services, not the salaries of teachers, and on defense services, not the salaries of the military. The only drawback to this approach is where government expenditures do not provide the services they are supposed to provide, or provide substandard services. Clearly it would be naive to insist on distributing nonexistent benefits to those for whom these expenditures were allegedly made.

A similar situation may arise when the government assumes the role of employer of last resort for certain categories of the labor force, such as graduates who cannot find jobs in a labor market already saturated with similarly trained people. Even if the additional disbursements made to these employees yield no services, at least those who receive the payments are better off with than without them. In a situation of severe unemployment and recession, it may thus be highly relevant in analyzing the effects of government expenditure to determine who—regionally, racially, or by income group—receives government appointments or contracts. Not only does the service benefit the ultimate beneficiaries, but the public servant or contractor realizes benefits from the employment opportunity which are often highly valued. As a recent study on Malaysia put it, “an investment project located on the East Coast and intended to benefit rural Malays may be constructed by a Chinese contractor from Kuala Lumpur. The significance of who carries out the government’s work is not lost on any group in Malaysia”. A similar comment could be made for countries in which the ethnic background of civil servants is an important selection criterion.

Public goods

The expenditures that cause the most difficulty in studies analyzing the distributional effects of government expenditure are those of a general nature and which provide public goods, i.e., goods such as defense or police protection available to everybody without discrimination and consumed jointly by all members of the community. The problem of allocating the benefits of these expenditures is that, although all families, in a sense, consume equal amounts of these goods, they do not necessarily place the same value upon them. No generally accepted method exists to determine who are the beneficiaries of these expenditures. Some researchers ignore this problem and assume that each family benefits equally from these general government expenditures; others assume that each family benefits from these expenditures in proportion to its overall income or its investment income; and others argue that one must consider the willingness of families with different levels of income to spend part of their income on public goods. Researchers taking the latter approach believe that this willingness rises more than proportionally with income so they conclude that the same amount of public goods available to every family, in fact, has a systematically greater benefit for upper-income than for lower-income families.

The upshot of all this is that the allocation of benefits among any subgroup varies sharply depending on which assumption is adopted. Many researchers have found them equally plausible, or implausible, and hence presented alternative benefit patterns from which the reader may choose. In the case of Brazil, for instance, the pattern of benefit incidence of general expenditures (Table 1) appears as (a) very pro-poor when each family is assumed to benefit equally from these expenditures (column 4); (b) very pro-rich when the willingness to spend money on public goods is assumed to vary directly with income (column 5); and (c) about proportional when each family receives general expenditure benefits in proportion to their income (column 3). Results for Puerto Rico and Colombia show a similar incidence pattern depending on the allocation formulas adopted.

Table 1.Benefit incidence in Brazil, 1962Benefits in per cent of income
General benefit expenditures

(method of allocation)
Total government expenditures

(including general benefit

expenditures allocated as indicated)
Willingness toWillingness to
spend money onspend money on
Income classesSpecificthese goodsthese goods
in thousands ofbenefitProportionalEqual pervaries directlyProportionalEqual pervaries directly
cruzeirosexpendituresto incomefamilywith incometo incomefamilywith income
Less than 9916.95.946.71.022.863.617.9
2,500 and over7.76.11.416.513.89.124.2
Source: Henry Aaron, “Estimates of the Distributional Impact of Brazilian Taxes and Expenditures,” Council of International Progress in Management (U.S.A.), Inc. (July-August 1968).
Source: Henry Aaron, “Estimates of the Distributional Impact of Brazilian Taxes and Expenditures,” Council of International Progress in Management (U.S.A.), Inc. (July-August 1968).

As yet no allocation formula for these general expenditures is very convincing. Incidence estimates for these expenditures are therefore only impressionistic. The resulting patterns, however detailed, do not really tell us how these expenditures influence the distribution of income but are rather quantifications of a set of initial assumptions about the incidence of general government expenditures. These quantifications cannot be any better than the assumptions themselves. Any incidence estimate which suggests that general expenditures improve the income distribution must thus be treated with extreme caution.

Specific programs

Specific benefit expenditures usually benefit clearly identifiable individuals. Their incidence ought, therefore, to be more readily assessable. An irrigation project in a particular area of the country, for instance, benefits the farmers of the area and possibly consumers of the products whose production is made either possible or cheaper. Despite the difficulties of equating the benefits received with their costs of production, specific government services are nevertheless a more gratifying object of study then general benefit expenditures. It is still extremely difficult, however, to determine just how much benefit the individual receives from a certain program, and how much the initial recipient of the government’s payment (e.g., a civil servant employed in the health sector) benefits from the employment opportunity given. Usually the researcher makes some arbitrary decisions and allocates benefits on the basis of a priori reasoning and casual empiricism.

On the whole, the results obtained from this procedure in the studies of developing countries show that specific benefit expenditures have a very pronounced regressive incidence pattern (in assessing expenditure, regressiveness indicates a pro-poor bias; in assessing taxation, a pro-rich bias). In other words, specific benefit expenditures represent a larger share of the income of the poor than of the rich. A typical result, for example, is that obtained in a study of Puerto Rico in which it was concluded that the lowest income class received benefits from specific government expenditures equal to their original income, while the top income class received specific benefits amounting to only 6.5 per cent of their income. These expenditures were therefore interpreted as quite regressive (pro-poor).

Different categories of specific expenditure—whether education, health, or agriculture—have their own particular incidence patterns, and the more interesting studies focus on these rather than on total expenditures. One of the most detailed studies of expenditure benefits was carried out by the World Bank on the educational sector in Colombia. This study concluded that the public financing of education in Colombia contributed to the redistribution of income from the rich to the poor. The estimated incidence of the different components of the educational expenditures was even more informative: it showed, for example, that this redistributive effect resulted entirely from the public financing of primary education. Public financing of secondary and higher education actually benefited the rich more than the poor.

To have and have not

Almost all expenditure benefit studies made so far concentrate on benefits expressed as a percentage share of the income of the recipients. But it is very important to remember that benefits representing a small share of a large income may well be larger in absolute amounts than benefits constituting a large share of a small income. The Colombian study, for instance, which compares benefits received from education expenditures with total taxes paid, may be said to mislead readers by claiming that benefits are distributed in a pro-poor or regressive fashion, when in fact upper-income families receive about six times as much in educational benefits than the poor (see Table 2). Focusing on relative percentage shares makes sense in the analysis of the tax burden distribution, where a higher percentage means that rich people pay more taxes than the poor, but when used in expenditure incidence studies they may be quite misleading. One ought, perhaps, to concentrate on the absolute amounts of benefits transferred to the various subgroups of the population; and only when the poor get more than the rich can one really talk about the redistribution of income through government expenditure.

Table 2.Distribution of educational benefits in Colombia
Incomebenefits asbenefits in
classper centpesos per
in pesosof incomefamily
0– 6,00013.1407
6,001– 12,0006.8618
12,001– 24,0007.01,244
24,001– 60,0005.92,202
Over 240,0000.82,555
Source: Calculated from data in Jean Pierre Jallade, Public Expenditures on Education and Income Distribution in Colombia, World Bank Staff Occasional Paper No. 8 (Baltimore: Johns Hopkins Press, 1974).
Source: Calculated from data in Jean Pierre Jallade, Public Expenditures on Education and Income Distribution in Colombia, World Bank Staff Occasional Paper No. 8 (Baltimore: Johns Hopkins Press, 1974).

Studies of the incidence of even such relatively specific public expenditure programs also give rise to many problems of implementation and interpretation. At best they can only approximate the effect of expenditures on income distribution. The main interest of such studies, in comparison to more ambitious and much vaguer attempts to examine the incidence of government expenditures as a whole, is that they can draw on much more relevant information and focus on specific policy instruments that are likely to act in the desirable direction. For example, the study referred to by John H. Adler (Finance and Development, September 1973) pointed out that projects for agricultural development, especially in the form of credit schemes and irrigation works, can go a long way toward the elimination of income inequalities. More such studies are needed to assist policymakers in choosing policies that are selective enough to reach the target group of the population whose social condition they want to improve.

When analyzing the distributional impact of specific expenditures it should be remembered that the intentions of the government which led to the initiation and/or expansion of certain programs are often not a good guide to the allocation of the benefits from these programs. While little research on this topic has been carried out for developing countries, the experience gained in the United States with programs which were initiated allegedly for distributional purposes is very suggestive in this respect. Farm programs, investment in higher education, and water resource developments are but three examples of programs which have developed a dynamism of their own and have turned out to have distributional consequences quite different from those apparently intended. In developing countries few programs have been initiated with the explicit purpose of redistributing income but various programs are commonly assumed to have this effect. Such evidence as exists suggests that a critical review of their distributional effects is urgently needed: a recent study on the incidence of agricultural price supports in Colombia showed that this program, far from benefiting the lower-income classes, has a definite pro-rich bias.

The politics of it all

Expenditure decisions in any country are taken primarily by those who exercise political power, the distribution of which is largely determined by those who hold greatest economic power. Radical changes in expenditure patterns toward the disadvantaged members of the society are not likely to occur overnight since such shifts imply disturbances in the equilibrium of political forces and endanger the positions of those groups that presently benefit most from the status quo. Short of a social revolution during which political power is redistributed and government policies may be drastically revised, income redistribution will in all probability continue to be a subsidiary policy goal attracting only limited resources. Notwithstanding pressures from international lending institutions and declarations of good intentions from national leaders, policies to redistribute incomes in favor of the less privileged members will be slow starters. It is, therefore, crucial that the limited efforts devoted to this goal be well directed. Efforts to analyze the distributional impact of specific government policies should be continued but with more concentration on those expenditures that are commonly thought to have a favorable distributional impact and that are most likely to be used as policy variables. Despite the severe statistical and conceptual limitations of such exercises, they should prove most useful in improving understanding of the distributional problem and of what can be done about it.

Other Resources Citing This Publication