Education and income inequality

The appropriate, but sometimes overlooked, starting point for a discussion of how education can contribute to a wider distribution of incomes in developing countries is that in most of these countries there exists a situation of income inequality. Unequal incomes result in unequal savings and investments. In order to redress this income inequality through education one and, if possible, two conditions should be fulfilled: low-income groups should be able to invest more in their own education than high-income groups; and/or the rate of return of their investment—that is, their ultimate earning power less, the costs of education—should also be at least as high as that of high-income groups.

But a study made of these problems in Brazil and Colombia during the early 1970s shows that low-income groups benefit less than the rich from education, and that government subsidization tends to exacerbate the inequity. For instance, the empirical data collected in Brazil show that adults in the highest income category—that of non-farm males—earn 13 times as much as their counterparts in the lowest category (farming females with a low socioeconomic background). They are also in a much better position to provide their children with basic education. They have an enrollment ratio of 78 per cent as against only 37 per cent for the farming females. Moreover, their children can also expect a rate of return on their investment which is usually higher than that expected by their counterparts in the lowest group (see Table 1).

Table 1

Brazil: earnings, enrollment ratios, and rates of return to education, in 1972

Source: Tables 1 and 4 in “Basic Education and Income Inequality in Brazil: The Long-Term View” Jean-Pierre Jallade, World Bank, Working Paper No. 268, June 1977.

¹Having a head of household earning less than 200 cruzeiros a month.

Table 1

Brazil: earnings, enrollment ratios, and rates of return to education, in 1972

		Monthly earnings (In cruzeiros)	Enrollment ratios in basic education	Social rates of return to:
				Primary education	Lower secondary education
			(In per cent)
Total males—nonfarm		873	78	23.5	13.1
	Males with a low socioeconomic background	247	63	22.7	10.5
Total females—nonfarm		380	77	21.2	12.6
	Females with a low socioeconomic background 1	187	64	30.6	11.2
Total males—farm		252	45	21.1	11.0
	Males with a low socioeconomic background	91	37	18.5	6.5
Total females—farm		112	45	13.9	10.4
	Females with a low socioeconomic background	64	37	15.7	11.5

Source: Tables 1 and 4 in “Basic Education and Income Inequality in Brazil: The Long-Term View” Jean-Pierre Jallade, World Bank, Working Paper No. 268, June 1977.

¹Having a head of household earning less than 200 cruzeiros a month.

View Table

Table 1

Brazil: earnings, enrollment ratios, and rates of return to education, in 1972

		Monthly earnings (In cruzeiros)	Enrollment ratios in basic education	Social rates of return to:
				Primary education	Lower secondary education
			(In per cent)
Total males—nonfarm		873	78	23.5	13.1
	Males with a low socioeconomic background	247	63	22.7	10.5
Total females—nonfarm		380	77	21.2	12.6
	Females with a low socioeconomic background 1	187	64	30.6	11.2
Total males—farm		252	45	21.1	11.0
	Males with a low socioeconomic background	91	37	18.5	6.5
Total females—farm		112	45	13.9	10.4
	Females with a low socioeconomic background	64	37	15.7	11.5

Source: Tables 1 and 4 in “Basic Education and Income Inequality in Brazil: The Long-Term View” Jean-Pierre Jallade, World Bank, Working Paper No. 268, June 1977.

¹Having a head of household earning less than 200 cruzeiros a month.

The Brazilian situation, which is probably typical of most Latin American countries, shows that education per se cannot reduce inequality in the long run. To what extent can government policies concerning the financing, pricing, and taxing of education services affect the impact of education on income distribution? The case for government involvement in education is usually made on two grounds: economic efficiency and social equity. In most developing countries, the subsidization of education is governed by the general and simple rule that everyone is equally entitled to the same amount of public subsidy for a given amount and type of education. In other words, education is subsidized, and, therefore, priced regardless of incomes. This is true for free public education and for fee-paying education as long as fees do not cover the total costs.

Needless to say, this pattern of subsidization has at best a “neutral” effect on incomes. In fact, it probably has an adverse effect because high-income groups tend to remain longer in the education system than low-income groups and therefore receive larger public subsidies. This may be particularly true in many developing countries where the distribution of educational opportunities, and subsidies, is very unequal. In some countries, high-income groups may pay for the education of their children via taxation when the tax system is sufficiently progressive (that is, when tax rates rise with incomes). But this is far from being the case everywhere, especially in the developing world where progressive direct taxation is a much less important source of revenue than indirect taxation, which is not progressive. Thus, the extent to which government involvement in the provision of education affects income inequality depends on the distribution of both taxes and education subsidies among income groups.

A detailed, empirical study of these problems was carried out in Colombia in 1970. The analysis of the distribution of education subsidies showed that in urban Colombia middle-income families (earning annually between 24,000 and 120,000 pesos) received higher subsidies per child enrolled than either very low- or very high-income families (see Table 2). When education subsidies were related to all children, whether enrolled or not enrolled, in each income group, they definitely appeared to benefit high-income groups more than low-income groups. The picture in rural Colombia was quite similar to that of urban Colombia—with the one big difference that education subsidies were, across the board, much lower in rural than in urban areas.

The extent to which the distribution of taxes across income groups offset the distribution of education subsidies is shown in Table 3. The table shows that, on average, Colombian taxpayers receive 33 pesos worth of education subsidies whenever they pay 100 pesos in taxes and that low-income groups receive back, in the form of education subsidies, a much higher proportion of their taxes than high-income groups. In fact, this proportion decreases regularly from low-income to high-income groups, which shows that not only do high-income groups pay for their education through their taxes but also that government involvement in the provision of education in Colombia contributes to redistribute income from high-income to low-income groups.

However, when each level of education is examined separately, it can be seen that only the public financing of primary education has a strong and positive effect on the distribution of income by redistributing income from the 13 per cent richer families to the 87 per cent poorer families, the cutoff point being 60,000 pesos a year. The re-distributive effect is strongly beneficial to the 40 per cent poorer families (earning under 12,000 pesos a year), more than three fourths of whom live in rural areas. These families receive 87 per cent of their taxes back in the form of public subsidies for primary education.

The picture that emerges from secondary education is quite different: here, the main beneficiaries are two middle-income groups, including about 48 per cent of all families. In other words, the public financing of secondary education redistributes income from both the 40 per cent poorest and the 13 per cent richest families to a sort of lower middle class, 80 per cent of whom are living in urban areas and whose incomes are in the 12,000–60,000 pesos range. The situation for higher education is very similar, except that the two income groups subsidized by both the poor and the rich are higher up in the income scale. Families in these groups represent close to one third of all families, they are almost exclusively urban and earn between 24,000 and 120,000 pesos. Thus a redistribution of income from the poor and the very rich to the upper middle class takes place through the public financing of higher education.

Table 2

Colombia: distribution of public subsidies for education among income groups, 1970

Source: Tables 3–10, 3–15, and 3–16 in Public Expenditure on Education and Income Distribution in Colombia Jean-Pierre Jallade, World Bank Occasional Papers No. 18, Johns Hopkins Press, 1974.

Table 2

Colombia: distribution of public subsidies for education among income groups, 1970

Income bracket (Pesos/year)		Subsidies per child enrolled (Pesos)	Subsidies per child in each income group (Pesos)
Urban Colombia
	0– 6,000	1315	640
	6,000– 12,000	1136	490
	12,000–24,000	1357	636
	24,000– 60,000	1691	852
	60,000–120,000	1521	848
	120,000–240,000	1201	784
	Over 240,000	986	605
	Total	1469	746
Rural Colombia
	0– 6,000	552	83
	6,000– 12,000	552	127
	12,000– 24,000	554	183
	24,000– 60,000	554	276
	60,000–120,000	333	333
	120,000–240,000	200	200
	Over 240,000	100	100
	Total	533	125

Source: Tables 3–10, 3–15, and 3–16 in Public Expenditure on Education and Income Distribution in Colombia Jean-Pierre Jallade, World Bank Occasional Papers No. 18, Johns Hopkins Press, 1974.

View Table

Table 2

Colombia: distribution of public subsidies for education among income groups, 1970

Income bracket (Pesos/year)		Subsidies per child enrolled (Pesos)	Subsidies per child in each income group (Pesos)
Urban Colombia
	0– 6,000	1315	640
	6,000– 12,000	1136	490
	12,000–24,000	1357	636
	24,000– 60,000	1691	852
	60,000–120,000	1521	848
	120,000–240,000	1201	784
	Over 240,000	986	605
	Total	1469	746
Rural Colombia
	0– 6,000	552	83
	6,000– 12,000	552	127
	12,000– 24,000	554	183
	24,000– 60,000	554	276
	60,000–120,000	333	333
	120,000–240,000	200	200
	Over 240,000	100	100
	Total	533	125

Source: Tables 3–10, 3–15, and 3–16 in Public Expenditure on Education and Income Distribution in Colombia Jean-Pierre Jallade, World Bank Occasional Papers No. 18, Johns Hopkins Press, 1974.

Table 3

Colombia: distribution of taxes and public subsidies for education among income groups, 1970 ¹

Source: Tables 3.19 and 3.20 in Jallade, op. cit.

¹Education subsidies are computed on the basis of enrollments and public expenditures by level and type (public or private) of education.

Table 3

Colombia: distribution of taxes and public subsidies for education among income groups, 1970 ¹

Income bracket (Pesos/year)	Number of households (In per cent)	Subsidies for education	Subsidies for primary education	Subsidies for secondary education	Subsidies for higher education
		(As percentage of taxes)
0– 6,000	19.0	117	109	9	0
6,000– 12,000	20.2	83	77	4	2
12,000– 24,000	24.9	72	49	18	5
24,000– 60,000	22.9	55	22	20	14
60,000–120,000	8.8	23	4	7	12
120,000–240,000	3.4	10	1	3	6
Over 240,000	0.8	2	—	1	1
Total	100.0	33	16	9	8

Source: Tables 3.19 and 3.20 in Jallade, op. cit.

¹Education subsidies are computed on the basis of enrollments and public expenditures by level and type (public or private) of education.

View Table

Table 3

Colombia: distribution of taxes and public subsidies for education among income groups, 1970 ¹

Income bracket (Pesos/year)	Number of households (In per cent)	Subsidies for education	Subsidies for primary education	Subsidies for secondary education	Subsidies for higher education
		(As percentage of taxes)
0– 6,000	19.0	117	109	9	0
6,000– 12,000	20.2	83	77	4	2
12,000– 24,000	24.9	72	49	18	5
24,000– 60,000	22.9	55	22	20	14
60,000–120,000	8.8	23	4	7	12
120,000–240,000	3.4	10	1	3	6
Over 240,000	0.8	2	—	1	1
Total	100.0	33	16	9	8

Source: Tables 3.19 and 3.20 in Jallade, op. cit.

¹Education subsidies are computed on the basis of enrollments and public expenditures by level and type (public or private) of education.

It is clear that the positive effect of the public financing of education on the distribution of income in Colombia is only due to the financing of primary education which strongly benefits the poor. This positive effect is partly but not wholly offset by the negative income distributive effects of the public financing of secondary and higher education which benefit most the lower and upper middle classes respectively.

Whether or not the Colombian case is “typical” of many other Latin American countries is an open question. It is clear, however, that any policy aimed at making a system of finance more equitable should act on either the structure of public subsidies or the structure of taxation.

Scaling subsidies to income

To overcome the better ability-to-pay for education of some as compared to others, public subsidies should be inversely related to incomes. No country has ever put this policy into practice, so it is hard to foresee its impact on, say, the demand for education, the quality of education, and the internal and external efficiencies of educational systems. But there exists one important situation—in Colombia—where education is subsidized and, therefore, priced differently for the poor and the rich, and this could yield useful clues about what would happen if subsidies were inversely related to incomes.

In many countries a fully subsidized public education sector coexists with a not-so-heavily subsidized private education sector. On the whole (with some important exceptions, such as Japan), public schools tend to recruit a student body mainly from low-income groups. Conversely, little subsidized, expensive private schools tend to cater to the needs of high-income groups.

Thus, it would seem that one way to make sure that education subsidies benefit low-income more than high-income groups would be to foster a private, little subsidized, education sector in which high-income groups can enroll their children.

This is, to a certain extent, shown by Colombia, which has undergone a certain “privatization” of some of the key levels of its education system during the past 30 years. (Only one half of university students were enrolled in public universities in 1970 as against over two thirds in 1940. In the same way, enrollments in public teacher training institutions have decreased from 80 to 70 per cent during the same period. Enrollments in private institutions have increased accordingly.)

Although this may promote equity in the short term, it would tend to harm low-income groups in the long term since the “private” education sector would provide a better education and higher ultimate earning capacity than the subsidized system. The existence of private education services for the rich has always faced fierce opposition from many policymakers. Their fear is that, as soon as the full cost—or something sufficiently close to it—of educational services are charged to some groups the quality of the most common type of education will deteriorate owing to the lack of a strong political constituency. Those with the highest purchasing power will foster the “best” service, which will yield the highest returns (in the form of examinations passed and, ultimately, earning opportunities), and those with a lower purchasing power will go for cheaper education. Thus, the “privatization” of education may serve to maintain, if not foster, long-term income disparities, especially if the size of the returns to education is positively associated with the importance of the private finance component in educational costs. In this case, the search for equity in the provision of education through an income related pricing system might run against long-term equity.

How far are the fears of these policymakers justified? Is it valid to assert that private education yields higher returns than public education? The only evidence available to support this assertion is circumstantial—rates-of-return calculations have, so far, never been carried out simultaneously and comparatively for public and private education. However, although the situation probably differs from country to country, one has to assume from an economic standpoint that those who seek and gain access to fee-paying schools in spite of sometimes fierce competition and dire financial strain, do so in order to improve on the rate of return which they could get from an equivalent education in a tuition-free school. In Latin American countries, the suspicion that the returns to private education do more than simply offset its higher costs is founded on the above-average ability of private schools to prepare students to gain access to the upper levels of the education system.

So, although the gradual “privatization” of education as income rises may help in promoting short-term equity among taxpayers in the provision of educational services, it still has to be proven that it does not contribute to inequity in the long term.

More progressive taxes

It seems that the only way to introduce more equity into the provision of educational services without the harmful effects of “privatization” is to increase the progressivity of the tax system. In a country like Colombia, this could be achieved through an additional tax on higher incomes, which would be earmarked for the financing of secondary and higher education. Such a tax would help to remove the adverse effects of the public subsidization of those levels of education on the distribution of income by increasing the tax payments of high-income groups. The objective of this tax would be to make sure that the rich will be at least paying for the subsidies which they receive. Of course, the tax rates corresponding to the various income groups could be manipulated in order to achieve any degree of income redistribution. In the long run, their gradual decrease could be geared to the gradual equalization of education subsidies across income groups as low-income groups gain access to the higher levels of education, without altering the rest of the tax system.

This Colombian example illustrates some of the issues concerning the impact of government policies on income inequality through the financing of educational services. This is, however, only the short-term aspect of the problem. In the long run, the concern for equity leads to an inquiry into the distribution of the returns to education among socioeconomic groups. Taxation of these returns in a progressive manner may be required in order to achieve a positive impact on income inequality.

A convenient way to assess how the various socioeconomic groups as a whole fare with regard to both the subsidization of education investment and the taxation of the returns to this investment is to adopt the viewpoint of the state, and include the entire population in the assessment. The taxes levied on the returns to education could be interpreted as a way for the government to get back part of the money spent in subsidizing access to education. Taxes would be the benefit stream of government subsidization of education, while the outlays incurred to subsidize individuals to reach a certain level of educational attainment would be the cost stream. The “net” amount of subsidies distributed by the government to each educated person would be assessed by the present value of all the taxes paid on the returns to this education by the educated person during his entire lifetime, minus the subsidies received to reach this level of educational attainment. The “net” subsidy per educated person would then be multiplied by the proportion of people reaching the level of education under consideration in each socioeconomic group to arrive at the comparative costs of education to the government for each group.

Rates of return to education and income distribution

Conventional “human capital” theory uses two rates of return to education to assess the overall degree of subsidization of education by the government—the private and the social rate. Both rates are computed on the incomes of the educated. The private rate is usually higher than the social rate—the private rate being based on after-tax earnings and private costs (excluding public subsidies), while the social rate is computed on the basis of before-tax earnings and total costs of education (including public subsidies). As a result, the difference between the private and the social rates reflects, on the benefit side, the tax stream paid by educated people and, on the cost side, the public subsidy required to reach the level of education under consideration. A big difference between the two rates means a high level of overall or “net” subsidization (high public subsidies to reach a certain level of education and low taxes afterward) while a small difference means that the subsidies received to reach a certain level of education are nearly offset by the taxes paid afterwards by educated individuals during their active lives.

However, traditional rate of return calculations are of little use for income distribution purposes because they are usually carried out at an aggregate level (using mean incomes for each age education group) in order to compare efficiency between education cycles—for instance by showing that the average rate of return in primary education is superior to that of secondary education. Such calculations fail to provide any clue about the resulting impact on income inequality. What is needed is a breakdown by socioeconomic groups so that one can compare the subsidies accruing to and taxes paid by the various socioeconomic groups for each level of education.

Such a disaggregate analysis was carried out in the case of Brazil (see the table).

The table shows that, by and large, educated people in disadvantaged groups are equally or more subsidized than those of other groups. In the same way people engaged in farming occupations appear to enjoy higher “net” subsidies than those in nonfarming occupations. It does appear, therefore, that government involvement in subsidizing the provision of education and taxing its returns is oriented in the right direction.

Source: Table 4.

		Difference between private and social rates of return to primary education	Difference between private and social rates of return to secondary education
Total males—nonfarm		1.2	0.8
	Males with a low socio-economic background	1.5	0.9
Total females—nonfarm
Total females—nonfarm		1.5	1.5
	Females with a low socio-economic background	4.6	1.4
Total males—farm
Total males farm		3.0	1.5
	Males with a low socio-economic background	4.0	1.0
Total females—farm
Total females—farm		2.9	3.0
	Females with a low socio-
	economic background	2.9	3.1

Source: Table 4.

View Table

		Difference between private and social rates of return to primary education	Difference between private and social rates of return to secondary education
Total males—nonfarm		1.2	0.8
	Males with a low socio-economic background	1.5	0.9
Total females—nonfarm
Total females—nonfarm		1.5	1.5
	Females with a low socio-economic background	4.6	1.4
Total males—farm
Total males farm		3.0	1.5
	Males with a low socio-economic background	4.0	1.0
Total females—farm
Total females—farm		2.9	3.0
	Females with a low socio-
	economic background	2.9	3.1

Source: Table 4.

A major deficiency of this approach, however, is that it focuses exclusively on the individuals who invest in education, and discounts nonparticipants. And it is not enough for redistributive purposes to know that the few disadvantaged who gain access to a certain level of education are more subsidized than the many coming from privileged groups. One also has to relate the levels of subsidization to the numbers able to take advantage of it. In other words, what is needed is what could be called, for want of a better word, a “redistributive” approach, that seeks to compare the “net” impact of government involvement in the financing and taxing of education across socio-economic groups as a whole, including those who invest and those who do not invest in education. The discussion of subsidies and taxes in Brazil in the latter part of the article, and the data in Table 4 provides an illustration of this approach.

Such an analysis was carried out for Brazil at the beginning of the 1970s. The distributive impact of subsidizing (and taxing the returns from) basic education are summarized in Table 4, which compares the situation of the educated from different income groups with the situation of the group as a whole. The table shows that the present values of “net” government subsidies allocated to each educated individual are higher for females and farm workers than for males and nonfarm workers respectively. Educated individuals in the subgroup “with a low socioeconomic background” are also usually getting higher subsidies than their counterparts in the group as a whole. When the analysis is limited to those participating in education it seems that low-income educated persons are more heavily subsidized by the government than high-income educated persons.

Table 4

Brazil: allocation of “net” government subsidies for basic education among socioeconomic groups

Source: Table 7 in Jallade, op cit.

Note: The discount rate to compute the present values is 20 per cent.

¹Positive present values means that the taxes recovered on educated individuals’ incomes offset the outlays incurred by the government to subsidize this type of education.

Table 4

Brazil: allocation of “net” government subsidies for basic education among socioeconomic groups

			Present value of “net” subsidies allocated by the government to each educated individual with:			Present value of “net” subsidies allocated by the government to each person in the group
		Enrollment ratios in basic education (In per cent)	Primary schooling	Lower secondary schooling	Total
			(In cruzeiros)
Total males—nonfarm		78	327 1	–1,640	–1,313	–1,024
	Males with a low socioeconomic background	63	148	–1,820	–1,672	–1,033
Total females—nonfarm		77	- 71	–1,950	–2,021	–1,556
	Females with a low socioeconomic background]	64	–112	–2,012	–2,124	–1,359
Total males—farm		45	–496	–2,370	–2,866	–1,290
	Males with a low socioeconomic background	37	–526	–2,420	–2,946	–1,090
Total females—farm		45	–558	–2,390	–2,948	–1,327
	Females with a low socioeconomic background	37	–560	–2,380	–2,940	–1,088

Source: Table 7 in Jallade, op cit.

Note: The discount rate to compute the present values is 20 per cent.

¹Positive present values means that the taxes recovered on educated individuals’ incomes offset the outlays incurred by the government to subsidize this type of education.

View Table

Table 4

Brazil: allocation of “net” government subsidies for basic education among socioeconomic groups

			Present value of “net” subsidies allocated by the government to each educated individual with:			Present value of “net” subsidies allocated by the government to each person in the group
		Enrollment ratios in basic education (In per cent)	Primary schooling	Lower secondary schooling	Total
			(In cruzeiros)
Total males—nonfarm		78	327 1	–1,640	–1,313	–1,024
	Males with a low socioeconomic background	63	148	–1,820	–1,672	–1,033
Total females—nonfarm		77	- 71	–1,950	–2,021	–1,556
	Females with a low socioeconomic background]	64	–112	–2,012	–2,124	–1,359
Total males—farm		45	–496	–2,370	–2,866	–1,290
	Males with a low socioeconomic background	37	–526	–2,420	–2,946	–1,090
Total females—farm		45	–558	–2,390	–2,948	–1,327
	Females with a low socioeconomic background	37	–560	–2,380	–2,940	–1,088

Source: Table 7 in Jallade, op cit.

Note: The discount rate to compute the present values is 20 per cent.

¹Positive present values means that the taxes recovered on educated individuals’ incomes offset the outlays incurred by the government to subsidize this type of education.

A different picture emerges when all individuals in each group—both educated and uneducated—are included in the analysis. For instance, in spite of their higher incomes, females engaged in nonfarm activities are more subsidized than farming males or females (but also more than their male, better-paid counterparts). In the same way, subgroups including only persons “with a low socioeconomic background” are, on an average, less subsidized by the government than larger groups in spite of their lower incomes. It seems thus that low-income groups are rather less subsidized than high-income groups—which means that the government is not, through its subsidies and tax policies, oriented toward distributing incomes in a more equitable way.

A more progressive taxing of the returns to education could be achieved through an “education tax” on the incomes of educated individuals. Such a tax would seek to lower the present value of “net” government subsidies accruing to high-income groups and increase this value for low-income groups. As the above analysis has shown, the degree of progressivity of tax payments would be determined on the basis of (1) the present “net” subsidy received by each educated individual in any given group and (2) the enrollment ratio for the same group. Progressivity in tax payments could, of course, vary according to the amount of income redistribution sought.

Progressive taxation of the returns to education through an “education tax” could be most simply achieved by adding or removing a few percentage points in the existing income tax rates of educated individuals in each socioeconomic group. Admittedly, this is not a conceptually perfect solution. The tax base would be absolute income instead of, as it should be, that particular fraction of additional income which is due to additional schooling. However, the advantages of coupling the “education tax” with the income tax would be important from the operational viewpoint. In addition to administrative simplicity, such a tax would probably be more easily accepted if it takes the form of a few additional points in the income tax rate structure without a change in the tax base, rather than if it were a new set of necessarily substantial tax rates applied to a small tax base, namely, additional incomes due to additional schooling. No overhauling of the existing tax system would be necessary and the new rates would be kept flexible in order to take into account changes in the subsidies received and taxes paid by each group.

The proposed “education tax” would also help to shape a pattern of incentives to acquire education that would be conducive to greater equality of educational opportunity, since the private returns from education of high-income groups would be reduced more than the returns of low-income groups. The proceeds of this tax would be used to increase the subsidization of education of low-income groups and improve the availability and quality of educational services available to them.

Nothing so far has been said about the political feasibility of these proposals. Apart from the obvious statement that such feasibility will vary greatly across levels of education within a single country, a possibly good test of a government’s willingness to proceed with these proposals and of their likely acceptance by the public is provided by the financing of other public goods in the country. If the financing of such goods as health care, public transportation, and subsidized housing is designed in such a way that its impact on the distribution of incomes goes in the right direction, the case for a progressive system of education finance should be easy to make. If the opposite situation prevails, the chances of education being singled out among other semipublic goods to receive distinctive treatment are weak.