Policy Research WORKING PAPERS Pubilc Sector Management and Private Sector Development Country Economics Department The World Bank January 1993 WPS 1069 Why Is There Proportionately More Enrollment in Private Schools in Some Countries? Estelle James Heavy enrollment in private secondary schools stems from limited public spending, which creates an excess demand from people who would prefer to use the public schools but are involuntarily excluded and pushed into the private sector. PoScy ResearchWorking Papers dis5niate the findings of workm prgrs and encouragethe exchange of ideas among Bank staff and altothersinuestedin developmentiiues.7fhesepapes, distributedby theResearchAdvisoty Staff, c thenames oftheauthors,eflect onty bheirvies,andshotdbeuse d and itedacordofingly. hefindins, itmapretaions,adconclusions aretheauho re own. eyshould not be attdbuted to the World Bank., its Board of Dincos, itS managemrnen, or any of itS member countries. Pollcy Research Pubb Sclo Mnagement and Private Sector Development WPS 1069 This paper - a product of the Public Sector Management and Private Sector Development Division, Country Economics Department - is part of a larger effort in the department to understand public-private relationship in the provision of social services. Copies of the paper are available free from the World Bank, 1818H StreetNW, Washington, DC20433. Please contact DanieleEvans, room N9-061,extension 37496 (January 1993, 34 pages). The proportion of students enrolled in private secondary schools in developing countries stems rather than public schools varies greatly among from limited public spending, which creates an countries. James tries to explain (1) the system- excess demand from people who would prefer to atically higher proportion of enrollment in use the public schools but are involuntarily private schools in developing than in developed excluded and pushed into the private sector. countries, at the secondary level, and (2) the seemingly random variation across countries Limited public spending on secondary within a given level of education and stage of education, in turn, is modeled as a collective development. decision which is strongly influenced by the numerous families that opt for many children, She argues that differentiated demand and and that consequently can only afford to invest nonprofit supply - both of which stem from small amounts in each child, in developing cultual heterogeneity, especially religious countries. heterogeneity - are the major explanations for variations in the proportion of private education The results of regressions that determine within a given stage of developmuent and educa- private-sector size recursively and simulta- tional level. neously with public educational spending are consistent with these hypotheses. By contrast, she hypothesizes that the proportionately heavy enrollment in private The Policy Research Working Paper Series disseminates the findings of work under way in the Bank. An objective of the series is to get these findings out quickly, even if presentations are less than fully polished. The findigs, interpretations, and conclusions in these papers do not necessarily represent official Bank policy. Produced by the Policy Research Dissemination Center Why Is There Proportionately More Enrollment in Private Schools in Some Countries? by Estelle James Estelle James is Senior Economist, World Bank, and Professor of Economics, SUNY, Stony Brook. The author wishes to thank the numerous people in the U.S. and abroad who helped her with the study that has been partially summarized in this paper. She especially appreciates the typing and related work by Suzanne Lane and the data analysis by Renqui Xiao, H. K. Lee, Amy Salsbury, and Johan van der Sluis. Financial support received from the Spencer Foundation, the Exxon Foundation, the Agency for International Development, and the Program on Nonprofit Organizations at Yale University is gratefully acknowledged. Anonymous referees also made very helpful comments in connection with earlier versions. Table of Contents I. Conceptual Framework 3 Basic Model 3 Key Variables in the Private Sector Equation 8 Collective Choice about EDSP 11 H1. Empirical Results 13 Methodology 13 Determination of %PVT in a Recursive Model 14 Determination of Public Educational Spending 16 Simultaneous versus Recursive Determination of EDSP and %PVT 18 The Political Economy of Public-Private Choices 19 Conclusion 20 Endnotes 23 Tables 27 Appendix A: Symbols and Data Sources 30 Bibliography 33 Impure public goods such as education yield both public and private benefits and hence can be financed through the public or private sectors. Even when government funding predominates, production can be carried out through public or private management. Thus, different combinations of public and private provision (funding and management) are feasible and are, in fact, observed in different countries. For example, the percentage of enrollments that are private at the primary and secondary levels covers the entire spectrum from 1% to 1001, as shown in Table 1. At higher educational levels the dispersion is also substantial. What demand and supply factors account for these differences across societies? How does the process of economic development affect the role of the private sector in education? To what degree can government policies influerce the outcome? This paper investigates these closely related questions. The answers are important since private schools may differ from public schools with respect to cost and quality and a system that is largely prlvate may provide a different educational service and distribution from one that is largely public. More specifically, I seek to explain (1) the systematically higher proportion of secondary school private enrollments (X PVT) in developing as compared with developed countries (mean - 31.32 for developing versus 21.4Z for developed countries in my sample); and (2) the seemingly random variation across countries within a given level of education and stage of development. I hypothesize that the large Z PVT at the secondary level in developing countries is due to limited public spending, which creates an "exceses demand" from people who would prefer to use the public schools but are unable to find a - 2 - place. The low public spending on secondary education is modelled as a collective decision which is strongly influenced by a coalition of high income, high tax families and low income families who have opted for quantity over quality of children, in developing countries. This "excess demand" explanation helps resolve a seening anomaly: why the private sector has grown relatively large in some developing countries where, paradoxically, the public sector is considered superior, and vice versa. In contrast, I argue that the seemingly random variation across countries within a given educational level and stage of development is due to differentiated demand and nonprofit supply, steiming mainly from cultural heterogeneity, especially religious heterogeneity. On the demand side, differentiated tastes about ideology lead people voluntarily to opt out of the public system even when space is available, to secure the kind of education they prefer. On the supply side, private schools are a convenient institution for diverse non-prof it-maximizing religious organizations to use in their competition for a larger market share of "souls." Part I develops the conceptual framework for analyzing private sector size and public spending on education. Part II presents the empirical results, using both a recursive model in which public spending is regarded as pre- cetermined and a simultaneous model in which size of the private sector and educational spending in the public sector are jointly determined. The recursive and simultaneous models lead to similar conclusions. Based on this sample of 50 developed and developing countries (all the countries for which data are available on private enrollments plus the most important explanatory variables), it appears that, if one knows the answers to a few key questions about a society, one can make a reasonably good prediction about the size of its private versus -3- public educational sectors. Moreover, the predictive model is the same for developed and developing countries, despite the large differences in their private sector size. But first, a caveat: the definition of "private" is by no means clear-cut in a situation where many "private" schools are heavily funded and regulated by the state. In most developing countries private schools depend mainly on private funding, but in many developed countries subsidies cover a large proportion of total expenses, and government control over hiring and firing of teachers, salaries and student admissions criteria accompany these subsidies. "Source of funding" and "decision-making authority" then yield different puLlic-private categories and many mixed rather than polar cases. In this paper private schools are defined as those which had private founders and continue to have private management, although varying amounts of public funding may be present.' I. Conceptual Framework Basic Model This section sets forth a theory in which the relative size of the private educational sector in a country (ZPVT) depends on its excess demand and differentiated demand for education, its supply of nonprofit entrepreneurship, and government policies that influence public and private supply. Demand. Consider a family utility function in which: Ui - Ui(QPUBi, QPVTi, EXTQi, Yi - Ti - PQPVTi) (I) That is, fismily welfare depends on its own publie and private education (QPUBi and QPVTi), other peoples' consumption of education (EXTQi)t and its disposable income to purchase other goods after paying taxes and private school tuition (Yi - Ti - PQPVTi). We assume here that public education is financed out -4- of taxes and uses non-price rationing, while private education finances and rations, at least partially, on the basis of price. We are examining the choice, aggregated over all families, of QPUB± versus QPVTi, for a given Y1, T1 and P, as well as the collective choice of public educational spending which influences T. As a first step, the family formulates its optimal consumption of public and private education, QPUB*i and QPVT*i. If public and private schools are perfect substitutes and P > 0, QPVT*i - 0. However, the supply of public places may be less than demand so some people are excluded, i.e. actual QPVSi < QPUB*i for some i. Then, if the private benefits from education perceived by family i are high enough to cover the price of private education (e.g. because family i's income is high), they will seek places in the private sector, as a second-best solution. I call this demand, stemming from families who would have preferred to enter the free or low-price public system, the excess demand motive for private education. (See Weisbrod, 1975 and 1977, for one of the earliest formulations of the excess demand hypothesis). Excess demand existed in many Western countries in the nineteenth century, before public systems became open access. I argue that it also exists in many developing countries today, and constitutes the major reason for large private sectors at the secondary and higher levels.2 The greater the total effective demand for education and the smaller the capacity of the public sector, the larger will be the excess demand for private education, ceteris paribus. Now suppose that public and private schools are imperfect substitutes, because people have diverse tastes about the kind of education to be consumed, but the public system is constrained to be relatively uniform. If family i prefers to enter the private sector, QPUB*i - 0. I call this demand, stemming from people who prefer the product variety offered in the private sector, the - 5 - differentiated demand for private education. Preliminary evidence from many countries indicates that mu-.. of this taste differentiation stems from religious, linguistic and nationality differences that concern group identification.3 The greater the cultural heterogeneity of the population and the more uniform the public educational system, the larger will be the differentiated demand for private education, both in developed and developing countries. Differential preferences about quality can also lead to the development of private schools. In particular, a low quality public system may stimulate the growth of a high quality private sector, meeting the demand of those willing and able to pay the price. If we assume that educational quality has a high income elasticity of demand and if the public sector provides a quality level that just satisfies the median family, greater income diversity within the population implies greater dissatisfaction among upper income people, who will seek superior education in the private market. 4 Private supRlv. Of course, the ability of people who are dissatisfied with the amount or type of public provision to find private alternatives depends on the supply behavior of private schools. This paper does not seek to fully explain this behavior but does analyze one important determinant -- the role played by the nonprofit sector in education. Private schools are often established as nonprofit organizations, i.e. as organizations that cannot distribute dividends or stcck that can be sold for capital gains. Indeed, nonprofit status is legally required for educational institutions in many countries, in part because nonprofits are considered more "trustworthy" (see James and Rose-Ackerman, 1986). This characteristic greatly influences the supply of private schools. For example, private schools may not spring up even though a potential profit exists, because there is no pecuniary return on equity - 6 - and non-profit-motivated capital and entrepreneurship are not available. On the other hand, nonprofit schools may spring up in situations where for-profi,s could not break even, because of their nonpecuniary goals and lower monetary cost functions due to donated capital, volunteer labor and tax advantages. We must therefore asks what are the motives of people who start nonprofit schools and what factore determine their availability? I argue that most founders of private nonprofit schools are ideol.ogical organizations, especially religious organizations. (See James and Rose-Ackerman, 1986, for a fuller development of this point.) Proselytizing religions such as Christianity have historically used schools as a mechanism for shaping values, socializing old members and attracting new ones; the Catholic Church has traditionally run its own school system around the world with these objectives. And competing ideologies have often been forced to start their awn schools, as a defensive strategy.5 The nonprofit form is used because these founders are interested in maximizing membership or faith rather than pecuniary profits. Therefore, for reasons of nonprofit supply as well as differentiated demand, I would expect the private educational sector to be larger in countries with many strong independent religious organizations competing Alor members and member loyalty, through their schools. Government golicies. Finally, it should be clear that many government policies influence the demand for and supply of private schools. For example, governments can and have prohibited private schools or have imposed costly requirements that have a similar effect (e.g., Catholic schools were strongly restricted in England and Holland during the eighteenth century, private schools were outlawed in Tanzania and Pakistan during the 1970., and extensive regulations deterred private school entrepreneurship in Sweden until very -7- recently). Governments have required nonprofit status of private schools or given tax privileges to nonprofit schools (the U.S. and Japan are two examples). Some governments (e.g. the U.S. and Switzerland) allow local control over public schools, which should increase diversity within the public sector, hence diminish the differentiated demand for private schools. It is very difficult to secure data on all these policies for the entire sample of countries. However, the effects of two important public policies, for which data could be secured, are analyzed in this paper: (1) the provision of subsidies to private schools, which increases the supply of private education; and (2) government spending on public schools, which increases their quantity and/or quality and decreases disposable income, hence decreases the demand for private education. In sum, these demand and supply forces are hypothesized to stem from per capita income (PCI), stage of development (DV and ADV), cultural heterogeneity (CULT HET), income diversity (INC DIV), public educational spending (EDSP), and public subsidies to private education (SUB). The reduced form equation is: Z PVT - f (PCI, DV, ADV, CULT HET, INC DIV, EDSP, SUB] (II) Public educational seendina. An imnediate problem concerns the endogeneity of EDSP, which may be determined by %PVT or by unobserved forces that also influence ZPVT. For example, people who intend to opt out to the private sector may push for a low spending public sector, so the effect of EDSP on %PVT may be biased in a negative direction by OLS. Conversely, an unobserved taste for education may lead both to a large public and private sector, in which case the public spending effect on ZPVT would be biased in a positive direction. To deal with these problems in the estimation of %PVT, I also develop a model determining EDSP and compare the predictions of the recursive and simultaneous models. More specifically, I assume: EDSP - g(SEC, DV, PCIs AGO-14, TOT, GOVSP, ZPVT) (III) where AGO-14 - proportion of the population aged 0-14, TOT - an index of totalitarian control, GOVSP - non-educational public spending as a proportion of GDP, and %PVT is included in the simultaneous model. The next two sections describe the key variables in the private sector equation (II) and provide the rationale for the collective choice about public spending in equation (III). (See Appendix A for definition of variables and data sources). Rev Variables in the Private Sector Equation Per caDita income and stage of develoDment. DV and ADV are used to represent developing and advanced industrial societies, respectively, and per capita income (PCI) picks up smaller income differences within each stage of development. These enter into equations II and III as indicators of the gross demand for education and the effective demand for differentiated education. If EDSP does not change in the course of economic growth, PCI would have a positive effect, DV a negative effect on %PVT and the interactions of heterogeneity with ADV would be positive in equation II. However, if EDSP increases with economic growth in equation III, this would have the opposite effect, so the net impact of growth on %PVT is ambiguous a priori and the regression results may vary depending on whether or not EDSP is controlled. One object of this paper is to examine this process, as it throws light on the relatively greater role of the private sector at the secondary level in developing countries. Cultural heteroaeneitv. Cultural heterogeneity enters into equation II as the major determinant of differentiated demand for and nonprofit supply of private education. To measure this, I secured the religious and linguistic breakdown of the populations of all the countries in my sample; indeed, the need for this information was the major factor that limited the sample size. I started with a meaoure of religious heterogeneity which weights all religions equally: RELIG - SKt ln 1/R;, where Ri - proportion of the population constituted by religion i. As the number of religions grow, so too does RELIG; the index is highest where the population is equally divided among a large number of religions.6 However, some religions are much more active proselytizers than others. As noted above, Christianity in general and Catholicism in particular have done so historically, using schools as a major competitive weapon, so I wanted to weight these groups more heavily in my "religious competition" index. I consequently constructed a "Christian" weight, WTCH - RCH or (I-RcH), whichever is smaller, and similarly, a Catholic weight, WTcA - RCA or (I-RcA), whichever is smaller. These weights increase as RCH (or RCA) increase until R - .5 is reached, after which they decrease; the weights are maximized when Christians (and Catholics) are a large minority or a small majority, the situation where their need to use schools as a competitive instrument is maximized. Therefore my index of religious heterogeneity and religious competition is REL - WTCH WTCA ZR; ln 1/1R. Both for demand and supply side reasons it is predicted to have a positive effect on ZPVT. Linguistic heterogeneity, LANG, was measured parallel to RELIG, as EL; ln 1/L*, where Li - proportion of the population speaking language i. Language may also serve as a proxy for more general cultural heterogeneity (e.g., based on nationality or ethnicity) and as such its effect is expected to be positive. Income Distribution. Comparable data on income distribution over large sets of countries are exceedingly difficult to find. I used the "Gini coefficient of sectoral inequality" (GINI) which is based on product per worker - 10 - across economic sectors. This index takes on larger values if product proportions are smaller than labor proportions for some sectors, indicating inter-sectoral inequality. Since it is calculated on the basis of large economic sectors rather than individuals or households it understates total inequality and, more importantly, distorts the relative inequality positions of different countries. As an alternative, I also used the proportion of national income received by the bottom 20% of households (DISTRIB), which is inversely related to individual inequality; however, such measures are known to be unreliable. Although inequality is predicted to have a positive sign on theoretical grounds given above, it is not clear whether these indices will be capable of detecting that effect. Subsidies. Most countries subsidize their private schools in some way but detailed data are generally not available and the existence of implicit tax subsidies further complicates the situation. Given this paucity of data, a dummy variable, SUB, is included for those countries (mostly advanced industrial societies) that cover more than 70% of the costs of their private schools out of public funds. Public Educational Spending. Many developing countries restrict access to their public systems at the secondary and higher levels, using non-price mechanisms to ration the limited number of places among the excess demanders, and this capacity constraint can clearly be relaxed by additional spending (EDSP). EDSP may also proxy public school quality (in which case a low EDSP leads to a differentiated demand for private school quality) but this relationship is ambiguous (see Hanushek, 1986). In this paper EDSP is expressed as a proportion of GDP and measures differences across societies in their share of GDP devoted to public educational finance. It is measured separately for the primary and - 17 - secondary levels and is expected to have a negative effect on %PVT.' Public educational spending is treated as pre-determined in equations (1-3), while its determination is simultaneously modelled in equation (4), Tables 2 and 3. Collective Choice About EDSP To fix ideas about the EDSP choice, let us suppose that only two alternatives are available--low public spending (EDSPL), which will provide a public school system large enough to accommodate only a minority of the age cohort and high public spending (EDSPm), which ¶11 accommodate a majority. I assume that EDSP is financed by taxes, TLX or Ts, in which everyone knows his share in advance, and is determined by a collective choice process in which: 1. If public provision is the only option (i.e. QPVT is required to be 0), families will be willing to expand EDSP ao long as their marginal (private and external) benefits exceed their marginal (tax) costs; and 2. Given a private alternative (i.e., QPVT can be - or > 0), families will additionally compare the net benefits to them of high public spending (EDSP) versus low public spending (EDSPW) plus optimal QPVT at EDSPL, and will choose to expand public educational spending if and only if they derive a greater "consumer surplus" from the former.8 Referring back to equation (I), it is clear that families favoring EDSP (who I call Group HI) are those who will thereby receive a positive redistribution because their probable benefits are greater and/or their costs are lower than under EDSPL due to a combination of the following factors: 1. Their tax share is low, possibly because their taxable income is low; 2. Their desired consumption of public schooling is high because they have school-age children who want to attend, they perceive a high return to education, - 12 - and they do not have a strong preference (on quality or ideological grounds) for private schooling; 3. Their probability of being admitted to the marginal public school places is high; 4. The external benefit they will derive from the educational consumption of others is high and is more easily achieved through the public sector where their tax share is matched by the tax share of others; in some cases this externality may depend on public control over detailed school decisions such as the beliefs and values that are imparted by schools.9 On the other hand, families favoring EDSPL (Group LO) are those who will be "redistributed away from" by EDSPW for the opposite reasons. In this paper I do not specify the collective decision process (e.g., I do not assume a majority voting scheme) but I do assume that as the relative size and political power of Group HI increases, the likelihood of EDSPH also increases, and vice versa.10 Therefore, the key variables hypothesized to determine EDSP in the recursive model are PCI, DV, the proportion of the population aged 0-14 (AGO-14), an index of totalitarian control (TOT), and non-educational government spending (GOVSP); in the simultaneous model, %PVT is added. AGO-14 is a proxy for effective demand, which is expected to raise desired public school expenditures at the primary level (particularly for large families whose tax share is below the private school price for multiple children), but which may not raise desired EDSP at the secondary level, because of the family's quantity-quality trade-off. TOT is a "totalitarian index," evidence of dictatorial power that may have a positive or negative effect on EDSP depending on whether the dictator is trying to maximize the utility of Group HI or Group LO (i.e., is a populist or elitist dictator) and whether he wishes to use schools - 13 - as an instrument for tightening his control."1 GOVSP captures the fact that some countries use public rather than private spending to finance other services; if the same group will benefit from public educational spending, this indicates they probably have the motivation and political power to enforce a high EDSP. %PVT is expected to have a negative effect if people who anticipate that they will attend the private system are in Group LO, ceteris paribus. ZPVT is obviously endogenous so 2SLS is used in the equations where it is included. II. Empirical Results Methodologv The analysis was conducted by pooling data from the primary and secondary levels in 12 developed and 38 developing countries, 100 observations altogether-- the largeet number for which I could get data on ZPVT and the most essential independent variables. My unit of analysis was the country, by educational level. Although the sample was not random it also was not biased in any obvious way and includes a wide variation in all variables as well as substantial representation from all geographic areas. To ensure that influential outliers were not strongly influencing my results in this small sample, I also reran my regressions omitting the observations with the 6 highest and 6 lowest ZPVT. While their omission affected the magnitude of the coefficients, it did not change their signs or the pattern of significant results. In estimating %PVT a linear probability model and logit analysis were both used, and yielded very similar conclusions. The former has the advantage that the coefficients are easier to interpret but it has the disadvantage that the predicted value of ZPVT may turn out to be >1 or <0 for some countries; however, this was not a big problem since it only occurred in two or three cases in all my regressions, and by very small amounts. Logit has the potentially greater - 14 - disadvantage that the estimated parameters are sensitive to small measurement errors if %PVT is close to 0 or 1, which holds for several countries in this study, and it assumes a smaller marginal effect at extreme values of ZPVT than in the middle range, which may be a misspecification. Since both methods yielded very similar conclusions, the linear probability results are presented in this paper and the logit results are available upon request. Determination of %PVT in a Recursive Model Table 2 presents the OLS and 2SLS results for ZPVT; the first 3 columns assume a recursive model in which EDSP influences %PVT and not vice versa; the last column uses a simultaneous model and is discussed in a later section. Equation (1) sets forth a simple model, based on the most clearly exogenous variables, in which ZPVT depends on stage of development (DV, SECDV), per capita income (PCI), level of education (SEC), and heterogeneity (REL, LANG, GINI, separately and interacted with ADV). This equation was designed to test whether heterogeneity is as important as expected on theoretical grounds (it is), to ascertain whether it has different effects in advanced (ADV) and developing countries (it doesn't), and to capture the systematically larger private sector at the secondary level in developing countries--as in the coefficient of 11.8 on SECDV (the secondary-developing interaction term). I hypothesized above that this difference stems mainly from an excess demand for quantity in developing countries, and this depends on public policies, particularly public educational spending. Therefore, equation (2) adds EDSP (and SUB) to the model. As expected, SUB has a positive and EDSP a strong negative effect. More important, the inclusion of EDSP causes the SECDV coefficient virtually to disappear. Equation (3) presents a more parsimonious version of this model, - 15 - eliminating the stage of development variables which have become redundant and focusing on the heterogeneity and policy variables which, both on a priori and ex poet grounds, are most important. Several conclusions can be drawn from these simple recursive equations: 1. On the basis of a few variables, which proxy excess demand, differentiated demand, nonprofit entrepreneurship and relevant public policies, we are able to explain over half the variation in percentage of enrollments that are private. 2. The most important explanatory factor is cultural heterogeneity, particularly religious heterogeneity. REL is always significant at the .1Z level; if REL increases by I standard deviation, ZPVT increases by 11 percentage points. This effect holds both for developing countries and advanced industrial societies, equally at the primary and secondary levels,12 evidence of the importance of religious entrepreneurship in private education. LANG, too, has a positive effect, although somewhat w.aker. 3. Income diversity (GINI), on the other hand, has an insignificant effect- -possibly because of the data problems discussed above. When I measured income diversity by DISTRIB instead of GINI it also had an insignificant effect. Nevertheless, if we remove the heterogeneity variables as a group from equation (3), the R2 falls from 52Z to 27Z, evidence of their key role. 4. Although basic cultural factors thus matter a great deal, public policies are also important. For example, SUB increases ZPVT ten percentage p&'nts, despite its crudeness as a variable. This effect is particularly important in developed countries, where mean SUB is higher. 5. More important, once public educational spending enters the equation (equ.3), the large difference between developed and developing countries in ZPVT - 16 - at the secondary level disappears. The F test shows that the stage of development variables (including their interaction terms) become insignificant, both individually and jointly [F(6,86)-1.15). That is, the same explanatory factors work for developed and developing countries; the large difference in private sector size at the secondary level is due almost completely to the large difference in public secondary education spending. On average, both sets of countries in my sample spend about 1.6% of their GDP on public education at the primary level. The advanced industrial societies spend more than that--2%--at the secondary level, where the technological cost requirements are higher. But the developing countries spend only .9% of their GDP on public secondary education, thereby eating an excess demand for quantity (and/or a differentiated demand for quality). If secondary EDSP in developing countries were increased by 1.1 percentage points (i.e., to the developing country mean), equations (2) and (3) tell us that ZPVT would fall 9-10,t roughly to parity with ZPVT in advanced industrial societies. Determination of Public Educational Sbending However, these results may be biased, if EDSP is really endogenous. Therefore, this section deals with the estimation of EDSP and the following section explores the simultaneous determination of EDSP and ZPVT. Column (1), Table 3, presents a simple OLS version of equation III designed to capture some of the most important variables influencing EDSP and, in particular, to reproduce the low level of educational spending at the secondary level in developing countries. To be consistent with the recursive model in columns (1)-(3), Table 2, ZPVT is omitted from this equation. As expected, SEC, PCI, AGO-14 and GOVSP have positive effects. TOT has a negative effect, consistent with a model of elitist dictatorial control." Stage of development - 17 - has virtually no effect at the primary level but SECDV has a significant coefficient of -1, thereby almost fully accounting for the fact that the mean EDSP at the secondary level in developing countries is 1.1% of GDP lower in developed than in developing countries. Column (2) seeks to explain where this large SECDV effect on EDSP is coming from. I conjectured above that a high proportion of school-age children (AGO-14) might have different effects at the primary and secondary levels, since it involves a quantity-quality trade-off. Large families, especially those with low incomes, might have a high desired consumption of public education at the primary level to make their children literate, but a low desired consumption at the secondary level because of a limited willingness to invest in each child and because their demand for other goods is more pressing.'4 Moreover, high income, high taxpaying families might be willing to subsidize the primary education of children from large poor families because they perceive externalities from having a literate citizenry with the "right" values and habits, but their external benefits and willingness to subsidize secondary education might be much lower, since that would facilitate labor market competition with their own children. Thus, while the high AGO-14 in developing countries increases the size of Group HI and EDSP at the primary level, it may not have this effect at the secondary level. To test whether this is the case, equation (2) adds an AGO-14*Secondary interaction term. As expected, this has a negative effect and almost completely dissipates the separate effect of SECDV. Finally, equation (3) presents a more parsimonious version of this model, omitting SECDV and DV, which are now redundant and, with no loss of explanatory power, focuses on the remaining important variables--SEC, PCI, GOVSP, TOT, and AGO-14. In all, 442 of the variance of EDSP is explained by these variables. - 18 - Simultaneous versus Recursive Determination of EDSP and %PVT Equation (4) in Tables 2 and 3 present the 2SLS version of this simultaneous model, based on the equations in column (3), with %PVT added to the EDSP equation. The equations are identified by the exclusion of the heterogeneity variables and SUB from the EDSP equation, AGO-14, GOVSP and TOT from the %PVT equation. Our results are virtually unchanged. Cultural heterogeneity remains the main determinant of %PVT for all educational levels and country groups, and EDSP is the second most important variable in Table 2, while AGO-14 and AGO-14wSEC play a key role in Table 3. %PVT is never close to significance as a determinant of EDSP. The reason may be that when high cultural heterogeneity leads to a high %PVT, this diminishes the private benefit of public educational spending among the minorities but increases the external benefit among the dominant group, who want to use the public schools as a means of inculcating a common language and values (see fn 9). The net effect on the relative size of Groups I and II and the equilibrium EDSP ie therefore negligible. Thus it appears that EDSP influences ZPVT but not vice versa; and the recursive OLS model of %PVT discussed earlier does not lead one astray. To recapitulate how this model works let us compare the derivation of EDSP and %PVT for an "average" developing country at the primary and secondary levels, respectively. Its low per capita income and government spending combined with its high totalitarian index, lead it to choose a low EDSP; this is offset at the primary but not the secondary level by its high proportion of school-age children. Thus this country ends up with high public educational spending at the primary level, much like that in developed countries, but relatively low public spending, hence a small public sector, at the secondary level. - 19 - At the same time, many people who are excluded from the small public secondary sector perceive benefits from education that exceed its priv.te price. This includes (1) high income, high tax share families from Group LO who prefer private education for their children; (2) high tax share families from Group LO who want a small public sector but would be willing to use it if admitted; and (3) some low income families from Group HI (i.e., those who chose quality over quantity of children) who want a large public sector but, having lost the collective choice Dattle, now exercise their personal choice in the private market. Table 2 shows that the low level of public secondary spending in developing countries greatly increases the excluded students from the latter two categories and hence their excess demand for private education, as a second choice. The Political Economy of Public-Private Choices At this point one might ask: Why are people not willing to spend publicly, if they are willing to spend privately, in excess-demand-drive private sectors? If the real cost of a private school place equals that of a public school place and if private enrollments are, on average, 31% of the total at the secondary level in developing countries, it follows that private spending augments public spending by 45%. If people are not willing to spend more in the public sector, why are they willing to spend so much in the private sector? Part of the answer is that costs per student are generally lower in the private sector. (See James, 1991a; James and Benjamin, 1988; Levy, 1986; and Jimenez, Lockheed, Luna and Paqueo, 1991). If low costs lead to low price in a private competitive market, this increases the number of families who will favor low public spending, even though they value education. But the more basic reason is that the distributional consequences of public - 20 - and private spending differ when taxes are not based on benefits. To see this, suppose that a country has chosen EDSPw' TLW (i.e., Group LO has prevailed) and consider a low-tax share, high-benefit member of Group HI who is excluded from the public schools. That family will purchase education in the private market. We know (by its membership in Group HI) that this family would have preferred EDSPe, T0, where its total costs would have been lower and/or its benefits higher (i.e., it would have received a positive redistribution). But it was outvoted by families in Group LO, whose marginal tax costs would have exceeded their marginal benefits if the public system expanded (i.e., they would have received a negative redistribution). Group LO apparently could not raise Group HI's tax share to match its benefits, but it could successfully impose a low EDSP, thereby forcing Group HI to spend more via fees in the private sector. (Of course, some members of Group HI will not spend more and will not acquire education, as they might have in a larger public system).1' Thus, if an equivalent amount of public spending replaced private spending, the distribution of costs and benefits would be quite different. Consequently, different groups will favor public versus private spending. I have tried to show why the size of Group HI versus Group LO changes over the course of economic development due to demographic change, thereby shifting the political equilibrium at the secondary level from EDSPL, Tf to EDSP", T3' and eliminating the excess demand motive for private educatiion. This explains why the private sector at the secondary level is systematically larger, even though it may be considered inferior, in developing countries; while developed countries are characterized by a larger public sector and a correspondingly smaller private sector, based on differentiated demand rather than excess demand. Conclusion - 21 - In summary, I have asked why different societies have made different choices about their reliance on public versus private provision of education. The relative size of the private sector was modelled as depending on excess demand, differentiated demand, the supply of nonprofit entrepreneurship, and government policies. I hypothesized that excess demand stemming from low public spending is the major explanation for the systematically larger size of the private sector at the secondary level in developing countries, while differentiated demand and nonprofit supply, both stemming from cultural heterogeneity, are the major explanations for variations in private sector size within a given stage of development and educational level. Regression analyses conducted across a pooled primary-secondary sample of 50 countries (100 observations) produced results that were consistent with these hypotheses. Religious competition and entrepreneurship have highly significant positive effects in all cases. Linguistic heterogeneity plays a positive (but somewhat lesser) role. These findings have important implications for the behavior of private schools. For example, they suggest that private schools may segment the population along religious, linguistic, nationality or ideological lines, because of the motivations of their nonprofit producers and consumers. While basic cultural forces thus play a large role, public policies can also influence the size and nature of the private sector. In particular, public educational spending, which increases the capacity (and possibly quality) of the public system, has a negative effect on %PVT. Since public educational spending is particularly low at the secondary level in developing countries, this result is consistent with the excess demand explanation for the large private sector there. Once public educational spending is taken into account, the same predictive model works for developed and developing countries, despite the large - 22 - differences in their private sector size. I also modelled the collective decision process that determines public educational spending. The coalition favoring higher public spending at the secondary level is predicted to increase with development, due to income and demographic changes, especially the decline in family size. As the public sector expands, people who have been involuntarily excluded can now find places, so excess-demand-driven private schools are crowded out. According to our estimates, this process ultimately reduces ZPVT to the level found in developed countries, which is explained mainly by cultural heterogeneity. While large excess-demand-driven private sectors can flourish in countries with limited public systems even if they are considered inferior, the differentiated-demand- driven private schools that survive in countries with open access public systems are likely to be considered superior (academically or ideologically), by the revealed preference of their consumers. - 23 - ENDNOTES 1. See James (1991b), for a more detailed discussion of these mixed cases, and the relationship between subsidies and regulations. 2. Examples are Kenya where the majority of secondary school enrollments were privately founded, Brazil and the Philippines where 80S of college enrollments are private. Among industrialized countries currently, Japan best fits the "excess demand" model at both the secondary and higher levels; over one-quarter of all high school - (upper secondary) students and three quarters of higher education students attend private institutions, mainly because of limited space in the preferred public schools and universities. See James (1986a and 1991a); James and Benjamin (1988). Also see West (1967 and 1970) for data on the U.S. and U.K. in the nineteenth century. 3. The many private schools and colleges that accommodate religious or linguistic minorities (e.g., schools for Muslims, Parsees, Sikhs in India, Chinese and Indians in Malaysia) are examples of private sector response to differentiated demand. Among Western countries, the best example of the cultural heterogeneity model is the Netherlands, where two-thirds of all students attend privately managed schools, a response to the pervasive religious cleavage which dominated that country at the turn of the century. For other examples of the importance of cultural heterogeneity see James (1984, 1986b, and 1987). 4. Two examples are Brazil and the Philippines, at the secondary level, where 25 and 38Z of enrollments, respectively, are in private schools, which are generally considered to be better than public secondary schools. But if greater income diversity means that upper income groups also control the government, they may use this power to choose a public system that is high in quality, low in - 24 - quantity and rationed to them. In that case, a large excess demand may develop, as at the higher education level in Brazil and the Philippines. Thus income dispersion is predicted to lead to a large private sector, but we cannot be sure a priori whether this will be due to excess demand for quantity or differentiated demand for quality. 5. For example, the caste groups in Southern India and the independence movements in India and Kenya before independence started their own schools, with the expressed intention of inculcating their own values and keeping their members out of the Western-dominated Christian schools. Other examples of the ideological/religious origin of private nonprofit schools are sectarian schools in the U.S. and U.K., schools run by Catholic orders in France and Latin America, Calvinist schools in Holland, orthodox Jewish schools in Israel, educational services provided by Muslim waqfs (religious trusts) in the Middle East and by missionaries in many developing countries. 6. For a discussion of the properties of this index see Theill (1972) and Allison (1978). For its use in an international comparative study of homicide see Hansmann and Quigley (1982). Obviously, this index is sensitive to the fineness with which. one disaggregates various religions. I used all the sub- categories found in my data sources which constituted more than .1% of the population, including "Other Religions" as a separate category. 7. The only eource giving education financial information for large sets of countries is the Unesco Statistical Yearbook. There are many practical problems with these data. They often give planned rather than actual expenditures, the7 sometimes exclude local government spending, the allocation between primary and secondary levels is admittedly imprecise, and in many cases they include government spending in public schools as well as subsid'.es to private schools. The number of years contained in the primary versus the secondary cycle varies among countries. To eliminate the bias this introduces into spending data in - 25 - cross-national analyses, I calculated an "expenditure per year" for each country and multiplied this by a standard 6-year dutation at each level. 8. It can easily be shown that the availability of a private alternative reduces the probability that families will favor EDSPhi. This is, of course, particularly true for families that have strong ideological or quality preferences for private education. 9. For example, external benefits may stem only from schools that instill a common language or that inculcate values such as nationalism or support for the existing political regime. The historical literature on the development of American public education focuses on the desire of the old-timers to control the language, ideology and values of the newcomers to the "melting pot" as one of the major motivating forces behind increased public funding and management of schools in the nineteenth century, particularly in the Northeastern part of the country. Control over the language of instruction has also been an important object of public educational spending in several African and Asian countries, and control over political ideology has been an impetus to public spending in Communist countries. Of course, people who have minority beliefs may oppose high EDSP and may make contributions to private schools for the same reason. 10. One would expect public spending to be highest in societies where political power is concentrated in Group HI, which uses EDSP as a mechanism for achieving a redistribution of real income from Group LO. For example, in Malaysia the politically dominant Malays are redistributing real income from the wealthy Chinese and Tamil communities, via their control over and preferred access to the public education institutions, especially at the higher educational level. 11. TOT is an index of political and civil rights as coded by Gastil, published by Freedom House and reprinted in Taylor and Jodice (1983). I used, alternatively, the 1975 score and the mean score for 1972-78. There is very little difference between the two and the latter are given in Table 3. The range - 26 - in this index is from 2 (highest political liberty and civil rights) to 14 (highest totalitarian control). See Appendix A on data sources for more details. 12. I tested whether heterogeneity had different effects at the primary and secondary levels by interacting REL, LANG and GINI with SEC. The interaction terms were always small and insignificant and their inclusion did not affect the other variables. 13. The negative coefficient suggests that many totalitarian regimes spend less on public education, because they are more immune to popular pressures. In this respect, my results are contrary to those of Lott (1990), possibly because of a different sample and a different variant of the dependent variable. 14. Many such families cannot afford the opportunity cost of secondary school. Often their children have dropped out of primary school, hence are not even candidates for secondary school. They may believe their children are unlikely to be admitted to a public secondary school. For empirical evidence on the quantity-quality trade-off and the negative relationship between family size and secondary school attendance in developing countries see Knodel (1990). 15. A similar explanation is likely to hold for excess-demand-driven higher education sectors. For a closely related analysis of the limited public spending on secondary and higher education in Japan see James, 1986a, James and Benjamin, 1988. TABLE 1: RELAtIVE ROLE OF THE PRIVATE SECTOR IN EDUCATION 12 Advanced Z Private X Private S Private 2 Private Industrial Primary Secondary Primary Secondary Societies (1) (2) (1) (2) Australia 20 26 Iran 8 17 Belgium 51 62 Jordan 30 7 Denmark 7 6 Morocco 5 8 England & Wales* 22 16 Saudi Arabia 3 2 France 15 21 Syria 5 6 Germany 2 9 Argentina 17 45 Italy 8 7 Bolivia 9 24 Japan ** 1 15 Brazil 13 25 Netherlands 69 72 Chile 18 23 New Zealand 10 12 Colombia 15 38 Sweden 1 2 Costa Rica 4 6 U. S. 10 9 Ecuador 17 30 Median 10 13.5 El Salvador 6 47 Mean 18.0 21.4 Guatemala 14 43 Haiti 42 76 M 38 Developing Honduras 5 51 Countries Jamaica 5 76 Mexico 6 25 Kenya 1 49 Panama 5 14 Lesotho 100 89 Paraguay 13 37 Sudan 2 13 Peru 13 37 Cameroon 43 57 Venezuela 13 17 Chad 10 6 India 25 52 Liberia 35 43 Indonesia 13 60 Niger 5 14 Philippines 5 38 Nigeria 26 41 Singapore 35 1 Togo 29 16 Thailand 11 32 Upper Volta 7 43 Median 11 27.5 Algeria 1 1 Mean 16.1 31.3 * These numbers include both the independent and voluntary aided sectors in the U.K. ** Data include upper and lower secondary. Figure for upper secondary is 281. Mean ZPVT - 22.71 (Mean %PVT - 16.52; %PVTSEC - 28.91) - 28 - Table 2 Dependent Variable: Proportion of Enrollments that are Private (ZPVT) _________ _ (1) (2) (3) (4) (5) 2 R2 .46 .56 .52 .52 __ C 27.2 38.4 19.9 21.6 _____________ _(2.07)' (2.15)0 (3.37)' (2.69 )b SEC 3.4 8.4 9.7 9.2 ____________ (.56) (1.48) (2.78) (2.44) _ SEC*DV 11.8 1.4 - - _____________ (1.66)d (*19) - - DV -15.4 -16.1 - - (1.13) (.97) - - PCI -2.8 -1.5 -.4 -.3 2.0 _____________ (1.96)0C (1.01) (.45) (.28) (2.4) REL 3.9 3.8 4.0 4.1 1.8 ____________ ! (3.81)* (4.99)' (6.27)' (6.41)' (2.8) LANG 5.2 4.3 6.3 6.4 .6 _____________ (1.51) (1.4) (2.04)0 (2.05)0 (.5) GINI -1.1 -.3 -.8 -.8 2.6 i _________G (.8) (.27) (.75) (.7) (1.6) REL*ADV .5 .4 - - (.25) (.27) - - LANG*ADV 13.1 16.1 - - (.72) (1.18) - - GINI*ADV -7.0 -14.1 - - (1.18) (1.6) _ _ SUB 8.4 10.2 10.1 .2 1 _________ (1.78)d (2.24)0 (2.24)0 (.4) EDSP _ -9.1 -8.1 -9.6 1.4 (3.03) (2.42)0 (1.72) (.6) Notes: Mean ZPVT = 22.7; (Mean ZPVTPRI - 16.5, ZPVTSEC - 28.9). t statistics are corrected for heteroscedasticity using White's method. See White, H. (1978). "A Heteroscedasticity Consistent Covariance Matrix and a Direct Text for Heteroscedasticity," Econometrica, 817-838. EDSPZ was also included in all equations to denote countries where the primary/secondary division of EDSP was missing and was imputed. Significance levels: a - .1Z Col. 1-3: OLS b - 12 4s 2SLS c - 52 5: Variable means and (standard d - 101 deviations) - 29 - Table 3 Dependent Variable: Public Educational Spending as a Proportion of GDP (EDSP) ________ - (1) (2) (3) (4) (5) .43 .44 .44 .41 _ C -.17 -.66 -.77 -1.01 ____________ (.28) (.97) (1.24) (1.46) SEC .38 1.36 1.54 1.58 _ ____ __ __(1.86)d (2.01)0 (3.55)' (3.68) SEC*DV -1.0 -.21 - - (4.23)' (.36) - - DV .23 -.16 - - (.58) (.35) - - PCI .05 .05 .06 .08 2.0 -___ (.79) (.78) (1.15) (1.36) (2.4) GOVSP .02 .02 .02 .02 18.9 (2.98)b (3.03)b (3.25) (3.83)0 (9.8) TOT -.03 -.03 -.04 -.03 7.7 (1.48) (1.5) (1.84)d (1.67)d (3.9) AGO-14 .03 .05 .05 .05 39.1 (2.19)* (2.65) (3.73)' (3.83)* (9.2) AGO-14*SEC _ -.04 -.05 -.05 _ ______ __ __ (1.52) (4.53)' (4.66)0 ZPVT _ .003 22.7 _ _ ~ (.74) (21.5) Notes: EDSP - 1.4 (mean EDSPPRI - 1.6, EDSPSEC - 1.2) EDSPZ was also included in all equations to denote countries where the primary/secondary division of EDSP was missing and was imputed. Significance levels: a - .1Z Col. 1-3: OLS b - 1Z 4: 2SLS c - 5 5: Variable means and (standard d - 101 deviations) - 30 Appendix A Symbols and Data Sources PCI - per capita income, 1975 in thousands of U.S. dollars, World Tables (Washington, D.C.: World Bank, 1988-89). Foreign exchange converted to U.S. dollars according to the World Bank Atlas method, to smooth the impact of temporary under or over-evaluation. (Purchasing power parity figures not available for most countries in sample). REL - an index of religious heterogeneity - 100(WTA WTc,gRln1/Rk) where R1 - proportion of population constituted by religion i. Calculated from data in Charles L. Taylor & Michael C. Hudson, Cross-National Aggregate Data for World Handbook of Political and Social Indicators (MRDF), (Ann Arbor: Center for Political Studies, University of Michigan, ICPSR, 1971); and H.W. Coxill & K. Grubb, World Christian Handbook (Nashville, N.Y.: Abingden Press, 1968). See text for discussion. LANG - an index of linguistic heterogeneity - EL1ln1/L1 where L, - proportion of population whose main language is i. Calculated from data in Charles L. Taylor & Michael C. Hudson, World Handbook of Political and Social Indicators II, Section V, RaW Data File: Fractionalization and Concentration Measure and Inequality Indices. (Ann Arbor: University of Michigan, 1970). GINI - Sectoral Gini, based on comparison between proportion of product coming from and proportion of labor engaged in each sector of the economy. Data for selected years in 1970s from Charles L. Taylor and David A. Jodice, World Handbook of Political and Social Indicators (New Haven: Yale University Press, 3rd ed., 1983). SUB - Dummy variable of 1 for countries that subsidized more than 70Z of the total cost of their private sectors in 1975. - 31 - TOT - Index of political and civil rights as coded by Raymond D. Gastil and published by Freedom House. This index is based on criteria such as the existence of elections, more than one political party, local governments, an independent judiciary and free media (press, radio & TV). 1975 data from Freedom at Issue, Freedom House, 1976, 11-20. Mean scores for 1972-78 from Taylor & Jodice, op. cit. Range is from 2 (highest political liberty and civil rights) to 14 (highest totalitarian control). GOVSP - Current general (central and local) government spending minus educational spending, as Z of GDP, 1973. Taylor & Jodice, op. cit. EDSP - Current educational spending on primary and secondary levels, respectively, 1975. To control for the fact that different countries have different years' duration for the primary and secondary cycles, I adjusted all to a standard 6-year duration. Division between primary and secondary imputed for Denmark, Australia, USA and New Zealand. Unesco Statistical Yearbook, 1980- 85. AGO-14 - S of population age 0-14, 1975. World Tables, World Bank & Johns Hopkins Press, Vol. II, Social Data, 1983. SEC - Dummy taking the value of 1 for secondary education. DV - Dummy taking the value of 1 for developing countries. ADV - Dummy taking the value of 1 for advanced industrial societies. %PVT - Z of total enrollments that were in private schools, selected years between 1975 and 1981, calculated separately for primary and secondary levels. Sources: Sweden (1978): Marklund, S. (1979). Educational Administration and Educational Development. Stockholm: University of Stockholm, Institute of International Education. Denmark (1981): Communications with Ministry of Education, Copenhagen. Rest of EuroRe (1980): Neave, G. (1985). "The non-state sector in education in Europe: A conceptual and historical analysis," Euronean Journal of Education. 20, 321-337; and Mason, P. (1983). Private education in the EEC. London: Independent Schools Information Service. - 32 - Australia (1980): Australlan School Statistics. Canberra: Commonwealth Schools Commission, 1984. Japan (1980): Mombusho. Tokyo: Ministry of Education, 1981. New Zealand (1978): Educational Statistics of New Zealand. Wellington: Dept. of Education, 1978. U.S.A. (1980): Digest of Educational Statistics. Washington, D.C.: National Center for Educational Statistices, U.S. Dept. of Education, 1982. India (1978): Fourth All-India Educational Survev. New Delhi: National Council of Educational Research and Training, 1978. Other DeveloRing Countries (1975). FinanciLna Education in Develoging Countries. Washington, D.C.: World Bank, 1986. - 33 - BIBLIOGRAPHY Allison, P.D. (1978). "Measures of Inequality," American Socioloaical RevieS, 43: 865-880. Hansmann, H. and J. Quigley (1982). "Population Heterogeneity and the Sociogenesis of Homicide," Social Forces, 61: 206-224. Hanushek, E. (1986), "The Economics of Schooling: Production and Efficiency in Public Schools," Journal of Economic Literature, 24: 1141-1177. James, E. (1991a). "Private Higher Education: the Philippines as a Prototype," Higher Education, 21: 189-206. James, Estelle (1991b). "Public Policies Toward Private Education," International Journal of Educational Research, 15: 359-376. James, E. (1987). "The Public-Private Division of Responsibility for Education: An International Comparison," Economics of Education Review, 6: 1-14. James, e. (1986a). "The Private Nonprofit Provision of Education: A Theoretical Model and Application to Japan," Journal of Comparative Economics, 10: 255- 276. James, E. (1986b). "The Private Provision of Public Services: A Comparison of Sweden and Holland," in The NonRrofit Sector in International PersDective, ed. E. James, New York: Oxford University Press. James, E. (1984). "Benefits and Costs of Privatized Public Services: Lessons from the Dutch Educational System," ComDarative Education Review, 28: 255-276. James, E. and G. Benjamin (1988). Public Policy and Private Education in Japan. London: Macmillan. James, E. and S. Rose-Ackerman (1986). The Nonprofit Enterprise in Market Economics. New York: Harwood Academic Publishers. Jimenez, E., M. Lockheed, E. Luna and V. Paqueo (1991). "School Effects and Costs for Private and Public Schools in the Dominican Republic," International Journal of Educational Research, 15: 377-391. Knodel, J., N. Havanon and W. Sittitra (1990). "Family Size and the Education of Children in the Context of Rapid Fertility Decline", Population and Development Review, 16: 31-62. Levy, D. (1986). Higher Education and the State in Latin America. Chicago: University of Chicago Press. Lott, John (1990). "An Explanation for Public Provision of Schooling: The Importance of ladoctrination," Journal of L*w and Economics, 33: 199-231. Theil, H. (1972). Statistical Decomposition Analysis. Amsterdams North Holland. - 34 - Weisbrod, B. (1975). "Toward a Theory of the Voluntary Nonprofit Sector in a Three-Sector Economy," pp. 171-191 in Altruism. Morality, and Economic Theory, ed. E. Phelps. New York: Russell Sage Foundation. Weisbrod, B. (1977). The Voluntary NonRrofit Sector, Lexington, Mass.: D.C. Heath. West, E.G. (1970). Education and the State. London: The Institute of Economic Affairs. West, E.G. (1967). "The Political Economy of American Public School Legislation," Journal of Law and Economics, 10: 101-128. Policy Research Working Paper Series Contact Tltle Author Date for paper WPS1047 C8te d'lvolre: Private Sector Enrique Rueda-Sabater November 1992 P. Infante Dynamics and Constraints Andrew Stone 37642 WPS1 048 Targets and Indicators in World George Baldwin November 1992 0. Nadora Bank Population Projects 31091 WPS1049 Money Demand and Seignorage- W;lliam Easterly November 1992 R. Martin Maximizing Inflation Paolo Mauro 31448 Klaus Schmidt-Hebbel WPS1050 Marginal Income Tax Rates and William Easterly November 1992 R. Martin Economic Growth in Developing Sergio Rebelo 31448 Countries WPS1051 The Legal Framework for Private Cheryl W. Gray November 1 992 M. Berg Secto; Activity in the Czech and 36969 Slovak Federal Republic WPS1052 A Reappraisal of How Oral Hoda Rashad November 1992 0. Nadora Rehydration Therapy Affected Mortality 31091 in Egypt WPS1 053 Development of the Zimbabwe Alex F. Zinanga December 1992 0. Nadora Family Planning Program 31091 WPS1054 Distributional Impact of Cash and Branko Mslanovic December 1992 S. Moussa In-Kind Social Transfers in Eastern 39019 Europe and RPssia WPS1055 Wealth, Weather Risk, and the Mark R. Rosenzweig December 1992 C. Banton Composition and Profitability of Hans P. Binswanger 34783 Agricultural Investments WPS1 056 Earnings and Education in Latin Geprge Psacharopoulos December 1992 L. Longo America: Assessing Priorities for Ying Chu Ng 39244 Schooling Investments WPS1057 Measuring the Incomes of Economies Socio-Economic Data December 1992 E. Zamora of the Former Soviet Union Division 33706 International Economics Department WPS1058 The Pricing of Country Funds and Ishac Diwan December 1992 R. Vo Their Role in Capital Mobifzation for Vihang Errunza 31047 Emerging Economies Lemma W. Senbet WPS1059 Political Economy of Policy Reform Ziya Onis December 1992 S. Gustafson in Turkey in the 1980s Steven B. Webb 37856 Policy Research Working Paper Series Contact Title Author Date for paper WPS1060 Economies of the Former Soviet Dmitr Steinberg December 1992 E. Zamora Union: An Input-Output Approach 33706 to the 1987 National Accounts WPS1061 Dynamic Response to Foreign Klaus Schmidt-Hebbel December 1992 A. Maranon Transfers and Terms-of-Trade Luis Serven 31450 Shocks in Open Economies WPS1062 Economic Development, Robert E. B. Lucas December 1992 WDR Environmental Regulation, and the David Wheeler 31393 International Migration of Toxic Hemamala Hettige Industrial Pollution, 1960-88 WPS1063 Mongolia: Privatization and System Cevdet Denizer December 1992 A. Oropesa Transformation in An Isolated Alan Gelb 39075 Economy WPS1064 More Evidence on Income George R. G. Clarke December 1992 R. Martin Distribution and Growth 39065 WPS1065 Strengthening Uganda's Policy Thomas Owen Eisemon January 1992 C. Cristobal Environment for Investing in John Sheehan 33640 University Development George Eyoku Franklin Van Buer Delane Welsch Louisa Masutti Nat Colletta Lee Roberts WPS1066 Pollution Control in a Decentralized Antonio Estache January 1992 A. Estache Economy: Which Level of Kangbin Zheng 81442 Government Should Subsidize What in Brazil WPS1067 Returns to Investment in Education: George Psacharopoulos January 1992 G. Psacharopoulos A Global Update 39243 WPS1068 Enterprise Reform in Eastern Europe Sweder van Wijnbergen January 1992 B. Pacheco 37033 WPS1069 Why Is There Proportionately More Estelle James January 1992 D. Evans Enrollment in Private Schools in Some 37496 Countries? WPS1070 Economic Fundamentals of Road Timothy Hau December 1992 Pricing: A Diagrammatic Analysis WPS1071 Congestion Charging Mechanisms: Timothy Hau December 1992 An Evaluation of Current Practice