Albert Hirschman's Exit-voice Framework and its Relevance to Problems of Public Education Performance in Latin America

Abstract

This paper applies Albert Hirschman's exit-voice framework to the problems of education coverage and quality in Latin America. It argues that the combination of low direct taxation and high levels of private primary enrolment provides exit options for the wealthy and reduces their incentive to exercise their “voice”, or protest mechanisms, in the face of poor education performance. It also argues that fragmented and clientelist political party structures limit the provision and monitoring of public education, and also reduce the political capacity of the poor to exercise their voice regarding public education coverage and quality. The main policy implication of the paper is that good governance in education cannot realistically be addressed without analysing how the structure of power and voice, and of conflicts of interest within civil society, affect the actual political pressures that state institutions face.

Notes

¹ The average Latin American adult in the richest 10% of the income distribution has seven more years of education than an adult in the poorest 30% (Hausmann & Székely, 1999).

² Low enrolment rates at the secondary level are partly the result of high repetition rates and late entry into primary school (de Ferranti et al., 2003, footnote 2, p. 21).

³ The data in this paragraph are from Lee & Barro (2001, Table 2, p. 474).

⁴ The data in this paragraph draw on de Ferranti et al. (2003, Table 4.4, p. 84).

⁵ Within Latin America, education spending in 1990 and 1995 was particularly high in Cuba (6.6 and 6.8%, respectively) and Jamaica (5.4 and 6.4%, respectively).

⁶ Within Latin America, a UNESCO study revealed that for primary education students, Cuban students perform best, with Argentina, Brazil and Chile the next best performers (Ratliff, 2003, p. 10).

⁷ The mean scores in mathematics for the following less-developed countries were: Korea (589), Taiwan (585), Estonia (531), Hungary (529), Malaysia (508), Iran (411), Indonesia (411), Tunisia (410), Egypt (406), Chile (387), the Philippines (378) and South Africa (264).

⁸ The mean scores for the following less-developed countries in science were: Taiwan (571), Korea (558), Estonia (552), Hungary (543), Latvia (543), Malaysia (510), Bulgaria (479), Iran (453), Indonesia (420), Chile (387), the Philippines (377) and South Africa (244).

⁹ The mean score on reading/literacy was 494, the OCED average was 500. The scores for the following less-developed countries were: Korea: 525; Latvia: 458; Russian Federation: 462; Hungary: 480; Mexico: 422; and Brazil: 396.

¹⁰ The 1998 UNESCO/OREALC exam in language, maths and science for third-graders for all Latin American countries found that Cuban students score significantly more highly than students in all Latin American counties (Gasperini, 1999, p. i).

¹¹ The sample of eight countries consisted of Argentina, Brazil, Colombia, Costa Rica, Mexico, Nicaragua, Peru and Venezuela (Kaufman & Nelson, 2004, Table 1,2, p. 10). There were important variations in spending across these countries. For instance, in 2000 central government spending in Costa Rica was the highest at 6.2% of GDP, Colombia's was 4.5% of GDP, with the Peru at the bottom end at 2.0% of GDP.

¹² Lee & Barro (2001, p. 487) presented the data on the superior performance of East Asian and Centrally Planned countries in comparison to Latin American countries in the period 1960–90, but acknowledged that “they do not attempt to explain why East Asian education systems perform better.”

¹³ For a discussion of type I and type II failures, see Khan (1995, pp. 72–73).

¹⁴ The cost of providing a single student with a free education in university education is estimated to be 50 times higher than the cost of supporting a student in primary school. This means that “for every student in a Latin American university, fifty poorer students will either not go to primary school or will attend one of inferior quality” (Ratliff, 2003, p. 14).

¹⁵ For a discussion of the data on income distribution in Latin America, see Inter-American Development Bank (1998) and Székely & Hilgert (1999).

¹⁶ The authors recognize two important methodological concerns in their argument. First, estimates of inequality are poor and exist for few countries. Second, it is difficult to identify causality because the provision of universal primary education has a powerful effect on the degree of inequality in the dimensions considered (wealth/income, human capital and political power) (Mariscal & Sokoloff, 2000, p. 198).

¹⁷ See Drèze & Sen (1989, pp. 226–253) on the advancement of education in Communist regimes (China, Cuba) and in capitalist authoritarian ones (i.e. Chile in the Pinochet period). In the latter case, the authors argue that political pressure from civil society and the long tradition of democratic and pluralist politics were important for the extension of social services even in an authoritarian setting. For Communist regimes in Eastern Europe, China and Cuba, success in extending the quantity and quality of education was integral to the construction of national identity (de Ferranti et al., 2004, p. 85).

¹⁸ On the other hand, rates of illiteracy in Mexico increased from 1990 to 2001 (Table 1).

¹⁹ Birsdall & Londoño (1998, p. 124) illustrated this point with a comparison of Brazil and Malaysia, two countries with similar income per capita, in 1989. Income distribution in Brazil is much more unequal: the income share of the bottom 20% is 2.4% in Brazil, and nearly double, at 4.6% in Malaysia. As a result the per capita income of the bottom 20% is US$1075 in Malaysia and US$513 in Brazil. They argued that given an income elasticity of demand for secondary education of 0.50 (a conservative figure), secondary enrolment in Brazil would have been 40% higher if income distribution in Brazil had been at the level of inequality in Malaysia.

²⁰ This second factor needs to be treated with caution. One, there was great progress in education coverage in the period 1950–80, when the nature of production was no less skewed toward large-scale plantation agriculture and mineral resource extraction than it was in 1980–2000, when the deterioration in the quality and coverage of education occurred in Latin America. Moreover, the relatively successful education performance in terms of quality and coverage in some natural resource-rich counties such as Malaysia, Botswana, Chile and Norway would not be explained with this analysis. Mineral abundance does not determine either income inequality levels or the composition and level of taxation and public education spending.

²¹ The relevance of exit options for households is dependent on two factors. First, the household earns enough income to afford private education. Most households cannot afford private education, particularly in low-income countries and/or in countries with very unequal distributions of income. Second, the state funds non-public education (religious schools and secular private schools) either through direct subsidies to non-public schools or through the introduction of voucher schemes. This, ceteris paribus, would allow a greater number of households to consider the option of private education.

²² It is important to point out that the division between “public” and “private” schools is not clear-cut, since direct public funding to government-dependent and independent private schools is prevalent in less-developed countries (UNESCO, 2004, p. 65).

²³ See Gasperini (1999).

²⁴ For a discussion of the now advanced countries' experiences, see Lindert (2004, pp. 87–170).

²⁵ See Di John (2006, pp. 12–16) for an analysis of why personal income tax collection is relatively low in Latin America in the period 1997–2002.

²⁶ Different patterns of English and Spanish colonialism and the institutions they left behind may have influenced differences in tax policy between the Caribbean and Latin America (Thirsk, 1997). The British Caribbean countries inherited legal institutions that enabled the development of more formal labour markets, which can explain, in part, the higher capacity of these countries to collect income tax compared to Central and South American economies (Stotsky & WoldeMariam, 2002). Additionally, Caribbean countries generally inherited parliamentary systems of governance, which may offer more feasible mechanisms of institutionalizing pacts with élites to pay taxes than in the generally more presidential systems in Latin America (Stotsky & WoldeMariam, 2002).

²⁷ It is also important to note that, contrary to the Mariscal and Sokoloff model, high levels of income inequality do not preclude government action to undertake progressive taxation, as countries with high levels of income inequality such as South Africa and Malaysia have been able to collect high levels of personal income tax.

²⁸ It is also worth mentioning that, in the 1970s, Chile had the highest share of personal income tax in the region (see further discussion of the Chilean case in Sections 5 and 6). Mexico and Costa Rica had the highest shares of personal income tax collection in the period 1985–88 and the second and third highest shares (after Chile) in the period 1975–78.

²⁹ See Di John (2006, pp. 12–16) for a discussion of the political economy implications of the reliance on indirect taxes (especially value-added taxes) in Latin America. It is important to note here that standard analysis of tax incidence indicates that who bears the ultimate burden of the tax may be substantially different from who pays the tax in the first instance. For example, a corporation may not pay the full amount of a corporate tax if it can shift some of that burden to consumers via higher prices or if it can force workers to accept a lower wage (Stiglitz, 1986, pp. 411–455).

³⁰ See World Bank (2004, pp. 80–86) for an analysis of the nature of political organizations and competition and their effects on the coverage and quality of pro-poor service delivery.

³¹ The Communist countries (Cuba, China, Soviet Union) also all had very disciplined and centralized party structures. Of course, in Communist countries the use of “voice” in the sense Hirschman uses the term was limited given the repression of political rights. However, one of the sources of legitimacy in such regimes was to provide high-quality education for all. In effect, the party and state in Communist countries acted as “voices” for the poor by providing high-quality education and did not permit the “exit” option of private sector education at any level of education. The almost universally superior performance of education in Communist countries does suggest that prohibiting exit options is more likely to force a regime to deliver high-quality state education to the population. In this sense, the source of state legitimacy and the ideology of the dominant political parties may be a more important factor for delivering high-quality public education than whether the regime is more or less democratic.

³² It is important to note that the shares of personal income tax declined in most countries in the region over the period 1975–2002 (Di John, 2006, pp. 12–16). While personal income tax shares declined in Mexico and Costa Rica over this period, both countries still had the highest shares of personal income tax collection (2.0 and 2.2% of GDP, respectively) in the region (the average personal income tax collection in Latin American was 1.2% of GDP) through the period 1985–88, as indicated in Table 6. In the period 1997–2002, Costa Rica's personal income tax share declined to 0.7% of GDP (slightly below the Latin American average of 1.0% of GDP). Since 1997, neither Chile nor Mexico has published disaggregated data on personal income tax, which is now included under the aggregate category of direct income tax (which combines personal and corporate income tax). However, the share of direct income tax in Chile and Mexico in the period 1997–2002 was 4.2 and 4.5% of GDP, respectively, which was above the Latin American average of 3.9% of GDP (Di John, 2006, Table 3, p. 14). The decline in personal income tax shares across the region, including countries that have social democratic parties, does not augur well for the construction of progressive welfare systems in Latin America (see Conclusion).

³³ The relatively low personal income tax collection in Brazil is all the more paradoxical since the Brazilian state has indeed achieved among the highest tax takes as a percentage of GDP in Latin America (and indeed among all less-developed countries) in the 20th Century, and in the period 1990–2004, has increased its take from 22% of GDP to over 30% of GDP (Di John, 2006, p. 10). As a result, the Brazilian state relies on a series of inefficient and regressive indirect taxes such as multi-tiered value-added taxes and financial transaction taxes (Di John, 2006).

³⁴ It will be interesting to see the effects the emergence of multi-party electoral competition will have on the Mexican public education system

³⁵ Tendler (2002) examined, with the use of surveys, the attitudes among business élites and found them to be unsupportive of initiatives to improve education. Despite the public discourse of the business sector in acknowledging the value of education as central to poverty reduction, business élites were more supportive of public investment in infrastructure and in improved security of property rights than in improving education (personal contact with Judith Tendler). The one country in East Asia that performs poorly relative to its income level is Thailand. Despite having a similar income per capita to Peru, workers' educational attainment in Peru is more than 1 year greater than is the case in Thailand (7.3 years compared with 6.1 years) (de Ferranti et al. (2003), p. 25). Thailand is also the country that has one of the highest levels of private enrolment in primary education in East Asia, as indicated in Table 3.

³⁶ For a discussion on the political economy of the Malaysian state's redistributive policies under the New Economic Policy (NEP), initiated in 1970, see Gomez & Jomo (1997).

³⁷ While econometric work may contribute to testing the hypotheses put forward here, the fact that so few Latin American countries participate in international tests (which is itself an indication of the lagging educational performance in the region) would make such an exercise difficult to undertake at this time.

³⁸ See Di John (2006).

Additional information

Notes on contributors

Jonathan Di John

I would like to thank Francisco Gutiérrez Sánin, Jo Beall, Rosemary Thorp, and Judith Tendler for helpful comments on an earlier draft. I would also like to thank the Crisis States Programme, London School of Economics for funding part of this research and my participation in the seminar: Development and Conflict: The Ideas of Albert Hirschman, Bogotá, Columbia, November, 2005. The usual disclaimers apply.