The Politics of Education Spending in Mexico,

Transcription

1 The Politics of Education Spending in Mexico, Douglas Hecock Department of Political Science University of New Mexico Albuquerque, NM Prepared for delivery at the 2003 meeting of the Latin American Studies Association, Dallas, Texas, March 27-29, 2003.

2 Abstract This paper examines the determinants of education spending at the sub-national level in Mexico from 1993 to I hypothesize that a number of variables will affect the variation in the national government s allocation of resources for education to the states, including economic need, demographic factors, partisan composition of the state legislatures, electoral competition at the state level, and the party affiliation of the states governors. I conduct a cross-sectional time series OLS regression on the 31 Mexican states from 1993 to I find that the only factor that affects allocation decisions is whether or not the governor is a member of the national ruling party, the Institutional Revolutionary Party (PRI). This finding, while disappointing with regard to the prospects for an educational system in dire need of improvement, does help to frame considerable opportunities for future research. In particular, research should be undertaken to examine the determinants of variation in the states share of spending on education. This will contribute to an understanding of the fiscal effects of trends towards decentralization in education. It will furthermore contribute to a body of literature that has thus far concentrated on cross-national rather than sub-national analysis of the determinants of social spending in the developing world. 2

3 After more than a decade of experience with policies stemming from neoliberal or free market-oriented economic reforms occurring throughout Latin America, it has become clear that this strategy by itself is not sufficient to ensure sustained, equitable economic development (Birdsall and Londoño 1998; Graham 2000). At the very least, these economic strategies must be complemented with significant reform to such areas of social policy as health care, social security, and education. Education is particularly important. In order to take advantage of the opportunities afforded by free markets and increased integration into the global economy, citizens must be prepared to compete with individuals who are products of advanced educational systems. With few exceptions, educational opportunities for the majority of people in Latin American countries are woefully inferior to those in Europe and the United States. One of the primary reasons for this divide is financial. In general, Latin American countries spend somewhat less on education than do developed countries as a percentage of GDP, and significantly less in real terms (PREAL 2000). While there are many factors that contribute to the quality of an educational system, it is clear that increased resources are a necessary, though not sufficient, component of any effort at improvement. To that end, this study examines the political determinants of varying levels of spending on education at the state level in Mexico. In particular, I examine the political characteristics of Mexican states that help to determine the amount of resources that are devoted to education, the bulk of which come from the central government in Mexico City. Primarily, I seek to understand whether these allocation decisions are based on the financial needs of the states or on more political calculations. Using cross-sectional time-series regression analysis, I examine spending on primary education from 1993 through 2000 in the 31 Mexican states. This is a first cut at research that I intend to expand significantly through intensive data gathering in the field. Based on the data that are available, I find that the primary determinant of varying levels of resources given to the states by the federal government is the presence of a state governor aligned with the ruling party, the Revolutionary Institutional Party (PRI). While this does not paint a terribly optimistic picture of the process of fiscal management in Mexico under the PRI, it does point to interesting opportunities for future research. First, in 2000, Mexico ended over 70 years of single party rule at the federal level with the presidential election of the National Action Party (PAN) candidate Vicente Fox. Future studies should examine the extent to which a more competitive national political arena affects decisions on resource allocation. Second, the educational system in Mexico is in the midst of a significant process of fiscal and administrative decentralization that began in earnest in Politics at the state level should have an increasing effect on varying levels of spending. This study builds on a body of research focused on the political determinants of social spending. Based on work that examined the levels of welfare spending in industrialized democracies (Cameron 1978; Castles 1982; Esping-Anderson 1985; Hicks and Swank 1992), recently scholars have examined levels of social spending in contexts of recession and market reform in developing countries, and Latin America in particular (Ames 1987; Sloan and Tedin 1987; Brown and Hunter 1999; Kaufman and Segura- 3

4 Ubiergo 2001). The work on industrialized welfare systems largely emphasized ideological and institutional explanations for varying spending levels, such as the strength of left-wing parties, union density (Hicks and Swank 1992), and constitutional structure (Huber, Ragin, and Stephens 1993). Studies of Latin America have found more explanatory power in other variables, including regime type (Brown and Hunter 1999) and trade openness (Kaufman and Segura-Ubiergo 2001). However, these findings are for aggregate levels of social spending. When broken down into spending on social security, health care, and education, they continue to be affected positively by democracy, (Avelino, Brown, and Hunter 2001) but health care and education spending levels seem unaffected by increased integration into the global economy (Kaufman and Segura-Ubiergo 2001). This paper takes a unique approach. First, both theoretically and empirically, there is reason to believe that the political forces that act upon some types of social spending are distinct from those that act on others. In disaggregating expenditure on education from that on other social programs, this problem is avoided. Furthermore, even within education there are divisions. Those political interests that support increases in resources for primary education are likely quite different than those that support more resources for secondary and/or higher education. Arguing that improvements in the former are likely to have the greatest impact on the most people, both in terms of economic and quality-of-life benefits, this paper focuses on primary education. Second, in addition to examining just a single component of social spending, this paper is innovative in its focus on expenditure at a level below the national government. Comparativists are increasingly seeing the advantages of sub-national analysis (Snyder 2001). In addition to the benefit of augmenting the number of cases available for study, sub-national comparisons diminish the problems of making generalizations from aggregate national data that conceal significant variation within countries. Furthermore, it is easier to control for socioeconomic, cultural, and historical variables, among others, at the sub-national rather than the national level. Mexico Mexico is a federal system comprised of 31 states and the Federal District (DF) of Mexico City. Each state has a legislature of between 20 and 75 seats, two thirds of which are elected via first past the post single member districts, and the remaining are elected through a formula of proportional representation. Each state also has a governor that is elected for six years and that cannot be elected to consecutive terms. The PRI, born out of the Mexican Revolution ending in 1917, dominated politics at the national level until the 1980s when it began to be challenged by two major parties, the Christian center-right PAN and the center-left Party of the Democratic Revolution (PRD). Despite its revolutionary roots, the PRI of the past few decades is considered center-right (Coppedge 1997, 1998); while still engaged in some populist political strategies, beginning in the 1980s it began to support a neoliberal economic agenda that culminated in the 4

5 implementation of NAFTA in The states have also experienced increased electoral competition at both the legislative and the executive levels. The Mexican education system has undergone successive reform movements since the Revolution. Consistent with revolutionary ideologies, most of these initiatives were aimed at increasing access to education. In the context of neoliberalism, however, there has been one overarching reform: a process of administrative and fiscal decentralization that began in 1978 was strengthened in 1992 with the Agreement for the Modernization of Basic Education. Despite being a nominally federal system, Mexico has long been relatively centralized under the corporatist system maintained by the PRI. This is changing rapidly, however, and education is one of the first formal decentralization projects. Although results of standardized achievement tests in Mexico that are directly comparable to those conducted in industrialized countries have not been made public, the indirect evidence suggests that Mexico (along with all of Latin America except for Cuba) lags far behind the developed world in terms of education quality. Colombia and Chile have announced their scores, and they rank near the bottom of the countries that participated. 1 In regional testing sponsored by UNESCO in 1998, Mexico ranked below both of these countries in 4 th grade mathematics achievement. Furthermore, Mexico has low secondary graduation rates even relative to countries with comparable levels of development (PREAL 2001). These indicators make clear the significant need for reform aimed at increasing educational quality. In a context of decentralization, however, it is important to identify the factors that differentiate Mexican states. Theoretical Considerations In many ways Mexico is an ideal laboratory in which to test, at the sub-national level, much of the work that has been done on social spending cross-nationally. As mentioned above, it is becoming increasingly decentralized, especially with regard to education. It has institutional structures at the state level that not only mimic those at the national level, but are quite comparable across states, something that clearly cannot be said of the cross-national studies within Latin America, let alone cross-regionally. Furthermore, while controlling for some historical and cultural variables, there is more than enough variation at the state level in the political variables that are hypothesized to affect levels of spending. That said, there are also some reasons as to why Mexico is less than ideal, the primary one being the difficulty in obtaining data for this level of analysis. Unfortunately, the data regarding spending that are available through Mexican governmental sources on-line and in US libraries is aggregate. It does distinguish educational resources by state, but it does not differentiate between that which is transferred from the federal government and that portion that is from the state itself. The 1 Chile ranked 35 th out of 38 countries in 1999 while Colombia ranked 40 th out of 41 countries in The program was the Third International Mathematics and Science Study (TIMSS-R). 5

6 data do exist in state budgets, but these are not on a single database and must be collected individually in Mexico. Nevertheless, these aggregate data can be used to begin hypothesis testing. The spending data available through the Mexican government s Geographic, Statistical, and Information Institute (INEGI) is total spending by state from 1993 through 2000 in current pesos (INEGI 2001b). I standardized the figures to constant 1995 pesos using GDP deflators (IMF 2002), and examine both per capita levels and changes in spending. In examining the spotty data that do differentiate between those funds coming from the state and those coming from the national government, I have been able to ascertain that the state portion is no more than 20% of overall spending, and that it is generally between 1 and 15% of the total. Therefore, while I would ultimately like to be able to replicate the cross-national studies of social spending at the sub-national level by looking at the factors that determine the levels of state resources that go to education, I can treat these aggregate figures as though they come from the national government. As such I can begin to understand whether the federal government disperses funds based on need or based on other considerations. First, the primary measurement that I use to determine the financial need of the state is its GDP per capita (INEGI 2001c). Once I have gathered state budget data, I will be able to improve this slightly by examining actual tax receipts, but GDP is a reasonable measure of the economic health of the state, and presumably, the government. States with lower levels of GDP per capita should expect greater levels of support from the federal government to offset their financial difficulties. H1: GDP per capita should have a negative effect on education spending. Another indicator of a state s need for resources for primary education is the percentage of the population under the age of 15 (INEGI 2001a). While this may seem like merely a control variable, one of the primary cues that national politicians have concerning state s requirements is population. H2: The percentage of the population under 15 years of age should vary positively with education spending. The primary political variables that are expected to have an effect on spending levels concern the partisan composition of the state governments (Crespo 1996, Banamex 1999, INEP 2003). First, the state legislatures are dominated by the three main parties. In most states the PRI still holds a majority of the seats, but the PAN has held more than 40% in a number of states, mostly in the northern part of the country. It is especially strong in states like Baja California and Nuevo Leon where in 2000 it held 57% of the seats. The PRD has been less successful at the state level, but its presence is felt everywhere. It is more popular in the poorer states of the south, and has had consistently strong showings in the 1990s in such states as Michoacán and Morelos where it has held over a third of the seats. 6

7 I expect the composition of the state legislatures to effect the extent to which the states are supported by the federal government. However, there are two distinct hypothetical outcomes. On the one hand, spending can be expected to be positively affected by the power of the PRI in a particular state. Simply, PRI dominated states can expect more resources than states where the PRI is relatively less strong. I test this using the percentage of seats held by the PRI. H3a: The percentage of seats held by the PRI will have a positive effect on education spending. On the other hand, it might be expected that the PRI at the national level are more strategic in the dispersion of patronage. The PRI may be expected to give greater resources to states where the PRI face significant electoral competition from other parties in order to bolster their support in these critical states. The national PRI controlled government would pay less attention to states where the party either has little hope of gaining a majority in the legislature (admittedly there are very few cases of this), or where they are firmly in control of the legislature. In this case, the relationship between PRI control of the legislature and spending would be curvilinear. I will test this hypothesis using two separate indicators. First I will use a measure of legislative competition. This will be operationalized as the absolute value of the effective number of parties minus two [ ( N s -2) ] (Laakso and Taagepera 1979; Mainwaring and Scully 1995). N s represents the effective number of parties and is calculated as follows: N s =1/?p i 2, where p represents the proportion of seats held by the i- th party (Mainwaring and Scully 1995). When there is a party system where two equally balanced parties vie for power, the effective number of parties will be two, while the indicator will be zero. The closer the indicator to zero, the more it approaches perfect two-party competition. Secondly, I will use a dummy variable for a legislature in which the PRI controls between 40 and 60 percent of the seats, indicating that the party s situation is somewhat precarious (though not hopeless), and that the state is deserving of resources from the national party hoping to maintain or broaden support there. H3b: State level legislative electoral competition will have a positive effect on education spending. The other primary political factor anticipated to affect levels of education spending is the party identification of a state s governor (Banamex 1999, INEP 2003). Indeed there is some reason to expect that this will be more important than the party composition of the legislature. The president of Mexico under the PRI was extremely powerful politically. He controlled the party and named candidates for the legislature, and therefore counted on complete support from the Chamber of Deputies. He also had considerable influence in the choice of gubernatorial candidates and in the dispersion of funds to the states. This led to loyalty, not only across institutional branches, but throughout the federal system. This began to change in the 1980s. While the majority of governors continue to be from the PRI, all three parties are represented. It is expected that PRI governors experienced preferential treatment from the national government in 7

8 the form of greater resources. PRI governors represent 85% of the state/years observations between 1993 and H4: The presence of a governor from the PRI should have a positive effect on education spending. Methodology and Analysis To test these hypotheses, I examine a pooled cross-sectional time-series (CSTS) data set that consists of the 31 Mexican states over a period from 1993 to INEGI (2001) does provide data on the dependent variable from 1988, but there are several methodological and data constraints concerning the points before First, Mexico converted to the new peso in 1992, and it is not entirely clear that the spending figures from are directly comparable with those from Furthermore, the aforementioned Agreement for the Modernization of Basic Education occurred in The following year there is a significant jump in spending (by more than a factor of 10 in most states), as well as an increased emphasis on primary education at the expense of secondary and higher education. Finally, consistent and comparable data on one of my primary variables, GDP, is not available at the state level before Accordingly, I test the cases from 1993 to There are several problems inherent in CSTS data. First, the time component of the dataset increases the likelihood that there is serial correlation wherein an observation in one year affects the observation in the next. This is dealt with in this study in two ways. First, I include a lagged dependent variable on the right hand side of the equation so that I minimize the data transformation necessary for the analysis. Second, I test two models: one that tests for effects on the level of education spending per capita (where I include a lagged dependent variable) and one that tests for effects on change in the level of spending per capita. In the second model, there should not be serial correlation because changes from year to year should be independent of one another. The second problem with CSTS data is that the panel component leads to problems with panel heteroskedasiticty. Following Beck and Katz (1995, 1996) I deal with this by employing Ordinary Least Squares (OLS) regression with panel corrected standard errors. Finally, CSTS data has potential problems with temporal and state-level fixed effects whereby there are effects specific to a particular panel or a particular year. The standard for dealing with fixed effects in the literature on spending is to include both state dummies and year dummies. The composite effect of the use of a lagged dependent variable, panel corrected standard errors, and state and year dummies is to increase the burden of significance of the independent variables. While this means that one can be reasonably certain of the reliability of a significant result, it runs the risk of artificially discarding what appear to be insignificant variables. As mentioned above, I attempt to deal with this partially by 8

9 testing a model of change in addition to one of levels of spending. Furthermore, I test the models both with and without the state and year dummies. For the sake of clarity of presentation, I report the results from condensed models that do not include the state and year dummies. This does not significantly affect the results. I do, however, include a year dummy for 1995 to account for the profound (though relatively brief) effect of the so called Peso Crisis that occurred at the end of In every state there is a steady increase in spending every year except for 1995 when there was a decrease in all states. I have two control variables. First, I control for the size of the population by converting the dependent variable from levels of spending to levels of spending per capita. Second, I control for economic shocks by including GDP growth. This was highly significant in earlier models that were run without the 1995 dummy. My model, therefore, is specified as follows: Education spending = β 1 (constant) + β 2 (POP<15) + β 3 (LNGDP) + β 4 (GDPGROWTH) + β 5 (PRILEGIS) + β 6 (PRI40-60%) + β 7 (COMPETETIVE) + β 8 (PRIGOVERN) β 9 (LAGSPENDING) + β 10 (1995) + ε The results are detailed in Table 1. First, I find no support for my first two hypotheses. The coefficients for both the percent of the population under 15 years of age and GDP per capita are insignificant. Indeed, the coefficient for GDP per capita is in the wrong direction. Therefore, it cannot be concluded that the national government bases its decisions as to where to allocate education resources based on states needs, at least as they have been defined with the available data. Likewise I find no support for hypotheses 3a and 3b regarding the partisan composition of the legislature. It cannot be asserted that the PRI gives more money to states where the PRI has strong control of the legislature. Nor can it be stated that the PRI gives more money to states where the PRI is engaged in high levels of electoral competition with other parties for legislative seats. This is true for both measurements of competition. The coefficient for the dummy variable representing legislatures where the PRI controls between 40 and 60 percent of the seats is actually negative. Hypothesis 4, however, is supported by the results. I find that there is a positive and highly significant relationship between the presence of a state governor affiliated with the PRI and the level of spending on education per capita. These results are nearly identical in Model 2 where I examine the determinants of changes in the levels of spending. This increases my confidence in the results regarding the governorship, as well as in the likelihood that my other variables are not falsely insignificant. 9

10 Table 1. Results of OLS regression with panel corrected standard errors on education spending in 31 Mexican states from Model 1 (Spending per Capita) Pop under (2.089) Logged GDP per Capita (10.657) GDP growth.0623 (.876) PRI Legislature (68.189) PRI 40-60% (14.443) Legislative Competition (17.76) PRI Governor 21.76* (8.091) Lagged Spending.993* (.065) * (37.314) Constant ( ) Adj. R Squared N *Significant at the.01 level Standard errors in parentheses Model 2 (Change in Spending).375 (.302) (1.951).0934 (.1534).0719 (9.505) (1.929) (2.583) 3.275* (1.205) * (6.049) (28.477) Interpretation and Implications for Further Research The results indicate that, even in the waning years of the PRI s political dominance at the national level, decisions regarding dispersion of funds for education to the states were politically motivated, and not driven by a states needs. Furthermore, this study demonstrates the extent to which the PRI s political machinery revolved around relationships with states executive branches and not state legislatures. Even in the context of a novel and seemingly genuine attempt at improving education nationally through devolving administrative and fiscal control of the educational system to the states, and even throughout the presidency of Ernesto Zedillo from who had overseen the watershed 1992 agreement as Secretary of Education, educational finance decisions were overwhelmingly political. While these findings are not entirely surprising, they are nevertheless disappointing with regard to the prospects for an educational system in dire need of improvement. They do, however, help to frame considerable opportunities for future research. First, given the results of this study, it will be important to determine how this process has changed in the context of a PAN president and increasingly competitive national legislative elections. Similarly, it will be instructive to determine if the pattern 10

11 revealed here continues in the case of a return to the presidency by the PRI in the 2006 elections. This will clearly require data points beyond 2000, but it is possible that qualitative field research will be useful in assessing the situation before the quantitative data are available. Second, this study should be replicated with improved data. Despite the fact that the vast majority of the resources in my dependent variable came from the national government such that I am confident in my results, the state portion of spending should ideally be removed. Additionally, the time period should be lengthened at least to the beginning of the Salinas administration in 1988 and into Fox s presidency. Moreover, indicators of state need should be improved. This may be quite difficult to obtain over time, but ideally an indicator of student performance such as standardized testing results that is available annually should be included in the model. Also, studies should be undertaken that compare these models with similar ones predicting spending on secondary and higher education. Third, improved data will allow a true replication of cross-national studies on social spending at the sub-national level. While I have diminished the importance of the states share of spending on education, decentralization over the past decade means that its significance is increasing. My analysis clearly demonstrates that state legislatures have little effect on how much the national government spends on education, but I anticipate that they have a very important effect on the level of state spending. It is likely here that one could see the effects of electoral competition, ideology, union density, and the structure of the economy, as have been found in the cross-national studies. Furthermore, the cultural and structural controls inherent to a sub-national analysis would strengthen the findings. Finally, one of the problems that has plagued the literature on social spending is the reliance on spending as a measurement of the health of social programs. There can be no doubt but that researchers are aware of this problem, and that the reason stems from the fact that these data are readily available and easily quantified. Most scholars would admit that these studies should be buttressed with qualitative analysis of the reforms that are taking place beneath the level of spending. But such qualitative analysis seems intractable at the cross-national level. This is not the case sub-nationally. Indeed, a federal system like Mexico offers an extraordinary opportunity to blend quantitative analysis of spending with more nuanced qualitative study of a much larger picture of change in education policy more generally. This study has shown the central role of politics in decisions regarding a particular aspect of social policy. Though one among several social policy areas in need of reform, education is particularly crucial. In the context of market oriented economic policies that emphasize individual skills and entrepreneurship, it is imperative that individuals be prepared to take part in local, national, and international economies. This can be achieved primarily through the expansion of access to education and through improvements in its quality. Until this is accomplished, Mexico will continue to rely on 11

12 low-cost, uneducated labor as a primary comparative advantage, and problems of inequality and poverty will remain. 12

13 References Cited Ames, Barry Political Survival: Politicians and Public Policy in Latin America, Berkeley: University of California Press. Avelino, George, David S. Brown, and Wendy Hunter Globalization, Democracy, and Social Spending in Latin America, , Paper presented at the American Political Science Association annual meeting, August 30- September 2, 2001, San Francisco, CA. Banamex México Sociál, Mexico, DF: Banamex, Estudios Sociales. Beck, Nathaniel and Jonathan N. Katz What to Do (and Not to Do) with Time- Series Cross-Section Data, American Political Science Review, 89(3): Beck, Nathaniel and Jonathan N. Katz Nuisance vs. Substance: Specifying and Estimating Time-Series-Cross-Section Models, Political Analysis, 6:1-36. Birdsall, Nancy and Juan Luis Londoño No Tradeoff: Efficient Growth Via More Equal Human Capital Accumulation, in Nancy Bridsall, Carol Graham, and Richard H. Sabot, eds. Beyond Tradeoffs: Market Reform and Equitable Growth in Latin America, Washington C: Inter-American Development Bank. Brown, David and Wendy Hunter Democracy and Social Spending in Latin America, , APSR 93 (December): Cameron, David R The Expansion of the Public Economy: A Comparative Analysis, APSR 72 (4): Castles, Francis G The Impact of Parties on Public Expenditures, in Francis G. Castles, ed. The Impact of Parties, Beverly Hills: Sage. Coppedge, Michael A Classification of Latin American Political Parties, The Kellogg Institute Working Paper #224, Notre Dame, IN: The Helen Kellogg Institute for International Studies. Coppedge, Michael The Dynamic Diversity of Latin American Party Systems, Party Politics, 4 (4): Crespo, José Antonio Votar en los Estados: Análisis Comparado de las Legislaciones Electorales Estatales en México, Mexico, DF: Fundación Friedrich Naumann and Centro de Investigación y Docencia Económicas (CIDE). Esping-Anderson, Grsta Politics Against Markets: The Social Democratic Road to Power, Princeton: Princeton University Press. 13

14 Graham, Carol From Safety Nets to Social Sector Reform: Lessons from the Developing Countries for the Transition Economies, in Joseph S. Tulchin and Allison M. Garland, eds. Social Development in Latin America: The Politics of Reform, Boulder, CO: Lynne Rienner Publishers, Inc. Hicks, Alexander M. and Duane H. Swank Politics, Institutions, and Welfare Spending in Industrialized Democracies, , APSR 86 (September), Huber, Evelyne, Charles Ragin, and John D. Stephens Social Democracy, Christian Democracy, Constitutional Structure, and the Welfare State, American Journal of Sociology, 99 (3), IMF International Financial Statistics Yearbook, Washington DC: International Monetary Fund. INEGI Anuario de Estadisticas por Entidad Federativa, Mexico, DF: INEGI. INEGI. 2001a. Anuario de Estadisticas por Entidad Federativa, Mexico, DF: INEGI. INEGI. 2001b. Sistema de Cuentas Nacionales de Mexico, Gobiernos Estatales: Cuentas de Producción, por Finalidad, , Mexico, DF: INEGI. INEGI. 2001c. Sistema de Cuentas Nacionales de Mexico, Producto Interno Bruto por Entidad Federativa , Mexico, DF: INEGI. INEP Mexico, DF: Insituto Nacional de Estudios Políticos, A.C. Kaufman, Robert R. and Alex Segura-Ubiergo Globalization, Domestic Politics, and Social Spending in Latin America: A Time-Series Cross-Section Analysis, , World Politics, 53 (4): Laakso, Markku and Rein Taagepera The Effective Number of Parties: A Measure with Application to Western Europe, Comparative Political Studies, 12 (1): Mainwaring, Scott and Timothy R. Scully, eds Building Democratic Institutions: Party Systems in Latin America, Stanford: Stanford University Press. PREAL (Partnership for Educational Revitalization in the Americas) Lagging Behind: A Report Card on Education in Latin America, Washington DC: PREAL and Inter-American Dialogue. Sloan, John, and Kent L. Tedin The Consequences of Regime Type for Public Policy Outputs, Comparative Political Studies, 20 (April):

15 Snyder, Richard Scaling Down: The Subnational Comparative Method, Studies in Comparative International Development, 36 (1):