The political cycle and the mexican economy

Transcription

1 The political cycle and the mexican economy Title The political cycle and the mexican economy Issue Date Publisher Instituto Tecnológico y de Estudios Superiores de Monterrey Abstract The purpose of this dissertation is to measure the influence of the political cycle in the Mexican economy, both at the national and subnational levels. Under the general framework of the political business cycle theory, and one of its variants, the political budget cycle, the influence of political cycles on economic activity are analyzed, as well as on aggregate demand, public expenditures and the exchange rate, at the national level, and public expenditure at subnational (state) level. The methodologies used are autoregressive models with dummy variables in the national case and panel data analysis with dummy variables in the subnational case. The period of analysis goes form 1934 to the present, according to the availability of information. At national level, evidence of significant influence of the political cycle is found in the general economic activity, sectoral activity, aggregate demand, public expenditures and the real exchange rate. At a state level, there is a significant increase in public expenditures in election years, as well as an increase in public wages in the years following elections. In the final chapters of the dissertation, the comparison of the results with those obtained in former research is discussed, and, finally some conclusions and lines of research are identified Item Type Tesis de doctorado Downloaded 12/10/ :27:00

2 Link to Item

3

4 INSTITUTO TECNOLOGICO Y DE ESTUDIOS SUPERIORES DE MONTERREY CAMPUS MONTERREY EGADE THE POLITICAL CYCLE AND THE MEXICAN ECONOMY A DISSERTATION CESAREO GAMEZ GARZA MONTERREY, N. L, MEXICO DECEMBER, 2006

5 Copyright by Cesáreo Gámez Garza 2006

6 THE POLITICAL CYCLE AND THE MEXICAN ECONOMY Cesáreo Gámez Garza, MA Dissertation Presented to the Faculty of the Graduate School of Business and Leadership of the Instituto Tecnológico y de Estudios Superiores de Monterrey in Partial Fulfillment of the Requirement for the Degree of Doctor in Philosophy Instituto Tecnológico y de Estudios Superiores de Monterrey December, 2006

7 THE POLITICAL CYCLE AND THE MEXICAN ECONOMY Approved by the members of the Dissertation Committee: Dr. Alejandro Ibarra. Main Advisor. Professor of Economics Dr. Pedro Villezca, Advisor, Professor of Econometrics Dr. Germán Otálora, Advisor, Professor of Management Dr. Alejandro Ibarra. Director of the Doctoral Program

8 A la memoria de mis padres, Arturo y Eustolia por todo lo que les debo AGRADECIMIENTO: A mi esposa Sonia, por su apoyo, comprension, paciencia, tolerancia, y sobre todo su presencia, sin la cual este proyecto de vida no tendría sentido. A Pedro Villezca, "autor intelectual" y a Ramon Guajardo, que me hizo una oferta que no pude rehusar. A la Direction de la Facultad de Economia, la Direction de Intercambio Académico y las autoridades de la UANL, que hicieron posible el financiamiento del programa doctoral. A los profesores del programa de Doctorado del EGADE, particularmente a German Otalora, por sus valiosas contribuciones al contenido del estudio. En forma especial, a mi asesor principal, Alejandro Ibarra, que con su dedication, profesionalismo y sincero entusiasmo por la investigacion fue un estimulo constante para la culmination exitosa de este trabajo. A todos, muchas gracias. CG Diciembre, 2006 iv

9 ABSTRACT The purpose of this dissertation is to measure the influence of the political cycle in the Mexican economy, both at the national and subnational levels. Under the general framework of the political business cycle theory, and one of its variants, the political budget cycle, the influence of political cycles on economic activity are analyzed, as well as on aggregate demand, public expenditures and the exchange rate, at the national level, and public expenditure at subnational (state) level. The methodologies used are autoregressive models with dummy variables in the national case and panel data analysis with dummy variables in the subnational case. The period of analysis goes form 1934 to the present, according to the availability of information. At national level, evidence of significant influence of the political cycle is found in the general economic activity, sectoral activity, aggregate demand, public expenditures and the real exchange rate. At a state level, there is a significant increase in public expenditures in election years, as well as an increase in public wages in the years following elections. In the final chapters of the dissertation, the comparison of the results with those obtained in former research is discussed, and, finally some conclusions and lines of research are identified. Key words: Political business cycle, political budget cycle, emerging markets, Mexico V

10 TABLE OF CONTENTS CHAPTER 1.- INTRODUCTION 1 CHAPTER 2.- THEORETICAL FRAMEWORK The traditional political business cycle The rational political business cycle The political budget cycle 9 CHAPTER 3.- EMPIRICAL EVIDENCE Evidence in developed countries Evidence in emerging countries Evidence in Mexico Political budget cycles at national level Political budget cycles at subnational level 25 CHAPTER 4.- THE POLITICAL CYCLE AND THE MEXICAN ECONOMY: SOME STILYZED FACTS sample sample sample sample sample Discussion 53 CHAPTER 5.- THE MODEL Previous work Assumptions Implications 62 vi

11 CHAPTER 6.- METHODOLOGY Panel data analysis Fixed effects models Limitations The model 67 CHAPTER 7.- THE DATA Fiscal variables Political variables Control variables Quality and limitations of the data 71 CHAPTER 8.- RESULTS Total expenditure (GASTOT) Wages (SAL) Subsidies (SUB) Public works (OB) Other expenses (OT) Discussion 84 CHAPTER 9.- CONCLUSIONS AND LINES OF RESEARCH 88 REFERENCES 94 APPENDIX: Descriptive statistics 101 BIOGRAPHICAL SKETCH 123 vii

12 CHAPTER 1.- INTRODUCTION The idea of this dissertation came from a very simple question: What are the effects of the political cycle in the Mexican economy? The influence of politics in the Mexican economy is a fact very acknowledged in Mexican political folklore and has plenty of anecdotes, colorful phrases (vivir fuera del prespuesto es vivir en el error, Presidente que devalua se devalua) and it has penetrated even in the popular language (el aho de Hidalgo, el "guardadito "). So, The research started by reviewing the literature of the political business cycle and the behavior of the main macroeconomic variables in Mexico under the assumptions of the so-called opportunistic branch of political business cycle theory, which will be described later in the dissertation. As the research went on, the focus of the study shifted from macroeconomic variables at national level to fiscal variables at state level, following the postulates of the so-called political budget cycle. The political budget cycle is a recent development of the literature on the political business cycle and it stresses the role of fiscal policy instruments and their utilization by opportunistic politicians and/or political parties in order to increase their probabilities to win elections. Specifically, the objective of the study is to test if state governments in Mexico manipulate public expenditures in general, and some items such as transfers and public works before elections in order to signal competence to the citizen and improve their chances for the party in power to stay in power. The study of the political budget cycle at 1

13 subnational level has some advantages, given that national elections are relatively infrequent events that provide few observations to econometric analysis. Moreover, analysis at the subnational level within one country provides analogous advantages to cross-country studies, in that there are many more degrees of freedom (Khemani 2000, p. 1). "The equilibrium political budget cycle theory suggests that it would be more promising to focus empirical research on testing for electoral cycles in taxes, transfers, and government consumption spending. For these variables, one can also look at data for state and local elections, instead of concentrating solely on the small number of observations available for national elections" (Rogoff, 1990, pp ). The political business cycle has been studied in emerging economies at the national level and in countries such as Russia, India, Argentina, Colombia and Portugal at subnational level, finding evidence of the opportunistic behavior of the governments. Thereafter, the main hypothesis of this dissertation is that state governments in Mexico take expansionary fiscal policies such as the increase in public spending before elections, and there is a contraction in public spending after elections have taken place. The main contribution of the dissertation is that this kind of analysis at subnational level has not been done in Mexico and the findings can contribute to the research and design of institutional arrangements and fiscal rules at the state level, as well as to the measure of the welfare losses caused by opportunistic deviations of fiscal policy. In addition, the evidence of political budget cycles at state level in Mexico has also important implications for planning and budgeting of firms, especially those in industries 2

14 related with public expenditures, such as the building industry and others. The study also contributes to theory by developing explanatory models of the political cycle theory at national and sub national levels. One of the main limitations of the study is the poor quality of the data of public expenditures at state level in Mexico. As it will be explained later in the dissertation, the data set available has plenty of omissions and missing values for different states and years, which makes it difficult to analyze the behavior of many items of public expenditure at a disaggregated level. The dissertation is organized as follows: In the second chapter the theoretical framework is presented, and the main currents of thought of the political business cycle theory are discussed. In the third chapter the empirical evidence of political business and budget cycles is reviewed: in developed countries, in emerging economies, at national and subnational levels. The fourth chapter presents some stylized facts about the influence of the political cycle in different macroeconomic variables in Mexico, using different samples and periods of observation. Significant evidence is found of the influence of the political cycle on the aggregate economic activity (GDP), which is stronger in those industries more related with the internal market, such as building and commerce. Evidence is also found of political influence in some components of the aggregate demand, as well as in the real exchange rate. Many of these finding were obtained during the author's course work, especially in the courses of Statistical Analysis, Corporate Finance and Econometrics, as well as in papers presented in academic events (Gamez 2003a, 2003b, 2004a, 2004b, 2005a, and 2005b) 3

15 In chapter five a formal model of the behavior of opportunistic governments in the face of elections is developed. It is demonstrated that, with asymmetric information about the incumbent government capabilities, opportunistic politicians try to "signal" their competence by the expansion of public expenditures before elections, in order to maximize the probabilities of electoral victory of the party in power. It is shown that the probability of tha existence of an opportunistic cycle depends on two parameters: The degree of control of the executive branch of the government over public expenditures, and the level of political education and information of the citizens. Chapter six explains the methodology used in the analysis at subnational level, which is panel data analysis with dummy variables and fixed effects. Chapter seven presents the data used in the study, definition of variables, sources and periods of observation, as well as the problems found and the limitations of the data. Chapter eight presents the econometric results. Evidence is found of an increase in public expenditures in electoral years and the year following elections, but not in the preelectoral year. The main effects of the political cycle are observed in public wages, and, somehow surprisingly, there is no evidence of political cycles either in transfer payments or in public works. In chapter nine the main conclusions of the study are outlined, as well as some interesting lines for future research. The data used in the study and some descriptive statistics are shown in the Appendix. 4

16 CHAPTER 2.- THEORETICAL FRAMEWORK The literature about the political business (PBC) cycle was developed in two waves. The first, which emerged in the middle of the seventies, uses traditional macroeconomic models in which governments can influence the macroeconomic conditions in a systematic and predictable way. The second wave incorporates rational expectations and game theory elements to explain the behavior of opportunistic governments, especially by taking advantage of asymmetries in information between government and citizens. A third wave has developed more recently, the so-called political budget cycle, which stresses the manipulation of fiscal policy, such as changes in taxes, debt, public expenditure and deficit as the main instruments by which opportunistic governments can influence the behavior of citizens and increase the probabilities of electoral victory. "A political budget cycle is a periodic fluctuation in a government's fiscal policies, which is induced by the cyclicality of elections" (Shi and Svensson, 2003, p.2). 2.1 The traditional Political Business Cycle In early formulations of this theory, opportunistic policymakers can take advantage of an exploitable Phillips curve and face naive voters who forget the past, are unaware of the policymaker's incentives and do not understand how the economy works. These models imply that the incumbent government artificially stimulates the economy immediately 5

17 before elections and eliminates the resulting inflation with a post electoral downturn or recession. The voters reward this behavior because elections take place when the economy is temporarily doing well and because they do not understand the opportunistic nature of this behavior so that they can be fooled time after time. This first stage, in turn, was developed in two branches, one that stresses the "opportunistic" behavior of governments, by assuming that politicians have no economic preferences of their own, but stay in power, whereas the other, known as "partisan", states that left-wing politicians and parties have certain economic preferences (high growth, low unemployment), different from the right-wing politicians or parties, who prefer low inflation and macroeconomic stability (Alesina et. al. 1999). Among the empirical implications of the traditional opportunistic PBC models are the following: - Expansion in the first or second years before the elections; GNP growth above normal, unemployment below normal in the election year. Inflation begins to increase immediately before or immediately after the election. Recession (or downturn) after the election, with gradual reduction of inflation. - No differences in policies and outcomes between different governments. - Incumbent reappointed when growth is high and unemployment low in election dates. 6

18 On the other hand, traditional partisan models imply that the unemployment rate is permanently lower, and the product growth and inflation are permanently higher during the tenures of office of left-wing governments than that of right-wing governments 2.2.-The rational Political Business Cycle The next stage in the literature surged in the middle of the eighties, with applications of game theory to the economic policy. These models incorporate rational expectations of the public and stress the capability of a rational public that may limit the degrees in which politicians can influence the economy. In this branch there are also the opportunistic and partisan modalities. The assumptions that voters are very naive, incapable of learning and prone to systematic mistakes in expectations were subject of severe criticism from the standpoint of rational expectations. Hence, this produced the second wave of the political business theory, the so-called rational approach. The new approach emphasizes the role of the policymaker's "competence," defined as their ability to reduce waste in the budget process (Rogoff 1990), to promote growth without inflation (Persson and Tabellini 1990) or to insulate the economy from random shocks (Cukierman and Meltzer, 1986). These models make the critical assumption that the policymakers are more informed than the voters about their own competence. By taking advantage of this informational asymmetry, and trying to "signal" as much competence as 7

19 possible, politicians (and policymakers) behave in ways leading to a political business cycle (Alesina et. al. pp ). The main empirical implications of the rational opportunistic models are: Short - run manipulations of policy instruments immediately before the elections: increase in deficits, inflation, and money growth in the two or three quarters before each election. - Tightening of monetary and fiscal policies after the elections. No systematic, multiyear effects on growth and unemployment except for, possibly, some minor effects immediately before the election. The incumbent is reappointed when growth is high and unemployment low in election years. The partisan wing of the rational PBC models states that: Short-run partisan effects after the elections: unemployment is temporarily lower than normal and growth is temporarily higher than normal for about two years after an electoral victory of the left; the opposite outcome results after an electoral victory of the right. Inflation is permanently higher when left-wing politicians are in office relative to when the right is in office. 8

20 The different models and representative authors are shown in the Table 1. Table 1 Political business cycle models Opportunistic Partisan Traditional Nordhaus(1975) Lindbeck(1976) Hibbs(1977) Rational Cukerman-Meltzer (1986) RogoffySiebert(1988) Rogoff(1990) Person ytabellini( 1990) Alesina(1987) Source: Alesina et. al. (1999) The Political Budget Cycle More recently, a different development of the PBC theory has been developed. It is focused on fiscal policy as the main instrument for governments to follow opportunistic behavior. The main argument is that models based on monetary surprises as the driving forces of the PBC are unconvincing explanations of either opportunistic or partisan cycles. Research is concentrated on fiscal policy as the driven force, with monetary effects being the result of accommodation of fiscal impulses. 9

21 The argument is two-fold: first, both theoretically and empirically, monetary-driven models are based on the manipulation of economic activity via monetary surprises before elections, but these are not very convincing. Specifically, activist monetary policy as a driving force and control of monetary policy by politicians are not sound assumptions, especially in countries with central bank independence, or at subnational level. Second, there appears to be a strong role for fiscal policies for several reasons: governments usually have a higher degree of control over fiscal rather than monetary instruments, fiscal policy has real effects even if anticipated, and it can affect voter behavior even if there are no aggregate effects (Drazen, 2000, p ). Drazen (2000) proposes a model that combines active fiscal policy and passive monetary policy, labeled the AFPM model, which addresses some of the objections posed on earlier PBC models. In addition to that, more attention has been given to the political budget cycle in emerging countries, especially in the so-called "new democracies", as are also former military dictatorships and the countries of the former Soviet Union. In these countries, voters have little experience in the democratic context, a situation that poses special circumstances for the exercise of opportunistic behavior of the politician. In order to infer correctly the competence of the incumbent politician, voters need to have both the necessary information to draw such inferences, and the ability to process that information correctly about the incumbent acts and types. This precludes the existence of institutions that would collect and provide the relevant information, as well as the capability 10

22 of the media to analyze and disseminate this information. In many developing countries or new democracies, these conditions are not met (Brender and Drazen, 2003). Shucknecht (1996) states that there is more room for manipulation in developing countries, as checks and balances are weaker and governments have more discretional power over fiscal policy instruments. He also argues that in developing countries expenditure policies, such as distribution of free or subsidized goods and services, or employment generation via public works programs are more effective that tax cuts to affect voter behavior. Rogoff and Siebert (1988) and Rogoff (1990) have developed a formal model of equilibrium of political budget cycles, based in the asymmetry of information between the government and the voters. This model has been extensively used in the study of political budget cycles at national and subnational levels. Shi and Svensson (2006) propose a model in which the size of the political budget cycle depends on the size of the politician's rents of remaining in power and the share of informed voters in the electorate. 11

23 CHAPTER 3.- EMPIRICAL EVIDENCE In this chapter, the empirical evidence of both the political business cycle and the political budget cycle are reviewed. It is presented in five parts. First, the evidence about the political business cycle found in developed countries is presented. In the second part the presence of this phenomenon in emerging countries is reviewed. The third part reviews the empirical evidence of the political business cycle in Mexico. The fourth and fifth parts of the chapter are dedicated to review the empirical evidence of the political budget cycle, both at national and subnational levels. The chapter closes with some concluding remarks and theory implications from the empirical evidence. 3.1 Evidence in developed countries Generally speaking, Alesina et. al (1999) find that in the United States and other developed countries the political business cycle takes the partisan form. They analyze U.S. quarterly data from 1947 to The economic data were inflation rate, output growth, unemployment as well as various measures of monetary and fiscal policy. The political data are election dates, dates of changes of administration, and the administration's political orientation. In this case, Democrats are identified as leftist-oriented party and Republicans as a right-wing party. With this information, they tested both opportunistic and partisan models under the following general form: 12

24 Y t =a + by,_ H, Where Y is the variable to measure, DUM is a political dummy, a, b and c are the parameters and e x is the error term. The political dummy is defined for the periods before and after elections in the test of opportunistic models and between leftist or right wing governments for the partisan models. They found that the data on the postwar United States generally support the rational partisan theory. Results are somewhat stronger on growth and unemployment than on inflation because several factors influenced the latter, such as oil shocks and the exchange rate regime. On the other hand, they found very little evidence in favor of the opportunistic political cycle model. They found no evidence that the economy grows faster or that the unemployment is lower during election years. In addition, monetary policy does not appear to be more expansionary in election years, and there is no evidence of a post-electoral increase in the inflation rate. In the case of electoral manipulation of fiscal policy, they found neither evidence that government transfers are greater close to election quarters nor that fiscal deficits are higher on election years. The same authors also studied the political business cycle in other OECD countries. They examined the OECD countries that have been democracies in the sample period The countries included were: Australia, Austria, Belgium, Canada, Denmark, Finland, France, Germany, Japan, Ireland, Italy, the Netherlands, New Zeeland, Norway, Sweden, Switzerland, the United Kingdom and the United States. 13

25 The economic data were quarterly observations of inflation rates, output growth and unemployment rates. The political data were the election dates, the dates of change of governments and the political orientation of the governments. The model used was a variant of the one used for the United States, with the addition of a proxy for the growth of the world economy, weighted by each country's share of GDP over the total. The results obtained were "remarkably consistent" with those obtained for the United States. Specifically they found: No evidence of a systematic opportunistic cycle, either for growth or unemployment. An electoral cycle on the inflation rate, consistent with the models of budget cycle of Rogoff and Siebert (1988). Evidence consistent with the rational partisan theory for growth, unemployment and inflation, particularly for a subset of countries with a bipartisan system or with clearly identifiable political changes from the left to the right and vice versa. No evidence of permanent partisan effects on growth and unemployment. Thus, a political cycle that seems to appear consistently in several countries is the following: Left-wing governments expand the economy when elected; for a period of about two years they succeed, then inflation expectations adjust and the economy returns to its natural rate of growth (Alesina et.al., 1999). Drazen (2000) identifies seven "regularities" which have been observed in the past 25 years, especially in developed countries: 14

26 Regularity 1: Aggregate economic conditions before an election, specifically per capita output or income growth, and to a lessen extent inflation, have a significant effect on voting patterns in the U.S and other countries. Regularity 2: There is no significant pre-electoral increase in aggregate economic activity prior to elections in either the US or the OECD countries. Regularity 3: In many OECD countries there is a clear post-electoral increase in inflation. In the US, there is evidence of such a post-electoral increase in inflation prior to 1979, but no evidence thereafter. Regularity 4: There is evidence of a pre-electoral increase in money growth rates in many countries. In the US, there is a pre-electoral effect from 1960 to 1980, but none thereafter. There is no evidence for the US of an electoral cycle on the Federal Funds rate. Regularity 5: There is a clear partisan effect on economic activity in the US, with economic activity being significantly higher under Democrats than Republicans in the first half of their terms. Regularity 6: There is no consensus on the role of monetary policy or inflation surprises in driving partisan effects, with views varying widely. Regularity 7: There is evidence of pre-electoral increases in transfers and other fiscal policy instruments in a number of countries. In the US, this appears to be strongest prior to 1980 (Drazen, 2000, pp ). Santa Clara and Valkanov (2003) studied the relationship between political cycles and the returns of the stock market in the United States. They analyzed monthly data from 15

27 1927 to 1998, a period which included 864 observations, 18 presidential elections, from which 10 were Democrat and 8 Republican. They tested the effects of a partisan political cycle on the monthly excess and real returns of a value weighted and equal-weighted portfolios of the Center for Research of Security Prices (CRSP) in the period. The main findings are the following: The excess return of the value-weighted CRSP portfolio over the one-month Treasury bill is on average 9% higher under Democrat than Republican administrations. The excess return of the equal-weighted is 16% higher during Democrat presidential terms. - The presidential cycle variables capture information about returns that is not correlated with business cycle variables. There is no evidence of large excess returns around election dates. Volatility is higher in Republican administrations. They also found that the difference in returns across presidential administrations consists largely in differences in unexpected returns rather than in expected returns Evidence in emerging countries In developing countries, or emerging markets, the study of the relationships among political events and macroeconomic and financial variables, emerged later. The study of the relationships between political cycles and economic variables in emerging countries gained relevance with the expansion of democratic regimes in the world that took place in the last decades of the XXth Century. 16

28 According with the World Bank (2000), the proportion of countries in the world governed by democratic regimes rose from 28% in 1974 to 61% in Among the factors that produced this phenomenon are the fall of the Soviet bloc and the end of military dictatorships in Africa and Latin America. The empirical studies of the political business cycle in emerging countries can be divided in two types: those focused in the study of a specific country and those which study the effect of the political cycle in an economic variable or set of variables for a group of countries. Among the first kind of empirical papers the following is found: Larrain and Assael (1994, 1995, and 1997) study the main variables of the Chilean economy in nine Presidential terms, from Pedro Aguirre Cerda ( ) to Patricio Aylwin ( ), finding evidence of the presence of the political cycle in some instrumental variables, such as the money supply, the public deficit and the exchange rate policy. The authors found significant increases in the money supply and the public deficit in the months before the elections, as well as the delay of exchange rate adjustments (devaluations of the currency) until the elections have taken place. Ogura (2000) analyzed the Brazilian economy from 1980 to 1999, using autoregressive models with dummy variables to test the postulates of the opportunistic political cycle model. This author finds evidence of increases in the growth rate of the GDP, real interest rates and credit supply before the elections, and increases in unemployment and inflation, as well as a higher rate of depreciation of the exchange rate in the months following the electoral processes. 17

29 Ergun (2001) uses the opportunistic model to analyze the effects of elections in the economic and financial variables of another emerging market; in this case, Turkey, using monthly data for the period between January 1985 and May, 1999, in which four national elections were held. Using autoregressive models with dummy variables for the months before and after elections, he found evidence of significant influence of the electoral processes in the behavior of inflation, interest rates, monetary aggregates and public expenditure. Among the empirical studies of the effects of the political cycles in specific economic variables, we find those related to the behavior of the exchange rate. Sibley (2001) employs event study methodology to investigate the impact of Presidential elections on nominal currency values for a group of 15 Latin American countries between 1980 and 1996, in which there were 41 elections. The results indicate that presidential elections are associated with a statistically significant decline in nominal currency values. A very interesting finding of the Sibley's work is the fact that the main effects on the exchange rate are placed not after the elections, but after the new government takes office. "Our empirical results indicate that the greatest currency depreciation tends to occur at or shortly following the time of inauguration, which is typically two or four months after the election outcome becomes known. This pattern suggests that actions taken by the new President may play an important role in many cases" (p. 19). 18

30 As a matter of fact, Sibley uses the 1994 election in Mexico as "prime example" of this pattern. The results found by Sibley suggest that the stage of the political cycle in emerging countries is an important factor influencing exchange rates and should be taken into account when making trade and investment decisions. Stein and Streb (2004) analyze the relationship between the timing of elections and the behavior of the exchange rate, specially the decisions to devalue the currency, a measure that has strong political repercussions. The authors study 26 Latin American countries between 1960 and 1994, in which there were 242 elections, 131 of which were Presidential elections. In each case, the authors observed the behavior of the exchange rate in a 19-month period, 9 before and 9 after the elections, considered the month 0. The study found evidence that in the months 2, 3 and 4 after the elections the depreciation rate of the currency is higher than in the rest of the period, which support the hypothesis that governments tend to delay devaluations till after the elections in order to increase the probability of victory of the party in power. Summarizing, according with the evidence reviewed, the opportunistic branch of the political business cycle is present in several emerging countries, especially in macroeconomic variables and some policy instruments, such as public expenditures and the exchange rate. 19

31 3.3 Evidence in Mexico Gamez and Botello (1987) studied the relationship between the Presidential cycle and some macroeconomic variables in Mexico. In this study they analyzed five Presidential terms, from Ruiz Cortines to Lopez Portillo ( ). The analysis was intuitive and it was developed in an "ad hoc" manner. The authors found evidence that the Presidential cycle influences significantly in the pattern of public consumption, exports and the aggregate income. The consumption function is free from Presidential influence and in the cases of imports and investment the evidence is inconclusive. Magaloni (2000) studies the influence of the political cycle in the Mexican economy between 1970 and 1998, using autoregressive models with dummy variables. She found increases in public expenditures, private consumption and the rate of growth of the economy before the elections, and a contraction of the economic activity after the elections. She also finds evidence of strategic delays of the adjustments in the exchange rate until the elections have taken place. A very interesting aspect of the work of Magaloni is the fact that she divides the period under study in two subsamples: the period of the so called "populist" governments, which are the regimes of Luis Echeverria ( ) and Jose Lopez Portillo ( ), and the "technocrats" or neoliberal governments, which include the regimes of Miguel de la Madrid ( ), Carlos Salinas de Gortari ( ) and Ernesto Zedillo ( ). 20

32 Magaloni identifies the populist governments as leftist and the neoliberals as right wing governments, incorporating elements of the partisan model to the analysis. She states that populist governments tend to favor economic growth and lower unemployment even at the price of higher inflation, and neoliberal governments tend to favor economic stability, even if it means a relative stagnation of economic activity. Under this criterion, she divides the period under study in two subsamples: the Populist" period, from 1970 to 1982, and the "technocrat" period, from 1983 to 1998, finding evidence of partisan behavior of these two wings of the PRI, consistent with the postulates of the partisan model. Gamez (2003) used the coincident and leading indexes of Mexico to examine one implication of the opportunistic traditional model in the Mexican case: the so-called "curse of the first year", the slowdown of economic activity which takes place after the new President takes office. He analyzed the last four changes of President in the country (1982, 1988, 1994 and 2000). Using ordinary linear regression analysis with dummy variables, he finds evidence of the presence of this phenomenon in the period under observation, although the estimations were affected by the presence of autocorrelation. Gamez (2004a) studies the effect of the opportunistic political cycle in a series of macroeconomic and financial variables for Mexico during the period. The variables included were: gross national product, coincident indicator, leading indicator, 21

33 inflation, nominal and real interest rates, nominal and real rate of depreciation of the currency, using an autoregressive model with dummy variables. The study found evidence of the presence of the opportunistic cycles in the variables related with economic activity (GDP, coincident and leading indicators), but not in the financial variables. Gamez (2004a, 2004b) tested some of the postulates of the opportunistic model of the political business cycle in Mexico during the last four Presidential elections (1982, 1988, 1994 and 2000). Analyzing information about economic activity and inflation between 1980 and 2004, evidence is found about one of the postulates of the model: the contraction of the economic activity after the elections, but there is not evidence of an expansion prior to neither the elections nor an increase of the inflation rate around the electoral dates. Gamez (2005a, 2005b) studied the effects of the political cycle on interest rates and the exchange rate in Mexico in the last three elections (1988, 1994 and 2000) using an autoregressive model with a set of 13 dummy variables, which cover the 13 quarters around elections. He found evidence of an increase in both the real interest rate and the rate of depreciation of the peso in the quarters 1 and 2 after the elections, which are the quarter in which the change of President takes place and the first quarter of the new administration. As can be seen after this general review of the literature, there are theoretical bases and empirical evidence of the influence of the political cycle in two kind of economic variables: the so-called outcome variables, such as the rate of growth of the GDP, 22

34 unemployment, inflation, interest rates; and the instrumental variables, which can be used by governments to influence the economy, such as monetary policy (money supply, credit), fiscal policy (taxes, public expenditure, public deficit and debt) and the exchange rate policy (adjustments to the exchange rate). Regarding the political events, those most studied in the literature are elections and changes of administration. It has been observed that elections are more important in developed countries, whereas in emerging economies both elections and changes of power influence the economy. Another important difference between developed and emerging countries is that in the former the patterns of behavior seem to follow the partisan rational model, which is based in the differences between the economic policies followed by different political parties and their effects on the economy. In emerging countries, on the contrary, follow the patterns postulated by the opportunistic model, in which the behavior of the economy is explained by the electoral cycles. 3.4 Political budget cycles at national level The empirical research of the political budget cycle is more recent and if has focused more in emerging countries. Here one mainly finds studies for groups of countries, and studies for single countries, in this case, Mexico. Schuknecht (1996, 2000) studies the fiscal policy instruments by which governments try to influence election outcomes in 24 developing countries for the period. The study finds that the main vehicle for expansionary fiscal policy is 23

35 increasing public expenditure rather that lowering taxes, and public investment cycles seem particularly prominent. Institutional mechanisms which constraint discretionary expenditure policies and which strengthen fiscal control are therefore worthwhile considering to prevent opportunistic policy making around elections. Brender and Drazen (2003, 2005) study a data set of 107 countries. They collect data on the central government balance, total expenditure and total revenue and grants for the period. Their main conclusion is that in a broad cross-section of democracies over the period , there exists a political cycle in the fiscal balance, although the strength of the cycle is sensitive to variable definitions, the time period and the set of countries included. "Our empirical results indicate quite clearly that the political budget cycle is a phenomenon of new democracies. The strong political cycle in these countries, which is characterized by increased expenditures in election years, account for the findings of political deficit cycles in larger samples including these countries...this finding suggests that fiscal manipulation may "work" because of lack of experience with electoral politics or lack of information that is produced in established democracies and that more experience voters use (Brender and Drazen, 2005 p. 1292)". Shi and Svensson (2002, 2005) studied a data panel set consisting of 85 countries over a 21-year period ( ). They find that there is evidence of political budget cycles. On average, fiscal deficit increases 22% in election years (1% of the GDP). They also find that political budget cycles are large in developing countries, but small or inexistent in developed countries. 24

36 Gonzalez (2000, 2002) studies the political budget cycle in Mexico between 1957 and The author finds evidence of manipulation of fiscal policy instruments in the periods prior to the elections. Among the fiscal instruments that follow an electoral cycle, she identifies public expenditure in general and some items as public investment, public works and transference payments. She also finds that the magnitude of the election cycle has been exacerbated during the country's most democratic episodes. 3.5 Political budget cycles at subnational level This phenomenon has been studied in Russia (Akhmedov, et. al. 2002, Akhmedov and Zhuravskaya, 2003). These authors study a comprehensive list of Russia's regional elections, using monthly panel data between 1996 and 2002, finding strong evidence of opportunistic political cycles in regional fiscal policies. They use panel data analysis with dummy variables. Among the main results, they found the following: There is strong evidence of opportunistic political cycles in regional fiscal policies, especially in spending on social programs, healthcare, education and industrial subsidies. - Budget cycle is very sizable and short-lived; large expansion and contraction in fiscal spending occur within two months of the election in both sides. - The magnitude of the cycle decreases with government transparency, level of local democracy and voter awareness. The cycle becomes smaller over time. 25

37 - Pre-election manipulation of fiscal instruments increases incumbent's chances of reelection. The results confirm theoretical findings that maturity of democracy, transparency and voter awareness are important in determining the scope for opportunistic cycles at the regional level. Khemani (2000) studies the effect of state elections on the policies of state governments in 14 major states of India, over the period Two levels of policy manipulation, fiscal policy and public service delivery are contrasted to distinguish between alternate models of policy cycles. The predictions of three models are tested: 1) populist cycles to poorly informed and myopic voters, 2) signaling models with asymmetric information, 3) a moral hazard model with high discount by political agents. She uses annual data of fiscal policy and political variables to study 116 state elections held in the period. The methodology used was autoregressive analysis with dummy variables. She found evidence of political cycles in fiscal policies in Indian states: taxes on producers are lower, public investment spending is higher, and road construction by public works departments is higher in election years. The substantial increase in new roads in election years, even after controlling for spending, implies that government management of public works improved. This is consistent with a moral hazard model of career concerns where politicians exert greater effort in public service delivery to influence voter's inference about their ability. Medina (2003) explores the existence of political cycles in fiscal policy variables at subnational level in Argentina during the period , using the date of elections as 26

38 explanatory variable. He finds evidence of electorally-motivated changes in the level of public expenditure, budgetary deficits and composition of public expenditures in Argentina. The empirical study is made using a panel data analysis with fixed effects for 22 Argentine provinces from 1985 to 2001, finding evidence of political cycles on fiscal policy variables around election dates. The results show that deficits and public expenditures increase in election years. Evidence also suggests that expenditures shift to more visible public investment and away from current consumption goods. Drazen and Eslava (2002), studied the political business cycle in Colombia on both the national and regional levels. For the regional level analysis, they used a panel of data with annual observations for each of the 32 states and their capital cities for the period There are 64 regional units for a total sample size of 960 observations. At the regional level, they found a significant political cycle for total government expenditure. This cycle is more pronounced in investment expenditure, with large (about 20-25%) and significant increases before regional elections. The effect of regional elections on public investment expenditures is worth stressing, as it gives support to the view that pre-electoral manipulation may be seen in public works at the local level. The paper also shows the great importance of using regional as well as national data in studying political business cycles. Goncalvez and Veiga (2004) tested the existence of rational political business cycles models using a large and unexplored data set of Portuguese municipalities. They use panel data on diverse items of public expenditures in the 279 mainland municipalities from 1979 to 2000, covering six electoral periods. 27

39 Empirical results provide clear evidence of political business cycles in Portuguese municipalities. The finding of such cycles at the local level provides support for models of rational opportunistic cycles like that of Rogoff and Siebert (1988). In such models, incumbent politicians manipulate economic policy instruments in order to reveal greater competence shortly before elections. There is evidence that municipal budget deficits and expenditures, especially investment expenditures, increase in election years, and sometimes in the year before. This opportunistic behavior focuses on investment expenditures that are highly visible to the electorate, as rural roads and public works. According to these authors, politically induced fluctuations in expenditures lead to inefficiencies in the allocation of resources. If such cycles could be prevented by new or tougher rules to the management of municipal finances, the result will be an improvement in overall welfare. After reviewing the empirical evidence, several trends in the study of the political business cycle in the last 30 years can be identified. In the first place, it has been a shift in the objects of study, from the economic outcomes, such as growth, inflation and unemployment to policy instruments, and even more, from the monetary to fiscal instruments, especially public expenditures, their size, trend and composition. A second trend is the stress of the study of both the political business cycle and the political budget cycle in emerging countries, especially the so-called "new democracies". By the way, Brander and Drazen place the beginning of democracy in Mexico in 1988 (Brender and Drazen, 2005, p. 1293). 28

40 A more recent trend is the study of the political budget cycle at subnational level. This kind of analysis has been done for some emerging countries, such as Russia, India, Colombia, Argentina and Portugal, to mention some, but it has not been done for Mexico, being this one of the contributions of this dissertation. 29

41 CHAPTER 4.- THE POLITICAL CYCLE AND THE MEXICAN ECONOMY: SOME STYLIZED FACTS Mexico has held regular elections and presidential regimes since 1934, with the administration of Lazaro Cardenas. This makes 12 presidential regimes (sexenios) without interruptions due to uprisings, military coups, civil wars or other disturbances that have affected many developing countries in this period. As a matter of fact, Mexico is the emerging country with the longest history of political stability, and this fact makes Mexico a good subject of analysis of the political business cycle in emerging economies. In this chapter the existence of a political economic cycle in Mexico is tested in several periods and with different variables, according to the availability of information in public sources, especially in the "Instituto Nacional de Estadistica, Geografia e Informatica (INEGI)" and Banco de Mexico. The purpose of the chapter is to show that the political cycle, specially the presidential cycle (sexenio) has influenced some of the main macroeconomic variables in Mexico as far as there is information available. The analysis is based in the assumptions of the opportunistic view of the political business cycle; that is, an expansion of the economy before elections, and a contraction after the electoral process has taken place. The chapter is presented in five sections: in the first one, the behavior of the total GDP in the period is examined; in the second, a sample of sectorial GDP from 1960 to 2003 is analyzed; after that the behavior of the real exchange rate from 1970 to 30

42 2004 is presented. The fourth section analyzes a sample of aggregate demand variables from 1980 to Finally, the behavior of some items of public expenditure at the federal level during the period is examined. An addition to the graphic presentation of the variables analyzed, regression analysis, using autoregressive models with dummy variables to represent the political cycle is included In the case of annual data, three dummy variables to represent the political cycle were defined: DAE = 1 in the year before elections = 0 otherwise DE = 1 in the year of elections = 0 otherwise DDE = 1 in the year after elections = 0 otherwise In the case of quarterly or monthly data, the political cycle is defined by a two sets of dummy variables, one to identify the period before elections and the other to identify the period after elections. The set will take the following form: Q (-6) = 1 in the sixth quarter before elections = 0 otherwise Q (-5) = 1 in the fifth quarter before elections = 0 otherwise Q (-4) = 1 in the fourth quarter before elections = 0 otherwise Q (-3) = 1 in the third quarter before elections = 0 otherwise Q (-2) = 1 in the second quarter before elections 31

43 = 0 otherwise Q (-1) = 1 in the first quarter before elections = 0 otherwise QO = 1 in the quarter when elections take place (July-September) = 0 otherwise Ql = 1 in the first quarter after elections (in this quarter the new President takes office) = 0 otherwise Q2 = 1 in the second quarter after elections = 0 otherwise Q3 = 1 in the third quarter after elections = 0 otherwise Q4 = 1 in the fourth quarter after elections = 0 otherwise Q5 = 1 in the fifth quarter after elections = 0 otherwise Q6 = 1 in the sixth quarter after elections = 0 otherwise This set of dummies will capture the effects of the two political events identified as relevant in the literature: the presidential elections and changes of administration. According to the postulates of the opportunistic political cycle theory the expected signs of the dummies representing the period before and during elections are positive, whereas for the dummies representing the period after elections a negative sign is expected. In order to run the regressions it is necessary to test if the series included are stationary. This was done using the augmented Dickey and Fuller test (Studenmund, p ). In all cases the series are stationary according to the ordinary criteria Sample In this sample, the behavior of the growth rate of the Gross Domestic Product is analyzed for the period This period includes 12 presidential administrations, 32

44 from Lazaro Cardenas ( ) to the first years of the administration of Vicente Fox, the current President of Mexico. In this period Mexico has experienced regular elections and changes of administration every six years. It is worth noting that in nine out of the last ten changes of president, a reduction in the growth rate of the GDP is observed in the first year of the new government. In three of these cases (1983, 1995 and 2001), this reduction has caused negative rates of growth, the so-called "curse of the first year". LONG RUN GDP 33

45 The regression analysis shows the following results: Dependent Variable: PIB Method: Least Squares Date: 02/10/05 Time: 08:59 Sample(adjusted): Included observations: 70 after adjusting endpoints Variable Coefficient Std. Error t-statistic Prob. C PIB(-l) DAE DE DDE R-squared Adjusted R-squared S.E. of regression Sum squared resid Log likelihood Durbin-Watson stat Mean dependent var S.D. dependent var Akaike info criterion Schwarz criterion F-statistic Prob(F-statistic) As can be seen, the coefficient of the dummy representing the year following elections, which is the first year of the new president, is negative and significant, confirming the slowdown of economic activity in the first year of the "sexenio". However, for the pre-electoral and electoral years, the signs are contrary to the expected and non significant Sample In this sample the behavior of the sectorial GDP in Mexico is analyzed from 1960 to 2003, period that covers 7 "sexenios". The main results are that the "curse of the first year" affects almost all the main sectors of the Mexican economy, but is especially strong in activities related with the internal markets, such as the construction industry and commerce. 34

46 GDP: SELECTED INDUSTRIES 2O.( ft It f O.c -5.L. ' J> -*» 10.0% 15.0% 20.0% % -MANUFACTURE 'BUILDING COMMERCE following: The regression for the for the Gross Domestic Product in the period (PIB) is the Dependent Variable: PIB Method: Least Squares Date: 02/10/05 Time: 09:52 Sample(adjusted): Included observations: 42 after adjusting endpoints Variable Coefficient Std. Error t-statistic Prob. C PIB(-l) DAE DE DDE R-squared Adjusted R-squared S.E. of regression Sum squared resid Log likelihood Durbin-Watson stat Mean dependent var S.D. dependent var Akaike info criterion Schwarz criterion F-statistic Prob(F-statistic)

47 For the manufactures sector (MAN): Variable C MAN(-l) DAE DE DDE R-squared Adjusted R-squared S.E. of regression Sum squared resid Log likelihood Durbin-Watson stat Dependent Variable: MAN Method: Least Squares Date: 02/10/05 Time: 09:53 Sample(adjusted): Included observations: 42 after adjusting endpoints Coefficient Std. Error t-statistic Mean dependent var S.D. dependent var Akaike info criterion Schwarz criterion F-statistic Prob(F-statistic) Prob For the construction industry (CONS): Variable C CONS(-l) CONS(-2) CONS(-3) DAE DE DDE R-squared Adjusted R-squared S.E. of regression Sum squared resid Log likelihood Durbin-Watson stat Dependent Variable: CONS Method: Least Squares Date: 02/10/05 Time: 09:54 Sample(adjusted): Included observations: 40 after adjusting endpoints Coefficient Std. Error t-statistic Mean dependent var S.D. dependent var Akaike info criterion Schwarz criterion F-statistic Prob(F-statistic) Prob

48 For the commercial sector (COM): Dependent Variable: COM Method: Least Squares Date: 02/10/05 Time: 09:55 Sample(adjusted): Included observations: 39 after adjusting endpoints Variable Coefficient Std. Error t-statistic Prob. C COM(-l) COM(-2) COM(-3) COM(-4) DAE DE DDE R-squared Adjusted R-squared S.E. of regression Sum squared resid Log likelihood Durbin-Watson stat Mean dependent var S.D. dependent var Akaike info criterion Schwarz criterion F-statistic Prob(F-statistic) As can be seen, the coefficient of the dummy representing the year after elections is negative in all cases, but it is stronger in the construction industry (CONS). The signs of the other two political dummies (DAE and DE) are non significant and, in many cases with a negative sign, which is contrary to the expected Sample In this section the real exchange rate index is analyzed during the period, in which there were 5 presidential administrations. This index is elaborated and published by Banco de Mexico on a monthly basis since January, 1970, calculated by using the differentials in consumer inflation between Mexico and a sample of 111 countries. 37

49 In addition to the index, the real annual depreciation rate of the peso is calculated, measured as the difference between the value of the index in a given month and the value of the same month of the former year, in percentage terms. REAL EXCHANGE RATE </ <f' </' </' </' </ <f <f </" </«/" </ <f </ </ </" </ </'</' </</ </ </" <?" <?"'' </' </" </ </' </" 38

50 REAL DEPRECIATION RATE 39

51 The results of the regression analysis of the real depreciation rate (VAN) are the following: Variable C VAN(-l) VAN(-2) Q-6 Q-5 Q-4 Q-3 Q-2 Q-1 Q0 Ql Q2 Q3 Q4 Q5 Q6 R-squared Adjusted R-squared S.E. of regression Sum squared resid Log likelihood Durbin-Watson stat Dependent Variable: VAN Method: Least Squares Date: 02/21/05 Time: 11:10 Sample(adjusted): 1971: :10 Included observations: 404 after adjusting endpoints Coefficient Std. Error t-statistic Mean dependent var S.D. dependent var Akaike info criterion Schwarz c;riterion F-statistic Prob(F-statistic) Prob As can be seen, in spite of the fact that in the period under study, Mexico experimented virtually all the spectrum of exchange rate regimes, (from fixed exchange rate in the beginning of the seventies to the floating exchange rate since 1995), there is a clear sexenal cycle in the real exchange rate, with a tendency to appreciate in the first quarter of the electoral year (Q-2) and abrupt adjustments after elections. 40

52 The coefficients of the annual depreciation rate (VAN) are positive and significant in (QO) and (Q2), which indicates strong adjustments of the exchange rate in the election quarter and in the first quarter of the new administration Sample In this sample the behavior of diverse components of aggregate demand in Mexico is analyzed, using quarterly data from 1980 to The period includes four presidential elections (1982, 1988,1994 and 2000). Information is included about private consumption (CPRIV), public consumption (CPUB), investment (INV) and exports (EXP). In all cases the annual rate of growth is analyzed. AGGREGATE DEMAND -PRIVATE CONSUMPTION PUBLIC CONSUMPTION 41

53 AGGREGATE DEMAND -INVESTMENT EXPORTS 42

54 The regression for private consumption (CPRIV): Variable C CPRIV(-l) CPRIV(-2) CPRIV(-3) Q-6 Q-5 Q-4 Q-3 Q-2 Q-1 Q0 Ql Q2 Q3 Q4 Q5 Q6 R-squared Adjusted R-squared S.E. of regression Sum squared resid Log likelihood Durbin-Watson stat Dependent Variable: CPRIV Method: Least Squares Date: 02/14/05 Time: 09:53 Sample(adjusted): 1981:4 2004:3 Included observations: 92 after adjusting endpoints Coefficient Std. Error t-statistic Mean dependent var S.D. dependent var Akaike info criterion Schwarz criterion F-statistic Prob(F-statistic) Prob

55 For public consumption (CPUB): Variable C CPUB(-l) CPUB(-2) Q-6 Q-5 Q-4 Q-3 Q-2 Q-l Q0 Ql Q2 Q3 Q4 Q5 Q6 R-squared Adjusted R-squared S.E. of regression Sum squared resid Log likelihood Durbin-Watson stat Dependent Variable: CPUB Method: Least Squares Date: 02/14/05 Time: 09:59 Sample(adjusted): 1981:3 2004:3 Included observations: 93 after adjusting endpoints Coefficient Std. Error t-statistic Mean dependent var S.D. dependent var Akaike info criterion Schwarz criterion F-statistic Prob(F-statistic) Prob

56 For gross formation of fixed capital (INV): Variable C INV(-l) INV(-2) INV(-3) Q-6 Q-5 Q-4 Q-3 Q-2 Q-1 Q0 Ql Q2 Q3 Q4 Q5 Q6 R-squared Adjusted R-squared S.E. of regression Sum squared resid Log likelihood Durbin-Watson stat Dependent Variable: INV Method: Least Squares Date: 02/14/05 Time: 10:05 Sample(adjusted): 1981:4 2004:3 Included observations: 92 after adjusting endpoints Coefficient Std. Error t-statistic Mean dependent var S.D. dependent var Akaike info criterion Schwarz criterion F-statistic Prob(F-statistic) Prob

57 For exports (EXP): Variable C X(-l) X(-2) Q-6 Q-5 Q-4 Q-3 Q-2 Q-i Q0 Ql Q2 Q3 Q4 Q5 Q6 R-squared Adjusted R-squared S.E. of regression Sum squared resid Log likelihood Durbin-Watson stat Dependent Variable: X Method: Least Squares Date: 02/14/05 Time: 10:08 Sample(adjusted): 1981:3 2004:3 Included observations: 93 after adjusting endpoints Coefficient Std. Error t-statistic Mean dependent var S.D. dependent var Akaike info criterion Schwarz criterion F-statistic Prob(F-statistic) Prob The cyclical effect is stronger in the investment component (INV) of the aggregate demand. It is observed a strong negative effect in (Q2), which is the first quarter of the new administration. In the case of exports, the effect is the contrary. In (Q0), the quarter of elections the rate of growth of exports increases significantly, reflecting the adjustments of the exchange rate. In the case of both private and public consumption, there is not a significant change around elections. 46

58 Sample In this sample, some items of federal expenditure are examined, using monthly data from 1987 to 2004, in which three elections and changes of President took place (1988, 1994 and 2000). This is a test of the presence of the political budget cycle in Mexico at the federal level. The data included are the real annual change rate of total expenditures (TOT), expenditures in public wages (SAL) and capital expenditures (CAPITAL). 47

59 FEDERAL EXPENDITURE 48

60 FEDERAL WAGES 50.0% - i 40.0% 30.0% -40.0% - - CAPITAL EXPENDITURE 49

61 The results of the regression analysis for the total federal expenditures (TOT) are the following: Variable C TOT(-l) TOT(-2) TOT(-3) Q-6 Q-5 Q-4 Q-3 Q-2 Q-1 Q0 Ql Q2 Q3 Q4 Q5 Q6 R-squared Adjusted R-squared S.E. of regression Sum squared resid Log likelihood Durbin-Watson stat Dependent Variable: TOT Method: Least Squares Date: 02/21/05 Time: 09:50 Sample(adjusted): 1987: :09 Included observations: 210 after adjusting endpoints Coefficient Std. Error t-statistic Mean dependent var S.D. dependent var Akaike info criterion Schwarz ( criterion F-statistic Prob(F-statistic) Prob

62 For the federal expenses in wages (SAL): Variable C SAL(-l) SAL(-2) Q-6 Q-5 Q-4 Q-3 Q-2 Q-l Q0 Ql Q2 Q3 Q4 Q5 Q6 R-squared Adjusted R-squared S.E. of regression Sum squared resid Log likelihood Durbin-Watson stat Dependent Variable: SAL Method: Least Squares Date: 02/21/05 Time: 09:55 Sample(adjusted): 1987: :09 Included observations: 211 after adjusting endpoints Coefficient Std. Error t-statistic Mean dependent var S.D. dependent var Akaike info criterion Schwarz criterion F-statistic Prob(F-statistic) Prob

63 For federal capital expenses and public works (CAPITAL): Variable C CAPITAL(-l) CAPITAL(-2) Q-6 Q-5 Q-4 Q-3 Q-2 Q-1 Q0 Ql Q2 Q3 Q4 Q5 Q6 R-squared Adjusted R-squared S.E. of regression Sum squared resid Log likelihood Durbin-Watson stat Dependent Variable: CAPITAL Method: Least Squares Date: 02/21/05 Time: 10:03 Sample(adjusted): 1987: :09 Included observations: 211 after adjusting endpoints Coefficient Std. Error t-statistic Mean dependent var S.D. dependent var Akaike info criterion Schwarz < criterion F-statistic Prob(F-statistic) Prob

64 It is possible to clearly identify a pattern in public expenditures around election dates. There is a significant increase in (Q-2), the first quarter of the electoral year, which is stronger in the capital expenditure, and a significant contraction in Q2, the first quarter of the following year, which is also stronger in the capital expenditure. In the case of public wages it is possible also to observe an increase in (Q-2), (QO) and (Ql), which are the first, third and fourth quarters of the electoral year, the last of the ending administration. There is also a significant contraction in (Q2), the first quarter of the new presidential regime. 4.6 Discussion As can be summarized, for the evidence presented here, the political cycle indeed affects the economic activity in Mexico, as far as there is information available for the main macroeconomic variables. The results are roughly consistent with the findings obtained in other emerging countries and in Mexico. As in Magaloni (2000), evidence is found of a contraction in economic activity after elections, especially after the change of administration, as well of a delay in the adjustment of the exchange rate until the elections have taken place. The expansion of public expenditures before elections identified in this work is also consistent with the findings of Gonzalez (2000, 2002), in particular the increase in public investment in election years. 53

65 In the Mexican case there is one point that makes the political business cycle different from the postulates of the theory and the evidence of other emerging countries: The fact that the main effects of the cycle are not observed before elections, but after them, especially after the new government takes office. Some explanations of this "curse of the first year" have been offered, related with the "learning curve" of the new public officials, as well of other administrative adjustments and delays in the first months of the new government, which cause a contraction in public expenditures, then transmitted to the rest of the economy by the multiplier effect. This effect has not been documented so far, and it constitutes a very interesting line of research. 54

66 CHAPTER 5.- THE MODEL In this section, the model that grounds the empirical investigation is presented. After reviewing the main previous work on this matter, the assumptions about the behavior of the participant agents (citizens, incumbent politicians) are defined, as well as the relationship among them. The equilibrium condition is derived, as well as the implications of it Previous work Most studies of the political budget cycle at subnational level are based on the rational opportunistic model proposed by Rogoff and Siebert (1988) and Rogoff (1990). In these models, the incumbent or the political party in power try to "signal" their competence to the voters, in a situation of information asymmetry, in order to improve their probabilities of victory in the elections. This is a dynamic signaling model in which both voters and politicians are rational, utility-maximizing agents. A political budget cycle arises due to temporary information asymmetries about the incumbent leader "competence" in administrating the provision of public goods. The incumbent leader has an incentive to bias preelection fiscal policies toward easily observed public goods, such as public works or transfer payments. In equilibrium, voters can deduce the leader's competence by the degree to which he distorts tax and expenditure policies. This equilibrium political business cycle model suggests that it would be more promising to focus empirical research on testing for electoral cycles in taxes, transfers and public expenditures. It is also suggesting to look of these kind of data for state and local 55

67 elections, instead of concentrating only in the small number of elections available for national elections (Rogoff, 1990, p ). Shi and Svensson (2005) develop a simple moral hazard model of electoral competition. The underlying feature of the model is the ability of the incumbent to manipulate policy instruments (which are observable to voters only with a lag) in order to bias the voter's inference process to his favor before elections. To increase the chances of reelection, or the winning of his party, the incumbent has an incentive to boost the supply of public goods prior to the election, hoping that voters would attribute the boost to his competence. In the model, all politicians, independently of their level of competence, have the same incentives, and these incentives depend on the institutional environment. Specifically, the more private benefits politicians gain when in power (ego rents) the stronger their incentives to influence the voter's perceptions Assumptions Borrowing ideas from these two models, in this case a model is proposed based on the following assumptions, to ground and justify the investigation: 1.- Citizens derive utility from the consumption of private and public goods, and the perceived competence of the government in power The utility function of a representative citizen is: 1) UC,=U(C,,G,) + e t 56

68 Where: C = Consumption of private goods G = Consumption of public goods, = = The competence of the government, perceived as the ability to provide public goods given a level of taxes 2. - Citizens face taxes Tt to determine their private consumption. 2) C,=Y,-T t 3. - Taxes are exogenous at the state level, and are supposed to be constant in the period 3) T,=T 4. - For the period in power (n), there is a budget balance in state public finances 4) ng, = nt 57

69 5. - The government in power (executive power) has some degree of discretionary influence (y) over the provision of public goods (G), which makes it able to provide discretionary public goods (DG) at time t 5) DG, = jg, Where y is the ability to influence public expenditures by the executive. Ay=l would mean that all public expenditures are under direct control of the executive. On the contrary, a y = 0 would mean that the executive has no influence at all in the public expenditures. In this context, the value of y will depend on: Institutional arrangements, such as the composition of the local Congress, the amount of public expenditures that are already "tied" by the federal government, and the ideology of both the state and federal government, among others. - The personal ability of the incumbent to gain discretionary power in the determination of public expenditures. In this model, it is assumed that: 0<y<l 6. - Politicians derive utility from consumption, and from "ego rents", which are the advantages associated with staying in power. 58

70 The utility function of the politician in power is: 6) UG, =UC, +P,X, Where: UG t = The utility of the politician in power at the time t UC t = The utility of the representative citizen at the time t X t = The advantages associated to being in power (ego rents) at the time t P t = The probability of staying in office in period t Competence is not observable; citizens infer it from the ability of the incumbent to provide public goods in the periods before elections. 7),=f{,_,+,_ 2 +,_ ) 8. - Citizens are shortsighted, and form their estimation of competence from the past actions of the incumbent, in such a way that they "discount" more heavily the more distant past facts. 8) e.^e.j+,_/+ + e t _f P = Discount rate of the representative citizen that will depend on: 59

71 Levels of education of the electorate Degree of political information and experience Quality of mass media In this model it will be assumed that: 0<p<l 9. - Incumbents try to maximize their probability of winning the election, and staying in power. The probability of winning the elections by the incumbent is a function of his (perceived) competence, as well as other factors, such as ideology, personal characteristics, and charisma, not related with the provision of public goods 9) P,=f(e,,L) 8 1 = Perceived competence of the incumbent at time t L = Personal characteristic of the incumbent (look) The probability of winning the election then becomes a function of the perceived competence of the incumbent at the time of elections (8 1 ) which, in turn, is a function of the past performance of the incumbent, in the form of (8) 60

72 Given that taxes (T) are constant, equation 8) becomes: 10) I =G,S+G I _/+ + G,_ f 11.- By assumption, the government in power can only manipulate a fraction of the provision of public goods (y G). Then, the function it wants to maximize is: 1 1 j t j /VJ"J_J T" J\J t _2 ' ' /vj r i n Subject to the budget restriction: This can be done by constructing a Lagrangian \l) L Y^t-x + r^i-2 + The first order conditions are: SL 61

73 SDG,_ 2 8L Solving this system, we have: And for every 0 < (3 < 1 yg,_ x Implications In this model the existence and magnitude of the political budget cycle would depend on two parameters: The degree of influence of the executive power in the allocation and timing of public expenditures (y), and the shortsightedness of the electorate, which, in turn, would depend on the levels of political information and education of the citizens, which is represented by (3. Under these circumstances, there are two extreme cases in which the opportunistic cycle would not take place: one is related to the institutional arrangements behind the 62

74 public expenditures and the other related with the information and awareness of the citizens. The first case can be represented when y = 0. In this case the executive power has no influence at all in the allocation of state expenditures, then, it will be simply not possible to manipulate the expenditures in order to increase the probabilities of an electoral victory, and there will not be an opportunistic budget cycle. The other extreme case is when P = 1. In this case HG,_ X = jg,_ 2 = /},_ Which means that the public expenditures would be distributed uniformly in the complete period. The intuition behind this is that if the electorate is well informed and aware of the possible manipulation of public expenditures in an opportunistic way, they will punish opportunistic behavior rather than reward it and then the main motivation that explains the opportunistic behavior of governments would disappear. 63

75 CHAPTER 6.- METHODOLOGY In this chapter, the methodology used in the analysis at the state level is presented. In the first section the general aspects of the panel data analysis are commented, as well as some of the applications of this analysis to the study of the political budget cycle. After that, some of the main characteristics of the fixed effects models are discussed as well as some of the limitations of this kind of analysis. The chapter closes with a description of the model used in this study Panel data analysis An important limitation of simple time-series analysis is that it requires many (at least thirty) observations to make reliable inferences. For annual economic series, as is the case of public expenditures at state level in Mexico, this means that the analysis has to be done for a period of time in which many other important elements, such as the political conditions, the institutional environment, the demographic composition of the population, the influence of mass media, to mention some, may have changed, and the comparison is not entirely valid. The use of panel data allows the analysis of periods of time relatively short, but with enough observations to make valid statistical inference. Panel data represents a mix of regression and time series analysis. As well as with many regression data sets, longitudinal data are composed by a cross section of subjects. Unlike regression data, with longitudinal data, subjects are observed over time. Unlike simple time-series data, with longitudinal panel data, many subjects are observed. By 64

76 observing a broad cross-section of subjects over time allows to study dynamic, as well as cross-sectional aspects of a problem. There are several advantages of panel data analysis, compared with either purely cross-sectional or purely time-series analysis. Two of the most important are the ability to study a dynamic relationship and to model the differences, or heterogeneity, among subjects (Frees, 2004, p. 2-3). The panel data analysis has been applied to the study of the political budget cycle in two levels: at national level there are the works of Shucknecht (1996, 2000), Brender and Drazen (2003, 2005), and Shi and Svensson (2002, 2005), mentioned in Chapter 3. However, one risk associated to the application of panel data analysis to a sample of countries is that the institutional environment around public finances may be different across countries, and then the comparison is not appropriate. In the case of the studies at subnational level, the institutional environment regarding public expenditures is more homogeneous. This methodology has been used by Khemani (2000) in India, Akhmedov et, al. (2002, 2003) in Russia, Medina (2003) in Argentina, Drazen and Eslava (2004) in Colombia and Goncalvez and Veiga (2004) in Portugal Fixed effects models There are two distinct approaches to model the quantities that represent heterogeneity among subjects. In one of them, the heterogeneity factor is treated as fixed, yet unknown, parameters to be estimated. In the other approach, the heterogeneity factors 65

77 are treated as draws of an unknown population and thus are random variables. In fixed effects models, there are two kind of parameters: on the one hand, there are the population parameters, which are common to each subject, and the subject-specific parameters, which vary with subjects. In many applications, the population parameters capture broad relationship of the sample studied and are the parameters of interest. The subject-specific parameters account for differences among subjects (Frees, p. 22). The main assumptions of the fixed effects linear longitudinal data model are the following: 1) Ey^Zp.+XJ 2) {x,,, x itk } and {z, n z ilk } are nonstochastic variables 3) Var>; ; = R(x) = R, 4) {y j } are independent random vectors 5) \y jt } are normally distributed (Frees, p. 50) Limitations Panel data analysis has also some drawbacks. Among them there is the case of missing observations, the data about some subjects in some periods of observation do not exist. Related to this is the possibility of selection bias, which means that in some cases, the absence of information is not due to a random process, but responds to a deliberate attitude of the subject. 66

78 In this study, as will be commented in Chapter 7, one of the main problems faced was the enormous amount of missing values in the disaggregate levels of state expenditures. These omissions could be related simply to the poor quality of the statistical recollection systems of the states, but it is also open the possibility of "deliberate" omissions, made in order to hide some inadequate or irregular behavior. There is another drawback which is common not only to panel data analysis, but to statistical analysis in general: their ability to detect causal relationships among variables. Statistical relationships must be backed by the relevant theory in order to infer causation. "Just as any statistical methodology, longitudinal data models in and of themselves, are insufficient to establish causal relationships among variables Rather, one thinks about the data and statistical models as providing relevant empirical evidence in a chain of reasoning about causal mechanisms. Although longitudinal data provide stronger evidence than purely cross-sectional data, most of the work on establishing causal statements must be based in the theory of the substantive field from which the data are derived" (Frees, p. 11) The model Following Brender and Drazen (2005) and Shi and Svensson (2006) The model proposed for this study takes the following form: + delec > + M > Where: 67

79 Fit = fiscal indicator of state i at time t X = vector of control variables ELEC = electoral dummies M = State fixed effect E = Error term In this study the unit of observation is the fiscal indicator of the state. The difference among states is represented by the sub-index "i", that goes from 1 to 31 to represent the Mexican states included in the study (Distrito Federal is not included). The sub-index "t" represent time, and goes from 1990 to The dependent variable will be the annual variations in public expenditures for the different states of the sample. The independent variables are the lagged dependent variable, to account for the inertia, some control variables, which are factors that can affect the behavior of the dependent variable, and political dummies to represent the political cycle, which are related with the hypothesis to be tested that will be specified in Chapter 8. This study will be a case of balanced data, because the same variables and the same quantity of observations are used for each subject. There are included 15 observations for five variables in each of the 31 states in Mexico, which makes a total number of 465 observations. 68

80 CHAPTER 7.- THE DATA For the empirical estimations three sets of data are used: fiscal variables, which are the dependent variable in the study, political data, related to the state election dates, and data about the control variables included in the study. 7.1 Fiscal variables The fiscal variables used in the study were obtained directly from the Institute Nacional de Estadistica, Geografia e Informatica (INEGI), the official agency of statistics in Mexico. This data set includes annual information of the income and expenditures of Mexican states from 1989 to In this period there were at least two (in some cases three) elections in each of the 31 states included in the data set. The data set was supposed to include information about nine categories of public expenditures. However, a close examination of the data revealed a lot of omissions, for many states and in many years. It was possible to construct only five series of expenditures for which there was information for all states and all the period, these are: - Total expenditure Wages and remunerations Subsidies, and transfer payments - Public works and social actions Other items, which include administrative expenses, payments of services, purchases and acquisitions, service of public debt and transfers to municipalities. 69

81 In the case of subsidies, there were four missing values (Chiapas 1996, Puebla 1995, Sonora 1995 and Veracruz 1997). The gaps were filled using the average of the adjacent years, as in Drazen and Eslava (2003, p. 7). The information is reported on nominal annual basis. The original data were deflated using the Indice Nacional de Precios al Consumidor (INPC) elaborated by the Banco de Mexico, to be expressed in 2002 constant pesos. The variables included in the estimations were the annual change in the real value of state public expenditures. The final database included five variables for 31 states from 1990 to 2004, which makes 465 observations Political variables The political variables included in the analysis are the election years in Mexican states, the year before and the year after. Three dummy variables are constructed in relation to election years: DAE representing the year before elections, DE representing election years and DDE representing the year after elections. The period of study includes 80 state elections Control variables According to the literature reviewed, control variables are related to the level of income of the population, the size of the state, the level of education of the electorate, the existence of relevant political information, and the quality of mass media, among others. 70

82 Khemani (2000) uses the state domestic product (SDP), the proportion of agriculture in SDP and the proportion of rural population in Indian states in her study. Ahkmedov et. al. (2002, 2003) use macroeconomic variables in Russia, such as GDP, inflation and public budget, and some indicators at the state level, such as the levels of education, urbanization and computerization of the population. In this investigation, the state GDP is used as a measure of the general economic conditions of the state. The INEGI provides information of annual state GDP in Mexico from 1993 to The GDP for the rest of the period ( ) was estimated retrospectively as a function of national GDP. Another control variable tested in this study was the rural proportion of the state population, using the estimations made by the Consejo Nacional de Poblacion (CONAPO). As will be explained in Chapter 8, the inclusion of this variable did not added significant explaining power to the model and was finally omitted. It was not possible to find adequate and complete series of information related to the level of education of the electorate, the availability of political information, and penetration of the media, that have been used to proxy the degree of awareness of the electorate to political events Quality and limitations of the data One of the findings of this study is the poor quality of the information about public finances at state level in Mexico. As mentioned before, the large number of omissions and missing values reduced drastically the number of series available for analysis in the period considered. 71

83 One of the main problems created by the poor quality of the data was that it made impossible to analyze the items of public expenditures at disaggregate levels, which very often is the level at which the political cycle is detected. Khemani (2000) could divide state expenditures into development and nondevelopment expenditure. Development expenditure includes recurrent expenditure on the maintenance of social and economic services, including large agricultural subsidies. Nondevelopment expenditure consists of spending on administrative and fiscal services, including compensation to government employees. Akhmedov et al. (2002, 2003) were able to analyze state expenditures in spending on education, healthcare, social disbursements, industrial subsidies, and mass media, among other and it was in this level of disaggregation where some of the more interesting finding were done. An important lesson derived from the present study is the need to improve the quality of the public finance data at subnational level in Mexico, in order to make possible to infer valuable and valid conclusions from the analysis of them. 72

84 CHAPTER 8- RESULTS In this chapter the estimated results of the model are presented. The regressions were made by using the econometric package E-Views 3.0. As stated in Chapter 5, the basic model to estimate is the following: < + delec < Where: Fi,t = fiscal indicator of state I at time t X = vector of control variables (PIB) ELEC = electoral dummies (DAE, DE, DDE) M = State fixed effect E = Error term The estimation is done for the five variables available for study: - Total expenditure (GASTOT) - Wages (SAL) - Subsidies (SUB) - Public works (OB) Other expenses (OT) In the case of the fiscal and control variables, they are included in real annual variation terms; that is: the real value of the variables (in 1993 pesos) in a given year, divided by the value of the variable in the previous year, expressed in percentage terms. The hypotheses to test are related with the electoral variables and are the following: Hypothesis 1: It is expected an increase of public expenditures in the year prior to elections, that is, it is expected that DAE to have a positive sign 73

85 Hypothesis 2: It is expected to have an increase in public expenditures in election years, which means that the expected sign of DE is positive. Hypothesis 3: It is expected a contraction of public expenditures in the year following elections, which means that the expected sign of DDE is negative The results of the estimations are the following: 74

86 8.1 Total expenditure (GASTOT) Variable GASTOT?(-1) PIB? DAE? DE? DDE? Fixed Effects AGS-C BC-C BCSUR--C CAMP--C COAH-C COL-C CHIA-C CHIH-C DUR--C GUA-C GUE-C HID-C JAL--C MEZ-C MICH-C MOR--C NAY-C NL--C OAX-C PUE-C QUER-C QR-C SLP-C SIN-C SON-C TAB-C TAM-C TLAX-C VER-C YUC-C ZAC-C Dependent Variable: GASTOT? Method: GLS (Cross Section Weights) Date: 11/21/06 Time: 18:10 Sample: Included observations: 14 Total panel observations 434 Convergence achieved after 7 iteration(s) Coefficient Std. Error _ t-statistic _ Prob

87 Dependent Variable: GASTOT? Weighted Statistics R-squared Adjusted R-squared S.E. of regression F-statistic Prob(F-statistic) Unweighted Statistics R-squared Adjusted R-squared S.E. of regression Durbin-Watson stat Mean dependent var S.D. dependent var Sum squared resid Durbin-Watson stat Mean dependent var S.D. dependent var Sum squared resid _ _ There is a significant increase it the rate of growth of total expenditures in the election years (DE) and the year following (DDE). The increase on the election years is according to the expected, but not the increase in the year following elections. This increase can be explained by the behavior of wages, as shown in the next section. This result means that public expenditures at state level in Mexico show an increase, "ceteris paribus" of 6.5% in election years and of 4.8% in the post-electoral year. The coefficient of DE is significant at the 99% level, whereas the coefficient of DE is significant at the 95% level. 76

88 8.2 Wages (SAU Variable SAL?(-1) PIB? DAE? DE? DDE? Fixed Effects AGS-C BC--C BCSUR-C CAMP-C COAH-C COL-C CHIA--C CHIH-C DUR-C GUA-C GUE--C HID--C JAL-C MEZ-C MICH-C MOR-C NAY--C NL--C OAX--C PUE--C QUER-C QR-C SLP-C SIN-C SON-C TAB-C TAM-C TLAX-C VER-C YUC--C ZAC-C Dependent Variable: SAL? Method: GLS (Cross Section Weights) Date: 11/11/06 Time: 08:26 Sample: Included observations: 14 Total panel observations 434 Convergence achieved after 8 iteration(s) Coefficient Std. Error t-statistic Prob

89 Dependent Variable: SAL? Weighted Statistics R-squared Adjusted R-squared S.E. of regression F-statistic Prob(F-statistic) Unweighted Statistics R-squared Adjusted R-squared S.E. of regression Durbin-Watson stat Mean dependent var S.D. dependent var Sum squared resid Durbin-Watson stat Mean dependent var S.D. dependent var Sum squared resid There is a significant increase in public wages in the year following elections (DDE), which is the first year of the new government. This result suggests that new state governments in Mexico increase substantially the wages for their teams when they take office. The coefficient of DE is significant at the 90% level and the coefficient of DDE is significant at the 95% level. 78

90 8.3 Subsidies (SUB) Variable SUB?(-1) PIB? DAE? DE? DDE? Fixed Effects AGS--C BC-C BCSUR-C CAMP-C COAH-C COL--C CHIA-C CHIH--C DUR-C GUA-C GUE-C H1D-C JAL-C MEZ--C MICH--C MOR-C NAY-C NL-C OAX-C PUE--C QUER-C QR-C SLP--C SIN-C SON-C TAB-C TAM-C TLAX-C VER-C YUC-C ZAC-C Dependent Variable: SUB? Method: GLS (Cross Section Weights) Date: 11/11/06 Time: 08:27 Sample: Included observations: 14 Total panel observations 434 Convergence achieved after 7 iteration(s) Coefficient Std. Error _ t-statistic _ Prob

91 Dependent Variable: SUB? Weighted Statistics R-squared Adjusted R-squared S.E. of regression F-statistic Prob(F-statistic) Mean dependent var S.D. dependent var Sum squared resid Durbin-Watson stat Unweighted Statistics R-squared Adjusted R-squared S.E. of regression Durbin-Watson stat Mean dependent var S.D. dependent var Sum squared resid There is no a significant change in the behavior of state subsidies and transfers in the years around elections. Although the signs of DE and DDE are the expected, none of them is significant at conventional levels. This result is completely different of what was expected, and can be explained by the poor quality of the data. 80

92 8.4 Public works (OB) Variable OB?(-1) PIB? DAE? DE? DDE? Fixed Effects AGS-C BC-C BCSUR-C CAMP-C COAH--C COL-C CHIA-C CHIH--C DUR-C GUA-C GUE-C HID-C JAL-C MEZ-C MICH--C MOR-C NAY--C NL-C OAX-C PUE-C QUER--C QR--C SLP-C SIN-C SON--C TAB-C TAM-C TLAX-C VER-C YUC-C ZAC-C Dependent Variable: OB? Method: GLS (Cross Section Weights) Date: 11/11/06 Time: 08:27 Sample: Included observations: 14 Total panel observations 434 Convergence achieved after 11 iteration(s) Coefficient Std. Error t-statistic _ Prob

93 Dependent Variable: OB? Weighted Statistics R-squared Adjusted R-squared S.E. of regression F-statistic Prob(F-statistic) Unweighted Statistics R-squared Adjusted R-squared S.E. of regression Durbin-Watson stat Mean dependent var S.D. dependent var Sum squared resid Durbin-Watson stat Mean dependent var S.D. dependent var Sum squared resid _ This is another surprising result. There is no electoral cycle at all in the behavior of public works and social actions in Mexican states around elections. The signs of the dummies corresponding the year before elections (DAE) and elections (DE) are contrary to the expected, and none of them is significant. 82

94 8.5 Other expenses (OT) Variable OT?(-1) PIB? DAE? DE? DDE? Fixed Effects AGS--C BC--C BCSUR--C CAMP-C COAH-C COL-C CHIA--C CHIH-C DUR-C GUA--C GUE-C HID-C JAL-C MEZ--C MICH--C MOR-C NAY-C NL--C OAX-C PUE--C QUER-C QR-C SLP-C S1N-C SON-C TAB-C TAM-C TLAX-C VER-C YUC-C ZAC-C Dependent Variable: OT? Method: GLS (Cross Section Weights) Date: 11/11/06 Time: 08:28 Sample: Included observations: 14 Total panel observations 434 Convergence achieved after 8 iteration(s) Coefficient Std. Error t-statistic Prob

95 Dependent Variable: OT? Weighted Statistics R-squared Adjusted R-squared S.E. of regression F-statistic Prob(F-statistic) Unweighted Statistics R-squared Adjusted R-squared S.E. of regression Durbin-Watson stat Mean dependent var S.D. dependent var Sum squared resid Durbin-Watson stat Mean dependent var S.D. dependent var Sum squared resid There is an increase in the rate of growth of the item classified as "other expenses" in electoral years, and the coefficient of DE is significant at a level of 90%. The coefficient of DAE has a negative sign, which is contrary to the expected, whereas DDE has the right sign, but is not significant. 8.6 Discussion As can be seen, the results obtained in the estimations confirm the expansion of public expenditures in election years, but not the increase in the year prior to elections, nor the contraction after the electoral processes. Regarding the composition of expenditures, there are no evidence of a significant electoral pattern in the behavior of subsidies and transferences (SUB) and public works (OB) in the years around elections. 84

96 One interesting result is the expansion of public wages in the electoral year and the year following elections. This result can be interpreted as an increase in the payroll by both the government in power by the time of elections and the new elected government in the first months of the new administration. Roughly speaking, the results support one of the hypotheses: the expansion of public expenditures in electoral year, but reject the other two: the expansion of expenditures in the year before elections and the contraction in the post-electoral year. There are several factors that can explain, at least partially, the results obtained: in the first place, the poor quality of data, which made impossible to test the political budget cycle at disaggregate levels. In addition, the omissions detected in the data base cast doubts about the quality of the remaining data, as well as their accuracy. A second factor is the fact that the original data (and their changes) are presented on annual basis, which makes it impossible to detect intra-year effects. As mentioned in Chapter 4, some of the main effects of the political cycle in public expenditures at federal level in Mexico are detected by analyzing quarterly data. This effect was also identified in Russia by Akhmedov et. al (2002, 2003). These authors found that budget political cycles at state level are very short-lived, and can only be measured by the use of monthly data. "Use of monthly panel data allowed us careful measurement of even very short cycles. This turned out to be important because most sizable increases in spending happen a month before and decreases a month after elections, thus, use of lower frequency data would have lead to substantial underestimation of cycles. To the best of our knowledge all 85

97 empirical studies on developing countries so far used quarterly or lower frequency data. Very short cycles, however, cannot be clearly seen in quarterly data because elections often take place in the middle of the quarter and the opposite-sign deviations from the trend around elections cancel out in data with low frequency" (Akhmedov et. al 2003, p. 4). A third point related to this is the fact that elections in Mexican states are not held at the same time. In some states elections take place in February, in others in March, July, September or November, to mention some dates. Given that the data are presented in annual basis, in some cases, if elections take place early in the year, let's say in February, the year registered as election year (DE) is actually the post election year (DDE). On the same basis, if elections take place later in the year (November), the dummy representing election year (DE), would actually be the year before elections (DAE). In these circumstances, some of the patterns of public expenditures before and after elections will mix, and the results can be misleading, or meaningless. One way to deal with this would be to split the sample according to the month in which elections take place, grouped by quarters, or semesters, and to test if it is possible to detect patterns of behavior different to those reported here. Another possible source of explanations is the omission of relevant variables in the model and the estimations. As mentioned before, in this work there were tested two control variables: real GDP and the share of rural population, mainly because there was information available for the period under study. However, some other control variables have been tested in former research. Among them are the proportion of agriculture in the state GDP (Khemani 2000), the levels of 86

98 education and computerization of the population (Ahkmedov et. al. 2003) and the age structure of the population (Goncalvez and Veiga. 2004). The inclusion of some proxies for these variables could improve the results of the estimations. 87

99 CHAPTER 9 - CONCLUSIONS AND LINES OF RESEARCH After reviewing the empirical evidence, as well as the theory behind it, and based in the results of this study, one can draw some conclusions about the political cycle and the Mexican economy. The first is that, generally speaking, there is a political cycle in the Mexican economy. At national level, the cycle is observed in the general economic activity and certain sectors and industries, as well as in some components of the aggregate demand. There is also a cyclical pattern in federal public expenditure and some items of it, which follow the general pattern proposed by the theory of the opportunistic political budget cycle model. At subnational level, there is also a political cycle in public expenditures, but it does not follow exactly the pattern proposed by the opportunistic model. There is an expansion of public expenditures in election years, but not a contraction in the following year. Although the results of the estimations must be interpreted with caution, especially due to the poor quality of the data, these results could suggest that the expansion of public expenditures in election years is not entirely motivated by opportunistic reasons (to "signal" competence), but by administrative factors or other kind of circumstances. In any case, this topic deserves more investigation. Related to the need of more research in this area, it is necessary to improve the quality of the data about public finance at subnational level. The experience of working with the information currently available was certainly frustrating, and many of the original ideas and hypothesis simply could not be tested, due to the fact that the necessary data were nonexistent or unreliable. 88

100 Regarding to the implications of these findings, they can be grouped in three types: those related to the political and institutional environment, those related to the economic and social consequences of the political business cycle and the implications of this phenomenon for business administration and operations. Among the political and institutional consequences, there is the first place the awareness of the use of public expenditures as political instruments and the existence of opportunistic motivations in the handling of fiscal policy instruments by different political actors at different levels. This possibility reinforces the need to strengthen the system of checks and balances in the use of public resources at the three levels of government in Mexico: federal, state, and municipal. Among the economic implications of the political business cycle is the potential welfare loss implicit in the "deviation" of public resources to objectives different of the one supposed to serve, which is the maximization of social welfare. "Economic costs of elections far exceed their direct official costs. Elections may result in inefficient policies that incumbent politicians undertake to manipulate public opinion in order to increase chances of electoral victory. These policies lead to temporal improvements in the pre-electoral economic situation at expense of long run economic consequences" (Akhmedov et. al. 2003, p. 1). Among the implication of the political business cycle in business administration, one of the most important issues is to study more closely the effects of the political cycle in business activity, especially in industries more related to public expenditure such as the building industry, and agricultural equipment, to mention some. 89

101 In spite of the fact that the political cycle is one of the facts which are regularly predictable in an environment characterized by uncertainty and volatility, as is the case in emerging countries in general and in Mexico in particular, it is not regularly incorporated to corporate planning practices. To take into account the political cycle in scenario formation, as well as in corporate and financial planning, would add a valuable element in the capability of corporations to deal with an uncertain future. It is possible, for example, to build "sexenal" (six-year) indexes of economic activity, both at the national or subnational levels. That would help in the projections of sales, costs, and other important concepts for corporate planning and administration. The incorporation of political elements in financial and operative projections would also improve the accuracy of cash flow previsions and project evaluation, as well as the identification of realistic alternatives for real option analysis and valuation. These elements are important not only for firms operating at national level, but also for multinational firms. The incorporation of political elements to the planning process of multinational firms would allow them to identify more precisely the risk associated with the political cycles in the countries they operate, and to design strategies of hedging, among other things. Another implication for business, related with organizational theory is the study of the "curse of the first year" under an organizational approach. One of the explanations of this phenomenon has to do with the effects of the changes in the complete directive staff of an organization, in this case, the Mexican government. 90

102 A crucial difference of the change of the directive staff of the government, differently to what happens in a private corporation, is the fact that changes in government staff are distributed regularly in time, which make them a predictable event. This may have important consequences for the functioning of the government and related organizations, in aspects such as planning, human resources formation, capital formation, to mention some, and constitutes an interesting field for organizational theory research. It is possible to identify several lines for future research in the field of the political business cycles, in emerging countries in general, and in Mexico in particular. In the first place, it is possible to deepen the study of the political budget cycle at the state level dividing the sample in some significant sub-samples, according to diverse criteria. One of them is the party in power when elections take place. Given that, in the period of observation analyzed here it was already present the "alternancia" at state level; that is, the fact that governors of different parties were in charge by the time of elections, a split of the sample according to the parties in power in election time would allow to test the postulates of the partisan model, which states that governments with different ideology would follow different patterns of economic behavior. The purpose will be to test if there are significant differences in the behavior of government on different parties in the handling of public expenditures around the election dates. This kind of analysis at subnational level has been done in Russia (Ahkmedov et al, 2002, 2003) and in Portugal (Goncalvez and Veiga, 2004), finding evidence of partisan behavior of the different governments. 91

103 Another split of the sample could be done under the "Congress composition" criterion. This would imply to separate the sample between governments whose party has majority in local Congress and those who have not. Ex-ante, it will be expected that governments with majority in Congress would have higher discretionary power over the use of public expenditures (a higher y, in the notation of the model), and higher degree of possibilities to use this power in an opportunistic manner. This pattern has been identified in Colombia (Drazen and Eslava, 2004) and it would be interesting to test if this happens in Mexico. It is also possible to split the sample between the state governments that belong to the same party in power at the federal level and those which are not. In the Mexican case, the "alternancia" at the Federal level is a relatively recent phenomenon, which limits this kind of analysis at the present. However, it may be an interesting line of research in the future. Another line of research is to investigate if opportunism "pays"; that is, if an opportunistic behavior of a government increases its possibility of electoral victory of the party to which it belongs. This effect was studied in Russia, finding evidence that, in fact, opportunism does increase the probabilities of victory of the party in power (Ahkmedov et al, 2002, 2003). It is also possible to investigate the presence of a political cycle in public revenues at state level. Given the fact that most of the state revenues come from the federal government, it would be interesting to study if these revenues follow a pattern related to the political cycle. 92

104 Another research line is to investigate if there is some influence of the political cycle in the state economy itself; in variables such state GDP, sector GDP, employment, wages and other indicators of economic performance. Ahkmedov et. al. (2002) studied the relationship between political cycles in economic growth, industrial production income, and profits of enterprises, finding evidence of significant influence in industrial production, wages and monetary income of the population (p. 18) It is also possible to deepen the study of the political budget cycle at municipal level. This has been done in other countries, such as Portugal (Goncalvez and Veiga, 2004), finding evidence of opportunistic use of public expenditures in that level. In order to do that, it will be necessary to have a reliable database at that level. The Instituto Nacional de Estadistica, Geografia e Informatica (INEGI) has a database of municipal data, the SIMBAD (Sistema Municipal de Bases de Datos). However, it is necessary to verify if the information is reliable and complete to run this kind of analysis. Finally, the study of the political business cycles in emerging countries constitutes a very interesting and promising line of research that can provide valuable elements for a better functioning of the political decisions, public finance analysis, corporate planning and administration, among other fields. It is a relatively new field of study and offers a broad set of opportunities for future research. 93

105 REFERENCES Akhmedov, Akhmed, Alexei Ravichev and Ekaterina Zhuravskaya (2002). Regional Political Cycles in Russia. Center for Economic and Financial Research. Akhmedov, Akhmed, and Ekaterina Zhuravskaya (2003). Opportunistic Political Cycles: Test in a Young Democracy Setting. Center for Economic and Financial Research Alesina, Alberto, Nouriel Roubini y Gerald D. Cohen. (1999). Political Cycles and the Macroeconomy. Cambridge.The MIT Press. Assael, Paola y Felipe Larrain (1994). El Ciclo Politico-Economico: Teoria, Evidenciay Extension para una Economia Abierta. Cuadernos de Economia, 92, pp Banco de Mexico. Indicadores Economicos. Banco Mundial (2000). En el Umbral del siglo XXI: Informe sobre el Desarrollo Mundial 1999/2000. Washington. Brender, Adi and Allan Drazer (2003). Where does the Political Budget Cycles Really Come From? NBER. 94

106 Brender, Adi and Allan Drazer (2005). Political Budget Cycles in new versus established democracies. Journal of Monetary Economics. 52, pp Drazen, Allan (2000). The Political Business Cycle After 25 Years. University of Maryland. Drazen, Allan and Marcela Eslava (2003). The Political Business Cycle in Colombia on the National and Regional Level. Archivos de Economia. Documento 215. Ergun, Mine (2002). Electoral Political-Business Cycles in Emerging Markets: Evidence from Turkey. Russian and East European Finance and Trade. 36 (6) pp Frees, Edward W. (2004). Longitudinal and Panel Data: Analysis and Applications in the Social Sciences. Cambridge University Press. Gamez, Cesareo and Jaime Botello (1987). La Influencia del Ciclo Presidencial en la Economia Mexicana: Un ejercicio econometrico con variables dummy, in El Dilema de la Economia Mexicana: Ensayos de interpretation. Mexico. Universidad Autonoma Metropolitana, Ediciones de Cultura Popular. Gamez, Cesareo (2003a). La Maldicion del Primer Ano: Evidencia estadistica. Paper presented in the course of Analisis Estadistico. Programa de Doctorado en Administracion. EGADE. Monterrey. 95

107 Gamez, Cesareo (2003b). Politica y Finanzas en Mercados Emergentes. Paper presented in the course of Finanzas Corporativas. Programa de Doctorado en Administracion. EGADE. Monterrey Gamez, Cesareo (2004a). Elecciones, Economiay Finanzas en Mexico ( ). Paper presented in the course of Econometria. Programa de Doctorado en Administracion. EGADE. Monterrey. Gamez, Cesareo (2004b). El Ciclo Politico Oportunistay la Economia Mexicana ( ). Paper presented in the II Coloquio Predoctoral Latinoamericano. XXXIX Asamblea Anual de CLADEA. Puerto Plata, Dominican Republic. Gamez, Cesareo (2005a). Elecciones y Finanzas en Mexico ( ). Paper presented in the IX Congreso Anual de la Academia de Ciencias Administrativas AC (ACACIA). Merida Yucatan. Gamez, Cesareo (2005b). Elections, Interest Rates and the Exchange Rate in Emerging Countries: The case of Mexico. Paper accepted to be presented in the 2005 Annual Meeting of the European Financial Management Association. Milan. 96

108 Goncalvez Veiga, Linda, and Francisco Jose Veiga (2004). Political Business Cycles at the Municipal Level. Universidade do Minho. Gonzalez, Maria de los Angeles (2000). On Elections, Democracy and Macroeconomic Policy cycles: Evidence from Mexico. Princeton University. Gonzalez, Maria de los Angeles (2002). Do Changes in Democracy Affect the Political Budget Cycle? Evidence from Mexico. Review of Development Economics. 6 (2) pp Instituto Nacional de Estadistica, Geografia e Informatica (INEGI). Banco de Informacion Economica (BIE). Instituto Nacional de Estadistica, Geografia e Informatica (INEGI). Sistema Municipal de Bases de Datos (SIMBAD). Khemani, Stuti (2000). Political Cycles in a Developing Economy: Effects of Elections in the Indian States. The World Bank. Development Research Group. Larrain, Felipe y Paola Assael (1995). Cincuenta Anos de Ciclo Politico-Economico en Chile. Cuadernos de Economia, 96, pp

109 Larrain, Felipe and Paola Assael (1997). El Ciclo Politico Economico en Chile en el Ultimo Medio Siglo. Estudios Piiblicos. (Primavera), pp Magaloni, Beatriz (2000). Institutions, Political Opportunism and Macroeconomic Cycles: Mexico Stanford University. Medina, Leandro Manuel (2003). Evidencia del Ciclo Electoral-Presupuestario en las Provincias Argentinas. Universidad del CEMA. Ogura, Laudo Massaharu (2000). Political Business Cycles in Brazilian Economy ( ). University of Brazilia. Department of Economics. Rogoff, Kenneth, and Anne Siebert (1988). Elections and Macroeconomic Policy Cycles. The Review of Economic Studies. 55 (1), pp Rogoff, Kenneth (1990). Equilibrium Political Budget Cycles. The Review of Economic Studies. 89, (l),pp Santa Clara, Pedro and Rossen Valkanov (2003). The Presidential Puzzle: Political Cycles and the Stock Market. The Journal of Finance, 58 (5) ). 98

110 Shi, Min and Jakob Svensson (2002). Political Budget Cycles in Developed and Developing Countries. Institute for International Economic Studies. Stockholm University. Shi, Min and Jakob Svensson (2003). Political Budget Cycles: A Review of Recent Developments. Institute for International Economic Studies. Stockholm University. Shi, Min and Jakob Svensson (2006). Political Budget Cycles: Do they differ across countries andwhy?. Journal of Public Economics. 90, pp Schuknecht, Ludger (1996). Political Business Cycles and Fiscal Policies in Developing Countries. Kyklos, 49 (2), pp ). Schuknecht, Ludger (2000). Fiscal Policy Cycles and Public Expenditure in Developing Countries. Public Choice. 102, pp Sibley, Mike (2001). The Impact of Presidential Elections on Currency Values in Latin America. Multinational Business Review, 9 (2). pp Stein, Ernesto H. And Jorge M. Streb (2004). Elections and the Timing of Devaluations. Journal of International Economics. 63, pp

111 Studenmund, A.H. (2001). Using Econometrics: A Practical Guide. Addison Wesley Longman. 100

112 APPENDIX Descriptive Statistics 101

113 l.-total expenditure (GASTOT) YEAR MEAN STDEV SUM OF THE STATES -7.8% 7.9% 29.4% 19.7% 23.9% -8.9% -3.7% 9.5% 15.8% 8.9% 11.4% 7.7% 4.1% 8.6% 4.9% 8.8% AGS. 12.3% 20.8% 6.3% 40.0% 15.8% -10.1% -1.9% 11.0% 28.9% 6.3% 15.0% 7.6% 2.3% 13.0% 3.8% 11.4% B.C. 5.4% 12.8% 29.8% -0.2% 9.2% -25.2% 10.9% 15.7% 15.0% 5.0% 15.2% -39.8% 3.8% 6.9% 8.2% 4.8% B.C.S. 13.6% 6.2% 56.9% 28.1% 10.8% -9.4% 4.3% 9.9% 26.1% -3.0% 21.1% 7.6% 4.8% 8.3% 11.5% 13.1% CAMP. 28.1% 34.1% 43.3% 39.9% 29.9% -20.3% -8.4% 13.7% 18.9% 10.3% 7.6% 2.1% -1.4% 12.1% 3.8% 14.2% YEAR MEAN STDEV COAH. 9.2% 0.9% 8.5% 84.9% 37.4% -5.6% -6.3% 7.4% 18.8% 3.8% 12.3% 10.0% 8.6% 1.2% 7.9% 13.3% COL. 19.9% -4.8% 12.1% 36.4% 28.8% -2.2% 23.6% -2.6% 11.1% 1.9% 21.1% 0.8% 4.0% 12.4% 1.0% 10.9% CHIS. 16.8%. 6.5% 12.6% 6.7% 119.7% -17.3% -6.2% 7.4% 24.4% 11.5% 12.4% 5.3% 6.2% 6.5% 10.8% 14.9% CHIH. 10.5% 6.2% 71.2% -17.3% 48.7% 6.3% 5.0% -10.7% 22.8% 7.4% 15.8% 8.9% 9.5% 10.3% 2.6% 13.2% DGO. 18.1% 10.1% 13.9% 11.8% 8.3% -15.9% 86.2% 8.1% 26.7% 11.3% 14.3% 11.6% 0.4% 12.3% -3.0% 14.3%

114 YEAR MEAN STDEV GTO. 19.3% 17.9% 12.5% 2.7% 77.5% -5.7% -1.9% 18.0% 21.5% 9.1% 14.3% 19.2% -8.5% 7.6% 5.4% 13.9% GRO. -8.8% 11.9% 99.2% 19.4% 8.2% -56.5% 100.8% 12.2% 32.6% 8.8% 9.0% 12.1% -0.5% 13.4% 1.1% 17.5% HGO. 7.5% 22.1% 33.5% 73.2% 18.6% -4.0% 1.1% 14.5% 31.1% -0.6% 11.6% 12.9% -2.4% 20.3% -2.9% 15.8% JAL. 13.8% -43.1% 189.8% 7.0% 16.0% -16.3% -61.8% 53.9% 21.4% 9.1% 19.7% 7.7% 3.2% 4.0% 5.4% 15.3% MEX. -1.4% 31.3% 22.9% 42.2% 9.9% 0.6% 3.4% 3.6% -4.0% 13.5% 13.8% 11.5% 16.2% 3.0% 3.6% 11.3% YEAR MEAN STDEV MICH. 17.5% 14.1% 22.0% 2.7% 116.6% -9.2% -1.5% 19.4% 22.9% 6.5% 18.6% 8.9% 4.1% 10.6% 1.3% 17.0% MOR. 20.0% 11.8% 3.5% 27.2% -2.8% 14.6% 1.4% 72.0% 0.2% 4.1% 12.1% 14.2% -3.7% 11.7% -1.6% 12.3% NAY % 5.6% 8.5% 90.0% 2.8% -4.2% -1.4% 26.8% 19.0% 3.4% 15.8% 10.0% 3.7% 1.9% 5.9% 11.7% N.L. 13.8% 4.8% 10.1% 22.3% 2.0% -14.9% -10.3% 27.0% -2.6% 7.4% -18.5% 6.9% 3.9% 2.6% 1.7% 3.7% OAX % 11.5% 12.2% -7.6% 27.5% 54.3% 9.4% -54.8% 13.2% 18.5% 21.3% 11.4% 5.4% 7.7% 9.1% 4.4%

115 YEAR MEAN STDEV PUE. 1.2% 22.9% 30.9% 19.3% 52.7% -2.6% -9.2% 34.1% 14.7% 14.5% 17.0% 8.1% 2.4% 16.5% -4.6% 14.5% QRO. 24.6% 18.8% 16.7% 20.4% 121.8% -10.7% -11.0% 7.8% 9.2% 11.7% 9.5% 9.5% 7.2% 8.4% 2.5% 16.4% Q. ROO 27.3% 11.1% 0.7% 106.3% 10.8% -16.0% 11.9% 19.2% 40.3% 4.9% 6.1% 12.1% 3.4% 7.1% 16.6% 17.4% S.L.P. 12.7% 15.4% 51.7% 75.5% 8.3% -14.1% -2.8% 11.8% 20.5% 12.3% 17.4% 2.7% 2.9% 18.7% 11.8% 16.3% SIN. -2.8% 30.2% 21.4% 28.0% 10.3% -8.1% -12.9% 15.8% 14.5% 7.9% 13.9% 10.3% 0.3% 12.0% 3.5% 9.6% YEAR MEAN STDEV SON. 14.4% 14.6% 33.8% 22.7% -9.0% -8.2% -10.0% 15.7% 5.0% 11.4% 15.0% 9.3% 1.7% 9.2% 13.1% 9.2% TAB. 3.8% 18.0% -25.2% 26.6% 27.5% -14.7% 4.2% 11.7% 30.9% 0.6% 11.4% 6.2% 2.6% 10.3% 9.6% 8.2% TAMPS. 4.5% 7.9% 10.8% 60.0% 16.0% -13.4% 0.8% 15.9% 14.9% 14.5% 11.1% 8.4% -1.4% 11.0% 17.0% 11.9% TLAX. 6.3% 17.0% 5.4% -19.0% 10.3% -5.8% 1.4% 1.8% 141.0% -0.3% 19.0% 2.9% 6.3% 9.2% 0.7% 13.1% VER. 4.2% 15.5% 2.8% 8.5% 56.3% -15.3% 17.9% 10.3% 19.6% 10.6% 6.9% 10.3% 5.8% 8.4% 6.6% 11.2%

116 YEAR MEAN STDEV YUC. 8.3% 5.8% 44.1% -2.8% 8.4% -11.2% -3.3% 11.7% 11.3% -1.5% 17.8% 122.3% -5.0% 15.6% -1.2% 14.7% ZAC. 8.6% 30.2% -11.6% -2.1% 17.4% 56.4% 2.0% 0.8% 41.2% 17.0% 6.1% 8.3% 4.6% 10.4% 5.6% 13.0% Wages (SAL) YEAR MEAN STDEV SUM OF STATES 6.7% 17.9% 27.2% 9.2% 27.7% 5.2% 16.8% -10.8% 16.2% 5.0% 10.8% -0.5% 8.3% 8.8% 4.1% 10.2% AGS. 12.7% 19.1% 34.9% -6.7% 2.1% -6.0% -15.7% -14.1% 35.3% 20.6% 21.0% 14.1% 6.6% 1.9% 7.5% 8.9% B.C. 10.1% 14.8% 9.8% 5.1% 0.7% 1.2% -13.2% -53.2% 160.1% 4.6% 8.9% 12.9% 8.6% 6.1% 4.2% 12.1% B.C.S. -9.7% 14.9% 7.0% 4.3% -15.9% -19.2% 5.9% -80.5% 383.7% -4.5% 7.9% 12.3% 8.9% 14.7% -30.1% 20.0% CAMP. 27.8% 99.7% 1.7% 0.8% -8.4% 199.3% 5.6% -37.2% 87.5% 6.1% 8.1% 6.5% 3.6% 5.0% 4.4% 27.4%

117 YEAR MEAN STDEV COAH. 4.2% 7.2% 17.4% 10.5% 13.2% 191.0% 6.4% -36.2% 92.2% 6.5% 8.4% 9.6% 6.7% -11.5% 17.7% 22.9% COL. 16.7% 9.8% 23.8% 2.2% -3.5% -13.2% -5.5% -67.6% 182.7% 11.0% 20.6% 11.1% 7.7% 5.3% 7.2% 13.9% CHIS. -3.9% 4.7% 10.0% 15.6% 17.4% 61.3% 45.5% 7.8% 14.5% -1.5% 16.9% -57.0% 13.2% 3.6% 7.3% 10.4% CHIH. 7.8% 24.7% 22.0% 11.5% 101.1% -12.1% -3.5% -78.8% 73.1% 20.6% -5.5% 9.5% 3.7% 5.7% 0.7% 12.0% DGO. 17.3% 17.3% 27.4% -6.5% 4.8% -8.4% -5.7% 2.1% 5.7% 34.9% 8.1% 14.1% 9.7% 3.7% 2.5% 8.5% YEAR MEAN STDEV GTO. 11.1% 29.3% 21.0% 13.3% 15.6% 114.8% -5.2% 8.0% 6.6% 7.3% 10.5% -66.0% 16.7% 20.4% -7.3% 13.1% GRO. 18.0% 24.1% 1.9% 10.1% 1.4% -7.3% -13.1% -29.6% 78.2% -6.8% 24.6% 22.8% 4.6% 10.6% 15.1% 10.3% HGO. 5.1% 17.5% 7.7% 18.6% 7.6% -13.6% 12.5% -76.1% 335.7% -13.6% 28.9% 6.2% -34.1% 50.2% 1.7% 23.6% JAL. 5.7% 13.9% 82.3% 12.4% 12.6% -10.8% -34.7% 110.2% 12.0% 6.9% 8.4% 9.8% 4.3% 2.9% 5.6% 16.1% MEX. 17.8% 15.1% 20.9% 10.0% 15.5% -5.0% 79.1% 16.4% -44.2% 7.6% 7.2% 10.9% 11.2% 10.3% 2.8% 11.7%

118 YEAR MEAN STDEV MICH. 23.4% 12.7% 16.2% 8.2% 335.3% -9.3% 2.6% 13.8% 13.8% 3.9% 11.3% 2.1% 10.3% 5.6% 3.4% 30.2% MOR. 20.1% 18.4% 16.9% 18.1% 28.4% 9.1% -26.0% -41.3% 42.3% -6.7% -18.6% 0.9% 14.4% 9.4% -15.5% 4.6% NAY. 3.8% -5.9% 17.3% 19.5% 7.8% -12.3% -8.1% 15.2% 6.2% 1.1% 16.3% 13.4% -6.6% 16.9% 15.6% 6.7% N.L. -1.0% 14.8% 33.8% 17.3% 4.9% 53.5% -10.8% -84.5% 647.3% 1.9% 11.2% -0.1% -51.8% 130.2% 2.1% 51.3% OAX. 15.5% 32.2% 55.4% 26.8% -6.4% -17.3% -3.6% 3.1% -0.8% 20.4% 9.9% 2.9% 10.3% -2.3% -0.4% 9.7% YEAR MEAN STDEV PUE. 2.1% 15.1% 34.4% 23.4% -9.5% -13.7% 212.0% -38.8% 94.3% 3.0% 15.5% -0.2% 8.2% 6.8% 5.1% 23.8% QRO. 16.4% 43.9% 16.9% 27.0% 13.9% -2.3% -13.0% 7.0% -8.5% 4.3% 13.7% 1.6% 12.6% 13.3% 6.7% 10.2% Q. ROO 26.2% -38.5% 55.4% 24.8% 18.5% -28.6% 7.3% -74.4% 383.7% 37.1% -13.5% 18.7% 9.9% 3.4% 2.1% 28.8% S.L.P. 5.3% 16.9% 59.7% -11.0% -46.3% -6.4% 195.8% -41.3% 114.3% 6.6% 17.6% 1.7% 2.5% 13.7% 14.6% 22.9% SIN. 0.6% 10.1% 21.2% 3.4% -1.7% -7.3% -4.5% -39.4% 79.1% 0.9% 10.3% 9.0% 1.3% 5.7% -3.1% 5.7%

119 ^»* PS. TLAX. -9.0%,7.4% -3.9% -4 9%, % 5-5% 2L3% 31-6% " % " 8-3 % 8.6%,3 5% KOO/ % - 4,.7% 8. 7 o /o f ^ % 25.9% I2. 6% ^ J ^ 1QQC ^-n, 1V.J/O 3 /.3% ; " 8-9% -5-0% -1.0% -, 5.3% -H.7% 385.8% -7.6%, 7.0% 0 0% ' T - 248% " 3-0% -1-5% 3.0ol 1998 L3% "20.4% -, 7.9%. 38 o /o 5?0/ % -61-5% 32.6%.350/ 85o/ % o /o 252% 3-5 o/ 2001,3.6%,40.3% 34.,% 67y 0 ^ % 103.2% 7.8% 4 5% 6i % 6.3% 5.4% 470/ 30o/ % 2.4% 7. 6 o /o ^o ; MEAN 4.o% 36.3% 8.8% 5 5% " y ^65 8 -^ YEAR MEAN STDEV YUC. 6.9% 4.3% 28.6% 32.0% 6.0% -12.6% 0.7% -34.9% 63.7% 2.0% 10.5% 22.3% 149.1% 1.8% 1.1% 18.8% ZAC. 7.7% 68.9% -52.8% 109.8% 65.5% -33.2% -14.9% 53.1% -54.2% 254.0% -2.3% -54.4% 1.4% 3.6% 1.2% 23.6%

120 3.- Subsidies (SUB) YEAR MEAN STDEV SUM OF STATES -1.6% 12.2% 38.8% 123.7% 32.6% -38.8% 31.7% 57.3% 64.8% 8.3% 20.2% 17.7% 2.3% 4.7% 6.9% 7.9% AGS. 8.5% 48.4% 61.9% 639.7% 13.2% -5.0% 3.3% 3.0% 53.4% 5.1% 16.8% 12.2% -1.4% 21.2% 10.5% 10.6% BC 17.3% 6.6% -5.9% 94.6% 27.6% -7.4% -34.5% 0.2% 336.5% 3.4% 19.3% 14.7% -3.8% 8.8% 1.6% 5.3% BCS -16.1% 22.7% 349.3% 109.2% 22.8% -17.9% 13.7% 4.7% 34.0% 6.6% 26.9% 8.8% 0.0% -0.5% 18.6% 6.7% CAMP. 10.7% 62.7% 43.9% 175.6% 53.0% -64.5% 3.0% 18.4% 16.7% 17.0% 42.9% 8.8% 9.4% 10.3% -16.8% 2.9% YEAR MEAN STDEV COAH % 67.5% 5.6% 333.4% 13.5% -85.4% -32.1% 72.9% 5.3% 48.3% 24.4% 20.0% 13.3% 73.9% -33.6% 18.4% COL. 41.7% 43.4% 4.9% 101.8% 98.3% -15.0% 5.8% 23.9% 25.0% 6.3% 25.7% 4.0% 1.8% 15.0% -0.1% 5.2% CHIS % 55.6% 19.9% 78.5% 555.3% -98.3% 819.8% 59.1% 34.9% 46.6% 9.0% 179.5% 6.8% 6.8% 8.5% 50.4% CHIH % 4.3% 3.6% 9.5% -15.8% 5.4% -10.1% % 8.2% 11.6% 18.3% 6.9% -6.3% 3.0% 7.7% 2.8% DGO. -8.1% 758.2% -68.0% 161.2% 8.2% -13.8% 538.9% -0.1% 58.1% 3.6% 9.9% 13.5% 1.5% 4.9% -7.0% 3.2%

121 YEAR MEAN STDEV GTO. 4.1% % -87.1% 16.0% 19.7% 13.6% 517.5% 60.6% 15.2% 30.0% 18.7% 161.5% 0.8% 0.3% 11.0% 43.4% GRO % 31.4% 896.1% 53.2% 21.3% -96.2% 56.1% % 65.8% -3.1% 20.0% 16.5% 0.8% 8.3% 7.1% 8.2% HGO. 25.2% -21.4% 65.0% 684.5% 13.4% -6.3% 5.2% -1.9% 33.0% 12.1% 9.3% 20.6% -2.2% 8.3% 9.1% 8.9% JAL. 9.7% 0.7% -74.0% -8.4% 91.6% 73.6% -90.0% % 70.7% 14.5% 82.8% 6.9% 2.5% -0.3% 6.4% 3.9% MEX. 17.8% -10.2% 9.8% -53.2% 117.0% 76.3% -8.7% 5.2% 494.9% 18.9% 19.3% -0.7% 8.4% -0.5% 4.5% 2.9% YEAR MEAN STDEV MICH. 2.7% -49.7% 57.4% 98.8% -50.2% -12.2% 0.5% 108.9% 28.9% -14.7% 49.7% 23.3% 9.3% 38.3% -26.4% 11.1% MOR. -7.0% 19.2% 0.0% 12.2% -6.1% -19.4% 8.8% 882.8% 29.6% 13.8% 13 7% 3.6% 1.8% 8.6% 0.1% 3.5% NAY % 47.7% 312.1% -51.2% 14.1% -10.8% 374.0% 18.0% 20.9% 9.4% 3.8% 24.4% 12.4% 0.6% 7.1% 11.1% NL 3.9% 16.9% 43.8% 15.6% -15.2% -10.1% 261.1% 59.8% -42.7% 19.9% 9.6% 0.7% 90.3% -38.7% 19.8% 18.0% OAX % 41.2% 24.4% -30.0% 396.3% 395.6% 5.2% 5.5% 39.5% 0.8% 23.4% 8.7% 0.5% 11.6% 8.6% 7.3%

122 YEAR PUE. QRO. Q. ROO SLP SIN % -4.7% -3.6% -74.5% -59.4% % 15.5% 76.2% -12.4% 22.0% % 44.1% 14.9% 283.0% 6.6% % 18.3% 108.4% % 497.7% % 498.6% -21.5% 15.8% 12.9% % -12.8% 166.3% -16.2% -82.6% % 51.2% -16.2% -8.9% -14.3% % 21.4% 31.2% 10.5% 483.7% % 3.6% 51.4% 44.2% 50.6% % 6.2% 12.6% 7.6% 7.0% % 6.0% 9.9% 15.6% 20.9% % 14.2% 6.7% 2.8% 8.7% % 6.7% 6.9% 2.7% 2.3% % 0.3% 4.0% 18.8% 11.8% % 0.6% 12.4% 14.2% -2.0% MEAN 5.0% 5.4% 7.5% 9.6% 5.2% STDEV YEAR MEAN STDEV SON. -8.7% 7.7% 49.4% % 4.7% -60.7% -91.6% % 1.9% 10.5% 12.8% 10.1% 0.8% 12.1% 22.5% 11.3% TAB. 3.6% -86.6% 45.8% -30.6% 69.4% -14.0% 1.8% -3.2% % 16.1% 12.4% -0.1% -38.7% 10.3% -7.7% -9.0% TAMPS. 7.2% 68.9% -43.3% 418.3% 14.7% -15.4% 1.0% 34.1% -9.9% 24.8% 11.2% 6.7% -0.3% 6.1% 8.0% 5.1% TLAX. 35.8% -56.6% 949.8% -89.1% 958.5% -29.9% -35.7% 32.3% 581.3% -6.7% 24.9% 10.6% -0.7% 9.4% 2.6% 5.5% VER. -8.5% -6.5% 50.1% -17.2% 47.9% -27.0% 132.0% 158.6% 44.9% -20.6% 32.0% 0.4% -6.1% -6.3% 58.7% 11.7%

123 YEAR MEAN STDEV YUC % 178.0% 283.6% -38.6% 1.9% -29.2% -6.1% 10.6% 18.7% 2.4% 14.5% 509.9% -59.2% 26.4% 2.6% 119.9% ZAC. 3.0% 4.7% 44.7% -24.0% -14.0% 15.0% 255.9% -7.5% 37.3% -7.2% 13.3% 22.9% 11.3% 6.1% 12.8% 13.3% Public works (OB) YEAR MEAN STDEV SUM OF STATES 6.0% 17.0% 7.1% -8.2% 27.2% -41.4% -14.1% -4.0% 21.3% 9.5% 17.0% 4.9% 2.0% 9.6% 2.6% 3.8% AGS. 33.7% 44.8% 1.8% -37.4% 37.2% -38.0% -14.3% -53.1% 86.9% 27.2% 44.7% -5.8% 25.2% -8.8% -48.1% 6.4% B.C % 33.3% 13.7% -0.9% -5.9% -30.0% -45.3% 22.4% 17.2% 49.0% 66.0% -49.1% 27.3% -8.9% 59.5% 8.7% B.C.S % 87.1% 235.1% -74.6% 82.1% -70.9% -9.3% -4.0% 124.9% -56.7% 203.7% 13.9% 89.2% 78.9% -26.1% 43.5% CAMP 53.6% -26.7% 92.4% -24.2% -7.4% -91.6% 282.2% 17.2% 219.8% 19.7% 66.8% -19.7% -18.3% -5.6% 6.5% 37.6%

124 YEAR MEAN STDEV YEAR MEAN STDEV COAH. 9.4% -42.4% 4.2% 17.9% 396.8% -38.7% -18.1% -68.2% 356.9% -19.8% -2.3% 29.5% 20.7% -30.6% 91.9% 47.1% GTO. 44.3% -21.4% 33.9% -2.8% 19.2% -33.1% -66.3% 59.4% 18.1% -46.5% 13.6% 61.4% -32.8% 28.4% 26.6% 6.8% COL. -9.7% 5.4% -5.5% 55.2% -32.8% 37.1% -34.6% 2.2% -54.3% 8.8% 159.8% -0.8% -11.3% 20.7% -19.2% 8.1% GRO. 28.3% -0.3% -7.2% -61.3% 147.3% 27.0% -40.7% 34.7% 16.2% 21.6% -0.4% -16.0% 9.7% -6.8% 22.8% 11.6% CHIS. 79.5% 7.8% 16.0% -16.1% 87.9% -77.9% 44.5% 37.7% 63.6% -12.6% -39.7% -30.8% 35.0% 47.3% 24.2% 17.8% HGO. 59.2% -39.6% 43.6% 13.8% 35.6% -10.4% 13.8% -23.9% 27.3% 46.7% 19.3% -22.0% 17.6% 45.3% -15.6% 14.0% CHIH % 25.5% 304.7% -62.9% -12.8% -35.7% 1.8% -10.3% 84.1% 32.1% 42.9% 29.1% -18.5% 39.8% -12.0% 25.5% JAL. 16.4% 23.6% -46.0% 45.3% 42.3% -68.4% 2.7% 44.5% 49.0% -2.2% -37.3% -36.9% 11.0% 79.8% 10.6% 9.0% DGO. 10.3% 31.1% -28.9% 78.2% -9.4% -74.0% 65.9% -42.4% 103.1% 37.8% 67.4% -8.7% -30.7% 106.9% -26.2% 18.7% MEX. -8.8% -5.9% 27.3% 9.8% 24.8% -17.8% -16.9% -30.4% -30.8% 24.4% 22.0% 49.4% -4.1% 8.0% -28.0% 1.5%

125 YEAR MEAN STDEV MICH. 21.2% 40.5% 38.8% 17.4% 18.7% -22.0% -29.6% 4.2% 15.1% 17.5% 34.6% 24.2% -7.7% -15.6% 85.8% 16.2% MOR. 33.4% -0.9% -20.5% 22.9% -17.1% 51.5% -44.7% 5.7% -8.1% -45.5% 73.6% 60.4% -30.0% 47.7% 28.5% 10.4% NAY % -10.8% -74.9% -12.9% 76.8% -56.6% -28.0% 612.1% 206.6% -25.3% 5.5% 19.5% -43.9% 0.5% -9.9% 42.5% N.L % 82.5% -22.2% 42.1% 14.7% -48.6% -50.8% -68.5% 62.0% 12.2% 134.6% 46.3% 97.2% -15.1% -37.0% 14.5% OAX. 9.7% 2.2% 5.1% 0.9% -2.4% -21.1% -6.8% 62.8% -32.7% -20.4% 45.6% 253.2% 20.0% -8.3% 6.4% 21.0% YEAR MEAN STDEV PUE. 4.6% 19.2% 130.8% -41.3% 32.8% -18.6% 4.2% -14.0% -18.1% -19.7% 51.8% 25.8% -38.3% 136.8% -61.9% 12.9% QRO. 48.3% 100.2% 5.6% 39.6% 97.4% -53.1% -15.6% 14.9% 0.2% -18.2% -4.6% 46.2% 7.9% 42.8% -45.2% 17.8% Q. ROO 40.3% 26.3% -2.1% 54.0% -55.3% -50.0% 221.3% -7.8% 65.9% -38.5% 80.9% 19.8% 4.1% 17.8% 3.9% 25.4% S.L.P % 41.7% 64.1% -37.9% 84.6% -25.2% -34.5% -42.4% -51.0% 119.3% 86.4% -33.4% -13.5% -38.4% 158.3% 17.8% SIN. -7.8% 92.5% 24.9% -41.5% -19.3% -73.2% 11.2% -43.1% 205.0% 77.5% -23.6% -0.3% -1.3% -52.9% 224.6% 24.8%

126 YEAR MEAN STDEV SON % 17.8% 16.4% 77.0% -37.9% -3.5% -49.9% 50.3% -36.1% 44.6% 28.7% 2.3% -8.7% 21.6% 32.7% 6.4% TAB. 11.7% 20.6% -33.5% 3.5% 88.2% -93.3% -10.7% -12.4% 78.9% 417.5% 27.1% -73.0% -77.4% -9.5% -12.4% 21.7% TAMPS % 19.1% 20.2% -18.1% 8.1% -14.5% 9.4% 7.7% 114.7% -20.2% -12.5% -12.1% 7.8% 8.1% 98.9% 13.2% TLAX. 3.1% 38.3% -54.1% 28.6% -66.6% -53.3% 195.3% -23.1% -47.0% 37.1% 14.2% -18.1% 153.9% 19.0% 18.0% 16.3% VER. 81.8% 8.8% -27.5% -20.3% 16.8% -34.6% 22.4% 28.0% -37.3% 5.4% 18.4% 10.6% 25.3% -36.0% -48.2% 0.9% YEAR MEAN STDEV YUC % -15.4% -10.7% 54.3% -13.0% -59.9% 138.0% 5.4% 12.6% 15.4% 48.0% 3.0% -34.1% 132.7% -37.4% 14.8% ZAC. -5.7% 75.5% -38.7% -30.2% 44.0% -66.5% 80.3% -36.7% 127.2% 4.4% -16.4% 8.7% 5.2% 10.9% 12.2% 11.6%

127 5.- Other expenses (OT) YEAR MEAN STDEV SUM OF STATES -18.4% -1.4% 39.5% 15.2% 17.6% 6.5% -21.7% 8.8% -14.8% 13.0% 0.9% 2.7% 3.8% 13.9% 3.6% 4.6% AGS. -0.8% -1.7% -6.6% -0.3% 9.3% 6.9% 0.9% 68.1% -10.0% -1.0% -2.3% 0.8% -1.1% 9.0% 10.5% 5.4% B.C. 5.3% 11.3% 40.0% -7.3% 9.9% -33.4% 32.5% 32.3% -22.3% 4.0% 12.1% -78.4% 11.8% 7.3% 17.9% 2.9% B.C.S. 50.0% -6.2% 10.1% 1.5% 2.6% 24.5% -10.1% 63.3% -5.8% -16.9% 8.2% 2.6% 10.4% 19.6% 18.3% 11.5% CAMP. 22.0% 40.1% 47.1% 37.5% 32.9% -15.0% -22.8% 48.1% -12.2% 8.6% -19.8% 3.3% -7.9% 26.1% 18.8% 13.8% YEAR MEAN STDEV CO AH. -8.0% -9.6% 4.3% 0.8% 19.6% 73.1% -15.0% 116.2% -44.9% -1.9% 22.1% -0.3% 4.4% 4.5% -0.6% 11.0% COL. 29.2% -34.0% 15.6% 18.5% 11.5% 19.8% 89.8% -6.9% -11.5% -8.3% 0.1% -10.8% 11.9% 8.0% 5.5% 9.2% CHIS. 21.7% 0.7% 10.8% 0.8% 73.3% 92.8% -51.6% -14.1% 25.5% 32.0% 25.3% 12.1% -1.2% 2.2% 13.5% 16.3% CHIH. 55.7% -24.4% 25.9% 21.1% 3.4% 95.6% 20.2% -41.4% 15.4% -19.5% 14.1% -2.2% 92.8% 10.5% 4.5% 18.1% DGO. 24.7% -54.0% 137.2% -15.7% 25.6% 5.9% 5.5% 41.3% -16.5% 11.6% 18.5% 12.0% -1.3% 20.9% 10.0% 15.0%

128 YEAR MEAN STDEV GTO. 6.4% -23.6% 81.7% -5.7% 223.0% -56.6% -8.4% 8.4% 65.9% 9.7% 17.0% 30.8% -30.1% 13.5% 0.1% 22.1% GRO % 10.9% 58.6% 9.2% -41.5% 24.5% 282.5% -45.5% -12.1% 37.7% -9.8% 8.9% -8.3% 34.5% -21.5% 19.0% HGO % 94.5% 32.6% -46.6% 33.8% 11.0% -19.1% 137.0% 14.6% -29.6% 9.6% 13.7% -2.4% 36.7% -23.7% 16.3% JAL. 15.5% -65.8% 417.4% 4.1% 14.5% -14.6% -70.2% -11.3% 10.0% 11.3% 6.7% 11.0% 2.0% 4.5% 3.6% 22.6% MEX % 115.0% 24.7% 99.7% -0.3% 4.5% -28.9% -3.0% -18.5% 11.1% 11.9% 20.3% 35.4% -0.4% 10.2% 17.4% YEAR MEAN STDEV MICH. 28.9% 74.1% 5.5% -62.3% 202.7% 3.8% 2.4% 11.8% 53.6% 21.9% 16.9% 10.2% -6.2% 13.4% -4.9% 24.8% MOR. 17.8% 20.5% 20.8% 41.3% -11.4% 7.1% 60.8% 25.8% -36.7% 4.4% 8.9% 30.7% -9.3% 9.7% -12.5% 11.9% NAY % 10.2% -37.7% 365.5% -1.6% 0.9% -59.1% 29.9% -11.0% 7.6% 47.3% -17.1% 6.9% -0.6% 1.8% 21.9% N.L. 25.6% -4.9% 8.7% 21.7% 1.1% -25.9% -31.4% 90.6% -20.5% 6.1% -47.1% 13.0% -18.7% 5.5% -1.4% 1.5% OAX % 10.0% 8.5% -11.6% 26.0% 33.3% 12.5% -82.7% -25.9% 96.1% 18.7% -5.4% 10.1% 9.0% 14.0% 2.4%

129 YEAR MEAN STDEV PUE. -0.5% 32.7% -24.1% 118.4% 100.1% 4.3% -72.5% 300.5% -46.1% 56.2% -2.2% 16.2% -1.7% 18.7% 1.9% 33.5% QRO. 28.2% -21.3% 19.5% 0.0% 94.1% 38.4% -57.8% -36.3% 75.4% 59.4% 21.3% -8.6% 5.7% 5.8% 40.5% 17.6% Q. ROO 36.0% 10.8% -20.8% 181.7% 46.6% -61.1% 41.5% 39.8% 5.8% -1.8% -7.8% 18.5% -5.3% 10.8% 35.2% 22.0% S.L.P. 81.6% 5.9% 28.1% 12.6% -16.2% -0.6% -12.2% 81.0% -24.3% 18.0% 8.8% 14.8% 6.6% 30.0% -4.9% 15.3% SIN. 88.9% 7.9% 23.0% -12.3% 45.5% 156.2% -16.4% -19.2% -43.6% -2.7% 16.2% 18.9% -4.9% 35.2% -1.0% 19.4% YEAR MEAN STDEV SON % -0.9% -24.9% -1.5% -23.5% 235.8% 42.5% -64.3% 34.6% 6.9% 18.8% 6.9% -0.5% 17.2% 6.4% 63.0% TAB. 1.5% 146.6% -28.6% 97.4% -6.7% 68.7% -42.1% 52.2% -20.3% -19.2% -5.4% 26.8% 3.9% 16.5% 32.2% 21.6% TAMPS. 23.1% -40.7% 100.2% -20.3% 31.7% -12.7% -0.3% -17.0% 43.5% 36.0% 28.8% 15.1% -13.6% 29.7% -5.1% 13.2% TLAX. 12.8% 2.6% 20.4% -14.3% -11.1% 32.7% -14.2% 2.9% 108.2% 9.8% 15.2% -9.4% 7.9% 8.8% -5.2% 11.1% VER % 15.4% 22.7% -48.8% 108.9% 2.7% 37.1% -36.5% 74.3% 53.5% -17.5% 11.2% 8.5% 45.2% -9.8% 15.6%

130 YEAR MEAN STDEV YUC. 43.0% -5.9% -8.7% -5.4% 27.7% 18.8% -20.4% 67.4% -14.7% -10.6% 16.6% 59.8% 3.7% 10.0% 4.9% 12.4% ZAC. 78.8% -27.2% 8.9% 48.0% -9.0% 589.0% -90.8% 58.0% 229.3% 30.9% 4.3% 19.6% -12.4% 27.3% -15.4% 62.6% State GDP (PIB) YEAR MEAN STDEV NATIONAL 5.2% 4.2% 3.5% 1.9% 4.4% -6.2% 5.2% 6.8% 5.0% 3.8% 6.6% 0.0% 0.8% 1.4% 4.2% 3.1% AGS. 4.9% 4.7% 2.9% -3.1% 7.5% -3.1% 9.4% 8.1% 6.1% 3.6% 12.7% 3.5% 3.3% 2.3% 3.4% 4.4% B.C. 5.7% 4.6% 2.3% -2.9% 7.0% -5.6% 7.9% 12.3% 4.5% 8.2% 10.8% -3.2% -2.2% 2.7% 8.9% 4.1% B.C.S. 2.7% 2.3% 1.3% -1.9% 3.1% -0.6% 8.9% 4.1% 0.7% 3.4% 7.0% 3.8% 0.6% 6.1% 4.0% 3.0% CAMP. 1.5% 1.2% 0.7% 0.7% 3.5% -3.5% 4.9% 3.0% 2.4% -2.4% 8.1% 4.7% 1.1% 7.7% 2.2% 2.4%

131 YEAR MEAN STDEV YEAR MEAN STDEV COAH. 3.8% 3.4% 1.8% -3.7% 3.4% -0.6% 11.2% 8.4% 5.7% 3.1% 4.5% -0.5% 5.7% 3.8% 6.1% 3.7% GTO. 3.3% 3.1% 1.6% -1.7% 4.8% -3.7% 9.1% 5.9% 6.6% 1.6% 7.4% 0.9% 4.9% 2.8% 6.0% 3.5% COL. 2.7% 2.5% 1.4% -3.6% 5.0% -3.8% 8.6% 3.4% 5.4% 6.0% 2.0% -3.5% 2.7% -0.3% 4.4% 2.2% GRO. 1.3% 1.1% 0.6% 0.4% 3.8% -4.7% 1.2% 1.6% 3.9% 3.1% 2.7% 0.6% -1.0% 1.0% 2.8% 1.2% CHIS. 2.0% 1.9% 0.9% -1.5% 4.1% -0.3% 1.0% 4.6% 4.7% 2.6% 4.0% 1.4% 3.0% 1.7% 4.9% 2.3% HGO. 2.6% 2.2% 1.2% 3.0% 2.1% -11.6% 8.9% 5.4% 7.8% 2.4% 4.2% -2.0% -0.5% 0.4% 5.4% 2.1% CHIH. 5.0% 4.1% 2.4% -0.9% 5.8% -6.4% 7.9% 7.8% 7.8% 6.6% 11.4% -3.7% 0.2% 4.0% 4.1% 3.7% JAL. 3.0% 2.8% 1.5% 1.8% 3.5% -7.9% 4.5% 6.2% 7.7% 4.2% 5.5% 0.4% 0.3% -0.3% 5.0% 2.6% DGO. 2.1% 1.7% 0.6% -0.6% 4.7% -3.8% 5.0% 2.6% 8.3% -0.2% 2.0% 3.8% 1.5% 6.1% 6.5% 2.7% MEX. 3.5% 3.1% 1.8% 0.6% 3.9% -8.9% 8.2% 8.8% 4.7% 3.7% 6.9% 1.1% -1.0% 0.1% 4.1% 2.7%

132 YEAR MEAN STDEV YEAR MEAN STDEV MICH. 2.8% 2.2% 1.1% -4.6% 6.3% -2.0% 3.3% 10.4% 0.1% 6.6% 1.2% -1.3% -0.7% 3.0% 4.8% 2.2% PUE. 4.6% 3.8% 2.6% -0.1% 4.1% -8.2% 8.9% 8.7% 7.7% 8.4% 4.3% 1.1% -1.2% 3.0% 0.5% 3.2% MOR. 2.5% 2.1% 1.2% 5.8% 1.5% -9.7% 3.6% 4.9% 6.7% 5.0% 4.9% 3.5% -0.9% 4.2% 3.1% 2.6% QRO. 5.6% 5.1% 2.9% -5.5% 8.1% -3.0% 8.6% 12.2% 9.0% 4.8% 7.2% 0.1% 2.8% 0.4% 5.4% 4.2% NAY. 1.5% 1.6% 0.7% 4.5% 2.0% -9.9% 2.6% 1.3% 7.1% 3.8% 2.1% 3.4% -2.2% -4.1% 6.1% 1.4% Q. ROO 3.5% 3.0% 1.2% -0.4% 4.7% -4.9% 6.9% 10.8% 6.6% -0.7% 5.6% 5.7% 1.1% 5.6% 9.0% 3.9% N.L. 3.9% 3.5% 1.8% 0.6% 5.5% -6.5% 4.9% 9.1% 7.1% 5.6% 7.6% -0.1% 3.2% 2.9% 6.5% 3.7% S.L.P. 2.7% 2.2% 0.9% 1.7% 7.1% -11.1% 6.1% 6.2% 6.2% 2.8% 6.1% 0.1% 0.8% 4.7% 8.2% 3.0% OAX. 1.5% 1.3% 0.7% 0.9% 3.3% -4.4% 2.0% 0.5% 4.0% 3.3% 4.8% 1.3% -1.2% 0.7% 3.0% 1.4% SIN. 1.5% 1.4% 0.5% 2.7% 0.2% -2.8% 1.7% 2.6% 2.5% 0.4% 8.3% 1.9% -1.2% 0.5% 6.2% 1.8%

133 YEAR MEAN STDEV YEAR MEAN STDEV SON. 3.8% 3.0% 1.5% -3.7% 6.8% -2.0% 3.4% 6.5% 6.2% 4.1% 7.3% 0.8% -4.5% 2.8% 7.2% 2.9% YUC. 3.3% 3.0% 1.6% 0.1% 6.1% -6.1% 4.8% 5.6% 5.4% 4.7% 8.5% 2.4% -0.1% 2.4% 6.0% 3.2% TAB. 1.6% 1.3% 0.7% -1.9% 3.4% -0.4% 0.8% 4.3% 0.3% 2.4% 4.9% 0.6% -1.8% 1.3% 3.8% 1.4% ZAC. 1.0% 1.4% 0.7% 1.3% 1.4% 1.0% -0.1% 1.3% 9.2% -2.0% 3.8% 3.0% 9.2% 3.3% 2.5% 2.5% TAMPS. 3.8% 3.1% 1.4% -0.9% 7.5% -5.6% 5.7% 5.6% 8.0% 6.2% 7.5% -2.5% 3.6% 5.7% 7.9% 3.8% TLAX. 3.9% 3.3% 1.7% -2.6% 4.6% -3.6% 8.6% 8.7% 2.8% 4.3% 6.7% 2.8% -2.2% 2.3% 6.7% 3.2% VER. 1.3% 1.1% 0.4% -2.5% 5.6% -2.2% 1.6% 3.3% 2.2% 0.3% 4.0% -0.4% 0.2% 2.1% 5.0% 1.4%

134