UC San Diego UC San Diego Electronic Theses and Dissertations

Title Citizens' evaluation of the president and democratic transition : determinants and effects of presidential approval in Mexico

Permalink https://escholarship.org/uc/item/6cq2d3ks

Author Gomez Vilchis, Ricardo Roman

Publication Date 2010

Peer reviewed|Thesis/dissertation

eScholarship.org Powered by the California Digital Library University of California

UNIVERSITY OF CALIFORNIA SAN DIEGO

Citizens’ Evaluation of the President and Democratic Transition: Determinants and Effects of Presidential Approval in Mexico

A Dissertation submitted in partial satisfaction of the Requirements for the degree Doctor of Philosophy

Political Science

Ricardo Roman Gomez Vilchis

Committee in charge:

Professor Samuel H. Kernell, Chair Professor Wayne A. Cornelius Professor Thaddeus B. Kousser Professor David R. Mares Professor Carlos H. Waisman Professor Christopher M. Woodruff

2010

Ricardo Roman Gomez Vilchis, 2010

The Dissertation of Ricardo Roman Gomez Vilchis is approved, and it is acceptable in quality and form for publication on microfilm:

Chair University of California, San Diego

2010

iii

DEDICATIO

To my family: my mother, Irene Vilchis; my sister, Carmelita; my three nieces, Camelis, Marisol, Karina; my nephew, David; Martita, the family’s nanny; and the memories of my late father Ricardo Gómez Robledo, and my late grand mother, Carmen Témbul, who made it all possible.

To the National Autonomous University of Mexico (UNAM) for teaching me to develop critical thinking, a feeling of dissatisfaction for injustice, and a rebellious spirit.

EPIGRAPH

I prefer to die standing than to live kneeling.

Ernesto Che Guevara

When the richness of some people is extreme, it is the reason for the poverty of many.

Ricardo R. Gómez Vilchis

TABLE OF COTETS

Signature Page……………………………………………………………………iii

Dedication………………………………………………………………………...iv

Epigraph……………………………………………………………...……………v

Table of Contents ………………………………………………………………..vi

List of Figures…………………………………………………………………...viii

List of Tables……………………………………………………………………...x

Acknowledgements………………………………………………………………xii

Vita…………………………………………………………………………….xviii

Abstract of the Dissertation……………………………………………..………xix

Introduction...... …....1 Works Cited ……………………………………………………………...….11

1. Presidential Approval at the Individual Level in Mexico………………..…..13 1.1. Introduction ……………………………………………………….…....13 1.2. The Literature Review…………………………………………………..18 1.3. The Context……………………………………………………………..24 1.4. Hypotheses……………………………………………………….……...37 1.5. Data and Method………………………………………………….……..39 1.6. Results……………………………………………………………….…..43 1.7. Conclusions………………………………………………………….…..72 1.8. Appendix 1……………………………………………………….……..74 1.9. Works Cited…………………………………………………….……….77 1.10. Web Sites………………………………………………………....……84

2. StateLevel Presidential Approval in Mexico……………….………….……85 2.1. Introduction………………………………………………………….….85 2.2. The New Political Geography of State Politics in Mexico…….…….….88 2.3. Socioeconomic Conditions across the States of Mexico………………100 2.4. The Variation of Presidential Approval across the States……………..106

2.5. Hypotheses……………………………………………………………..113 2.6. Data and Method……………………………………………………….115 2.7. Results……………………………………………...…………………..125 2.8. Conclusion……………………………………………...……………...136 2.9. Appendix 2.1 ……………………………………………………...... …137 2.10. Appendix 2.2…………………………………………………....….…145 2.11. Works Cited……………………………………………………….….153 2.12. Web Sites……………………………………………………….…….155

3. Presidential Approval in Mexico: MultiLevel Analysis……...... 156 3.1. Introduction…………………………………………………………….156 3.2. Findings and Limitations of the TimeSeries Analysis………...……....159 3.3. The Literature Review and Rationale………………………………….163 3.4. Hypotheses………………………………………………………….….174 3.5. Data and Method…………………………………………………….…175 3.6. Results……………………………………………………………….…183 3.7. Conclusion………………………….………………..…….…………..204 3.8. Appendix 3……………………………………………………………..207 3.9. Works Cited……………………………………………………………211 3.10. Web Sites …………………………………………………………….212

4. Presidential Approval and Roll Call Voting in Mexico…………………….213 4.1. Introduction………………………………………………………..…...213 4.2. The Review of the Literature…………………………………………..219 4.3. The ExecutiveLegislative Relationships in Mexico………………...... 228 4.4. Hypotheses……………………………………………………………..233 4.5 Data and Method………………………………………………………..235 4.6 Results…………………………………………………………………..241 4.7. Conclusion………………………………...…………………………...255 4.8. Works Cited …………………………………………………………...258 4.9. Web Sites ………………………………………………………...……261

Conclusion……………………………………………………………………...262 Works Cited……………………………………………….……………….271

vii

LIST OF FIGURES

Chapter 1: Figure 1.1: The Advancement of the Opposition in the Chamber of Deputies….27

Figure 1.2: Leading Obstacles to Achieving Democracy in Mexico (1998)…….29

Figure 1.3: Perceptions of the Main Task of Democracy in Mexico (1998)…….30

Figure 1.4: Reported LocalJurisdiction Crime for Mexico, 19912001 (per 100,000 Inhabitants) …………………………………………………………….30

Figure 1.5: Presidential Election Results in 1994 and in 2000…………………..32

Figure 1.6: Why Mexican Voters Cast Their Ballots for President in 200 when Changes was the Main Reason to Vote ………………………………………….35

Figure 1.7: Changing Visions of Democracy in the 2000 Presidential Election...36

Figure 1.8: Levels of Trust in the Mexican Presidents during Their First Year of Tenure……………………………………………………………………………37

Chapter 2: Figure 2.1: Political Parties across the States of Mexico 19942006……………89

Figure 2.2: State Governments of Mexico Controlled by the Three Main Political Parties: PAN, PRD, and PRI (2006)………………………………………..……91

Figure 2.3: Citizens’ Perceptions of the Political Actors’ Responsibility for the Economic Crisis……………………………………………………………….…99

Figure 2.4: Level of Unemployment across the States of Mexico in 2005…..…101

Figure 2.5: Variation of Unemployment in Four States of Mexico 19942005...101

Figure 2.6: Level of Inflation in Mexico across the States in 2005…………....102

Figure 2.7: Variations of Perceptions of Crime as the Main Concern for Mexican from 1995 to 2006…………………………………………………………..…..103

viii

Figure 2.8: Convicted Criminals by State in Mexico (2005 )……………….….104

Figure 2.9: Variations of Crime in Four States of Mexico 19942005…………105

Figure 2.10: Presidential Approval by State in Mexico 2005…………………..106

Figure 2.11: Presidential Approval (2005) by Party Lenses (States Governed by PAN, PRD and PRI) …………………………………………………………...112

Chapter 3: Figure 3.1: Changing Visions of Democracy in 2000………………………….170

Figure 3.2: Citizens’ Trust in Three Mexican Presidents after Their First Year of Tenure………………………………………………………………………..…171

Figure 3.3: Convicted Criminals by State in Mexico (2005)……………….…..172

Figure 3.4: Complaints against the Civil Service by State in Mexico (2005).…173

Chapter 4: Figure 4.1: Presidential Approval in Mexico before and after the 2000 Democratic Transition……………………………………………………………………….225

Figure 4.2: Presidential Approval vs. Presidential Success in the House before and after the 2000 Democratic Transition…………………………………………..226

Figure 4.3: Composition in the Chamber of Deputies by Partisanship…………227

LIST OF TABLES

Chapter 1: Table 1.1: Determinants of Presidential Approval before and after the Transition in Mexico (19942006)………………………………………………………..…44

Table 1.2: Determinants of Presidential Approval, PreDemocratic Era and Democratic Period, Compared…………………………………………….……..51

Table 1.3: Changes in Probabilities of the Determinants of Presidential Approval in Mexico…………………………………………………………………...……57

Table 1.4: Effects of Perceptions of Crime, Corruption, Inflation, and Unemployment on Presidential Approval in Mexico (19942006), Using the 2000 Election Presidential Election as a Breaking Point, Logit Models………..…..…61

Table 1.5: Effects of Perceptions of Crime, Corruption, Inflation, and Unemployment on Presidential Approval in Mexico (19942006), Using the 1997 Intermediate Election as a Breaking Point, Logit Models……………………….65

Table 1.6: Effects of Perceptions of Crime, Corruption, Inflation, and Unemployment on Presidential Approval in Mexico (19942006), Using the 2003 Intermediate Election as a Breaking Point, Logit Models…………………….…69

Chapter 2: Table 2.1: Percentage of Presidential Approval across the States 2005………..109

Table 2.2: Explaining StateLevel Presidential Approval in Mexico: Variables……………………………………………………………………..…121

Table 2.3: Effects of People’s Perceptions of Crime, Unemployment, and Inflation across the States and SubNational Conditions on StateLevel Presidential Approval (19942005) TimeSeriesCrossSectional Models.……130

Table 2.4: InteractiveTest Model…………….…………………………..…….133

Table 2.5: Comparing the Distribution of the Population of Mexico (0105) with the Distribution of the Surveys during the PostPRI Era (0105)………………137

Table 2.6: Effects of Perceptions Aggregated at the State level and SubNational Conditions on StateLevel Presidential Approval (19942005), Including

Perceptions of Corruption and Complaints against Member of the Civil Service ………………………………………………………………………………….145

Table 2.7: Effects of Changes in Perceptions Aggregated at the State Level, and Changes in SubNational Conditions on StateLevel Presidential Approval (1994 2005), TimeSeriesCrossSectional Models……...……………………………149

Chapter 3: Table 3.1: Explaining Presidential Approval (MultiLevel): Variables…..……179

Table 3.2: Multilevel Models Using StateYears Random Effects (Random Intercepts)………………………………………………………………………188

Table 3.3: Multilevel Models Using StateYears Random Effects (Random Intercepts) Results Expressed as Marginal Effects…………………..…………193

Table 3.4 Multilevel Models Using StateYears Random Effects (Random Intercepts) Only Including States with the Most Stable Levels of Crime…...…199

Table 3.5: Determinants of Presidential Approval (9405). Multilevel Models Using StateYear Random Effects (Random Intercepts), Results Expressed as Marginal Effects, Excluding Corruption Variables……………….……………207

Chapter 4: Table 4.1: Explaining Roll Call Voting in Mexico: Variables…………………240

Table 4.2: Effects of Presidential Approval on Roll Call Voting in Mexico (9806) Multilevel Models Using Deputies Random Effects (Random Intercepts)…….246

Table 4.3: Effects of Presidential Approval on Roll Call Voting in Mexico (9806) Multilevel Models Using Deputies Random Effects (Random Intercepts), Results Expressed as Marginal Effects…………………………………………...……..252

ACKOWLEDGEMETS

(From the Grassroots of Bella Vista to La Jolla )

After studying for almost six years at UC San Diego, more than a thousand words cross my mind. I cannot lie… that is why I am not a politician. I am a sincere and true man, which is my way at the end. My heart is divided… and

I will start with the easiest part of the story.

First, I thank the members of my committee. My advisor and chair, Sam

Kernell has offered continuous support and encouragement every step of the way.

Wayne Cornelius, a foremost authority on Mexican politics, has been very patient with me. I do not hesitate to say that my dissertation was improved by Wayne’s insights and comments. David Mares has been my friend and, with his suggestions and critiques, worked as if he were a cochair of the entire dissertation. I am greatly indebted to him. His great support and friendship are two of the best memories of my intellectual life at UC San Diego. Thad Kousser made econometrics accessible by encouraging me to represent my problems in quantitative terms, and by enthusiastically seeing my work through. Nobody explains econometrics in as easy and friendly a way as Thad can. During my work at the Center for U.SMexican Studies as a research assistant, Chris Woodruff has been one of the nicest people that I have met at UCSD. His comments and

xii

suggestions from an econometrics perspective allowed me to conduct better and more accurate models. Carlos Waisman encouraged me to extend my argument in a comparative perspective, focusing my attention on Latin American democracies.

Happiness and sadness were parts of my life at UCSD. I would like to acknowledge some of my classmates who helped me a lot as a graduate student:

Craig Burnett, Matt Chielders, Christian Donath, Jeremy Horowitz, Ryan

Jablonski, Emily Matthews, Jennifer Piscopo, and Samuel Seljan. I also thank

Professor Zoltan Hajnal, whom I consider the best Professor in political behavior at UCSD. For those who made my life difficult, those who neither “saw” my raised hand nor “listened” to my questions in class, those who treated me differently than my classmates, I have only one word… thanks. I feel stronger.

I also would like to acknowledge my debt to two of my classmates, who really became my friends, Jacob Allen and Adam Brown. Jacob has been the nicest person that I have met in the U.S. His wonderful family, Linn, Jacob’s wife, and his four children, Campbell, Annie, Matthew, and Benjamin adopted me as a member of the Allen family. Jacob helped me a lot, and his patience and friendship were determinant in my academic life in San Diego. Adam was always ready to answer my questions. He explained me with clarity when I needed his help.

At UCSD, I had fantastic moments with people who offered me their friendship. Ricardo Ramírez, my Tocayo, the “BibleGuy”, has been an example

xiii

about faith and patience. I also had the opportunity to meet Pedro Camacho, a workmate in the Department of Literature, his support and friendship, were determinant in my most difficult time at UCSD. I will always be grateful to him.

It will be impossible for me to forget the “WineTestingParty Nights” with my friend Ricardo Fagoaga, my other Tocayo , in his department at Mesa Housing.

There, I had the opportunity to temporarily forget all the pressure that I felt as a

PhD student. At Ricardo’s place, I had the chance to meet great friends: Bárbara,

Jesús, Pavito, Zack, Jorge, Pepe Lugo, among others. Raquel, my tennistable mate, has been a wonderful and incredible friend during my last year as a PhD student. Antonieta Mercado, who studied at the National Autonomous University of Mexico, UNAM, was a true friend. We share our love for the UNAM and a feeling of dissatisfaction for social injustice. I will always remember Kayla, who gave me one of my best and happiest memories in the U.S…

I was very lucky to take classes at UCSD Extension to improve my

English. There, I met the best teachers and nicest people of UCSD. Bob

McKinney, David Fein and Kim Lawgali helped me a lot, and answered all my questions and doubts. I am going to miss them. Their classes made my life easy and enjoyable at UCSD.

I feel special admiration for all the Mexican people whom I met at UCSD.

Most of them work at the Price Center in the fast food: José Vidal, Luis, David,

Doña Maky, Francisco, Ricardo, Félix, Marcela, Cathy, “Wicho”, among others.

I consider all of them my buddies. They make this country, the U.S., a better

xiv

place. Without their work, this country would face many challenges. I am really thankful to Germán Vázquez, the Manager of El Tacone, a Pumafan, like me, and an excellent friend. I will always remember Doña Mary, who helped me a lot and prayed for me when I had difficult times at UCSD. They are people like me.

The system was unfair with them. They came from lowincome families, from common neighborhoods, from the grassroots. Any place of a common neighborhood, which you may call Bella Vista, where I come from.

Several institutions funded portions of this work. The resources provided by The National Council on Science and Technology (CONACYT), and UC

MEXUS allowed me to study at UCSD for five years. The Dirección General de

Relaciones Internacionales de la Secretaría de Educación Pública (DGRISEP) also supplied support for three years during my PhD in political Science. UC

MEXUS Foundation and UCSD’s Center for Iberian and Latin American Studies offered critical dissertation grants which helped me to assess the project’s feasibility and to visit Mexico several times. The Center for U.S.Mexican Studies awarded me with a visiting research fellowship for the 20082009 academic year.

At the Center, I had the opportunity to receive helpful insights and friendship from Alberto Díaz Cayeros, Graciela Platero, James Samstad, Mauricio Benítez,

Luz Marina Airas, and Gabirela Torres. I consider all of them my friends.

I am also thankful for having worked as a teaching assistant (TA) in the

Language Program of the Linguistics Department at UCSD. As a Spanish TA, I had the opportunity to meet students from different backgrounds. Moreover,

working as a TA during my last year at UCSD allowed me to stay in the U.S. without problems of financial support. I sincerely appreciate the support and help from Alicia Muñoz, Academic Coordinator of Spanish, Grant Goodall, Director of the Language Program, and Vicki King, Linguistics Language Program

Assistant Director, who trusted in my teachingskills as a TA.

I am indebted to the people who helped me to collect the data for this dissertation. Most of the data of Presidential approval in Mexico was facilitated by Banco de Infomación para la Investigación Sociales (BIIACS). The survey data of presidential approval in 2006 was provided by Francisco Abundis,

Director of Parametria. Alejandro Moreno facilitated some of the figures of presidential approval and citizens’ perceptions of crime, corruption, unemployment, and inflation that came from Diario Reforma. David A. Shirk,

Director of Transborder Institute at the University of San Diego, provided the data about the levels of crime in Mexico. The data of corruption in Mexico came from Secretaría de la Función Pública. I am also thankful to Khemvirg Puente and Franciso Cantú, their comments and pieces of advice were very helpful to write the first drafts of the fourth chapter.

I also would like to acknowledge my debt to Ángel Gerardo Trejo, his friendship and support were determinants in most of the difficult moments that I had as a PhD student.

None of my achievements would have been possible without the support of my family. The love of my mother, Irene Vilchis; my sister, Carmelita; my

xvi

three nieces, Camelis, Marisol, Karina; my nephew, David; Martita, the family’s nanny; and the memories of my late father, Ricardo Gómez Robledo, and my late grand mother, Carmen Témbul have provided the strength and energy to undertake most of what I have accomplished and thanks to their caring and support I have come thus far. This dissertation and the last six years of my life are dedicated to them.

xvii

VITA

1996 Bachelor of Arts, Journalism and Communication, National Autonomous University of Mexico (UNAM)

1996 Gabino Barreda Medal (award UNAM)

2001 Master of Arts, Political and Social Studies (UNAM)

2002 De la Violencia a la egociación: el EZL y el Gobierno Federal, Master’s thesis. UNAM

20012004 Lecturer in Political Science at the UNAM

20042005 Research Assistant at the Center for U.S.Mexican Studies

20062010 Spanish TA Linguistics Department at UCSD

2009 Associate in Political Science, UCSD

2010 Doctor of Philosophy, University of California, San Diego

2010 “Riesgo y Subpolítica: la Estrategia del EZLN desde la Perspectiva de Ulrich Beck” (forthcoming in Etnicidad y Conflicto en las Américas. Activismo Político Vol. II.)

xviii

ABSTRACT OF THE DISSERTATIO

Citizens’ Evaluation of the President and Democratic Transition: Determinants and Effects of Presidential Approval in Mexico

Ricardo Roman Gomez Vilchis

Doctor of Philosophy in Political Science

University of California, San Diego, 2010

Professor Samuel H. Kernell, Chair

xix

Political institutions affect presidential approval, and the political regime works as the market when people evaluate the president. What determines presidential approval in democratic systems does not necessarily explain approval in authoritarian regimes. I use Mexico as a case study in comparative politics to analyze the determinants and effects of presidential approval before and after a democratic transition.

The analysis follows four steps. First, I examine the determinants of presidential approval at the individual level in Mexico, paying special attention to citizens’ perceptions of rising crime and corruption as key determinants of voters’ evaluation of the president after the democratic transition. Second, I extend my argument by analyzing the determinants of presidential approval as the sub national level. Third, I conduct a multilevel analysis in order to show that after the 2000 presidential election, the turning point of the democratic transition, perceptions of crime and corruption strongly affected voters’ evaluation of the president. And fourth, I examine the effects of presidential approval on roll call voting in Mexico before and after the transition.

I show that since the 2000 democratic transition, perceptions of rising crime and corruption have become more salient, and citizens have begun to evaluate the president in terms of his determination to address these two problems. Moreover, I find that while presidential approval had positive effects on roll call voting before the democratic period, after the 2000 presidential

election, the effects of presidential approval were positive for members of the president’s party, but negative for opposition deputies. In the new Mexican democracy, opposition deputies considered that the best strategy to advance in their political careers was to embarrass and frustrate the president’s plans in

Congress.

xxi

ITRODUCTIO

Does the political regime affect citizens’ evaluation of the president? Are voters more sensitive to political issues when they approve of the president in an emerging democracy than in a semiauthoritarian system? Can the political arena work as the market by either favoring citizens’ expectations of the president or discouraging voters’ support for the president? What determines presidential approval in democratic systems does not necessarily explain approval in authoritarian regimes. Political institutions affect presidential approval, and the political regime works as the market when people evaluate the president. From the demand side, citizens have expectations about the government, and from the supply side, the president searches for support by addressing people’s main concerns.

After more than 70 years of the Institutional Revolutionary Party (PRI) in office, Mexico experienced a democratic transition. The political arena turned into a different and new politicalmarket because of changing citizens’ perceptions of the president. Because of this process of democratization, the influence of public opinion of the president altered Mexican politics in two ways: the determinants of presidential approval were affected after the 2000 transition

1 2

and public opinion of the president changed the relationship between the executive and the legislative powers.

Although this dissertation relies on empirical evidence from survey data for the case of Mexico, the analysis is broadly comparative. As a case study

Mexico provides an excellent opportunity to improve our understanding of presidential approval in comparative politics. First, the Mexican case permits an exploration of how citizens evaluate the president during a semiauthoritarian regime. Since authoritarian governments attempt to manage the flow of political information to the public, these governments rarely permit the conduct of survey research. However, the availability of survey data gathered during a semi authoritarian period allows me to analyze presidential approval during two distinct periods: the last years of the PRI and the democratic period. The second reason to study Mexico is that after the transition to democracy, the new government has been in office for some years. This condition of political stability provides an opportunity to examine citizens’ perceptions of the president in a mediumlength democratic period. Finally, the 2000 democratic transition itself and the data that I have collected during the past five years allow me the use of democratization as the treatment of a quasiexperimental design, which tests the determinants of presidential approval before and during the democratic era in

Mexico.

Providing insights into how public opinion of the president affected

Mexican politics after the 2000 transition, this dissertation is guided by one

research question: how has the influence of public opinion of the president in the last years affected Mexican politics? My general hypothesis is that Mexican politics has changed in the last decade because of the influence of public opinion.

First, Mexican citizens have evaluated the president differently since the 2000 presidential election, the turning point of the democratic transition. Crime and corruption have become more salient and citizens have begun to evaluate the president in terms of his determination to address these two problems. Second, public opinion of the president has influenced the executivelegislative relationships differently since the 2000 transition. Since Vicente Fox’s ascension to power in 2000, opposition deputies have used embarrassing the president in

Congress when he enjoys high levels of approval ratings as a main political strategy to advance in their political careers. The higher the popularity of the president becomes, the more likely that the presidential bills will be rejected in

Congress.

My general argument is that the 2000 presidential election worked as a critical election affecting political behavior. First, because of the victory of a non

Priista presidential candidate after 70 years, the 2000 election altered citizens’ perceptions of the president. More precisely, the 2000 democratic transition is a determinant political event in Mexican politics that modified how citizens viewed the president. Mexican people began to have more expectations of a greater competence of the next Mexican president in managing not only economic but also political issues (Magaloni and Poiré, 2004). Second, after the transition,

public opinion of the president had different effects on opposition deputies’ political behavior. During the PRI regime, the opposition supported presidential bills when the president enjoyed high approval rating. However, during the democratic period opposition deputies considered the best strategy to advance in their political careers was to embarrass and frustrate the president’s plans in

Congress. Therefore, the higher the approval of the new president became, the more incentives that opposition legislators had to make the president’s life difficult in the Chamber of Deputies.

Conceptualizing Crime and Corruption in the Mexican Political Spectrum

As in other emerging democracies, crime and corruption have become salient in Mexico as relevant issues for an eventual consolidation of the democratic system. Empirical data indicate that some of these new democracies have failed to reduce these problems. In Latin America, most of these democratically elected governments have done a poor job of addressing crime and citizen security. El Salvador, Guatemala, and Colombia have among the highest murder rates in the world today, and crime is a serious problem in virtually every major Latin American city (Hagopian and Mainwaring 2005). In the Mexican case, Cornelius and Shirk (2007) suggest that democratization coincides with a significant increase in crime rates in the last decades. Furthermore, according to

the Transparency International Corruption Score, some of these new Latin

American democracies are among the most corrupt nations (Nef 2001).

In the new democracies of Latin America, public tolerance for governmental lack of efficiency in addressing political problems has varied, which has affected the eventual consolidation of the democratic systems. Some analyses show that LatinAmerican citizens strongly associate political demands as the main duties of democratic regimes (Camp 2001; Klesner 2001). Moreno’s work (2001, 43) demonstrates that while in the U.S. Americans consider elections to be the main task of democracy, in Latin American the story is different.

Chileans and Costa Ricans regard addressing crime as the main task of a democratic system. In Mexico opinions are divided: 33 percent view elections as the main task of democracy, while 31 percent consider combating crime the main duty of a democratic government. Mexican people perceive crime as a central political issue for an eventual democratic consolidation, rather than regarding crime only a social problem.

In relation to corruption, Latin American citizens believe that one of the main challenges of the political spectrum is the lack of efficient anticorruption policies. Costa Ricans perceive corruption as the main obstacle for a democratic consolidation at a rate of 46 percent, Chileans at a rate of only 20 percent, and

Mexicans at a rate of 42 percent (Clark 2001, 87; Klesner 2001, 127). Moreover, in Mexico 76 percent states that almost all government officials are corrupt and accept bribes (Clark 2001, 87; Klesner 2001, 127). Stories of Mexican politicians

and their “vulnerability” to corruption explain why Mexican people strongly associate corruption with politics, and mainly with the PRI . In 1982, Arturo

Durazo Moreno, head of the Police Department and transit of Mexico City, was arrested and accused of illegal acquisition of nearly 30 million dollars. In May

1987, Jorge Diaz Serrano, former head of the Mexican state oil monopoly Pemex

(Petróleos Mexicanos ), was sentenced to 10 years in jail on charges of personally enriching himself during his tenure (Agustín 1990). Therefore, Mexican voters mainly consider corruption to be a salient political variable when they evaluate the president.

Contributions of the Dissertation

Few studies have analyzed presidential approval outside the U.S. Most of these analyses focus their attention on well consolidated democracies such as

France, Germany, and the U.K. (Powell and Whitten 1993; Norpoth 1994; Lafay

1994). Democratization and its effects on presidential approval have been ignored in these studies. No analysis explaining whether or not people change their expectations about the president’s responsiveness after a democratic transition exists in the literature. No study examines whether democratization affects issue salience. Thus, one of the main goals of this dissertation is to provide a consistent explanation of the conditions under which political rather than economic issues become more relevant for approval after democratization takes place.

From comparative politics, most scholars have analyzed presidential approval focusing their studies on one kind of political regime, either an authoritarian system (Geddes and Zaller 1989; Zhong et al. 1998; Weyland 2000) or an emerging democracy (Cuzan 1997; Duch 2001; Mishler and Willerton

2003). These analyses overlook central questions in comparative politics: Does a democratic transition affect how citizens evaluate a new government in office?

Does a change in the political regime affect the effects of presidential approval on roll call voting? I attempt to answer these questions using presidential approval data in Mexico.

One of the main theoretical contributions of this dissertation is to provide an analysis of the circumstances that can affect citizens’ expectations of the responsiveness of the government. In an emerging democracy, political issues become relevant. Voters believe that the new democratically elected president must be worried about people’s main concerns because keeping citizens satisfied by the implementation of the government’s policies is the best way to get political support. This is in contrast to an authoritarian regime that uses manipulation of election results, vote buying, fraud, and clientelism as main strategies to stay in office (Cornelius, 2004).

This analysis also attempts to offer empirical evidence about a central normative debate in political science involving the meaning of democracy. The findings of this dissertation reveal that democracy involves both elections and citizens’ main concerns. A competitive political struggle is one of the main

requirements of a democratic system (Schumpeter 1976). However, after a new president is democratically elected, citizens’ expectations from the government’s policies show that for voters democracy means both elections and the president’s effort to address citizens’ concerns.

Organization of the Dissertation

To support my general hypothesis that Mexican politics has changed in the last decade by the influence of public opinion, I do four empirical analyses divided into four chapters. In the first three chapters, the dissertation is focused on the determinants of presidential approval in Mexico before and after the democratic transition. In the last chapter, I examine the effects of presidential approval in Mexican politics during and after the PRI regime.

The first chapter provides an analysis of presidential approval in Mexico at the individual level before and after the transition. The eleven national surveys on which this study is based were organized by the Mexican Office of the

Presidency 1 (MOP) and by the private polling firm, Parametria . In total the surveys contain almost 18,000 household interviews, and 600 telephone interviews from November 1994 to November 2006. This period covers three

1 The name of the presidential pollster in Mexico. This office conducts all public opinion research for the Mexican President. Polls are founded by the federal budget through the Office of the President. Surveys are confidential for governmental use. However, Asesoría Técnica de la Presidencia de la República made them public after Salinas and Zedillo finished their tenures as presidents. Data of all surveys is now public and available at the Banco de Información para la Investigación Aplicada en Ciencias Sociales, BIIACS .

different presidents (Carlos Salinas de Gortari, Ernesto Zedillo, both Priista presidents, and the Panista Vicente Fox). A logistic analysis is conducted, allowing the measurement of the probability that voters approved of the president before and after the Mexican transition.

The second chapter examines presidential approval at the statelevel. The forty three surveys on which this study is based were organized by MOP. Thirty three of them are national and ten state surveys. In total the surveys contain

74,150 household interviews, from 1994 to 2005. I conduct a timeseriescross sectional analysis to analyze statelevel presidential approval in Mexico before and after the transition by aggregating individual survey data at the level of the states.

In the third chapter, I conduct a multilevel analysis of presidential approval in Mexico as a robust test of the first two chapters. The fourteen national surveys on which this study is based were also organized by MOP. In total the surveys contain more than 20,000 household interviews, from November 1994 to

November 2005.

The fourth chapter addresses the second part of my hypothesis, the effects of public opinion of the president on Mexican politics. I analyze the impact of public opinion of the president on the executivelegislative relationships before and after the transition. The surveys on which this study is based were organized by MOP. In total the surveys contain more than 20, 000 interviews, from 1998 to

2006, covering the legislatures 57 th , 58 th and 59 th . Most of the data come from

national facetoface household surveys and other from nationaltelephone surveys. To analyze roll call voting in Mexico, I created an original database with the roll call voting of more than 1, 500 Mexican deputes, and including more than

64,000 observations. I conducted a multilevel analysis by including aggregate and individual level variables.

The conclusion summarizes the main findings of my dissertation and lists some possibilities for future research.

Works Cited

Agustín, José. 1990. Tragicomedia Mexicana. Volumen II México: Planeta

Camp, Roderic. 2001. Citizen Views of Democracy in Latin America. The United States: University of Pittsburgh Press

Clark, Mary A. 2001. “Costa Rica: Portrait of an Established Democracy”. Citizen Views of Democracy in Latin America. Ed. Roderic Ai Camp. The United States: University of Pittsburgh Press. 7390

Cornelius, Wayne and David Shirk. 2007. Reforming the Administration of Justice in Mexico. San Diego, California: University of Notre Dame Press and Center for U.SMexican Studies.

, 2004. “Mobilized Voting in the 2000 Elections: the Changing Efficacy of Vote Buying and Coercion in Mexican Electoral Politics” Mexico’s Pivotal Democratic Election. Ed. Jorge I. Dominguez and Chappell Lawson: Stanford: Stanford University Press: 4766.

Cuzan, Alfred. 1997. “Presidential Popularity in Central America”. Political Research Quarterly , Vol. 50, No. 4”: 833849.

Duch, Raymond. 2001. “A Developmental Model of Heterogeneous Economic Voting in New Democracies. American Political Science Review, Vol. 95, No. 4: 895910.

Geddes, Barbara and John Zaller 1989. “Sources of Popular Support for Authoritarian Regimes”. American Journal of Political Science , Vol. 33, No. 2 (May): 319347

Hagopian Frances and Scott P. Mainwaring. 2005. The Third Wave of Democratization in Latin America. Advances and Setbacks. Cambridge, New York: Cambridge University Press.

Klesner, Joseph. 2001. “Legacies of Authoritarianism: Political Attitudes in Chile and Mexico”. Citizen Views of Democracy in Latin America. Ed. Roderic Ai Camp. The United States: University of Pittsburgh Press.

Lafay, JeanDominique. 1994. “Political Dyarchy and Popularity Functions: Lessons from the 1986 French Experience”. Economics and Politics. The Calculus of Support. Ed. Helmut Norpoth, Michael S. LewisBeck, and JeanDominique Lafay. The United States: The University of Michigan Press.

Magaloni, Beatriz and Alejandro Poiré. 2004. “The Issues, the Vote, and the Mandate for Change” Mexico’s Pivotal Democratic Election. Ed. Jorge I. Dominguez and Chappell Lawson: Stanford: Stanford University Press: 293320

Mishler, William and John P. Willerton. “The Dynamics of Presidential Popularity in PostCommunist Russia:. The Journal of Politics, Vol. 65 No. 1: 111141.

Nef, Jorge. 2001. “Government Corruption in Latin America”. Where Corruption Lives . The United States of America: Kumarian Press.

Norpoth, Helmut. 1994. “The Popularity of the Thatcher Government: A Matter of War and Economy”, Economics and Politics. The Calculus of Support. Ed. Helmut Norpoth, Michael S. LewisBeck, and JeanDominique Lafay. The United States: The University of Michigan Press.

Powell, Bingham and Guy D. Whitten. 1993. “A CrossNational Analysis of Economic Voting: Talking into Account of the Political Context”. American Journal of Political Science, Vol. 37, No. 2: 391414.

Weyland, Kurt. 2000. “A Paradox of Success? Determinants of Political Support for President Fujimori”. International Studies Quarterly, Vol. 44, No. 3 (September): 481502.

Zhong, Yang, Jie Chen, and John Scheb . 1998. “Mass Political Culture in Beijing”. Asian Survey, Vol. 38. No. 8: 763783.

1 PRESIDETIAL APPROVAL AT THE IDIVIDUAL LEVEL I MEXICO

1.1 Introduction

Political institutions affect presidential approval, and the political regime works as the market when people evaluate the president. From the demand side, citizens have expectations about the government, and from the supply side, the president searches for support by addressing people’s main concerns. What determines presidential approval in democratic systems does not necessarily explain approval in authoritarian regimes. Mexico and its democratic transition in

2000 allow me to analyze the determinants of presidential approval in two different political contexts: the semiauthoritarian in transition PRI regime

(Institutional Revolutionary Party) and the emerging democracy. This chapter shows that political behavior does not occur in a vacuum and that the political regime impacts presidential approval. Since the 2000 transition, corruption and crime 2 have become salient and have had stronger effects on citizens’ expectations of the president.

2 As this dissertation explains, I consider corruption and crime as political issues because empirical evidence shows that in emerging democracies citizens associate combating both crime and corruption as two of the main tasks of democracy.

13 14

This chapter analyzes presidential approval in Mexico before and after the

2000 presidential election, the turning point of the democratic transition. It covers three presidents: Carlos Salinas de Gortari (19881994), Ernesto Zedillo (1994

2000), and Vicente Fox (20002006). Unfortunately, it is not possible to include observations for every year. During the Salinas and Zedillo tenures the public’s perception of economic issues was not regularly surveyed. In the case of Fox, in

2005 questions about citizens’ perceptions of crime and corruption were not included in the surveys. Therefore, I include observations for one year of

Salinas’s 3 tenure (1994), two of Zedillo’s (1995 and 2000), and five of Fox’s

(2001, 2002, 2003, 2004, and 2006). I divide this study into two main periods.

The first is the predemocratic era in Mexico when the PRI was in office, which includes the tenures of Salinas and Zedillo. The second is the democratic term when Fox, a nonpriista president, ascended to power. Two research questions lead my analysis: How different are the determinants of presidential approval in two different institutional frameworks, the semiauthoritarian PRI regime, and the emerging Mexican democracy? Why are the determinants of presidential approval in Mexico different after the 2000 transition?

From American politics (Clarke and Stewart 1994; Kenski 1977;

MacKuen, Erikson, and Stimson 1992; Markus 1988; Thomas 2003) we know

3 In the case of Salinas, before 1994 questions about perceptions of unemployment and inflation, my control variables, were not included in the surveys. Thus, it was not possible to include observations before 1994.

that voters’ perceptions of the economy and economic issues such as unemployment and inflation affect presidential approval. Moreover, a popular president has more opportunities to bargain with Congress and to persuade citizens (CanesWrone 2004; 2006; Edwards 1989; 1980; Kernell 1997; Neustadt

1991; Pritchard 1983).

However, few studies have analyzed presidential approval outside the U.S.

(Arce 2003; Cuzán 1997; Powell and Whitten 1993; Weyland 2000; Zaller and

Geddes 1989; Zhong et al. 1998). Most of these studies focus their attention on well consolidated democracies (Powell and Whitten 1993; Norpoth 1994; Lafay

1994). Furthermore, analyses of presidential approval both in authoritarian systems (Weyland 2000; Geddes and Zaller 1989; Zhong et al. 1998) and in emerging democracies are rare (Cuzan 1997; Mishler and Willerton 2003; Arce

2003). Democratization and its effects on presidential approval are absent in these studies. No analysis explaining whether people change their expectations about the president’s responsiveness after a democratic transition exists in the literature.

We also lack consistent explanations of the conditions under political issues become relevant for approval rather than economic issues after democratization takes place.

As a case study, Mexico provides an opportunity to examine how citizens evaluate the president in two different political contexts, before and the after the

2000 democratic transition. The Mexican case may shed light on how differently people weigh both economic and political issues in two institutional frameworks.

Citizens’ support for the president may be a key determinant for the consolidation of a new democracy. A popular president may be able to implement the necessary policies to strengthen the emerging democratic regime because he is in a better position to bargain with congress.

2003). These analyses overlook central questions in comparative politics: Does a democratic transition affect how citizens evaluate a new government in office?

Does the emergence of a new democracy affect voters’ perceptions of the responsiveness of the political regime? Can democratization affect issue salience?

Are political issues more salient than economic issues after a democratic transition takes place? I answer these questions using presidential approval data collected before and after the 2000 democratic transition in Mexico, which include more than 18,000 observations.

The main proposition in this chapter is that after the 2000 democratic transition issues such as crime and corruption become relevant and have a stronger impact when citizens evaluate the president. Edwards (1995) states that

“for an issue to have a significant influence on evaluations of the president, it must be salient to people and people must evaluate the president in terms of his performance regarding”. Therefore, the popular salience of issues directly affects

the public’s evaluation of the president. In the case of Mexico, empirical evidence reveals that citizens strongly associate democracy with combating corruption and crime. Two years before the transition in 1998, 42 percent of the Mexican people believed that corruption had been the major obstacle to democracy in Mexico. In the same year, 31 percent considered combating the crime the main task of democracy by 31 percent (only 2 percent lower than electing governors) (Camp

2001). These findings suggest that, perhaps, during the PRI regime Mexicans focused their attention on economic rather than political issues. From the citizens’ perspective, the incumbent party did not consider anticrime policies as a central issue to staying in power. Moreover, they viewed the PRI government as a synonym of corruption itself. In contrast, after the 2000 presidential election, with the emergence of a new democratic system, voters have begun to strongly associate issues such as corruption and crime when they evaluate the president

(Camp 2001). The main theoretical implication of this chapter is that democratization can affect citizens’ expectations about the responsiveness of the government. In an emerging democracy, political issues become relevant. Voters consider that the new democratically elected president must be worried about people’s main concerns because keeping citizens satisfied by the implementation of the government’s policies is the best way to get political support, which is in contrast to an authoritarian regime that uses manipulation of election results, vote buying, fraud, and clientelism as main strategies to stay in office (Cornelius,

2004),

This chapter is divided as follows: The first section of this article provides background information on the determinants of presidential approval from the perspective of American politics, and a review of the comparative politics literature. The second section examines the context of Mexican politics before the

2000 transition. Third, I formulate my hypotheses, and explain the data and methodology. Fourth, I present a logistic analysis to test my main argument about the stronger effects of perceptions of crime and corruption on presidential approval after the 2000 transition. Finally, the conclusion summarizes the main findings of this analysis.

1.2 The Literature Review

In American politics the economy as the main determinant of presidential approval has been the focus of most scholars (Brodi 1991; Clarke and Stewart

1994; Fiorina 1981; Kenski 1977; Key 1966; Kiewiet 1983; Kinder and Kiewiet

1979; MacKuen, Erikson and Stimson 1992; Markus 1988; Rudolph, 2003).

Citizens’ evaluations of the president are very sensitive to the variation of real economic conditions (Brodi 1991; Fiorina 1981; Kenski 1977; Key 1966; Kiewiet

1983; Kinder and Kiewiet; 1979). Others argue that citizens’ perceptions of the economy and economic issues such as unemployment and inflation affect voters’ behavior when they approve the president (Clarke and Stewart 1994; MacKuen,

Erikson and Stimpson 1992; Markus 1988; Rudolph; 2003). Although many scholars advance interesting and significant insights, they focus their attention on

the impact of economic factors and overlook the political context. Some scholars have noticed this limitation in the study of American politics and have argued that how citizens perceive the president is affected not only by the economy but also by political variables. Cohen (2002), Cohen and Powell (2005) Kernell (1978;

1997), Ostrom and Simon (1985; 1989), state that foreign policies and going public activities have strong effects on presidential approval. Other analyses show that media coverage, citizens’ trust in the media, and political knowledge can work as the best predictors of presidential approval (Miller and Krosnick

2000; Nadeau et al. 1999). Moreover, political speeches (Ragsdale 1987), political scandals (Newman 2002), political conditions such as divided government

(Nicholson, Segura and Woods 2002), and the presidents’ attributes as a political leader (Mayer 2004; Thomas, Sigelman, and Baas 1984; Newman 2002) may explain the variation of presidential popularity. These studies reveal that both the economy and political variables should be considered as key determinants for presidential approval. The key question for this article is how these two kinds of perspectives, an economic explanation and a political view, can be applied to understanding presidential approval in contexts different than that of American politics.

The empirical evidence suggests that in wellconsolidated democracies, contexts similar to the U.S., citizens focus their attention on the economy when they approve of the president. Powell and Whitten (1993) find significant effects of unemployment and inflation on presidential approval in three stable

democracies: the U.S., Great Britain, and Germany. Cohen (2004) argues that in established democracies, Canada, France, Italy, Germany and the U.S., citizens use prospective assessments of the economy when they evaluate the president.

Norpoth’s analysis (1994) shows that both the economy and the Falklands War explain the popularity of the Thatcher’s government. Lafay (1994) states that citizens’ perceptions of the economy and economic issues such as unemployment and inflation had strong effects on the French Prime Minister’ popularity from

1978 to 1987. However, the story would not be the same in political contexts different than the U.S.

Inspired by the Third Wave of Democratization (Huntington, 1992), empirical investigations on the determinants of presidential approval in new democracies have started to appear only recently. Mishler and Willerton’s (2003) analysis shows that not only people’s perceptions of the state of the economy, but also authoritarian culture and values are the key determinants for the post communist Russian democracy. Duch (2001) argues that citizens’ levels of information about the democratic process in their own countries affect how citizens evaluate the president in two emerging democracies, Hungary and

Poland. Nerrebi and Klor (2007) show that the antiterrorist policies impact voters’ behavior when they evaluate the president in Israel.

Latin America scholars have found that both the economic and political variables matter because the political context affects people’s perceptions of the executive (Arce 2003; Gelineau 2007; GomezVilchis 2008; Geddes and Zaller

(1989). Arce’s analysis (2003) reveals that when Peruvians evaluate the president, they focus their attention on the government’s antiguerrilla policies. Gélineau

(2007) explains that in Argentina, Brazil, and Venezuela when citizens evaluate the president’s party their perceptions of the economy matter. However, those effects are restricted by the political context, when voters can easily identify the president’s responsibility. Geddes and Zaller (1989) find that those citizens who are the most exposed to media effects and who have the least political knowledge are most susceptible to government influence. GómezVilchis (2008) shows citizens’ perceptions of the responsiveness of the political regime affects presidential approval in Mexico.

Empirical data indicate that some of these new Latin American democracies have failed to reduce problems such as crime and corruption

(Hagopian and Mainwaring 2005; Camp 2001). First, most Latin American governments have done a poor job of addressing crime and citizen security. El

Salvador, Guatemala, and Colombia have among the highest murder rates in the world today, and crime is a serious problem in virtually every major Latin

American city (Hagopian and Mainwaring, 2005). In the Mexican case, Cornelius and Shirk (2007) suggest that democratization coincides with a significant increase in crime rates in the last decades. Second, according to the Transparency

International Corruption Score, some of these new Latin American democracies are among the most corrupt nations (Nef 2001).

Chileans and Costa Ricans regard addressing crime as the main task of a democratic system. Mexicans’ opinions are divided: 33 percent view elections as the main task of democracy, while 31 percent consider combating crime the main duty of a democratic government. In relation to corruption, 46 percent of Costa

Ricans perceive corruption as the main obstacle for a democratic consolidation, while in Chile only 20 percent; and in Mexico 42 percent (Clark 2001, 87;

Klesner 2001, 127). Moreover, Latin American people strongly associate politicians with corruption. 75 percent of Costa Ricans, 58 percent of Chileans, and 76 percent of Mexicans state that almost all government officials are corrupt and accept bribes (Clark 2001, 87; Klesner 2001, 127).

Although this chapter is broadly comparative, the analysis relies on empirical evidence from survey data for the case of Mexico. I chose to analyze this country for several reasons: First, the Mexican case permits an exploration of how citizens evaluate the president during a semiauthoritarian regime. Since authoritarian governments attempt to manage the flow of political information to

the public, these governments rarely permit the conduct of survey research.

However, the availability of survey data allows this study to analyze presidential approval during the last years of the PRI era: 1994, 1995 and 2000. The second reason to study the Mexican case is that after the transition to democracy, the new government has been in office for almost ten years. This condition of political stability provides an opportunity to examine citizens’ perceptions of the president in a mediumlong time democratic period. This study only covers the years: 2001,

2002, 2003, 2004 and 2006 because of the availability of the data. Finally, the

2000 democratic transition itself and the data that I have collected during the past five years allow me the use of democratization as the treatment of a quasi experimental design, which tests the determinants of presidential approval before and during the democratic era in Mexico.

Although corruption and crime have been two of the most salient issues in

Mexican politics, on the whole scholars have focused their studies on the economy (Buendía 1996; Domínguez and McCann 1995; Magaloni 2006;

Villarreal 1999). In one of the seminal studies of presidential approval in Mexico,

Davis and Langley (1995) reveal that Mexican citizens evaluate the president based on partisan cues and their perceptions of anticorruption policies. This important finding in presidential approval has been ignored in recent analysis because scholars have focused their attention on people’s perceptions on economic issues. These studies have significantly improved our understanding of presidential approval in Mexico, yet they ignore the effects of the political

variables on approval. Buendía (1996) and Magaloni (2006) argue that real variation of economic conditions affect presidential approval in Mexico.

Domínguez and McCann (1995) and Villarreal (1999) explain that citizens’ assessments of the economy and economic policies have strong effects on people’s political opinions. My understanding is that crime has neither been used as a key nor as a control variable. This analysis tests the impact of both crime and corruption, two relevant issues in Mexican politics and in other emerging democracies, on presidential approval before and after a democratic transition.

1.3 The Context

Political regimes can work as the market, creating commitments and expectations from both the demand and the supply sides (Przeworski 1991). No market is exactly equivalent to another one because each market has specific institutions and regulations that affect political actors’ behavior. An autocracy is one kind of market in which vote buying, clientelism, and fraud are part of the government’s supply. When these supplymechanisms do not work, an autocracy can eventually use repression to stay in power. In contrast, a democracy is another kind of market in which credible commitments are used to address citizens’ main concerns because the only way to stay in office is through competitive elections and political struggle (Schumpeter, 1976). Successful democratic governments are those that put together attractive policies and offer credible commitments for the electorate (Beetham, 1993).

In the case of Mexico, the semiauthoritarian PRI regime and the emerging

Mexican democracy created two different kinds of supplymechanisms. In the first case, the PRI stayed in office for more than 70 years using vote buying, fraud, clientelistic policies, and selective repression as main supplymechanisms.

The PRI government established close relationships with peasants, organized labor, and the military, the three major sectors of the Mexican society. The peasant sector endorsed PRI candidates for public office in exchange for distribution of land. Organized labor expressed their support to the PRI through mass demonstrations, campaign rallies, and voter registration drives in exchange for nonwage benefits such as subsidized food, clothing, housing, health care, and transportation. The Mexican military endorsed the PRI government in crisis situations through the use of armed repression of dissident groups, counterinsurgency campaigns against rural guerrillas, and the breaking of major labor strikes. In return for their loyalty to the PRI, the Mexican military received regular salary increases and a variety of generous nonwage benefits such as housing, medical care, loans, and subsidized consumer goods (Cornelius and

Craig 1991, 8595). These interchangerelationships of political support and the allocation of private benefits through clientelistic strategies allowed the PRI to stay in office for over 7 decades.

The PRI succeeded in attending to most of the citizens’ demands. Between

1940 and 1960 the Gross Domestic Product (GDP) more than tripled, and during the 1960s Mexico achieved a solid per capita growth rate averaging 3.3 percent

per year. It was this performance that came to be known as the “Mexican miracle”, an exemplary combination of economic progress and political stability within the developing world (Cornelius, Gentleman, Smith 1989, 4). However, the situation began to change in the early 1970s when production declined. Then, in the early 1980s, the collapse of the oil caused the economic crisis in Mexico to deepen dramatically. As a consequence of the economic problems and of citizens’ dissatisfaction with the government, the PRI regime had difficulty staying in office. The 1988 election concluded with a controversial result (Eisenstadt 2004).

A proPRI federal electoral commission headed by the interior secretary, the president’s most important cabinet secretary, charged with keeping internal order, declared that Carlos Salinas was the new president of Mexico in 1988. After that election, burned ballots appeared floating in rivers by the hundreds, and the mass media reported that the votetallying computers “mysteriously” crashed

(Eisenstadt 2004, 7). Manuel Bartlett, Minister of interior during the 1988 elections, admitted some years later that there had never been a computer system crash, as publicly claimed. Rather, he had shut down the vote tally on orders from

President de la Madrid, to ensure a Salinas victory. Because of the erosion of the

PRI, the opposition was able to advance in the political arena through electoral reforms 4 (1977, 1986, 1990, 1994, and 1997) (Woldenberg, Salazar and Becerra

2000), creating a supply and a demand for democracy in Mexico. The

4 Most of these reforms attempted to strengthen political competition through the implementation of mechanisms for proportional representation that eventually gave the opposition the opportunity to advance both in the Senate and the Chamber of Deputies.

advancement of the opposition favored the conditions for a different kind of market in the Mexican political arena, and citizens did not see the PRI as the best supplier for their demands. Figure 1.1 shows the advancement of the opposition in the Chamber of Deputies since 1976.

90 80 70 60 50 PRI 40 Opposition

Seat Shares Seat 30 20 10 0 1976 1979 1982 1985 1988 1991 1994 1997 2000 2003 Elections

Figure 1.1 : The Advancement of the Opposition in the Chamber of Deputies. Figure Done by the Author with Data of the Chamber of Deputies

The political context had created a supply and a demand for democracy in

Mexico before the 2000 presidential elections. From the supply side, in 1994, opposition parties jointly controlled 11 percent of Mexico’s thirtytwo gubernatorial posts, 40 percent of the seats in the national Chamber of Deputies, and 26 percent of the Senate positions. In June 2000, just weeks before the election, opposition parties controlled 34 percent of governorships, 52 percent of deputy positions, and 41 percent of Senate seats (Camp, 2004; Domínguez, 2004).

Even the PRI contributed to establishing a supply for democracy. In the 2000

presidential election, Priistas established an open primary system for nominating the 2000 PRI’s presidential candidate, breaking the tradition in which the outgoing president had the right to handpick the presidential candidate (Lawson,

2004a).

From the demand side, citizens started to associate democracy with political issues such as corruption and crime. The PRI government made its best effort to convince voters that the PRI was able to address these concerns.

However, the empirical evidence suggests that as in the market, citizens did not perceive the government’s efforts as credible commitments, and they viewed the opposition as a better possible supplier. In the case of corruption, Figure 1.2 shows that before the 2000 presidential election 42 percent of Mexican people strongly associated corruption as the main obstacle to democracy in Mexico.

Moreover, they considered that the PRI itself represented the main symbol of corruption. In 1998 during the PRI regime, 76 percent of Mexicans perceived that most government officials were corrupt and accepted bribes (Klesner 2001).

Stories of Mexican politicians and their “vulnerability” to corruption explain why

Mexican people strongly associate corruption with politics, and mainly with the

PRI . In 1982, Arturo Durazo Moreno, head of the Police Department and transit of Mexico City, was arrested and accused of illegal acquisition of nearly 30 million dollars. In May 1987, Jorge Diaz Serrano, former head of the Mexican state oil monopoly Pemex ( Petróleos Mexicanos ), was sentenced to 10 years in jail on charges of personally enriching himself during his tenure.

45 42 40 35 30 25 25 22 Obstacles to Democracy 20

Percentage 15 11 10 5 0 Corruption The Political Others Government Parties

Figure 1.2: Leading Obstacles to Achieving Democracy in Mexico (1998) Source: Hewlett Foundation/MORI Internacional, 1998 (Camp 2001)

In the case of crime, Figure 1.3 reveals that combating crime (31 percent) is almost as important as electing governors (33 percent) as the main task of democracy. The possible explanation for this concern is a significant increase in crime rate in the 1980s and 1990s. The number of crime reports filed nationwide has dramatically increased in the last two decades, as Figure 1.4 shows (Zepeda

2007, 134).

35 33 31 30

25 19 20 17 Tasks of Democracy 15 Percentage 10

0 Electing Combating Distributing Others Governors Crime Wealth

Figure 1.3: Perceptions of the Main Tasks of Democracy in Mexico (1998). Figure Done by the Author. Source: Hewlett Foundation/MORI Internacional, 1998 (Camp 2001)

1800 1600 1400 1200 1000 Reported Crimes 800 600

Reported Crimes Reported 400 200 0 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 Years

Figure 1.4: Reported LocalJurisdiction Crime for Mexico, 19912001 (per 100,000 Inhabitants). Data from Zepeda (2007).

Data indicating that citizens associated crime with the PRI do not exist.

However, the empirical evidence suggests that before the 2000 transition voters considered the incumbent party’s anticrime policies inefficient or at least that the

PRI government did not perceive crime as a central issue to address. Perceptions of crime in Mexico as the main task of the government increased sharply from 2 percent in 1995 to almost 20 percent in 2000 ( Banco de Información para la

Investigación Aplicada en Ciencias Sociales, BIIACS). 5 Therefore, the empirical evidence indicates that before the 2000 transition the incumbent party, the PRI, was not successful in developing efficient anticrime policies in Mexico, while at the same time Mexican citizens regarded crime a central issue that the government had to face. The competitiveness of the 2000 presidential election, the role of more independent mass media, and the salience of change as an issue in the 2000 presidential campaigns were other features of the political context in

Mexico before the 2000 transition that eventually encouraged citizens’ demands for democracy.

In contrast to the 1994 presidential election in which Ernesto Zedillo, the

PRI candidate, had a clear advantage since the beginning of the election over the other two candidates (Diego Fernández de Cevallos, candidate of the National

Action Party, PAN, and Cuauhtémoc Cárdenas, candidate of the Party of the

Democratic Revolution, PRD), the 2000 presidential election was very close and affected citizens’ perceptions of the next president. The two leading candidates,

5 The data are available at the web page of the BIIACS: http://www.biiacs.cide.edu/

Francisco Labastida (PRI) and Vicente Fox (PAN) were ranked in a dead heat in the last weeks of the presidential contest. In Schumpeter’s’ perspective (1976), a competitive struggle, the first requirement of a democratic regime, made Mexico ready for a democratic transition. Therefore, voters knew in the last week of June that, for the first time in Mexico, a competitive electoral landscape provided a realistic opportunity for an opposition candidate to win, especially since Labastida himself openly admitted earlier in the campaign that he could lose (Camp, 2004).

Figure 1.5 shows the final results of the 1994 and 2000 presidential elections.

60 48.7 50 43.4

40 36.8

30 25.9 Electoral Result

Percentage 20

0 PAN PRI PAN PRI 1994 2000

Figure 1.5 6: Presidential Election Results in 1994 and in 2000. Source Federal Electoral Institute (IFE)

6 In the 2000 presidential election, Vicente Fox was the candidate of the Alliance for Change, integrated by the PAN and other political parties.

The development of a more independent mass media in Mexico made citizens less vulnerable to vote buying and more sensitive to media effects (the two debates of the 2000 presidential election) when they picked a presidential candidate (Lawson 2002). According to Cornelius (2004) the efficacy of vote buying and coercion declined in the 1990s, and it was rather low in the 2000 presidential election. Moreover, the opposition took advantage of the independent mass media and alerted people to the Priísta strategy of vote buying with suggestions like this one: “Take the gift, but vote as you please” (Cornelius 2004).

The two televised presidential debates also had effects on public opinion. The debates boosted audience impressions of Fox’s leadership abilities, solidified his own partisan base, and helped him lure away some voters from Labastida.

According to Fox’s internal tracking poll, for instance, his support grew by 4 percent in the week following the first debate (Lawson 2004b).

The 2000 presidential election was framed as being about “change”. The salience of “change” as a national issue in Mexican politics in the 2000 presidential elections raised Mexican citizens’ expectations about the determination of the next president to address not only economic but also political problems. All candidates, including the PRI candidate (Francisco Labastida) accepted this framing. The PAN presidential campaign (candidate Vicente Fox) identified democracy in terms of change, meaning alternation in power and distinguished its candidate as the only one capable of beating the PRI.

Cuauhtémoc Cárdenas, the PRD candidate, attempted to identify change as a kind

of alternation in policy rather than in party terms. The Priista strategy identified change as an internal reform of the PRI as its slogan shows “A new PRI, closer to you” (Bruhn, 2004). If “change” was the most salient issue in the 2000 presidential election, it was natural that Mexican citizens looked for an opposition candidate capable of bringing this change to Mexico.

An electoral context characterized by the salience of change as the main issue in political campaigns encourages voters to develop expectations for a more effective government. Magaloni and Poiré (2004, 315) argue that in the 2000 presidential election the salience of change, as a main national issue in Mexican politics, was based mostly on the expectation of a greater competence of the next

Mexican president “in managing the economy, fighting crime, and reducing corruption”. Whenever a voter identified crime as the main problem in Mexican politics, the probabilities of voting for the PRI candidate, Francisco Labastida, were reduced by 17 percent (Magaloni and Poiré, 2004). Moreover, when people identified the desire for change in Mexico as the main reason to vote for a presidential candidate in the 2000 election, the candidate who benefitted the most was Vicente Fox, as Figure 1.6 reveals:

70 66

40 Voting for President 30

Percentage 18 20 15

10 1 0 Fox (PAN) Labastica (PRI) Cardenas (PRD) Others

Figure 1.6: Why Mexican Voters Cast Their Ballots for President in 2000 when Change was the Main Reason to Vote. Figure Done by the Author with Data of the Postelection Poll by Reforma, July 2000, and Camp (2004, 33)

The 2000 presidential election significantly affected citizens’ perceptions of both democracy and the president. Camp’s analysis (2004) reveals that only 40 percent of the people thought that Mexico was a democracy in February 2000, four months before the presidential election; in contrast, in July 2000 some days later of the election 63 percent of the citizens believed that Mexico was a democratic regime. Figure 1.7 presents this change in citizens’ perceptions of democracy:

70 63 60

50 40 40 Perceptions of Mexico as a 30 Democracy

Percentage 20

0 February July Before and After the 2000 Presidential Election

Figure 1.7: Changing Visions of Democracy in the 2000 Presidential Election. Source: Camp (2004), Mexico 2000 Panel Study (Domínguez and Lawson, 2004). The question asked was: “Do you think Mexico is a democracy?”

This change in citizens’ perceptions affected how people perceived the president. Figure 1.8 shows that Mexican voters began to trust more in the president after the democratic transition. The levels of citizens’ trust in Vicente

Fox are clearly higher than those of Salinas and Zedillo after the first year of tenure of each president.

70 58 60

50 40 40 32 Trust 30

(Percentage) 20

Presidential Trust Raiting Presidential Trust 10

0 Salinas (1989) Zedillo (1995) Fox (2001) Mexican Presidents

Figure 1.8: Levels of Trust in the Mexican Presidents during Their First Year of Tenure. Figure Done by the Author. The Data Come from Banco de Información para la Investigación Aplicada en Ciencias Sociales www.biiacs.cide.edu and Parametría www.parametria.com.mx

1.4 Hypotheses

According to Moreno (2009) the 2000 presidential election worked as a critical election, which affects voting behavior. After the 2000 election, when the

PRI lost the presidency for the first time in seven decades, the regime cleavage

(PRI vs. antiPRI) started to fade, allowing for a redefinition to left and right positions around other issues of economic and social contents (Moreno 1999;

2006). I extend Moreno’s arguments to formulate my hypotheses. If Moreno is right and the 2000 electoral process worked as a “critical election” affecting voting behavior in Mexico, we can expect that the 2000 election also altered citizens’ perceptions of the president. More precisely, the 2000 transition is a

“determinant political event” in Mexican politics that modified how citizens viewed the president. Mexican people started to see the president in a different

way. They had more expectations of a greater competence of the next Mexican president in managing not only economic but also political issues (Magaloni and

Poiré 2004). In Mexico and in other new democracies, the political context has encouraged citizens to associate democracy with anticrime and anticorruption policies. In the first case, citizens in Latin America regard combating crime as one of the most important tasks of democracy (Camp 2001). In the second case, Latin

American countries, Mexico and Chile among others, (Klesner 2001) consider corruption to be one of the main obstacles to democracy. Therefore, my expectations are twofold.

H.1 All else being equal, the effect of citizens’ perceptions of crime on

presidential approval after the 2000 transition should be stronger than

those during the PRI regime. After the 2000 transition, Mexican citizens

had more expectations about the determination of the new president to

address crime. During the 1980s and 1990s, when the PRI stayed in office,

crime began to increase sharply in Mexico. Although there is not empirical

evidence that suggests whether Mexican people associated crime with the

PRI regime, it is likely that during the PRI hegemony, citizens perceived

the incumbent party’s anticrime policies as inefficient or at least that the

PRI government did not consider crime a central issue to address. In

contrast, Mexican people believed that one of the main tasks of the new

democratically elected government was combating crime and the new

president could eradicate this problem from Mexican politics.

H.2. All else being equal, the effects of citizens’ perceptions of corruption on

presidential approval after the 2000 transition should be stronger than

those during the PRI regime. After the 2000 democratic transition,

Mexican citizens started to have more expectations about the

determination of the new, democratically elected, president to address

corruption in Mexico. While Mexican people considered the PRI regime

the main symbol of corruption, they thought that the new president could

eradicate this problem from Mexican politics.

1.5 Data and Method

The eleven national surveys on which this study is based were organized by the Mexican Office of the Presidency 7 (MOP) and by the private polling firm,

Parametria . When the questionnaires were designed within the Presidency, the surveys were conducted by Opinion Profesional S.A. de C.V., a private company specializing in policy polling and focus groups 8. Field workers were not informed of the client’s identity; they identified themselves to respondents only as

7 The name of the presidential pollster in Mexico. This office conducts all public opinion research for the Mexican President. Polls are founded by the federal budget through the Office of the President. Surveys are confidential for governmental use. However, Asesoría Técnica de la Presidencia de la República made them public after Salinas and Zedillo finished their tenures as presidents. Data of all surveys is now public and available at the BIIACS . 8 Unfortunately, in my analysis it is not possible to control for a possible polling firm effect on presidential approval (Cohen, 2003). I created a dummy variable, called Parametría , to address this problem, but because of the little variation (I have only tow polling firms) the variable was dropped. However, a correlation test shows that presidential approval and the dummy variable Parametría are weakly correlated (.002). Therefore, the effect of polling firm on approval should be weak.

employees of Opinion Profesional . Parametría also designed and conducted questionnaires.

In total the surveys contain almost 18,000 household interviews, and 600 telephone interviews from November 1994 to November 2006. This period covers three different presidents (Carlos Salinas de Gortari, Ernesto Zedillo, both Priista presidents, and the Panista Vicente Fox). When the surveys were organized by

MOP, households were located for interviews through a multistage, stratified, and probability sampling procedure. At each household, an interview of about 40 minutes was conducted with one respondent, selected at random, with about an equal proportion of men and women. When Parametría organized the surveys, households were located for interviews through a multistage, nonstratified, and probability sampling procedure. At each household, an interview of about 25 minutes was conducted with one respondent, selected at random. There is one poll for Salinas’ presidency (in 1994) and two surveys for Zedillo’s presidency (1995 and 2000), all conducted by Asesoría Técnica de la Presidencia de la República .

Finally, there are eight surveys for Fox’s administration, six conducted by

Asesoría Técnica de la Presidencia de la República (2001, 2002, 2003, 2004) 9, and two conducted by Parametría (2006). Because every survey is an independent representative sample of the population, I pool all the eleven surveys in a single dataset. As a result, I run regressions that include between 6,000 and

18,000 observations. Pooling independent cross sections across time has the

9 Three surveys were conducted in 2001.

advantages of increasing sample size, obtaining more precise estimators, and more robust test statistics.

Governmentsponsored surveys can sometimes be criticized for loading questions in ways that elicit responses favorable to the political objectives of the president. However in the surveys organized by the Mexican Office of the

Presidency respondents were offered explicit “pro and con” choices on issues, including the performance of the president, citizens’ view of the political context

(corruption and crime), and their perceptions of the economy and economic problems (unemployment and inflation).

A logistic analysis is conducted, allowing the measurement of the probability that voters approved of the president before and after the Mexican transition (1994, 1995, 2000, 2001, 2002, 2003, 2004, and 2006) 10 . To measure presidential approval as dependent variable, I use a standard question: “Do you approve or disapprove the way that (Salinas, Zedillo, or Fox) is handling his job as president?” 11 My key explanatory variables are citizens’ perceptions of crime and corruption. People were asked to evaluate, respectively, the situation of crime and corruption in Mexico, respectively. The analysis also includes a series of

10 Unfortunately, it is not possible to include observations for every year. During the Salinas and Zedillo tenures the public’s perception of economic issues was not regularly surveyed. In the case of Fox, in 2005 questions about perceptions of the economy were not included in the surveys. Furthermore, in the case of Salinas, before 1994 questions about perceptions of unemployment and inflation, my control variables, were not regularly surveyed. Thus, it was not possible to include observations before 1994. 11 Some of the surveys have a wide range of options. People could respond “totally approve”, “partly approve”, “partly disapprove” and “totally disapprove”. Others have a narrow range options: “approve” or “disapprove”. I recoded the former surveys to correspond with the latter ones in order to increase my number of observations.

control variables that have been widely used in prior presidential approval research in Mexico: party identification (ID), citizens’ perceptions for economic issues (unemployment 12 and inflation 13 ) and the economy (Domínguez and

McCann 1995; Magaloni 2006; Moreno 2003; 2009; Villarreal 1999). I decided to use retrospective perceptions of the economy as control variables because, when the questions about corruption and crime were included in the surveys, people were not asked to evaluate either the general situation of the economy or their personal economic situation using a prospective view during most of the years of the PRI era. I also use as controls sociodemographic variables such as income, education, age and gender (see Appendix 1 for precise questions and coding process). Because the population may have different distributions in different time periods, I also create dummy variables for all but one year

I expect that when citizens think that corruption and crime increase they disapprove of the way the president is handling his job. Perceptions of corruption and crime are negatively associated with presidential approval. My expectations about unemployment and inflation are similar to those of crime and corruption.

Citizens punish the president with a low approval rating when they believe that unemployment and inflation are increasing. I anticipate that partisanship has a

12 I use citizens’ perceptions of poverty as a proxy of perceptions of unemployment in the survey conducted in 1995. In the rest of the surveys whenever perceptions of unemployment and poverty were asked the correlation between them was 85 percent approximately. Therefore, this strong association allows me to assume that these two variables are measuring something similar. 13 Fox’s tenure (20002006) was highly evaluated by the public because of low inflation rate. This is the reason why I use perceptions of Fox’s capacity for managing the economy as a proxy of perceptions of inflation in the surveys conducted in 2006.

positive effect on presidential approval. Members of the president’s party support his performance. Moreover, I predict that citizens’ perceptions of both their personal economic situation and the economy in Mexico affect positively presidential approval (see Appendix 1 for precise questions and coding process).

1.6 Results

Table 1.1, 1.2, and 1.3 present the results of the regression analysis. Table

1.1, Model 1 tests a general baseline model solely with the key explanatory variables, perceptions of crime and corruption. Model 2 introduces some of the control variables, perceptions of unemployment, inflation, country’s economic situation, personal economic situation, and partisanship. Model 3 includes the sociodemographic control variables. Taken together, and consistent with the economic voting literature, the results of Model 3 indicate that over the period of analysis, before and after the 2000 democratic transition, rising citizens’ perceptions of rising crime and corruption decreases presidential approval. The same is true for perceptions of unemployment and inflation. People’s evaluations of both country’s economic situation and personal economic situation have positive effects on presidential approval. Partisanship has strong impact on presidential approval. Model 3 shows that citizens’ perceptions of both political and economic problems matter for presidential approval in Mexico.

Table 1.1

Determinants of Presidential Approval before and after the Transition in Mexico (19942006) Source: www.biiacs.cide.edu and www.parametria.com.mx

Independent Model 1 (1994 Model 2 (1994 Model 3 (1994 Variables 2006) 2006) 2006) Perceptions of Rising .73*** .28*** .26*** Crime (.02) (.03) (.04)

Perceptions of Rising .72*** .34*** .32*** Corruption (.02) (.03) (.03)

Perceptions of Rising .4*** .42*** Unemployment (.03) (.04)

Perceptions of Rising .4*** .43*** Inflation (.03) (.04)

The Country’s .7*** .68*** Economic Situation (.03) (.04)

Personal Economic .36*** .36*** Situation (.03) (.03)

Partisanship 1.6*** 1.6*** (.05) (.05)

Lower Middle Class .001 (.06)

Middle Class and .03 Rich (.07)

YoungMature .12** People (.05)

Mature People .04 (.06)

The Elderly People .13 (.08)

Table 1.1, Continued

Independent Model 1 (1994 Model 2 (1994 Model 3 (1994 Variables 2006) 2006) 2006) Primary .03 (.11)

JuniorHigh School .03 (.12)

SeniorHigh School .14 (.13)

University or More .19 (.13)

Male .14*** (.04)

Pseudo R² .13 .33 .33

18,303 16,158 15,191 *p< .10, ** p < .05, ***p < .01. Note: The dependent variables is presidential approval at the individual level. I include year dummies except for one year in the estimation of the models but not presented in the table. Some of them are significant.

While the prior discussion suggests that perceptions of crime and corruption affect presidential approval, it is important to examine whether people evaluated the president differently in two different politicalinstitutional contexts, a semiauthoritarian regime in transition and an emerging democracy. As Table

1.2 shows, there were important differences in the way voters assessed two Priista presidents, Salinas and Zedillo, and a president who came from the opposition,

Fox, which is consistent with theory. During the predemocratic period, Model 4, when Mexican people thought that crime increased presidential approval did not suffer significant changes. However, perceptions of crime turned out to have significant effects on approval after the 2000 democratic transition, Model 5. The interactive variable PRI Era X Perceptions of Crime, Model 6 14 in Table 2, shows that the effects of perceptions of crime are significantly different between these two periods. This finding indicates that the institutionalpolitical context affects the way in which citizens evaluate the president. As H.1 predicts, in a semi authoritarian regime, voters do not think that combating crime is one of the main concerns of an authoritarian systems. In contrast, citizens strongly associate crime as one the main tasks to be addressed by the government in an emerging democracy.

The story of corruption is a little different than that of crime. In both periods, before and after the transition, citizens’ perceptions of corruption strongly affected presidential approval. However, as the interactive variable PRI

Era X Perceptions of Corruption, Model 6 in Table 1.2, shows the effects of perceptions of corruption are significantly different between these two periods.

They became stronger after the transition. This finding indicates that corruption was a relevant issue for Mexican people even before the democratic transition.

14 In Model 6, I test the hypothesized differences between these two periods, the predemocratic period and the emerging Mexican democracy, by including a battery of interaction effects for each of the key independent variables. Every interactive variable is the product of the variable PRI (the nondemocratic age) and the key explanatory variables. Every time the interactive variable is significant the analysis shows that the different impact of the variable tested is significant by itself, and not by the different number of observations in models 4 and 5.

The results suggest that Mexican citizens thought that during the PRI era the president could address this problem even though he was a member of the PRI.

Some empirical evidence in Mexican politics gives support for this explanation.

It was Miguel de la Madrid (19821988), a Priista president who started a plan against corruption through the campaign Renovación Moral. It was his successor,

Carlos Salinas, who developed a plan for restructuring the economy, providing social programs, and attacking corruption in government and some labor unions.

Moreover, Davis and Langley (1995) find that perceptions of corruption have significant effects on presidential approval during the PRI era. Regardless the real determination of both Priista presidents to address corruption, the findings of this analysis reveal that corruption was a relevant issue when citizens evaluated the president before the 2000 transition. The key point for this analysis is that although people’s view about corruption matters in both periods, it had stronger effects after the PRI era. As H.2 predicts, after the 2000 transition, Mexican people believed that a new president, from a party different from the PRI, would eradicate corruption. Thus, after 2000 perceptions of corruption have become more relevant and have had stronger effects on presidential approval.

Model 6 reveals that the effects of both perceptions of unemployment and inflation are significantly different between these two periods, before and after the transition, but in opposite direction. While perceptions of unemployment had stronger effects in the democratic period, perceptions of inflation were stronger before the transition. This finding is challenging to explain. The impact of

perceptions of unemployment is consistent with earlier studies. Magaloni and

Poiré (2004, 315) find that in the 2000 presidential election Mexican voters had more expectations of a greater competence of the next president in managing the economy, fighting crime, and reducing corruption. However, perceptions of inflation, with stronger effects on approval during the PRI era, suggest that only some economic issues, those that have been salient for the public, may exert more impact after a democratic transition. This is the case of unemployment but not of inflation. Empirical data reveal that unemployment has been more relevant than inflation for Mexican people in the last years. 15 The effects of some sociodemographic variables are significantly different between the predemocratic era and the democratic period. In Model 6, the interactive variables of both income and age 16 show that wealthy people (the variables Lower Middle Class,

Middle Class, and Rich) and old citizens (the variables YoungMature, Mature, and Elderly People) generally support more strongly the president before the transition. During the PRI regime the wealthy were satisfied with the status quo and approved the president. However, with the emergence of a democratic system, the wealthy people became more critical of the president because they thought that a bad executive would affect their profits. The elderly enjoyed some of the benefits of the PRI regime, economic growth and political stability. Thus,

15 The Reforma ews Paper published survey results in 2004 and 2005 in which unemployment were more salient than inflation in Mexico, as a national concern. This evidence is consistent with the data that it is possible to get from the BIIACS ’s web page: http://www.biiacs.cide.edu/

16 The base line group of income is the people who have a salary lower than 1 minimum salary. The base line group is the youngest people in the surveys, from 18 to 25 years old.

they approved of Priista presidents and did not approve of a president who came from the opposition. The effects of gender and education are not significantly different between these two periods. In the case of gender 17 , the reason is that this variable had similar effects before and after the transition. Men approved of the president less than women. In the second case, education may be associated with other variables such as ideology that were not possible to include in the analysis.

17 The base line group is the people who do not have scholarly education.

Table 1.2

Determinants of Presidential Approval, PreDemocratic Era and Democratic Period, Compared

Independent Model 4 Model 5 Model 6 Variable PreDemocratic Democratic Era (Interactive Period (9400) (0106) Test) (9406) Perceptions of .02 .46*** .46*** Rising Crime (.05) (.05) (.05)

Perceptions of .2*** .42*** .42*** Rising Corruption (.04) (.05) (.05)

Perceptions of .26*** .48*** .48*** Rising (.06) (.05) (.05) Unemployment Perceptions of .6*** .33*** .33*** Rising Inflation (.05) (.04) (.04)

The Country’s .42*** .96*** .96*** Economic Situation (.05) (.05) (.05)

Personal Economic .5*** .32*** .32*** Situation (.04) (.05) (.05)

Partisanship 1.7*** 1.4*** 1.4*** (.07) (.08) (.08)

Lower Middle Class .27*** .13 .13 (.1) (.08) (.08)

Middle Class and .41*** .32*** .32 Rich (.11) (.09) (.09)

YoungMature .3*** .03 .03 People (.08) (.08) (.08)

Mature People .12 .17* .17* (.09) (.09) (.09 )

Table 1.2, Continued

Independent Model 4 Pre Model 5 Model Variables Democratic Democratic Era Interactive Test Period (9400) (0106) (9406) The Elderly People .17 .26*** .26*** (.15) (.11) (.11)

Primary .16 .09 .09 (.2) (.14) (.14)

JuniorHigh School .02 .01 .01 (.2) (.15) (.15)

SeniorHigh School .06 .15 .15 (.21) (.16) (.16)

University or More .23 .17 .17 (.21) (.16) (.16)

Male .15*** .17*** .17*** (.06) (.06) (.06)

PRI Era .18 (.28)

PRI Era X .44*** Perceptions of Crime (.07)

PRI Era X .22*** Perceptions of (.07) Corruption PRI Era X .22*** Perceptions of (.07) Unemployment PRI Era x .25*** Perceptions of (.08) Inflation PRI Era X The .54*** Country’s Economic (.07) Situation

Table 1.2, Continued

Independent Model 4 Pre Model 5 Model 6 Variables Democratic Democratic Era (Interactive Period (9400) (0106) Test, 9406) PRI Era X Personal .17*** Economic Situation (.07)

PRI Era X .3*** Partisanship (.11)

PRI Era X Lower .41*** Middle Class (.13)

PRI Era X Middle .74*** Class and Rich (.15)

PRI Era X Young .34*** Mature People (.11)

PRI Era X Mature .29** People (.13)

PRI Era X The .43** Elderly People (.19)

PRI Era X Primary .26 (.25)

PRI Era X Junior .006 High School (.26)

PRI Era X Senior .09 High School (.26)

PRI Era X .06 University or More (.27)

Table 1.2, Continued

Independent Model 4 Pre Model 5 Model 6 Variables Democratic Democratic Era (Interactive Period (9400) (0106) Test, 9406) PRI Era X Male .02 (.08) Pseudo R² .32 .36 .35 N 6755 8436 15,191 *p< .10, ** p < .05, ***p < .01. Note: The dependent variables is presidential approval at the individual level. I include year dummies except for one year in the estimation of the models but not presented in the table. Some of them are significant.

Although the raw coefficients presented in Table 1.2 are essential in determining the accuracy of my hypotheses, they are quite difficult to interpret in nonlinear specifications. To address this problem and to highlight substantive findings, I present results of simulations based on the models specified above.

Table 1.3 presents the estimated probabilities of presidential approval in two different periods, before and after the 2000 democratic transition. I include only the most important variable for this analysis. These results suggest that after the democratic transition, perceptions of both crime and corruption became more relevant as determinants of presidential approval. Keeping all other explanatory factors at their median, when citizens thought crime increased in Mexico, the likelihood of approving the president decreased by 1 % before the transition, and

by 16.5 % after the PRI regime. The story of perceptions of corruption is similar.

Controlling for the rest of the variables, when citizens believed that corruption increased, the likelihood of approving the president decreased by 9 % before the emergence of the new Mexican democracy, and by 15 % after democratization. In the cases of unemployment and inflation, Table 1.3 indicates that while perceptions of inflation had stronger effects during the PRI era, perceptions of unemployment were more powerful in the democratic period.

This analysis also reveals that citizens’ perception of the economy did not have the same impact after democratization in Mexico. While perceptions of personal economic situation had stronger effects on presidential approval before the transition, the impact of country’s economic situation became stronger after the emergence of the new Mexican democracy. This result suggests that democratization makes people more sociotropic than egocentric. One part of the supply of the old PRI regime was the benefits that citizens could get through clientelism. Clientelistic strategies allowed the PRI to stay in office and strengthened citizens’ egocentric view. People were more worried about their personal economic situation than country’s economic situation. After the democratic transition, Mexican citizens started to focus their attention on the country rather than their personal economic status. Finally, Table 1.3 reveals that partisanship lost some power as a predictor of presidential approval after the 2000 transition. During the PRI semiauthoritarian regime, being a member of the president’s party increased the probabilities of approving the president by 32

percent, while in the postPRI era the probabilities of approving of the president were only 16 percent.

Table 1.3

Changes in Probabilities of the Determinants of Presidential Approval in Mexico When this Shifts from… to… Changes in Changes in Characteristic Probabilities Probabilities PreDemocratic Democratic Era Period Perceptions of Crime 1% decreases 16.5% decreases Rising Crime decreases (0) increases (2)

Perceptions of Corruption 9% decreases 15% decreases Rising decreases (0) Corruption increases (2)

Perceptions of Unemployment 12.3% decreases 17% decreases Rising decreases (0) Unemployment increases (2)

Perceptions of Inflation 27% decreases 12% decreases Rising Inflation decreases (0) increases (2)

Personal Worse (0) 23% increases 12% increases Economic Better (2) Situation The Country’s Worse (0) 20% increases 33% increases Economic Better (2) Situation Partisanship Non Member of the 32% increases 16% increases President’s Party (0) Member of the President’s Party (1)

The main theoretical proposition of this chapter is that since the 2000 democratic transition, perceptions of both crime and corruption have become more salient and have had stronger effects on presidential approval. Two counter hypotheses can question the main argument of this chapter. First, the media coverage on crime and corruption before the transition encouraged voters to evaluate the president differently after the 2000 presidential election. Thus, perceptions of crime and corruption had stronger effects on presidential approval.

I test this counterhypothesis in Table 1.5. Second, Fox’s presidency focused a lot of public attention on crime and corruption, making these issues more salient in citizens’ mind subsequently. Therefore, crime and corruption became more salient and had stronger effects on approval. I test this counterhypothesis in Table 1.6.

The empirical evidence reveals that the changes in perceptions of crime and corruption as relevant issues to evaluate the president had the strongest and most significant effects when the study uses the 2000 democratic transition as the turning point of presidential approval in Mexico. The analysis uses observations for all the years from 1994 to 2006, except for 1998 and 1995. Some of the control variables, the public’s perceptions of the economy, were not regularly surveyed. Thus, it was not possible to include them in the test. The models of tables 1.4, 1.5, and 1.6 include both the surveys that the analysis used in Table 1.2 and those that it was possible to include when some of the control variables were taken out.

To test the counterhypothesis that media coverage on crime and corruption affected how citizens evaluated the president, I use as the breaking point of my analysis the 1997 intermediate elections, the first time that Mexico had divided government. In the 1990s, independent publications emerged and flourished, affecting citizens’ perceptions of Mexican politics (Lawson 2002).

Changes in media started in the early 1990s. If media coverage affected citizens’ perceptions of the president, its effects should be seen after the mid1990s. I use the year of 1997 as the breaking point of the analysis. To test the counter hypothesis that Fox’s presidency focused a lot of public attention on crime and corruption, and these issues became more salient in citizens’ mind subsequently, I use as the breaking point of this study the year of 2003. Fox ascended to power in

2000. The effects of Fox’s agenda setting on citizens’ approval of the president should be seen in a middlelong time period, after the first part of his tenure.

Therefore, I use the 2003 intermediate elections as the breaking point.

Tables 1.4, 1.5 and 1.6 show that when I use as the breaking point of the analysis the 2000 presidential election, the effects of the interactive variable The

2000 Presidential Election X Perceptions of Crime , in Table 1.4, Model 9, are stronger (.42) than those when the breaking points are the 1997 (.35 in Table 1.5,

Model 12) and the 2003 (.17 in Table 1.6, Model 15) intermediate elections. The story corruption is even more interesting. The effects of the interactive variable

The 2000 Presidential Election X Perceptions of Corruption are stronger (.16 in

Model 9) than those when the breaking point is the 1997 intermediate election

(.13 in Model 12). Moreover, when the breaking point is the 2003 intermediate elections, the effects of perceptions of corruption are not significantly different between the two periods (before and after 2003), as the interactive variable The

2003 Intermediate Election X Perceptions of Corruption , in Table 1.6, Model 15, shows.

The results of this chapter should be strengthened by including objective indictors. Because of the little variation of the yearly levels of crime, corruption, unemployment and inflation in the period of study, it was not possible to include these variables in this chapter. The analysis suffered from a problem of collinearity when I included national objective indicators. When I controlled for objective conditions, some of the variables were dropped. Although the correlation among the objective indicators was not high, some the variables were expressed as a liner function of the others. I attempt to overcome this problem by including state objective indicators in the next two chapters.

Table 1.4, Effects of Perceptions of Crime, Corruption, Inflation, and Unemployment on Presidential Approval in Mexico (19942006), Using the 2000 Presidential Election as the Breaking Point, Logit Models

Model 7 (9400) Model 8 (01 Model 9 PRI in Office 06)Democratic InteractiveTest Period Model (9406) Perceptions of .17*** .59*** .59*** Rising Crime (.03) (.04) (.04)

Perception of .29*** .45*** .45*** Rising Corruption (.03) (.04) (.04)

Perceptions of .74*** .55*** .55*** Rising (.04) (.04) (.04) Unemployment

Perceptions of .71*** .54*** .54*** Rising Inflation (.04) (.04) (.04)

Partisanship 1.65*** 1.8*** 1.7*** (.05) (.06) Socio demographic Variables Lower Middle .14** .15*** .02 Class (.06) (.06) (.04)

Middle Class and .11 .26*** .08* Rich (.07) (.06) (.04)

Primary .02 .02 .02 (.12) (.12) (.08)

JuniorHigh .02 .1 .04 School (.12) (.12) (.09)

SeniorHigh .12 .12 .001 School (.13) (.13) (.09)

Table 1.4, Continued

Model 7 (9400) Model 8 (0106) Model 9 PRI in Office Democratic InteractiveTest Period (9406) University or .23* .003 .11 More (.13) (.13) (.09)

YoungMature .02 .09 .02 People (.05) (.06) (.04) Mature People .06 .2*** .12*** (.06) (.07) (.05)

The Elderly .08 .29*** .18*** People (.09) (.09) (.06)

Male .03 .1** .06** (.04) (.04) (.03)

InteractiveTest Variables The 2000 .71*** Presidential (.06) Election (PE)

The 2000 PE X .42*** Perceptions of (.05) Crime

The 2000 PE X .16*** Perceptions of (.05) Corruption

The 2000 PE X .19*** Perceptions of (.06) Unemployment

The 2000 PE X .16*** Perceptions of (.05) Inflation

Table 1.4, Continued

Model 7 (9400) Model 8 (0106) Model 9 PRI in Office Democratic InteractiveTest Period (9406) The 2000 PE X .11 Partisanship (.08)

N 12615 11820 24435

Pseudo R² .26 .30 .29 *p< .10, ** p < .05, ***p < .01. Note: The dependent variables is presidential approval at the individual level. I include year dummies except for one year in the estimation of the models but not presented in the table. Some of them are significant.

Table 1.5, Effects of Perceptions Crime, Corruption, Inflation, and Unemployment on Presidential Approval in Mexico (19942006), Using the 1997 Intermediate Election as a Breaking Point, Logit Models

Model 10 (9497) Model 11 (9806) Model 12 InteractiveTest Model (9406)

Perceptions of .17*** .52*** .52*** Rising Crime (.03) (.03) (.04)

Perceptions of .27*** .42*** .42*** Rising Corruption (.04) (.04) (.03)

Perceptions of .75*** .43*** .44*** Rising (.04) (.04) (.04) Unemployment Perceptions of .78*** .44*** .44*** Rising Inflation (.04) (.04) (.03)

Partisanship 1.86*** 1.54*** 1.5*** (.06) (.05) (.05)

Sociodemographic Variable Lower Middle .19*** .13*** .01 Class (.07) (.06) (.04)

Middle Class and .19*** .2*** .06 Rich (.08) (.06) (.04)

Primary .12 .07 .009 (.13) (.11) (.08)

JuniorHigh School .08 .04 .01 (.13) (.12) (.09)

SeniorHigh .02 .07 .05 School (.14) (.12) (.09)

Table 1.5, Continued

Model 10 (9497) Model 11 (9806) Model 12 InteractiveTest Model (9406) University or .09 .31*** .22** More (.14) (.12) (.09)

YoungMature .1* .12** .01 People (.06) (.05) (.04)

Mature People .05 .27*** .12*** (.07) (.06) (.05)

The Elderly .08 .33*** .16*** People (.11) (.08) (.06)

Male .02 .09** .06* (.05) (.04) (.03) InteractiveTest Variables The 1997 .78*** Intermediate (.06) Election (IE) The 1997 IE X .35*** Perceptions of (.05) Crime The 1997 IE X .13*** Perceptions of (.05) Corruption The 1997 IE X .32*** Perceptions of (.06) Unemployment The 1997 IE X .34*** Perceptions of (.06) Inflation The 1997 IE X .32*** Partisanship (.07)

Table 1.5, Continued

Model 10 (9497) Model 11 (9806) Model 12 InteractiveTest (9406) N 10585 13850 24435

Pseudo R² .3 .26 .28

*p< .10, ** p < .05, ***p < .01. Note: The dependent variables is presidential approval at the individual level. I include year dummies except for one year in the estimation of the models but not presented in the table. Some of them are significant.

Table 1.6, Effects of Perceptions of Crime, Corruption, Inflation, and Unemployment on Presidential Approval in Mexico (19942006), Using the 2003 Intermediate Election as a Breaking Point, Logit Models

Model 13 (1994 Model 14 (2004 Model 15 2003) 2006) InteractiveTest Model (1994 2006) Perception of .36*** .54*** .54*** Rising Crime (.03) (.06) (.07)

Perception of .3*** .41*** .41*** Rising Corruption (.02) (.07) (.07)

Perceptions of .57*** .56*** .56*** Rising (.03) (.07) (.07) Unemployment Perceptions of .57*** .58*** .57*** Rising Inflation (.03) (.06) (.06)

Partisanship 1.62*** 2*** 1.97*** (.04) (.1) (.1)

Lower Middle .06 .35*** .0006 Class (.04) (.11) (.04)

Middle Class and .06 .46*** .02 Rich (.05) (.12) (.04)

Primary .02 .1 .02 (.09) (.2) (.08)

JuniorHigh .03 .04 .02 School (.09) (.2) (.08)

SeniorHigh .07 .17 .04 School (.1) (.21) (.09)

University or .2* .18 .2** More (.1) (.23) (.09) YoungMature .03 .17* .0002 People (.04) (.1) (.04)

Table 1.6, Continued

Model 13 (1994 Model 14 (2004 Model 15 2003) 2006) (19942006) Mature People .11** .16 .11*** (.05) (.11) (.04)

The Elderly .04 .36** .09 People (.07) (.15) (.06)

Male .06* .07 .06* (.03) (.07) (.03) InteractiveTest Variables The 2003 .26*** Intermediate (.06) Election (IE)

The 2003 IE X .17** Perceptions of (.07) Crime

The 2003 IE X .1 Perceptions of (.08) Corruption

The 2003 IE X .009 Perceptions of (.08) Unemployment

The 2003 IE X .009 Perceptions of (.06) Inflation

The 2003 IE X .35*** Partisanship (.11)

Table 1.6, Continued

Model 13 (1994 Model 14 (2004 Model 15 (1994 2003) 2006) 2006) N 19953 4482 24435

Pseudo R² .26 .3 .27 *p< .10, ** p < .05, ***p < .01. Note: The dependent variable is presidential approval at the individual level. I include year dummies except for one year in the estimation of the models but not presented in the table. Some of them are significant.

1.7 Conclusion

The main theoretical goal of this chapter has been to explain how political institutions affect presidential approval. I have presented empirical evidence to show that the political regime works as the market when citizens evaluate the president. From the demand side, citizens have expectations about the government, and from the supply side, the president’s behavior to address these concerns is restricted by the political regime. In Mexico since the 2000 presidential election, citizens’ main political concerns have become more relevant and have had stronger effects on presidential approval. I have used Mexico as a study case to analyze how democratization affects presidential approval.

I have argued that in Mexico since the 2000 transition political issues, crime and corruption, have become salient. Citizens consider combating crime to be the second most important task to be addressed by democracy, and view

corruption as the main obstacle for a democratic consolidation (Clark 2001, 87;

Klesner 2001, 127; Moreno 2001, 43). Thus, after the 2000 presidential election, the turning point of the transition, Mexican citizens have begun to evaluate the president in terms of his anticorruption and anticrime policies, and performance.

This analysis reveals that the effects of citizens’ perceptions of both rising corruption and crime in the democratic period were stronger and significantly different than those during the PRI semiauthoritarian system. As a robust test, I conducted an analysis by using the 1997 and 2003 intermediate elections as breaking points, respectively. When I used the 2000 presidential election as the breaking point of the analysis, the effects of both perceptions of rising crime and corruption were stronger than those when the breaking points were 1997 and

2003, respectively.

Beyond Mexico, this chapter makes two theoretical contributions to the study of presidential approval in comparative politics. First, political behavior does not occur in a vacuum. The political context affects the way in which citizens evaluate the president. Second, cases such as Mexico reveal that our understanding about emerging democratic systems should be reevaluated because as this analysis shows democracy is not only about elections but also about people’s main concerns.

1.8 Appendix 1 18 :

DEPEDET VARIABLE:

Presidential Approval

En general, ¿está de acuerdo o en desacuerdo con la manera como está gobernando el presidente? (1) Acuerdo, (2) Acuerdo en parte, (3) Desacuerdo en parte, (4) Desacuerdo. Do you approve the way the president is handling his job? (1) Agree, (2) Partially agree, (3) Partially disagree, (4) Disagree [Replies are recoded; 1 and 2 = 1 (People approve the president); 3 and 4 = 0 (People do not approve the president)

IDEPEDET VARIABLES:

Perceptions of Crime in Mexico

Desde que Salinas (Zedillo o Fox) es presidente, ¿cree usted que ha aumentado o disminuido el crimen?(1) Aumentado (2) Igual (3) Disminuido Since Salinas (Zedillo, or Fox) is president, do you think that crime has increased or decreased in Mexico? (1) Increased (2) The same (3) Decreased [Replies are recoded: 1= Decreased; 2= The same; 3= Increased]

Perceptions of Corruption in Mexico Desde que Salinas (Zedillo o Fox) es presidente, ¿cree usted que ha aumentado o disminuido la corrupción? Aumentado (2) Igual (3) Disminuido Since Salinas (Zedillo, or Fox) is president, do you think that corruption has increased or decreased in Mexico? (1) Increased (2) The same (3) Decreased [Replies are recoded: 1= Decreased; 2= The same; 3= Increased]

Perceptions of Unemployment in Mexico 19 Desde que Salinas (Zedillo o Fox) es presidente, ¿cree usted que ha aumentado o disminuido el desempleo? Aumentado (2) Igual (3) Disminuido Since Salinas (Zedillo, or Fox) is president, do you think that unemployment has increased or decreased in Mexico? (1) Increased (2) The same (3) Decreased

18 It is important to mention that there are some wording differences in the surveys. 19 I use citizens’ perceptions of poverty as a proxy of perceptions of unemployment in the survey conducted in 1995.

[Replies are recoded: 1= Decreased; 2= The same; 3= Increased]

Perceptions of Inflation in Mexico 20 Desde que Salinas (Zedillo o Fox) es presidente, ¿cree usted que ha aumentado o disminuido la inflación? Aumentado (2) Igual (3) Disminuido Since Salinas (Zedillo, or Fox) is president, do you think that inflation has increased or decreased? (1) Increased (2) The same (3) Decreased [Replies are recoded: 1= Decreased; 2= The same; 3= Increased]

Perception of Country’s Economic Situation Comparada con el año anterior, ¿cómo cree usted que está la situación económica del país actualmente, mejor o peor? (1) Mejor (2) Igual (3) Peor Compared with last year, do you think that the current economic situation of the country is better of worse? (1) Better (2) The same (3) Worse [Replies are recoded 1 = Worse; 2= The same; 3= Better]

Perception of Personal Economic Situation ¿Cree usted que su situación económica personal es mejor o peor que hace un año? (1) Mejor (2) La misma (3) Peor Do you think your personal situation is better or worse than a year ago? (1) Better (2) The same (3) Worse [Replies are recoded 1 = Worse; 2= The same; 3= Better]

Partisanship (Matching the President’s party) ormalmente usted se considera, ¿priísta, panista o perredista? (1) Panista, (2) Priísta, (3) Perredista, (4) De otro partido, (5) o se identifica con ningún partido. Do you consider yourself as (1) Panista (2) Priista (3) Perredista (4) For any other party (5) Independent [Replies are recoded as a dummy variable: 1 if the voter belongs to President’s party; 0 if the voter does not. In 1994, 1995, and 2000, 1 if the voter is Priista , 0 if this is not the case. From 2001 to 2006, 1 if the voter is Panista , 0 if this is not the case.

Income ¿Cuál es su rango de ingreso familiar? (1) 0 a 1 Salarios Mínimos (SM) (2) 1 a 3 SM (3)3 a 5 SM (4) 5 o más SM What is the range of your family income? (1) 0 to 1 Minimum Salaries (MS) (2) 1 to 3 MS (3) 3 to 5 MS (4) 5 or more MS

20 I use perceptions of the president’s capacity for managing the economy as a proxy of perceptions of inflation in the surveys conducted in 2006.

[Responses are recoded: 2 Lower middle class; 3 and 4 Middle class and rich] The reference group is poor. It was not possible to create more categories because in some surveys people had only three possible answers to choose. (1) 0 a 1 Salarios Mínimos (SM) (2) 1 a 3 SM (3)3 o más SM

Education

¿Hasta qué año escolar estudió usted? (1) inguno (2)Primaria incompleta (3) Primaria completa (4) Secundaria incompleta (5) Secundaria completa (6) Preparatoria incompleta (7) Preparatoria completa (8) Universidad incompleta (9) Universidad completa y más What year of school did you reach? (1) None (2) Primary incomplete (3) Completed primary (4) Junior High Scholl incomplete (5) Completed Junior High School (6) Senior High Scholl incomplete (7) Completed Senior High School (8) Incomplete university (9) Completed university or more. [Recoded as: 1 None; 2 and 3 Primary; 4 and 5 Secondary; 6 and 7 Senior High School; 8 and 9 University or more] The base line group is the people who do not have scholarly education.

Age ¿Cuántos años cumplidos tiene usted? How old are you? [Recoded as: From 26 to 40 years old, YoungMature People; from 41 to 60 years old, Mature People; 61 or more, Elderly people] The reference group is people from 18 to 25 years old.

Gender (Male) The interviewers supplied information in this category (1) Male (2) Female Recoded as: 1 Male; 0 Female

1.9 Works Cited

Arce, Moises. 2003. “Political Violence and Presidential Approval in Peru”. The Journal of Politics 65 (May): 572583.

Beetham, David. 1993. “Four Theorems about the Marked and the Democracy”. European Journal of Political Research, Vol. 23: 187201.

Brody, Richard A. 1991. Assessing the President. Stanford, California: Stanford University Press.

Buendía, Jorge. 1996. “Economic Reform, Public Opinion and Presidential Approval in Mexico 19881993”. Comparative Political Studies, Vol. 29, No. 5: 566591.

Camp, Roderic. 2001. Citizen Views of Democracy in Latin America. The United States: University of Pittsburgh Press.

2004. “Citizen Attitudes toward Democracy and Vicente Fox’s Victory in 2000” Mexico’s Pivotal Democratic Election. Ed. Jorge I. Dominguez and Chappell Lawson: Stanford: Stanford University Press: 2546.

CanesWrone, Brandice. 2004. “The Public Presidency, Personal Approval Ratings and Policy Making”. Presidential Studies Quarterly, Vol. 34, No. 3: 477492. 2006. Who Leads Whom? Chicago: The University of Chicago Press.

Clark, Mary A. 2001. “Costa Rica: Portrait of an Established Democracy”. Citizen Views of Democracy in Latin America. Ed. Roderic Ai Camp. The United States: University of Pittsburgh Press. 7390

Clarke, Harold and Marianne Stewart. 1994. “Prospection’s Retrospections, and Rationality: The Bankers Models of Presidential Approval Reconsidered”. American Journal of Political Science, Vol. 38, No. 4: 11041123.

Cohen, Jeffrey E. 2002. “The Polls: PolicySpecific Presidential Approval, Part 2”. Presidential Studies Quarterly, Vol. 32, No. 4: 779788.

, 2004, “Economic Perceptions and Executive Approval in Comparative Perspectie”. Political Behavior, Vol. 26, No. 1: 2743.

, and Richard Powell. 2005. “Building Public Support from the Grassroots Up: The Impact of Presidential Travel on StateLevel Approval” Presidential Studies Quarterly 35, No. 1 (March): 1127.

Cornelius, Wayne, Judith Gentleman and Peter H. Smith. 1989. Mexico’s Alternative Political Futures. San Diego, California, Center for U.S. Mexican Studies.

Cornelius, Wayne, and Ann L. Craig. 1991. The Mexican Political System in Transition. Monograph Series 35. San Diego, California, Center for U.S. Mexican Studies.

Cornelius, Wayne. 2004. “Mobilized Voting in the 2000 Elections: the Changing Efficacy of Vote Buying and Coercion in Mexican Electoral Politics” Mexico’s Pivotal Democratic Election. Ed. Jorge I. Dominguez and Chappell Lawson: Stanford: Stanford University Press: 4766.

, and David Shirk. 2007. Reforming the Administration of Justice in Mexico. San Diego, California: University of Notre Dame Press and Center for U.SMexican Studies.

Cuzan, Alfred. 1997. “Presidential Popularity in Central America”. Political Research Quarterly , Vol. 50, No. 4”: 833849.

Davis, Charles L. and Ronald E. Langley. 1995. “Presidential Popularity in a Context of Economic Crisis and Political Change: The Case of Mexico”. Studies in Comparative International Development, Vol 30. No. 3: 2448.

Domínguez, Jorge I. and James McCann. 1995. “Shaping Mexico’s Electoral Arena: The Construction of Partisan Cleavages in the 1988 and 1991 National Elections”. The American Political Science Review, Vol. 89, No. 1: 3448

Domínguez Jorge, and Chappell Lawson. 2004. Mexico’s Pivotal Democratic Election. Stanford: Stanford University Press

Domínguez, Jorge. 2004. “Conclusion: Why and How Did Mexico’s 2000 Presidential Election Campaign Matter?” Mexico’s Pivotal Democratic Election. Ed. Jorge I. Dominguez and Chappell Lawson: Stanford: Stanford University Press: 321345.

Duch, Raymond. 2001. “A Developmental Model of Heterogeneous Economic Voting in New Democracies. American Political Science Review, Vol. 95, No. 4: 895910.

Edward, George. 1980. Presidential Influence in Congress. The United States: Library of Congress.

1989. At the Margins. Presidential Leadership of Congress. New Haven and London: Yale University Press.

1995. “Explaining Presidential Approval: The Significance of Issue Salience”. American Journal of Political Science , Vol. 39: 108134.

Eisenstadt, Todd. 2004. Courting Democracy in Mexico. USA: Cambridge University Press

Fiorina, Morris. 1981. Retrospective Voting in American ational Elections. New Haven and London: Yale University Press.

Geddes, Barbara and John Zaller 1989. “Sources of Popular Support for Authoritarian Regimes”. American Journal of Political Science , Vol. 33, No. 2 (May): 319347

Gélineau, François. 2007. “Political Context and Economic Accountability: Evidence From Latin America”. Political Research Quarterly. Vol. 69, No. 3: 415428.

GómezVilchis, Ricardo R. 2008. “Responsiveness of the Political System and Presidential Appoval in the Mexican Transition to Democracy.” Presented at the Annual Meeting of MPSA National Conference.

Hagopian Frances and Scott P. Mainwaring. 2005. The Third Wave of Democratization in Latin America. Advances and Setbacks. Cambridge, New York: Cambridge University Press.

Huntington, Samuel P. 1992. The Third Wave. The United States: University of Oklahoma Press.

Kenski, Henry. 1977. “Inflation and Presidential Popularity”. The Public Opinion Quarterly, Vol. 41, No. 1 (Spring): 8690.

Kernell, Samuel. 1978. “Explaining Presidential Popularity. How Ad Hoc Theorizing, Misplaced Emphasis, and Insufficient Care in Measuring One’s Variables Refuted Common Sense and Led Conventional Wisdom Down the Path of Anomalies”. The American Political Science Review, Vol. 72, No. 2 (June): 506522.

, 1997. Going Public. The United States of America. Congressional Quarterly Press.

Key, V. O. 1966. The Responsible Electorate. Cambridge, Massachusetts: Harvard University Press.

Kiewiet, Roderick. 1983. Macroeconomics and Micropolitics. Chicago and London: The University of Chicago press.

Kinder, Donald and Roderick Kiewiet. 1979. “Economic Discontent and Political Behavior”. American Journal of Political Science, Vol. 23, No. 3: 495 527.

Lawson, Chappell. 2002. Building the Fourth Estate: Democratization and the Rise of a Free Press in Mexico. London, England: University of California Press

, 2004a. “Introduction” Mexico’s Pivotal Democratic Election. Ed. Jorge I. Dominguez and Chappell Lawson: Stanford: Stanford University Press: 124.

, 2004b. “Mexico’s Great Debates: the Televised Candidate Encounters of 2000 and Their Electoral Consequences” Mexico’s Pivotal Democratic

Election. Ed. Jorge I. Dominguez and Chappell Lawson: Stanford: Stanford University Press: 211241

MacKuen, Michael B., Robert S. Erikson, and James A. Stimson. 1992. “Peasants or Bankers? The American Electorate and the U.S. Economy”. American Journal of Political Science, Vol. 86, No. 3: 597611.

Magaloni, Beatriz. 2006. Voting for Autocracy. The Untied States: Cambridge University Press.

Markus, George. 1988. “The Impact of Personal and National Economic Conditions on the Presidential Vote: A Poole CrossSectional Analysis”. American Journal of Political Science, Vol. 32 No. 1: 137154.

Mayer, Jeremy. 2004. “The Contemporary Presidency: The Presidency and Image Management”. Presidential Studies Quarterly , Vol. 34, No. 3: 620631.

Mishler, William and John P. Willerton. “The Dynamics of Presidential Popularity in PostCommunist Russia:. The Journal of Politics, Vol. 65 No. 1: 111141.

Miller, Joanne and Jon Krosnick. 2000. “The News Media Impact on the Ingredients of Presidential Evaluations”. American Journal of Political Science, Vol. 44, No. 2: 301315.

Moreno, Alejandro. 1999. Political Cleavages. The United States of America: Westview Press.

, 2001. “Democracy and Mass Belief Systems in Latin America”. Citizen Views of Democracy in Latin America. Ed. Roderic Ai Camp. The United States: University of Pittsburgh Press. 2750..

, 2003. El Votante Mexicano. México: Fondo de Cultura Económica.

, 2006. “Changing Ideological Dimension of Party Competition in Mexico, 19902006”. Presented at the Annual Meeting of the American Political Science Association (APSA).

, 2009. La Decisión Electoral. Votantes, Partidos y Democracia en México. México: Miguel Ángel Porrúa.

Nadeau, Niemi et al. 1999. “Elite Economic Forecasts, Economic News, Mass Economic Judgments, and Presidential Approval.” The Journal of Politics, Vol. 61, No. 1: 109135.

Nef, Jorge. 2001. “Government Corruption in Latin America”. Where Corruption Lives . The United States of America: Kumarian Press.

Nerrebi, Claude and Esteban F. Klor. 2007. “The Impact of Terrorism on Voters’ Prefernces”. Presented at the Annual Meeting of MPSA National Conference.

Neustadt, Richard 1991. Presidential Power and the Modern Presidents. The United States of America: The Free Press.

Newman, Brian. 2002. “Bill Clinton’s Approval Ratings: The More Thinks Chage, the More They Stay the Same”. Political Research Quarterly, Vol. 55, No. 4: 781804.

Nicholson, Stephen, Gary M. Segura, and Nathan D. Woods. 2002. “Presidential Approval and the Mixed Blessing of Divided Government”. The Journal of Politics, Vol. 64, No. 3 (August): 701720.

Ostrom, Charles and Dennis M. Simon. 1985. “Promise and Performance: a Dynamic Model of Presidential Popularity”. American Political Science Review. Vol. 79, No. 2: 334358.

, 1989. “The Impact of Televised Speeches Foreign Travel on Presidential Approval”, The Public Opinion Quarterly, Vol. 53, No. 1: 5882.

Pritchard Anita. 1983. “Presidents Do Influence Voting in the U.S. Congress”. Legislative Studies Quarterly, Vol. 8, No. 4: 691711.

Przeworski, Adam. 1991. Democracy and the Market. New York: Cambridge University Press.

Ragsdale, Lyn. 1987. “Presidential Speechmaking and the Public Audience 1987”. The Journal of Politics, Vol. 49, No. 3: 704736.

Rudolph, Thomas. 2003. “Who’s Responsible for the Economy”. American Journal of Political Science, Vol. 47. No. 4: 698713.

Schumpeter, Joseph A. 1976. Capitalism, Socialism and Democracy. New York: HarperPerennial.

Thomas, Dan, Lee Sigelman and Larry Baas. 1984. “Public Evaluations of the President: Policy, Partisan, and ‘Personal’ Determinants”. Political Psychology, Vol. 5, No. 4: 531542.

Villarreal, Andrés. 1999. “Public Opinion of the Economy and the President among Mexico City Residents: The Salinas Sexenio”. Latin American Research Review , Vol. 34, No. 2: 132151.

Weyland, Kurt. 2000. “A Paradox of Success? Determinants of Political Support for President Fujimori”. International Studies Quarterly, Vol. 44, No. 3 (September): 481502.

Woldenberg, José, Ricardo Becerra, and Pedro Salazar. 2000. La Mecánica del Cambio en México. México: Cal y Arena

Zepeda, Lecuona. 2007. “Criminal Investigation and the Subversion of the Principles of the Justice System in Mexico”. Reforming the Administration of Justice in Mexico . Ed. Wayne Cornelius and David Shirk: University of Notre Dame Press: 133152.

Zhong, Yang Jie Chen and John Scheb. 1998. “Mass Political Culture in Beijing”. Asian Survey, Vol. 38. No. 8: 763783.

1.10 Web Sites

Banco de Información para la Investigación Aplicada en Ciencias Sociales (BIIACS) http://www.biiacs.cide.edu

Cámara de Diputados, Chamber of Deputies in Mexico, http://www.camaradediputados.gob.mx

Instituto Federal Electora (IFE), Electoral Federal Institute, http://www.ife.org.mx

Periódico Reforma, http:// www.reforma.com

2 STATELEVEL PRESIDETIAL APPROVAL I MEXICO

2.1 Introduction

Mexican citizens evaluate the president differently after the 2000 democratic transition at the individual level. In the previous chapter, I used individual survey data to state that perceptions of crime and corruption have stronger effects after the 2000 presidential election, the turning point of the democratic transition in Mexico. After 2000, citizens have displayed more confidence in the president’s determination to address crime and corruption. In this chapter, I argue that Mexican voters evaluate the president differently after the democratic transition not only at the individual but also at the state level.

Scholars (Buendía 1996; Davis and Langley 1995; Magaloni 2006;

Villarreal 1999) have showed that Mexican presidents (Carlos Salinas, 1988

1994; Ernesto Zedillo, 19942000; Vicente Fox, 20002006) have enjoyed high levels of approval rating across the time. However, these studies implicitly suggest that presidential approval is homogeneous in Mexico, ignoring that the extreme socioeconomic conditions across the states can create different contexts in which citizens evaluate the president. In 2005, according to figures provided by

85 86

the INEGI (the National Institute of Statistics, Geography and Informatics 21 ) the number of unemployed people in the Federal District (D.F.) was almost 44 times bigger than that of Baja California Sur. Moreover, while the number of convicted criminals was huge in 2005 in some states such as the D.F. (20,055), Baja

California (10,821), Tamaulipas (10,127), in others the number of convicted criminals was in some way modest. This was the case of Campeche (773),

Tlaxcala (942) and Baja California Sur (1078). 22 This extreme variation of social and economic conditions across the states may encourage citizens to pay attention to state factors when they evaluate the president. Voters may use a closer reference than that of the national context when they approve of the president.

The American literature shows that state factors affect presidential approval (Beyle, Niemi, and Sigelman 2002; Cohen 2003; Cohen and Powell

2005). Beyle, Niemi, and Sigelman (2002) find that presidential approval levels widely vary from state to state in the U.S. Cohen and Powell (2005) argue that state partisanship and statelevel economic conditions affect presidential approval across the states. However, the necessary question for this chapter is whether sub national conditions have the same impact on statelevel presidential approval in different institutional contexts, a semiauthoritarian regime and an emerging democracy. To study statelevel presidential approval in Mexico can improve our

21 www.inegi.gob.mx 22 Justice in Mexico Crime Statistics Database, prepared by the TransBorder Institute at the University of San Diego, data facilitated by Professor David A. Shirk.

understanding of citizens’ approval of the president in comparative politics. This chapter analyzes whether citizens’ perceptions of political and economic problems, aggregated at the state level, and subnational conditions such as the level of crime, unemployment, and inflation by state have different effects after the democratic transition. Theoretical reasons suggest that state factors produce different effects after the democratic transition.

During the PRI hegemony, Mexican citizens either blamed or rewarded the president for bad or good results in addressing both political and economic problems. The reason was that the PRI controlled not only the federal government but also the majority of the governorships. It was until 1989 when Ernesto Ruffo, the PAN candidate (National Action Party) in Baja California, was the first non

PRI governor in office since the Mexican Revolution. With almost all the governorships controlled by the ruling party, the PRI, state conditions had strong effects on statelevel presidential approval. During the PRI regime the federal government and the subnational governments were somewhat nested, and there was a main and easily “identifiable” responsible for negative political and economic outcomes in a state, the President. In contrast, with more members of the opposition as state executives in different governorships, it was more difficult for citizens either to blame or to reward the president for bad or good outcomes in a state after the democratic transition. In this chapter, I argue that subnational

conditions and voters’ perceptions of crime, unemployment, and inflation began to have fewer effects on statelevel presidential approval after the PRI regime 23 .

This chapter has two main objectives: (1) To analyze presidential approval across the states in Mexico. (2) To show that at the state level Mexican citizens evaluate the president differently after the transition.

This chapter is divided as follows. The first section analyzes the main political changes in state politics in Mexico during my period of study, both before and after the transition. The second section examines both the variation of socioeconomic characteristics and presidential approval across the states of

Mexico. Third, I formulate my hypotheses, and present the data and methodology of this chapter. Finally, I present an econometric analysis, and discuss the implications of this chapter.

2.2 The ew Political Geography of State Politics in Mexico

The political geography of Mexico has changed significantly in the last two decades. In this section I pay attention to the main political changes across the states of Mexico. First, this section describes the advancement of the opposition in the political arena with an emphasis on statepolitics in Mexico.

Second, I study the implications of these political changes and present some

23 In the section of Data and Method, I explain why I do not include the variable corruption in the final results of this chapter. However, I present some models in the appendix 2.2 that include both citizens’ perceptions of corruption and objective indicators of corruption.

empirical evidence, survey data, to construct the rationale of this chapter. Third, this section analyzes the main factors that explain these changes in the political geography of state politics in Mexico. I use the state of Baja California and the

D.F. as two paradigmatic cases.

The opposition has spread its influence across the states of Mexico in the last decades. During many decades of the predemocratic period, the PRI could maintain an absolute domain across the states. The ruling party controlled all the governorships. It was until 1989 when the opposition, the PAN, won a governorship for the first time. Figure 2.1 shows the advancement of the opposition across the states of Mexico before and after the 2000 transition.

100 89 90 80 66 70 60 53 47 50 Percentage 40 34 30 20 11 10 0

I n n o RI o PR PRI P siti po positi p p O Opposition O 1994 2000 2006

Figure 2.1: Political Parties across the States of Mexico 19942006. Figure Done by the Author. The Data Come from Domínguez and Lawson (2004, 26)

The advancement of the opposition at the state levels reveals that Mexico is a completely different country politically speaking. While during many decades of the PRI regime, citizens lived in a singlepartysystem, in which the PRI controlled the federal and the subnational governments, in the last decades

Mexican voters have begun to live in a new political geography in which there are different political actors.

Figure 2.2 shows the presence of the three main political parties, the PRI, the PAN, and the PRD (Party of the Democratic Revolution) across the states of

Mexico. It is clear that Mexico is a consolidated threepartysystem in which the

PRI keeps influence across the states, the PAN is competitive in the north region of the country, and the PRD in the south. As Figure 2.2 indicates , in the North and West areas –where the Catholic Church remains influential and local business elites have long opposed centralized control from Mexico City –the PAN constitutes the main opposition force. In the poorer, more rural South, the leftist party the PRD remains the principal challenge to PRI rule. Mexico comprises a pair of essentially twoparty systems (PRDPRI in the South, PANPRI in the

North) (Lawson 1999, 149).

Figure 2.2 State Governments of Mexico Controlled by the Three Main Political Parties: PA, PRD and PRI 24 (2006)

24 Patricio Patrón Laviada governed Yucatán in 2006 (the states in vertical lines in Figure 2.2). Laviada was the first governor of the state of Yucatán who emanated from the PAN, through a coalition with the PRD, the PT (Labor Party) and the PVEM (The Green Party of Mexico)

The advancement of the opposition in the political arena at the state level constitutes one of the main features of the new Mexican political geography. This chapter examines the main factors that explain these changes in Mexican politics.

I focus my attention on Baja California and the D.F.

National and local factors explain this process of democratization and the advancement of the opposition (the PAN and the PRD) across the states in the last decades. National factors involve social, economic, and political components that have a general impact in all the federal entities of Mexico. Local factors refer to the particularity of each subnational context.

Three main national factors explain the democratization process and the advancement of the opposition at the subnational level in Mexico. First, the restructuring of the Mexican economy that began in Mexico in the 1980s seriously affected the political relationship between the PRI government and key sectors such as unions and peasants. The PRI could not provide private benefits to these groups, as the ruling party used to do. In these circumstances there were not reasons to support the PRI regime for many Mexicans. Second, as part of the economic reform, the PRI government had to recognize some electoral victories of the opposition at the statelevel in order to appear democratic in the eyes of their trading partners (Pastor, 1990). Third, the controversial 1988 presidential election gave the victory to the PRI candidate Carlos Salinas (19881994); however, Salinas was perceived as an illegitimate president at least for some social and political sectors in Mexico. In order to get some “recognition” inside

the country from some political and social groups, Salinas had to recognize the presence and electoral victories of the opposition across the states (Eisenstadt

2004).

Local factors refer to the specific political context that allowed the advancement of the opposition across the states in Mexico. This chapter analyzes in detail two paradigmatic cases: Baja California (the first Panista governorship) and the D.F. (the first Perredista victory at the subnational level).

Three main local factors explain the victory of Ernesto Ruffo as governor of Baja California (the first non priista governor): the long history of the PAN in the state, the dynamism of the state’s economy and society, and the occurrence of splits within the traditional alliances (of the PRI) among state elites (Rodríguez and Ward 1994, 32).

In Baja California, as in several other border states, the PAN has a long history of being the principal opposition. Since the 1960s, the PAN showed some electoral presence in Baja California. After four decades of organizing the opposition in the state, the PAN was the party with the most recognition among the electorate and had firmly established links with the business community and the urban middle class (Rodríguez and Ward 1994, 33). The second factor, the dynamism of the Baja California’s economy and society, made this state share some of the main ideological values of the PAN such as the liberalization of the economy, competition and free market. The third factor is the occurrence of a split within the PRI government and the business community’s reaction of Baja

California because of the bank nationalization of 1982. The business community’s dissatisfaction with the PRI translated into increased levels of electoral support for the PAN.

After the 1989 elections in Baja California, the PAN could spread its influence mainly in the north region of Mexico. In 2006, as Figure 2.2 shows the

PAN stayed in office in Baja California 25 (Eugenio Elorduy is the governor), governed San Luis Potosí (by Marcelo de los Santos), Jalisco (by Franciso Javier

Ramírez), Guanajuato (by Juan Manuel Oliva), Querétaro (by Franciso Garrido),

Aguascalientes (by Luis Armando Reynoso). Moreover, in the central region of

Mexico the PAN controlled Morelos (governed by Marco Antonio Adame) and

Tlaxcala 26 (governed by Héctor Ortiz Ortiz).

In the case of the Federal District, two main reasons explain the victory of

Cuauhtémoc Cárdenas, the PRD candidate to Head of Government of the D.F. in the 1997 elections in Mexico City, the undecided voters’ support in favor of

Cárdenas in the last months previous the election and citizens’ disappointment with the former PRI governments.

While Cárdenas increased his vote share by 26 points in the last four months of the campaign, the PRI and the PAN, by contrast, lost a roughly equal portion of voters each (12 and 14 percent respectively (Lawson 1999, 151). Who switched to the PRD? In a panel study, Lawson (1999) shows that three categories

25 In coalition with the PVEM. 26 In coalition with the PT.

of voters switched to Cárdenas: First, 4 percent of undecided voters supported

Cárdenas. Second, PRI supporters changed their minds in the course of the campaign and voted for the PRD (about 8 percent). Finally, 10 percent of

Panistas benefited Cárdenas with their voters.

The second reason that explains Cardenas’s victory refers to Mexican voters’ dissatisfaction with the former Priista governments that implemented unpopular economic reforms during the 1980s and the beginning of the 1990s.

Moreover, the ruling party could not address citizens’ concerns about insecurity.

In a panelstudyanalysis Lawson (1999, 152) reveals that opposition to neoliberal economic policies, lowerclass status, perceptions of rising street crime, disenchantment with existing institutions, were all associated with improved opinions of the PRD.

Since the 1997 elections in Mexico City, the PRD has been able to spread its influence mainly in the south of Mexico. In the previous pages, Figure 2.2 shows that in 2006, the PRD governed Michoacán (by Lázaro Cárdenas Batel);

Guerrero (by Zeferino Torrreblanca); Chiapas 27 (by Juan Sabines Guerrero). All of these states are in the south of Mexico. Moreover, the PRD governed Zacatecas

(by Amalia García) and Baja California Sur (by Narciso Agúndez) in the north, and stayed in office in the D.F. (with Marcelo Ebrad as Head of Government)

This second chapter pays attention to this question: How have changes in state Mexican politics affected citizens’ perceptions of the president across the

27 The PRD won the lection in coalition with the PT and Convergencia.

federal entities? Cohen (2003), and Cohen and Powell (2005) claim that state factors affect statelevel presidential approval. However, this statement can neglect that in a context different than the U.S. the influence of subnational conditions on approval may not be the same before and after a democratic transition. During the PRI regime, the ruling party had an almost absolute domain, controlling both the federal government and the subnational governments. In the last two decades, the opposition has advanced in the political arena. In 2006 the opposition controlled almost 50 percent of the governorships. These differences in the political contexts suggest that state factors may affect statelevel presidential approval differently after the transition.

Scholars (CanesWrone 2006; Kernell 1978; 1997; Page 1978; Page and

Bouton 2006) have argued that presidential approval is very sensitive to people’s perceptions about the president’s responsibility to address political and economic issues. At the same time, one of the main findings in presidentialapproval analysis is that in a more complex political context citizens experience a hard time to blame the president for negative outcomes (Hansen 1999; Nicholson, Segura, and Woods 2002; Rudolph 2003). According to Nicholson, Segura and Woods

(2002) and Rudolph (2003), in contexts of unified governments citizens easily blame the president for a negative outcome in the economy. Only one political party can be held responsible for unsatisfactory performance. In contrast, in a divided government, voters tend to divide the responsibility of this negative outcome into the executive and the legislative branches. Divided Government

provides ambiguous and conflicting information about which level of government, either federal or state, to hold accountable for government performance. Presidents are less likely to be held accountable for government’s failures when the state governor’s party is different than that of the president’s.

These findings of American politics shed light to construct the rationale of this chapter. While during the PRI regime, the political context was simple with the ruling party controlling the federal government and the subnational governments, after the predemocratic period citizens have begun to approve of the president in a more complex political framework. With the presence of the opposition controlling almost 50 percent of the governorships across the states in

Mexico, at the aggregate level it has been more difficult for Mexican citizens to blame the president for a negative outcome or to reward him for a good performance. Who may be responsible for an increase either in crime or unemployment in a state, the president or the governor?

Figure 2.3 offers empirical evidence for the rationale of this chapter.

While during the PRI regime citizens had a clear tendency to blame the President for negative economic outcomes (45 percent), in the postPRI era Mexican voters began to alter their political behavior. Only 23 percent of people identified the president as the main responsible for the economic crisis in 2004.

90 77 80 70 58 60 50 45 Percentage 40

Percentage 30 23 20 10 0 The President Other Actors The President Other Actors PRI Regime 1995 Post-PRI Era 2004

Figure 2.3: Citizens’ Perceptions of the Political Actors’ Responsibility for the Economic Crisis. Figure Done by the Author. The Data Come from the BIIACS, www.biiacs.cide.edu . The question asked was: “Who do you think is responsible for the current economic crisis?”

During the predemocratic period, Mexican citizens considered the

President as the main responsible for both positive and negative outcomes. This was a consequence of the almost absolute domain of the PRI that controlled the federal government and the subnational governments. In this context, during the

PRI regime, when the federal government and the subnational governments were nested, state factors had very strong effects on statelevel presidential approval.

An increase in a federal entity either in crime or in unemployment affected citizens’ evaluation of the president across the states. In contrast, in the last decades other political actors became important in the political arena. The federal government controlled by the president began to share its responsibility with other political actors. The federal government and the subnational governments were

100

not nested anymore across the states. After the postPRI era, a more complex political context emerged with more political actors, and citizens had more problems to identify the president as the main responsible for local problems.

2.3. Socioeconomic Conditions across the States in Mexico

The fist chapter of my dissertation showed that crime, unemployment, corruption, and inflation affect presidential approval in Mexico. Some indicators suggest that the salience of these issues may not be the same across the states.

Objective indicators such as the level of unemployment and inflation by state suggest that economic conditions vary across the federal entities of Mexico.

Figure 2.4 shows that the variation of unemployment across the states of Mexico is significant. While the level of unemployment in Aguascalientes in 2005 was 4.9 percent, in Baja California this figure was in some way modest, 1.4. 28

Furthermore, Figure 2.5 indicates that the variation of unemployment was not the same across the states of Mexico. In the D.F. and Nuevo León, two of the entities with the highest level of unemployment, the variations were bigger than those in

Baja California and Yucatán.

28 Figure done by the author with data collected from the ENOE, Encuesta acional de Ocupación y Empleo, the National Survey of Employment and Occupation, organized by the INEGI (www.inegi.gob.mx ). The INEGI data are based on four national surveys conducted in 2005.

101

3 Unemployment in a State

Percentage 2

0 1 3 5 7 9 11 13 15 17 19 21 23 25 27 29 31 States

Figure 2.4: 29 Level of Unemployment across the States of Mexico in 2005. Figure Done by the Author. The Data Come from the National Institute of Statistics Geography and Informatics (INEGI)

0.09 0.08 0.07 0.06 B. California 0.05 Yucatan 0.04 N. Leon

Percentage 0.03 D.F. 0.02 0.01 0 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 Years

Figure 2.5: Variation of Unemployment in Four States of Mexico 19942005. Figure Done by the Author with Data from the INEGI

29 Aguascalientes (1), Baja California (2), Baja California Sur (3), Campeche (4), Coahuila (5), Colima (6), Chiapas (7), Chihuahua (8), Distrito Federal (9), Durango (10), Guanajuato (11), Guerrero (12), Hidalgo (13), Jalisco (14), Estado de México (15), Michoacán (16), Morelos (17), Nayarit (18), Nuevo León (19), Oaxaca (20). Puebla (21), Querétaro (22), Quintana Roo (23), San Luis Potosí (24), Sinaloa (25), Sonora (26), Tabasco (27), Tamaulipas (28), Tlaxcala (29), Veracruz (30), Yucatán (31), Zacatecas (32).

102

Given that inflation has a national effect across the states, its variation is not extreme as that of unemployment. Figure 2.6 reveals that the level of inflation across the federal entities of Mexico was not homogeneous in 2005.

0.06

0.05

0.04

0.03 Inflation Level by State

Percentage 0.02

0.01

0 1 3 5 7 9 11 13 15 17 19 21 23 25 27 29 31 States

Figure 2.6: 30 Level of Inflation in Mexico across the States in 2005 Source: Prepared by the author with Data from the Banco de México 2005

Both economic conditions and political problems vary across the states of

Mexico. The variation of the level of crime has become extreme in the last years.

In Mexico in the last decades, citizens have consistently ranked crime as one of their main concerns. Figure 2.7 shows the variation of perceptions of crime as the most important problem. Although the authorities have attempted to address crime and insecurity, citizens experience enormous frustration with the inability

30 Vide footnote 29 to match the numbers of the states with their names.

103

of Mexican government and law enforcement officials to effectively guarantee public security. Even after the 2000 elections –the watershed event that confirmed

Mexico’s transition to democracy the public’s general perception is that lawlessness prevails because government and law enforcement officials remain indifferent and inept (Cornelius and Shirk 2007).

25 22 20 20.8 18.46 17.11 15 15.3 12.11 Perceptions of Crime 10 Percentage 6.7 5 2.05 0 1995 1996 1997 1998 1999 2000 2003 2006 Years

Figure 2.7: Variations of Perceptions of Crime as the Main Concern for Mexican Citizens from 1995 to 2006. Figure Done by the Author. The Data Come from the BIIACS, www.biiacs.cide.edu , (1995, 1996, 1997, 1998, 1999, 2000), Diario Reforma www.reforma.com (September 2003), Parametría www.parametria.com.mx (2006) .

Figure 2.8 reveals that although crime is a national concern by Mexican citizens, its great variation across the states suggests that voters may have different perceptions about crime at the subnational level. For example, while the number of convicted criminals per 100 inhabitants in 2005 was dramatic in some states such as Baja California (.38) and Tamaulipas (.34), in others federal entities the number of convicted criminals per 100 inhabitants is somewhat modest. This

104

is the case of Guerrero (.05) and Hidalgo (.06). Moreover, Figure 2.9 indicates that the annual crime rate is not the same across the federal entities of Mexico. In

Tamaulipas and Baja California, two of the entities with the highest levels of crime, their variations of crime rate are bigger than those of Hidalgo and

Guerrero.

0.4 0.35

0.3

0.25

0.2 Convicted Criminals 0.15 0.1

100 Inhabitants Inhabitants by 100 State 0.05

Num. Criminals of Convicted per 0 1 3 5 7 9 11 13 15 17 19 21 23 25 27 29 31 States

Figure 2.8 31 Convicted Criminals by State in Mexico (2005) Source: Justice in Mexico Crime Statistics Database, prepared by the TransBorder Institute at the University of San Diego

31 Vide. Footnote 29 to match the numbers of the states with their names.

105

0.45 0.4 0.35 0.3 Hidlago 0.25 Guerrero 0.2 Tamaulipas

Inhabitants 0.15 B. California 0.1 0.05 0 Num. Criminals of Convicted for 100 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 Years

Figure 2.9: Variations of Crime in Four States of Mexico19942005. Figure done by the author with data of the Justice in Mexico Crime Statistics Database, prepared by the TransBorder Institute at the University of San Diego

I have explained the variation of the main economic and sociopolitical conditions at the subnational level in Mexico. The evidence shows that there is an extreme variation across the federal entities. States are different from both an economic and a political perspective.

106

2.4 The Variation of Presidential Approval across the States

Presidential approval has significantly varied across the states in Mexico.

One example of this variation, as Figure 2.10 shows, is Vicente Fox’s approval in

2005. At that time, Fox’s approval reached a positive rating of 59 percent.

However, this national level of approval did not show the variation of approval across the Mexican states. While in Sonora, presidential approval was almost 80 percent, in the D.F. and in Tabasco, Fox had very poor approval rates of 30 and

28.8 percent, respectively (Mitofsky, 2005).

90 80 70 60 50 Approval 2005 40

Percentage 30 20 10 0 States-Level Presidential Approval States-Level

1 3 5 7 9 1 7 1 7 1 1 13 15 1 19 2 23 25 2 29 3 States

Figure 2.10 32 : Presidential Approval in Mexico by State 2005. Figure Done by the Author. Source: Consulta Mitofsky( 2005). The question asked was: “Do you approve the way the President is handling his job?”

32 Vide. Footnote 29.

107

Mitofsky 33 (2005) shows that presidential approval in Mexico is clearly heterogeneous across the 32 federal entities. These results reveal 5 different levels of presidential approval in Mexico: very high approval, high approval, medium approval, low approval, and very low approval.

As Table 2.1 reveals, in the very high approval group, Fox overcame 70 percent of approval in 2005. This groups includes four states, three of them in the

North of Mexico (Sonora, 79.1, Baja California, 71.7, and Durango 71.2 percent) and one in the South (Campeche, 72.3 percent).

In the second group, the high approval group, Fox’s approval varied between almost 70 and more than 60 percent across 11 states. It is possible to identify 68.5 (San Luis Potosí), 68 (Colima), 67.9 (Coahuila), 66 (Baja California

Sur), 64.6 (Guanajuato), 64.5 (Quintana Roo), 64.4 (Aguascalientes and Nayarit),

63.3 (Zacatecas), and 64.1 percent (Tamaulipas).

Table 2.1 shows that most of the states are concentrated in the medium approval group, 12 states in total. Fox enjoyed the following levels of approval in these states: 60.8 (Chihuahua), 59.2 (Puebla), 58.1 (Oaxaca), 57.8 (Hidalgo), 57.4

(Michoacán), 57.3(Chiapas), 57.1 (Nuevo León), 57.1 (Tlaxcala) 57 (Guerrero),

56.9 (Jalisco), 56.3 (Queretaro), and 55.9 percent Veracruz. As the level of approval decreases, the presence of states from the north also diminishes. In this

33 Mitofsky did not indicate either if the survey was representative at the state level or if they used a poststratification technique.

108

group, there are only two states of the north region of Mexico, Chihuahua and

Nuevo León. The rest of the states come from either the center or the south of

Mexico.

The low approval group includes four states: 53.9 (Yucatán), 52 (the State of México), 51.7 (Sinaloa), and 47.2 percent (Morelos). Only Sinaloa belongs to the north region of Mexico.

In the very low approval group there are only two federal entities with the following approval rating: 30 and 28.8 percent in the D.F. and Tabasco, respectively.

109

Table 2.1 Percentage of Presidential Approval across the States 2005 The Data Come from www.consulta.com.mx

110

Very High Level Approval Group

Sonora 79.1

Campeche 72.3

Baja California 71.7

Durango 71.2

High Level Approval Group

San Luis Potosí 68.5

Colima 68

Coahuila 67.9

Baja California Sur 66

Guanajuato 64.6

High Level Approval Group, Continued

Quintana Roo 64.5

Aguascalientes 64.4

Nayarit 64.4

Tamaulipas 64.1

Zacatecas 63.3

111

Table 2.1, Continued

Medium Approval Group

Chihuahua 60.8

Puebla 59.2

Oaxaca 58.1

Hidalgo 57.8

Michoacán 57.4

Chiapas 57.3

Nuevo León 57.1

Tlaxcala 57.1

Guerrero 57

Jalisco 56.9

Querétaro 56.3

Veracruz 55.9

Low Approval Group

Yucatán 53.9

México 52

Sinaloa 51.7

Morelos 47.2

112

Table 2.1, Continued

Very Low Approval Group

Distrito Federal 30

Tabasco 28.8

Presidential Approval at the National 59.1 Level Source: Consulta Mitofsky (2005), “Aprobación Presidencial por Entidad Federativa en 2005”.

In some states partisanship encourages citizens to take positions about the president and his performance. Figure 2.11 indicates that on the whole in states governed by the PAN, the president was highly approved (the approval rate was

67.1 percent)

70 67.1 61.4 60 51.1 50

40 Approval by Party Lenses

30 Percentage

0 PAN PRD PRI

Figure 2.11: Presidential Approval (2005) by Party Lenses (States Governed by PA, PRD and PRI) Source: Consulta Mitofsky( 2005).

113

However, the empirical evidence also shows that in other states the president was evaluated beyond citizens’ party lenses. For example, in Sonora

(79.1 percent) and Campeche (72.3 percent), states governed by the PRI, the president enjoyed the highest approval. Moreover, in states such as Morelos

(47.2), Yucatán (53.9), Queretaro (56.3) and Jalisco (56.9 percent), Fox did not reach a very high level of approval even though these states were governed by the

PAN.

Party lenses explain part of the story of presidential approval in Mexico at the state level, but not the complete story. Moreover, this variation of presidential approval across the states cannot be understood by national factors because their effects spread across all the country. In Mexico, a nation characterized by extreme variations of economic and social conditions, different levels of unemployment and crime, subnational contexts encourage citizens to evaluate the president by state factors.

2.5 Hypotheses

The empirical evidence offered in this chapter shows that in a simple political context citizens easily blame the president for negative outcomes. In contrast, when the political context turns complex voters have difficulties to blame the president for bad results. In the Mexican case, during the pre democratic period, the PRI controlled almost all offices. The public easily identified the president as the main responsible. In contrast, after the PRI regime with more members of the opposition as state executives in different

114

governorships, the president’s responsibility became murkier. Subnational conditions began to have fewer effects on statelevel presidential approval.

My general hypothesis is that after the transition citizens began to interact with a complex political context, with more members of the opposition as state executives, and had difficulties to identify the president as the main responsible for negative outcomes. Therefore, the effect of subnational conditions on state level presidential approval became weaker. Therefore, my expectations are twofold:

H1: The effects of objective indicators, crime, unemployment, and inflation by state on statelevel presidential approval during the predemocratic period are stronger than those in the new Mexican democracy. During the PRI regime, the ruling party controlled almost all the office. The public easily identified both the PRI and the president as the main responsible for negative outcomes. After the PRI era, citizens began to interact with a more complex political context, the opposition controlled almost 50 percent of the governorships, and Mexican people had difficulties to identify the president as the main responsible for economic and political problems.

H2: Given that the effects of subnational conditions on statelevel presidential approval lost power after the PRI regime, the impact of citizens’ perceptions, aggregated at the state level, of crime, unemployment and inflation, on approval in the democratic period turned weaker than those during the PRI era. During the predemocratic period the federal government and the sub

115

national governments were somewhat nested. The ruling party controlled most of the governorships. Thus, Mexican citizens easily blamed the president for negative outcomes in a state.

2.6 Data and Method

The lack of statesurvey data focused on presidential approval has been a problem to analyze the variation of citizens’ evaluation of the president across the sates (Cohen, 1996). To study statelevelpresidential approval has been challenging even in the U.S. The 1996 presidential election created an explosion in polling and many state polls became available. In the last years, scholars

(Beyle, Niemi, and Sigelman 2002; Cohen 2003; Cohen and Powell 2005) have used different techniques to analyze the variation of presidential approval across the states in the U.S.

Cohen (1998) develops a forecasting model to predict presidential election results across the states. His analysis (1998) uses statelevel public opinion polls.

The main methodological limitation of this study was that all the observations were gathered in 1996. Using better data than those of Cohen’s (1998), Cohen and

Powell (2005) analyze the variation of presidential approval in the U.S. across the states. These scholars use presidential job approval data from 1981 to 1999.

However, the problem of their methodology is that only few states were polled frequently. Cohen and Powell (2005) employ a crosssection study, and omit the variation of approval across the time.

116

In Mexico the lack of statesurvey data focused on presidential approval has restricted approval analysis. Scholars have chosen to conduct analyses either at the individual (Davis and Langley 1995; Villarreal 1999; Domínguez and

McCann 1995) or at the national level (Buendía, 1996; Magaloni, 2006). This chapter uses individual survey data aggregated at the state level to analyze the variation of presidential approval across the states.

The forty three surveys on which this study is based were organized by the

Mexican Office of the Presidency (MOP). Thirty three of them are national surveys. The national survey conducted in December 2004 was complemented by ten state surveys carried out in Baja California, Chiapas, Chihuahua, Coahuila,

Estado de México, Nuevo León, Puebla, San Luis, Tamaulipas, and Veracruz.

The surveys carried out from 1994 to 2000 were conducted by Opinion

Profesional S.A. de C.V., a private company specializing in policy polling and focus groups. Field workers were not informed of the client’s identity; they identified themselves to respondents only as employees of Opinion Profesional . It is not possible to know if the surveys organized by the Office of the Presidency from 2001 to 2005 were conducted by Opinion Profesional 34 .

34 The survey data from 2001 to 2005, organized by the Mexican Office of the Presidency, come from the CIDE, Centro de Investigación y Docencia Económicas, Center for Research and Teaching in Mexico. The CIDE got these data through the IFAI, Instituto Federal de Acceso a la Información , the Federal Institute of Access to Information. The IFAI is an agency responsible to facilitate access to the information of the Mexican government. One of the limitations of the surveys carried out from 2001 to 2005 is that they do not include the weight that corresponds to each observation. Although the CIDE attempted to address the limitation asking for help to the current staff of the Mexican Office of the Presidency, the CIDE could not do it because nobody of the current staff of the Mexican Office of the Presidency was part of the staff of the former administration. Moreover, the information about the sample frame and the survey design is not

117

In total the surveys contain 74,150 household interviews, from 1994 to

2005. This period covers three different presidents (Carlos Salinas de Gortari,

1994; Ernesto Zedillo, 19952000, both Priista presidents; and the Panista

Vicente Fox, 20012005). Households were located for interviews through a multistage, stratified probability sampling procedure. At each household, an interview of about 4560 minutes was conducted with one respondent, selected at random, with about an equal proportion of men and women.

I conduct a timeseriescrosssectional analysis to measure statelevel presidential approval in Mexico from 1994 to 2005. My original purpose was to examine the variation presidential approval from 1989 to 2006. However, three limitations restrict my study. First, the questions necessary to measure citizens’ perceptions of political and economic problems were not included in all the surveys. Second, data of some of the state indicators before 1994 were not regularly reported. This is the case of state GDP. INEGI reports data of state

GDP from 1993 to 2009; however, before 1993 data of the state GDP were published every five years. Third, I do not have survey data of presidential approval in 2006 to aggregate at the state level. Despite these restrictions, survey data of presidential approval from 1994 to 2005 have not been systematically analyzed before. Moreover, it is the first time that an analysis is focused on presidential approval in Mexico at the state level.

available. The former staff of the Mexican Office of the Presidency did not give this information to the IFAI that facilitated the data to the CIDE.

118

The main limitation of my analysis is that given the lack of statesurvey data focused on presidential approval, I aggregate individual survey data at the state level. To reduce the effects of this limitation in this study, two kinds of post stratification techniques are used. First, the surveys carried out from 1994 to 2000 used stratified random sampling, and include the weight that corresponds to each observation according to the socioeconomical characteristics of a particular stratum. I call these weights, Weight One. I use Weight One to make the distribution of the population of the states included in the surveys identical to the distribution of the real population of each state of Mexico from 1994 to 2000.

Weight One allows me to calculate aggregate averages in a state for a given survey. To get the average level of approval of the president in a state, I calculate the weighted average of what each respondent thinks, with respondents who have higher weights, those who come from strata less likely to be surveyed, and with respondents who have lower weights, those who come from strata more likely to be surveyed. After I calculate the weighted average of what each respondent thinks, I aggregate these data at the state level. 35 I use the number of respondents who were surveyed in a state by year, as a second poststratification technique that I call Weight Two. I used Weight two to run a “weighted regression”, and to reflect the fact that I am more confident in the aggregate total for the states with

35 Some surveys include the relative weight but other the expansion weight. When it was necessary, I converted the expansion weight into the relative weight in order to get more accurate results.

119

more respondents. 36 I use the same procedure to calculate citizens’ perceptions of the president’s policies of to address crime, unemployment, and inflation at the aggregate level. These techniques are used to calculate state partisanship. The weight employed in the surveys organized from 2001 to 2005 is not included in the database. Therefore, it is impossible to know the specific weight that corresponds to each specific observation, and use the first poststratification technique, Weight One. However, it is likely that these surveys did not use any kind of poststratification technique. As you can see in the Appendix 2.1 of this chapter the distributions of most states included in these surveys are similar to those of the real population across the states of Mexico from 2001 to 2005. Thus,

I use only the second poststratification technique, Weight Two, to analyze surveys organized from 2001 to 2005.

To measure my dependent variable (statelevel approval of the President),

I aggregate individual survey data at the state level. I use the standard “Do you approve the way the president is handling his hob?” Some of the surveys have a wide range of options: people could respond “totally approve”, “partly approve”,

“partly disapprove” and “totally disapprove”. Others have a narrow range options:

“approve” or “disapprove”. I recoded the former surveys to correspond with the latter ones in order to increase my number of observations.

As explanatory variables, I use state objective indicators and citizens’ perceptions of the crime, unemployment, and inflation aggregated at the state

36 Stata allows me to calculate both weights.

120

level. I do not include corruption in this analysis because of one methodological concern. It is not mandatory that the states report the number of complaints against the civil service involved with corruption. This limitation and the modest number of observations in the models may affect the analysis. Thus, the variables perceptions of rising corruption and objective indicators of corruption were taken out of the study. However, I include both variables in the Appendix 2.2. The results on the whole are similar. To measure objective indicators, I employ state

GDP per capita, the levels of unemployment, inflation, and crime by state. The data of unemployment were gathered from Anexo Estadístico de la Presidencia de la República. Sexto Informe de Gobierno 2006, the EEU (the Encuesta acional de Empleo Urbano , the National Survey of Urban Employment )37 , the EE

(Encuesta acional de Empleo, the National Survey of Employment), and the

ENOE ( Encuesta acional de Ocupación y Empleo, the National Survey of

Employment and Occupation) these surveys were organized by the INEGI

(Institutio acional de Geografía e Informática, the National Institute of

Statistics, Geography and Informatics). The data of crime (number of convicted criminals by state) comes from the TransBorder Institute at the University of San

Diego. To employ a better measure in my analysis, I calculate the number of convicted criminals for every 100 inhabitants by state using census data of the

Mexican population. The analysis also uses a dummy variable to measure the

37 The Encuesta acional de Desempleo Urbano collects information from the main cities of each state of Mexico.

121

effect of state government controlled by a party different from the president’s. To measure state partisanship and citizen’s perceptions’ of crime, unemployment, and inflation I aggregate individual survey data at the state level. To measure state partisanship, I examine whether respondents’ party matches with the president’s party. I also control for the effects of a partition between the federal and the state government, when the state governor’s party is different than that of the president’s.

Table 2.2 Explaining StateLevel Presidential Approval in Mexico: Variables

Concept Operationalization

Dependent Variable

StateLevel Presidential Approval % of people who support the President by state each year (from 1994 to 2005). The analysis employs individual survey data aggregated at the state level. I use the standard question included in all the surveys: “Do you approve the way the president is handling his job?”

Independent Variables:

122

Table 2.2, Continued

Concept Operationalization

Lag State Approval The lag of statelevel presidential approval of the previous year.

Divided Government I expect that Divided Government is Dummy variable, 1 when the state positively associated with statelevel governor’s party is different than that of presidential approval. Divided the president’s. 0 if this is not the case. Government provides ambiguous and conflicting information about which level of government, either federal or state, to hold accountable for government performance. Presidents are less likely to be held accountable for government’s failures when the state governor’s party is different than that of the president’s. PRI (Party Control of the President) Dummy variable, 1 during the PRI Hegemony (9400). 0 after the PRI era (0105) PRI*Lag State Approval Interactive variable (the product of the variable PRI and the Lag) State GDP Per Capita The data come from the INEGI. To calculate the state GDP per capita, this chapter used the XI Censo General de Población y Vivienda 1990 , the Conteo de Población y Vivienda 1995 , the XII Censo General de Población y Vivienda 2000 , and the II Conteo de Población y Vivienda 2005

123

Table 2.2, Continued

Concept Operationalization State Crime The number of convicted criminals yearly for every 100 inhabitants by state. I collected the number of convicted criminals by state in Mexico from the Justice in Mexico Crime Statistics Database, prepared by the TransBorder Institute at the University of San Diego. This chapter used the XI Censo General de Población y Vivienda 1990 , the Conteo de Población y Vivienda 1995 , the XII Censo General de Población y Vivienda 2000 , and the II Conteo de Población y Vivienda 2005 . To conduct a more accurate analysis, for the intermediate years I estimated a projection of the population using the data of the Censos and the Conteos .

State Unemployment (Source To calculate the unemployment level by INEGI) state in Mexico, I used the ENOE, the I expect that state unemployment National Survey of Employment and and statelevel presidential Occupation, (2005) and the ENE, the approval should be negatively National Survey of Employment (1998 associated. 2004). I employed both surveys to calculate the level of unemployment from 1998 tp 2005 by state, using the economically active population (PEA) and the number of people unemployed. To calculate the level of unemployment by state before 1998, the analysis used the Anexo Estadístico de la Presidencia de la República. Sexto Informe de Gobierno 2006, which gives information of unemployment across the most important cities in Mexico. These data were collected from the ENEU, the National Survey of Urban Employment (Data of state unemployment in Mexico before 1998 does not exist). I calculated the level of unemployment by state before 1998 using the weighted average of unemployment in the main cities of a state.

124

Table 2.2, Continued

Concept Operationalization State Inflation Rather than paying attention to the I expect that state inflation should have states, the Banco de México calculates a negative effect on presidential the level of inflation of the most approval across the states. important cities. Thus, I calculated state inflation using the weighted average of the inflation of the cities that are part of a state. Perceptions of rising crime. I employed individual survey responses I predict that this variable should be aggregated at the state level. The negatively associated with statelevel analysis uses this question: “Since presidential approval. Whey citizens Salinas (Zedillo or Fox) is president, do think that crime increases, they you think that crime has increased or disapprove the president. decreased in Mexico?” 1= Decreased; 2= The same; 3= Increased Perceptions of rising unemployment. I aggregated individual survey I expect that this variable should have responses at the state level. The negative effects on presidential question used in this study is: “Since approval across the states. Perceptions Salinas (Zedillo or Fox) is president, do of rising unemployment decrease state you think that unemployment has level presidential approval. increased or decreased in Mexico?” 1= Decreased; 2= The same; 3= Increased Perceptions of rising inflation. I employed survey data aggregated at I anticipate that should be a negative the state level. The chapter used this relationship between statelevel question: “Since Salinas (Zedillo or presidential approval and this variable. Fox) is president, do you think that inflation has increased or decreased in Mexico?” 1= Decreased; 2= The same; 3= Increased State Partisanship (matching the The analysis used individual survey president’s party) data aggregated at the state level. I I predict that this variable should have employed a dummy variable : 1 if the positive effects on statelevel voter belonged to President’s party; 0 if presidential approval. the voter did not. From 1994 to 2000, I coded 1 if the voter was Priista ; from 2001 to 2005, I used 1 if the voter was Panista

125

2.7 Results

Table 2.3 presents the results of the regression analysis. Model 1 tests a general baseline model of statelevelpresidential approval from 1994 to 2005, the predemocratic era and the democratic period. The results of Model 1 show that the variable Lag State Approval had a positive significant impact of statelevel presidential approval. An increase in statelevel presidential approval was associated with an increase of approval the following year, from 1994 to 2005.

During the predemocratic period, Model 2, the Lag had the same direction, it was positive. However, after the transition, Model 3, the term of the Lag turned negative. This shift is challenging to explain:

Model 1 indicates that during these twelve years of analysis (19942005) an increase in statelevel presidential approval was associated with an increase of approval next year. However, it is likely that the states that strongly approved the priista President, Ernesto Zedillo, in 2000, did not support the new panist a executive, Vicente Fox, in 2001 and approval fell down. At the same time, the states that did not approve Zedillo in 2000 began to approve the PAN President,

Vicente Fox, in 2001. If this is true, the interactive term (the product of the Party

Control of the President, variable PRI, with the Lag State Approval) should be positive because during almost all the years there is a dominant positive relationship (an increase in statelevel presidential approval is associated with an increase of approval the next year), the only exception is the period from 2000 to

2001 in which there must be an inverse relationship (the term of the Lag should

126

be negative, an increase in statelevel presidential approval in 2000 was associated with a decrease of approval in 2001). As the rationale predicted, Model

4 shows that the interactive term (PRI X Lag State Approval) is positive.

Model 1 reveals that when the variable Divided Government takes a value of 1 (the federal government and the state government are divided. The governor’s party is different than that of the president’s) statelevel presidential approval in a state increased by .05 percentage points. When the president’s party is the same as the governor’s, citizens easily identify the president as the main actor responsible for negative outcomes. In contrast, when one party controls the federal government and other the subnational government, citizens blame for negative outcomes both the president and other political actors. Divided

Government provides ambiguous and conflicting information about which level of government, either the federal or the state, to hold accountable for government performance. Presidents are less likely to be held accountable for government’s failures when the state governor’s party is different than that of the president’s.

Model 1 also shows that state GDP per capita affected statelevel presidential approval in the direction that I expected. The higher the GDP per capita in a state, the higher statelevel presidential approval becomes. The rest of the subnational conditions did not have significant effects on approval across the states. In contrast, Model 1 also reveals that citizens’ perceptions of rising crime, unemployment, and inflation across the states have significant effects on state level presidential approval. A oneunit increase in perceptions of both rising crime

127

and rising inflation in a state decreased statelevel presidential approval by .07 percentage points. A oneunit increase in perceptions of rising unemployment in a state decreased statelevel presidential approval by .08 percentages points.

In contrast to Model 2, Model 3 shows that the variable Divided

Government had significant effects only after the transition. During the PRI regime the federal and the subnational governments were in some way unified.

Most state governors were members of the president’s party. After the pre democratic period, the political geography changed and the federal government and the subnational governments were divided. Most state governors were not members of the president’s party. Thus, more than one political party could be held responsible for negative outcomes. This partition between the federal and the subnational governments had positive effects on presidential approval.

Presidents are less likely to be held accountable for government’s failures when the state governor’s party is different than that of the president’s.

The results of Model 2 and Model 3 suggest that consistent with

Hypothesis 1 38 , the effects of subnational conditions during the predemocratic era were stronger than those in the democratic period. Except for state GDP per capita, the effects of objective indicators during the PRI era, Model 2, were stronger than those after the transition, Model 3. The story of the perception variables is similar to that of subnational conditions. The effects of perceptions

38 H1: The effects of objective state indicators are stronger during the PRI era than those of the democratic period.

128

of crime, unemployment, and inflation were stronger before the democratic transition, Model 2, than after the predemocratic period, Model 3. This finding is consistent with Hypothesis 2 39 . In contrast, state partisanship had stronger effects before than after the transition.

This analysis should test whether the different effects of subnational conditions and citizens’ perceptions aggregated at the state level during the PRI era and in the democratic period are significant by themselves. The reason of the different impact of the variables between these two periods may be associated with the different number of observations during the predemocratic era (86) and the democratic period (144). In Model 5 every interactive variable is the product of the variable PRI (the nondemocratic age) and the key explanatory variables.

Every time the interactive term is significant the analysis shows that the different impact of the variable tested (either perception variables or subnational conditions) is significant by itself, and not by the different number of observations in Model 2 and Model 3.

In Table 2.4, Model 5 tests the hypothesized differences between these two periods, the predemocratic period and the merging Mexican democracy, by including a battery of interaction effects for each of the key independent variables. As Model 5 reveals, except for state crime the effects of the rest of the variables were not significantly different between these two periods. The only

39 Given that the effects of subnation conditions on statelevel presidential approval lost power after the PRI regime, the impact of citizens’ perceptions, aggregated at the state level, of crime, unemployment, and inflation in the democratic period turned weaker than those during the PRI era.

129

interactive variable that became significant in Model 5 was PRI X Num.

Convicted Criminals. The other interactive variables show that subnational conditions and citizens’ perceptions aggregated at the state level did not have significantly different effects between these two periods, the predemocratic period and the emerging Mexican democracy 40 .

40 In Appendix 2.2, Table 2.7, I run some models using the effects of changes in subnational conditions and changes in perceptions aggregated at the state level on statelevel presidential approval.

130

Table 2.3 Effects of People’s Perceptions of Crime, Unemployment and Inflation across the States and Subational Conditions on StateLevel Presidential Approval (19942005), TimeSeriesCrossSectional Models

131

Model 1 (94 Model 2 (Pre Model Model 4 05) (Base Democratic 3(Democratic (Kernell’s Line Model) Period 9400) Period 0105) Test, 9405) Lag State .12 ** .16* .16** .14** Approval (.05) (.08) (.07) (.06)

PRI .3*** (.05) PRI X Lag State .41*** Approval (.07) Subational Conditions: Divided .05*** .01 .05* .03*** Government (.01) (.02) (.03) (.01)

Num. Convicted .4 .61 .24 .48* Criminals for (.27) (.44) (.57) (.25) Every 100 Inhabitants by State State GDP per .00002*** .00002 .00003** 6.8e06 Capita (6.84 e06) (.000015) (.00001) (7.2e06)

Unemployment 1.16 .6 .5 .8* Level by State (.83) (1.2) (1.42) (.7)

Inflation by .04 .12 .01 .007 State (.04) (.18) (.01) (.01) Perception Variables: Perception of .07*** .14*** .12*** .13*** Rising Crime (.03) (.03) (.05) (.02) Aggregated at the State Level Perception of .08*** .11** .04 .1*** Rising (.03) (.05) (.07) (.03) Unemployment Aggregated the State Level

132

Table 2.3, Continued 41

Model 1 (94 Model 2 (Pre Model 3 Model 4 05) (Base Democratic (Democratic (Kernell’s Line Model) Period, 94 Period, 01 Test 9405) 00) 05)

Perception of .07** .13** .09 .02 Rising Inflation (.03) (.06) (.07) (.03) Aggregated at the State Level

State .56*** .46*** .66*** .56*** Partisanship (.06) (.1) (.08) (.06)

Constant .17 .37 .27 .17 (.16) (.33) (.4) (.15)

R² .77 .92 .77 .81 N 230 86 144 230 *p<.10, **p<.05, ***p<.01 Note: The dependent variable is statelevel presidential approval. Robust standard errors in parentheses. I include year dummies except for one year and state fixed effects in the estimation of the models but not presented in the table. Some of them are significant.

41 In this model and the rest in this chapter I check for robust standard errors.

133

Table 2.4 InteractiveTest Model

134

Model 5 (9405)

Lag State Approval .009 (.06) Subational Conditions:

Divided Government .06*** (.02)

Num. Convicted Criminals for Every .6** 100 Inhabitants by State (.25)

State GDP per Capita .00001 (8.76e06)

Unemployment Level by State 1.5 (.97)

Inflation by State .004 (.007)

Perception Variables Aggregate at the State Level: Perception of Rising Crime Aggregated .08* at the State Level (.05)

Perception of Rising Unemployment .05 Aggregated at the Sate Level (.07)

Perception of Rising Inflation .07 Aggregated at the State Level (.08)

State Partisanship .56*** (.07)

135

Table 2.4, Continued

InteractiveTest Variables: Model 5 (9405)

PRI .2* (.12) PRI X Divided Government .04 (.03)

PRI X Num. Convicted Criminals .33** (.16)

PRI X State GDP per Capita 2.6e06 (1.7e06)

PRI X Unemployment Level by State .83 (1.1)

PRI X Inflation by State .22 (.18)

PRI X Perceptions of Rising Crime .004 (.05)

PRI X Perceptions of Rising .1 Unemployment (.08)

PRI X Perceptions of Rising Inflation .08 (.1)

PRI X State Partisanship .05 (.11) Constant .03 (.18) R² .8

N 230

*p<.10, **p<.05, ***p<.01 Note: The dependent variable is statelevel presidential approval. I include year dummies except for one year and state fixed effects in the estimation of the models but not presented in the table. Some of them are significant.

136

2.8 Conclusion

This chapter has examined data on statelevel presidential approval in

Mexico from 1994 to 2005, before and after the democratic transition. The results show that as in the U.S. case, in Mexico subnational conditions and perceptions aggregated at the state level affect statelevel presidential approval. However, in the Mexican case, the empirical evidence suggests that the perception variables have more explanatory power than subnational conditions. Except for state GDP per capita, the effects of objective indicators do not have generally significant effects on approval. In contrast, perceptions of crime, unemployment, and inflation across the states affect significantly statelevel presidential approval, especially during the predemocratic period.

However, in most of the cases the effects of the explanatory variables were not significantly different between these two periods, the semiauthoritarian

PRI regime and the democratic period. Except for state crime, the effects of the rest of the explanatory variables were not significantly different before and after the transition. These nonstatistically significant results can be explained by the limitation of the data, the surveys of this analysis are not representative at the state level and the models use modest number of observations. Most of the surveys did not include observations for all the states of Mexico. I attempt to overcome these problems with a multilevel analysis in chapter three.

137

2.9 Appendix 2.1

Table 2.5 Comparing the Distribution of the Population of Mexico (0105) with the Distribution of the Surveys during the PostPRI Era (0105) 42

42 I calculate the population of Mexico from 2001 to 2005 using the Censo General de Población y Vivienda 2000 and the II Conteo de Población y Vivienda 2005. To calculate the population of 2001, 2002, 2003 and 2004, I estimate a projection of the yearly population race based on the years of 2000 and 2005.

138

States Distribution of Distribution Distribution Distribution the Mexican of Survey 1, of Survey 2, of Survey 3, Population 2001 2001 2001 2001 1. Aguascalientes .94 .67 .67 2. Baja California 2.46 2.67 2.67 3. Baja California .46 .67 .67 Sur 4. Campeche .70 .67 .67 .66 5. Coahuila 2.5 2.67 2.60 6. Colima .54 .67 .67 7. Chiapas 3.55 3.33 3.34 3.9 8. Chihuahua 3.17 3.33 3.34 1.9 9. Distrito Federal 10.2 10.6 10.6 13.9 10. Durango 1.44 1.33 1.34 11. Guanajuato 4.57 4.67 4.67 12. Guerrero 2.87 2.67 2.67 4.64 13. Hidalgo 2.25 2.0 2.00 14. Jalisco 6.52 6.67 6.6 7.30 15. Estado de 13.20 12.67 12.55 11.02 Mexico 16. Michoacan 3.9 4.0 4.01 4.6 17. Morelos 1.58 2.0 2.0 1.9 18. Nayarit .94 .67 .67 5.97 19. Nuevo Leon 4.28 4.0 4.07 6.57 20. Oaxaca 3.32 3.33 3.34 6.44 21. Puebla 4.9 4.67 4.67 13.07 22. Queretaro 1.39 1.33 1.34 23. Quintana Roo .89 .67 .67 24. San Luis Potosi 2.27 2.0 2.0 4.64 25. Sinaloa 2.64 2.67 2.67 26. Sonora 2.38 2.67 2.67 .66 27. Tabasco 1.9 2.0 2.0

139

Table 2.5, Continued

States Distribution of Distribution Distribution Distribution the Mexican of Survey 1, of Survey 2, of Survey 3, Population 2001 2001 2001 2001 28. Tamaulipas 3.0 3.33 3.34 5.24 29. Tlaxcala .979 .67 .67 30. Veracruz 7.17 7.33 7.34 5.91 31. Yucatan 1.8 2.00 2.00 1.33 32. Zacatecas 1.33 1.33 1.34

States Distribution of Distribution Distribution Distribution the Mexican of Survey 4, of the of Survey 1, Population 2001 Mexican 2002 2001 Population 2002 1. Aguascalientes .94 .95 2.5 2. Baja California 2.46 2.63 5.0 3. Baja California .46 .48 1.25 Sur 4. Campeche .70 .66 .71 1.21 5. Coahuila 2.5 2.43 1.71 6. Colima .54 .55 .92 7. Chiapas 3.55 3.9 3.63 2.08 8. Chihuahua 3.17 2 3.16 6.0 9. Distrito Federal 10.2 13.9 10.2 10.96 10. Durango 1.44 1.42 1.25 11. Guanajuato 4.57 4.52 4.75 12. Guerrero 2.87 4.64 2.8 5.0 13. Hidalgo 2.25 2.22 2.75 14. Jalisco 6.52 7.3 6.51 7.7 15. Estado de 13.20 11.02 13.43 10.33 Mexico 16. Michoacan 3.9 4.6 3.79 3.75 17. Morelos 1.58 1.9 1.58 2.5 18. Nayarit .94 5.9 .93 2.5 19. Nuevo Leon 4.28 6.5 4.26 5.3

140

Table 2.5, Continued

States Distribution of Distribution Distribution Distribution the Mexican of Survey 4, of the of Survey 1, Population 2001 Mexican 2002 2001 Population 2002 20. Oaxaca 3.32 6.4 3.24 2.08 21. Puebla 4.9 13.07 4.91 3.33 22. Queretaro 1.39 1.44 1.25 23. Quintana Roo .89 1.0 .83 24. San Luis Potosi 2.27 4.6 2.24 1.67 25. Sinaloa 2.64 2.59 1.25 26. Sonora 2.38 .66 2.36 1.25 27. Tabasco 1.9 1.88 1.25 28. Tamaulipas 3.0 5.2 3.0 2.08 29. Tlaxcala .979 .98 .83 30. Veracruz 7.17 5.9 7.0 4.17 31. Yucatan 1.8 1.33 1.76 1.25 32. Zacatecas 1.33 1.3 1.25

States Distribution of Distribution of Distribution of the Mexican Survey 1, 2003 Survey 2, 2003 Population 2003 1. Aguascalientes .96 1.29 1.33 2. Baja California 2.7 2.59 2.66 3. Baja California Sur .49 .65 .66 4. Campeche .70 .65 .66 5. Coahuila 2.42 2.59 2.66 6. Colima .56 .65 .66 7. Chiapas 3.66 3.88 3.99 8. Chihuahua 3.17 3.24 3.32 9. Distrito Federal 10.22 8.41 8.64 10. Durango 1.41 1.94 1.99 11. Guanajuato 4.5 4.53 4.65 12. Guerrero 2.76 2.59 2.66 13. Hidalgo 2.21 2.59 2.66 14. Jalisco 6.5 6.47 6.65

141

Table 2.5, Continued

States Distribution of Distribution of Distribution of the Mexican Survey 1, 2003 Survey 2, 2003 Population 2003 15. Estado de Mexico 13.53 12.94 13.30 16. Michoacan 3.74 3.88 3.99 17. Morelos 1.58 1.29 1.33 18. Nayarit .92 1.29 1.33 19. Nuevo Leon 4.25 3.88 3.99 20. Oaxaca 3.21 3.24 3.32 21. Puebla 4.92 5.18 5.32 22. Queretaro 1.46 1.94 1.99 23. Quintana Roo 1.05 1.29 1.33 24. San Luis Potosi 2.22 2.27 2.33 25. Sinaloa 2.57 2.59 2.66 26. Sonora 2.35 1.94 1.99 27. Tabasco 1.87 1.94 1.99 28. Tamaulipas 3.0 2.59 2.66 29. Tlaxcala .98 1.29 1.26 30. Veracruz 6.92 7.12 4.65 31. Yucatan 1.77 1.94 1.99 32. Zacatecas 1.29 1.29 1.33

142

Table 2.5, Continued

States Distribution of Distribution Distribution Distribution the Mexican of Survey 1, of Survey 2, of Survey 3, Population 2004 2004 2004 2004 1. Aguascalientes .96 .67 .67 .67 2. Baja California 2.77 2.0 2.0 2.0 3. Baja California .51 .67 .67 .67 Sur 4. Campeche .71 .67 .67 .67 5. Coahuila 2.41 2.0 2.0 2.0 6. Colima .56 .67 .67 .67 7. Chiapas 3.7 4.0 4.0 4.0 8. Chihuahua 3.17 3.33 3.34 3.33 9. Distrito 10.24 8.67 8.67 8.67 Federal 10. Durango 1.39 1.33 1.33 1.33 11. Guanajuato 4.47 4.67 4.67 4.67 12. Guerrero 2.74 3.33 3.34 3.33 13. Hidalgo 2.20 2.67 2.67 2.67 14. Jalisco 6.50 6.67 6.67 6.67 15. Estado de 13.63 13.33 13.34 13.33 Mexico 16. Michoacan 3.7 4.0 4.0 4.0 17. Morelos 1.58 1.33 1.33 1.33 18. Nayarit .91 1.33 1.33 1.33 19. Nuevo Leon 4.24 4.0 4.0 4.0 20. Oaxaca 3.18 3.33 3.34 3.33 21. Puebla 4.92 5.33 5.34 5.33 22. Queretaro 1.48 1.33 1.33 1.33 23. Quintana Roo 1.09 .67 .67 .67 24. San Luis 2.21 2.67 2.6 2.67 Potosi 25. Sinaloa 2.55 2.67 2.67 2.67 26. Sonora 2.34 2.67 2.67 2.67 27. Tabasco 1.87 2.0 2.0 2.0 28. Tamaulipas 3.0 3.33 3.34 3.33 29. Tlaxcala .99 .67 .67 .67 30. Veracruz 6.85 6.67 6.67 6.67 31. Yucatan 1.77 2.0 2.0 2.0 32. Zacatecas 1.27 1.33 1.33 1.33

143

Table 2.5, Continued

States Distribution of Distribution Distribution Distribution the Mexican of Survey 4, of the of Survey 1, Population 2004 2004 43 Mexican 2005 Population 2005. 1. Aguascalientes .96 1.23 .97 .55 2. Baja California 2.77 2.84 1.82 3. Baja California .51 1.23 .51 .55 Sur 4. Campeche .71 1.23 .71 .91 5. Coahuila 2.41 2.41 2.55 6. Colima .56 1.23 .56 .55 7. Chiapas 3.7 3.73 2.92 8. Chihuahua 3.17 3.17 3.83 9. Distrito 10.24 18.52 10.26 10.01 Federal 10. Durango 1.39 2.47 1.38 1.64 11. Guanajuato 4.47 8.64 4.48 4.56 12. Guerrero 2.74 4.94 2.71 3.65 13. Hidalgo 2.20 4.94 2.19 2.73 14. Jalisco 6.50 12.35 6.49 4.38 15. Estado de 13.63 13.72 11.83 Mexico 16. Michoacan 3.7 7.41 3.66 4.92 17. Morelos 1.58 2.47 1.58 1.46 18. Nayarit .91 2.47 .91 1.09 19. Nuevo Leon 4.24 4.23 3.65 20. Oaxaca 3.18 6.17 3.15 4.01 21. Puebla 4.92 4.93 3.83 22. Queretaro 1.48 2.47 1.49 .91 23. Quintana Roo 1.09 1.23 1.14 .67 24. San Luis 2.21 2.19 2.55 Potosi 25. Sinaloa 2.55 4.94 2.53 4.92 26. Sonora 2.34 4.94 2.33 1.82 27. Tabasco 1.87 3.7 1.86 1.82

43 This national survey was complemented with 11 state surveys done in the states of Baja California, Chiapas, Chihuahua, Coahuila, Estado de México, Nuevo León, Puebla, San Luis Potosí, Tamaulipas, and Veracruz.

144

Table 2.5, Continued 44

States Distribution of Distribution Distribution Distribution the Mexican of Survey 4, of the of Survey 1, Population 2004 2004 Mexican 2005 Population 2005 28. Tamaulipas 3.0 3.0 2.73 29. Tlaxcala .99 1.23 .99 1.09 30. Veracruz 6.85 6.78 7.84 31. Yucatan 1.77 3.7 1.78 1.64 32. Zacatecas 1.27 2.47 1.26 2.55

44 The data of the distribution of the Mexican Population come from the INEGI.

145

2.10 Appendix 2.2

Table 2.6 Effects of Perceptions Aggregated at the State level and Sub ational Conditions on StateLevel Presidential Approval (19942005), Including Perceptions of Corruption and Complaints against Members of the Civil Service

146

Model 1 Pre Model 2 Model 3 Democratic Democratic (Interactive Test) Period (9400) Period (0105) Lag State .22** .12 .07 Approval (.1) (.08) (.06) Subational Conditions Divided .02 .08 .02 Government (.09) (.06) (.04)

Num. Convicted .06 .77 .5 Criminal for Every (1.03) (.57) (.36) 100 Inhabitants by State Num. Complaints .2 2.95 1.7 against Members (.43) (2.4) (1.8) of the Civil Service for Every 1000 Inhabitants by State State GDP per .00001 .00008*** .00003** Capita (.00004) (.00002) (.00001)

Unemployment .64 .6 .6 Level by State (2.2) (1.5) (1.2)

Inflation by State .31 .93 2.8*** (.36) (.98) (1.09) Perception Variables at the State Level Perceptions Rising .23*** .3*** .4*** Crime (.07) (.11) (.15)

Perceptions of .14 .12 .1 Rising Corruption (.12) (.11) (.16)

147

Table 2.6, Continued

Model 1 Pre Model 2 Model 3 Democratic Democratic (Interactive Test) Period (9400) Period (0105) Perception of .22*** .06 .04 Rising (.08) (.12) (.15) Unemployment

Perception of .19* .01 .08 Inflation (.11) (.1) (.13)

State Partisanship .4*** .66*** .7*** (.16) (.11) (.17) PRI .32 (.2) PRI X Divided .04 Government (.06)

PRI X Num. .48** Convicted (.21) Criminals

PRI X Num. 2.04 Complaints (1.71) against Members of the Civil Service PRI X State GDP 3.67e06 per Capita (2.4e06)

PRI X .2 Unemployment (1.5) Level by State

PRI X Inflation by 3.04*** State (1.18)

PRI X Perceptions .22 of Crime (.16)

148

Table 2.6, Continued 45

Model 1 Pre Model 2 Model 3 Democratic Democratic (Interactive Test) Period (9400) Period PRI X Perceptions .008 of Corruption (.19)

PRI X Perceptions .26 of Unemployment (.17)

PRI X Perceptions .29* of Inflation (.16)

PRI X State .27 Partisanship (.19)

Constant .34 1.04** .07 (.55) (.43) (.26) R² .94 .92 .9

N 52 66 118 *p<.10, **p<.05, ***p<.01 Note: The dependent variable is statelevel presidential approval. I include year dummies except for one year and state fixed effects in the estimation of the models but not presented in the table. Some of them are significant.

45 The data about citizens’ complaints against the civil service were provided by the Secretaría de la Función Pública through the Instituto Federal de Acceso a la Información (IFAI)

149

Table 2.7 Effects of Changes in Perceptions Aggregated at the State Level, and Changes in Subational Conditions on StateLevel Presidential Approval (19942005), TimeSeriesCrossSectional Models

150

Model 1 Pre Model 2 Model 3 Democratic Democratic (Interactive Test) Period (9400) Period (0105) Lag State .05 .07 .15 Approval (.33) (.11) (.09) Subational Conditions Divided .09 .07 .02 Government (.08) (.04) (.03)

Increase in the % .03 .31*** .26*** of the Num. (.1) (.11) (.1) Convicted Criminal for Every 100 Inhabitants by State Increase in the % .11 .7 .5 of State GDP per (.66) (.5) (.4) Capita Increase in the % .003 .05* .05 of Unemployment (.03) (.03) (.03) Level by State Increase in the % .24*** .0003 .0007** of Inflation by (.07) (.0003) (.0003) State Perception Variables Aggregated at the State Level Increase in the % .02 .03 .04 of Perceptions of (.04) (.04) (.03) Crime Increase in the % .08*** .001 .009 of Perception of (.02) (.03) (.03) Unemployment Increase in the % .17* .04 .03 of Perception of (.1) (.03) (.03) Inflation

151

Table 2.7, Continued

Model 1 Pre Model 2 Model 3 Democratic Democratic Interactive Test Period (9400) Period (0105) Increase in the % of .07 .03* .04** State Partisanship (.05) (.02) (.02)

PRI .02 (.04)

PRI X Divided .02 Government (.05)

PRI X Increase in .15 the % of Convicted (.13) Criminals PRI X Increase in .53 the % of State GDP (.69) per Capita PRI X Increase in .03 the % of (.04) Unemployment Level by State PRI X Increase in .21*** the % of Inflation (.05) Level by State PRI X Increase in .02 the % of (.04) Perceptions of Crime PRI X Increase in .07* the % of (.04) Perceptions of Unemployment PRI X Increase in .18*** the % of (.06) Perceptions of Inflation

152

Table 2.7, Continued

PRI X Increase in .05 the % of State (.04) Partisanship Const .69** .58*** .48*** (.3) (.1) (.08) R² .88 .6 .64 47 128 175 *p<.10, **p<.05, ***p<.01 Note: The dependent variable is statelevel presidential approval. I include year dummies except for one year and state fixed effects in the estimation of the models but not presented in the table. Some of them are significant.

153

2.11 Works Cited

Beyle, Thad, Richard G. Niemi, and Lee Sigelman. 2002. “Gubernatorial, Senatorial, and Statelevel Presidential Job Approval: The U.S. Officials Job Approval Ratings (JAR) Collection”, State Politics and Policy Quarterly , Vol. 2, No. 3 (Fall): 215–229.

Buendía, Jorge. 1996. “Economic Reform, Public Opinion, and Presidential Approval in Mexico, 19881993”. Comparative Political Studies, Vol. 29 No. 5 (October): 566591.

CanesWrone Brandice. 2006. Who Leads Whom? Presidents, Policy, and the Public. Chicago: The University of Chicago Press.

Cohen, Jeffrey. 1998. “StateLevel Public Opinion Polls as Predictors of Presidential Election Results”. American Politics Quarterly, Vol. 26, No. 2 (April): 139159.

Cohen, Jeffrey, Jon R. Bond, Richard Fleisher, and John A. Hamman. 2000. “Statelevel Presidential Approval and Senatorial Support”. Legislative Studies Quarterly, Vol. 25, No. 4: 577590.

. 2003. “The Polls: StateLevel Presidential Approval: Results from the Job Approval Project”, Presidential Studies Quarterly, Vol. 33, No. 1 (March): 211220.

, and Richard Powell. 2005. “Building Public Support from the Grassroots Up: The Impact of Presidential Travel on StateLevel Approval” Presidential Studies Quarterly 35, No. 1 (March): 1127.

Cornelius, Wayne, Judith Gentleman and Peter Smith.1989. Mexico’s Alternative Political Futures. San Diego, California: Center for U.S.Mexican Studies.

Davis, Charles and Ronald E. Langley. 1995. “Presidential Popularity in a Context of Economic Crisis and Political Change: The Case of Mexico”. Studies in Comparative International Development 30 Fall: 2448.

Domínguez, Jorge and Chappell Lawson (2004). Mexico’s Pivotal Democratic Election. Stanford, California: Stanford University Press.

154

Hansen, Susan B. 1999. “Life Is Not Fair: Governors’ Job Performance Rating and State Economics”. Political Research Quarterly, Vol. 52, No. 1 (March): 167188.

1997. Going Public. ew Strategies of Presidential Leadership. Washington, D.C., Congressional Quarterly Press.

Lawson, Chappell. 1999. “Why Cárdenas won: The 1997 Elections in Mexico City”, Toward Mexico’s Democratization: Parties, Campaigns, Elections, and Public Opinion. Ed. Jorge I. Domínguez and Alejandro Poiré: Routledge: New York: 147173

Magaloni, Beatriz. 2006. Voting for Autocracy. The Untied States: Cambridge University Press.

Mitofsky. 2005. “Aprobación Presidencial por Entidad Federativa en 2005”, www.consulta.com.mx.

Nicholson, Stephen, Gary M. Segura, and Nathan D. Woods. 2002. “Presidential Approval and the Mixed Blessing of Divided Government”. The Journal of Politics, Vol. 64, No. 3 (August): 701720.

Page, Benjamin. 1978. The Choices and Echoes in Presidential Elections. Chicago. The University of Chicago Press.

Page, Benjamin and Marshall M. Bouton. 2006. The Foreign Policy Disconnect. What Americans Want from Our Leader but Don’t Get. Chicago: The University of Chicago Press.

155

Rodríguez, Victoria and Peter M. Ward.1994. Political Chage in Baja California. San Diego, California: Center for U.S.Mexican Studies, University of California, San Diego.

Rudolph, Thomas. 2003. “Who’s Responsible for the Economy? The Formation and Consequences of Responsibility Attributions”. American Journal of Political Science, Vol. 47, No. 4 (October): 698713.

Villarreal, Andrés. 1999. “Public Opinion of the Economy and the President among Mexico City Residents: The Salinas Sexenio”. Latin American Research Review , Vol. 34, No. 2: 132151.

2.12 Web Sites

Banco de Información para la Investigación Aplicada en Ciencias Sociales (BIIACS), http://www.biiacs.cide.edu

Banco de México, http:// www.banxico.org.mx

Instituto acional de Geografía y Estadística (IEGI), http:// www.inegi.gob.mx

Mitofsky. 2005. “Aprobación Presidencial por Entidad Federativa en 2005”. www.consulta.com.mx

Parametría, http://www.parametria.com.mx

Reforma, http:// www.reforma.com.mx

TransBorder Institute at the University of San Diego, http://www.sandiego.edu/peacestudies/tbi

3 PRESIDETIAL APPROVAL I MEXICO: MULTI LEVEL AALYSIS

3.1 Introduction

Mexican citizens evaluate the president differently after the democratic transition both at the individual and at the state level. The first chapter shows that the effects of citizens’ perceptions of both corruption and crime in the democratic period are stronger than those during the PRI semiauthoritarian system. Since the

2000 transition political issues, crime and corruption, have become salient.

Citizens consider combating crime to be the second most important task to be addressed by democracy, and view corruption as the main obstacle for a democratic consolidation. Thus, after the 2000 presidential election, the turning point of the transition, Mexican citizens have begun to evaluate the president in terms of his anticorruption and anticrime policies, and performance. The second chapter reveals that some of the subnational conditions had stronger effects on statelevel presidential approval before the transition. The effects of the number of convicted criminals for every 100 inhabitants by state during the PRI regime were significantly different and stronger than those in the postPRI era.

The first and the second chapters demonstrate that the factors that affect presidential approval have a different impact after the 2000 presidential election.

156 157

The first chapter reveals that the effects of perceptions of rising crime and corruption on presidential approval during the democratic period are stronger than those of the predemocratic era. In contrast, the second chapter shows that sub national conditions have stronger effects on approval during the PRI hegemony.

Therefore, the multilevel analysis of this chapter allows me to offer more accurate results including two kinds of data, citizens’ characteristics and objective indicators.

The aggregate analysis of the second chapter shows that presidential approval varies across the states in Mexico. This finding reveals that approval can be affected not only by people’s perceptions but also by subnational contexts.

However, the key point to examine is how all these nested factors, citizens’ characteristics (perceptions of political and economic problems, and sociodemographic background) and subnational conditions work when citizens evaluate the president. Do citizens’ characteristics matter more than subnational conditions? Or do subnational contexts work as key explanatory variables of presidential approval? To what degree do these two different levels complement each other when individuals are the unit of analysis? Can subnational contexts work as treatments in the analysis?

In comparative politics, scholars (Duch and Stevenson 2005; Huber,

Kernell and Leoni, 2005; Jones, Johnston and Pattie 19992; Long and Shivel

158

2005; Steenbergen and Jones 2002) have found that the factors that affect political behavior are nested. These studies illustrate the methodological convenience of using multilevel models in the analysis of presidential approval. The multilevel models allow me to formulate two main research questions in this chapter: (1) To what degree do subnational conditions and citizens’ characteristics work as accurate predictors in the analysis of presidential approval in Mexico? (2) Do citizens’ characteristics and subnational conditions have the same impact on approval after the transition?

This chapter has two main objectives: (1) To conduct a multilevel analysis of presidential approval in Mexico. (2) To demonstrate that citizens’ perceptions and subnational contexts affect voters’ evaluation of the president differently after the 2000 presidential election. This chapter is divided as follows.

The first section summarizes some of the findings and limitations of the previous chapter. The second section reviews the literature and develops the rationale of this chapter. Third, I formulate my hypotheses, and explain the data and methodology. Fourth, I present a multilevel analysis. Finally, the conclusion summarizes the main findings of this study.

159

3.2 Findings and Limitations of the TimeSeriesCrossSectional Analysis

The timeseriescrosssectional analysis of the second chapter displays two main findings: subnational conditions and citizens’ perceptions aggregated at the state level affect statelevel presidential approval in Mexico. However, except for the number of convicted criminals per 100 inhabitants by state, the effects of the independent variables were not significantly different between these two periods of analysis, the PRI regime and the postPRI era. Furthermore, it was not possible to see how subnational contexts interact with citizens’ characteristics (people’s perceptions of both political and economic problems, and citizens’ sociodemographic background) when both variables have effects on presidential approval. I summarize and examine the findings and limitations of the time seriescrosssectional analysis of the second chapter. These results are useful for the multilevel study of this chapter.

The main restriction of the timeseriescrosssectional analysis is the absence of individual level variables that reflect citizens’ perceptions and voters’ sociodemographic background. The previous limitation can be summarized as three main problems in the analysis of the second chapter: the nature of aggregating individual survey data at the state level is different than that of citizens’ political behavior, the absence of multiple levels of information, and the lack of a micro unit of analysis. I explain these restrictions in detail:

160

(1). Although aggregating individual survey data at the state level allows me to study the variation of presidential approval across the states, aggregating data is a different process than that of citizens’ evaluation of the president

(Paterson and Goldstein 1991, 389). The process of approval is done by people not by the states. Aggregatelevel analyses do not include some of the most important and relevant explanatory variables, citizens’ perceptions and sociodemographic background. (2) The timeseriescrosssectional models of the second chapter show that both citizens’ opinions aggregated at the level of the states influence political behavior and the context affects voters’ decisions.

However, it was not possible to analyze the real interaction between these two levels of information, i.e. subnational conditions and citizens’ attitudes (Jones,

Johnston, and Pattie 1992, 378). Voters make political decisions using more than one level of information. When they evaluate the president’s performance all these resources that come from different levels of information (individual and state levels) are nested. Therefore, the analysis of approval should include sub national indicators and citizens’ characteristics (voters’ sociodemographic background and perceptions). (3) Aggregating individual survey data at the state level may alter the meaning of the core of the analysis. In a strict sense nothing is saying directly to political behavior, the micro unit of analysis (Snijders and

Bosker 1999, 3). The process of aggregating citizens’ opinions on crime, corruption, unemployment and inflation at the state level refers to a political and

161

geographical area, rather than voters’ perceptions when they evaluate the president.

The best methodological alternative to overcome the previous limitations is a multi level analysis. Four main reasons support the previous statement:

Multilevel analysis is useful to examine an accurate relationship between X and

Y, a multilevel analysis permits to examine the interaction between the different sources and levels of information, multilevel models are helpful to express the micromacro relationships in terms of the context, and a multilevel analysis is necessary to formulate research questions from different perspectives. I explain in detail these four benefits of the multilevelanalysis:

(1). Multilevel models are necessary and helpful to show the real relationship between Y and X (Snijders and Bosker 1999, 1315). When citizens evaluate the president, they use different sources that are interconnected.

Individual perceptions interact with subnational conditions, when citizens approve the president. This complex process of making decisions and evaluations can be examined only by multilevel models. Moreover, multilevel analysis can be better for including the necessary Xs, explanatory variables, which may affect Y, reducing the risks of omitted variable bias. (2) The use of multilevel models allows me to examine the interaction between two different levels of information

(Snijders and Boksker 1999, 14), subnational contexts and people’s attitudes, which affect presidential approval. For example, even though the first chapter demonstrates that citizens’ perceptions of rising crime affects presidential

162

approval, the effect of this variable cannot be the same in all the states. I expect that while in the D.F. (with a high level of crime), citizens’ perceptions of rising crime should have strong effects on approval, the story should be different in

Aguascalientes (a state with a low level of crime). In Aguascalientes, the level of crime should have weak effects on approval. (3) The multi level analysis can express the micromacro relationships in terms of characteristics on the context, which takes the form of macro level variables (DiPrete and Forristal 1994, 333).

Citizens’ perceptions of the president’s policies do not occur in a vacuum. The specific context can make these perceptions either relevant or irrelevant. What we can expect of citizens’ perceptions of the president is related with the subnational context that takes the form of macro economic and political indicators such as the level of unemployment, inflation, corruption and crime by state. (4) A multi level analysis allows the researcher to analyze a problem from different perspectives and make multiple research questions (Sttenbergen and Jones 2002, 227). In this chapter, I attempt to answer these questions: What is the importance of each of these two levels (citizens’ characteristics and subnational contexts) for understanding presidential approval ? What predictors of both levels can explain presidential approval in Mexico? What level has more impact on approval? Is the impact of each level the same before and after the transition? The previous questions reveal the nature of approval which is very sensitive and easily affected by factors that are nested, similar to that of political behavior (Duch and

163

Stevenson, 2005; Huber, Kernell and Leoni, 2005; Jones, Johnston and Pattie,

19992; Long and Shively, 2005; Steenbergen and Jones, 2002;).

3.3 The Literature Review and Rationale

Recent studies in comparative politics (Jones, Johnston, and Pattie, 1992;

Steenbergen and Jones, 2002; Duch and Stevenson, 2005; Huber, Kernell, and

Leoni, 2005; Long and Shively, 2005) suggest that the nature of political behavior is more complex that the traditional wisdom assumed. Jones, Johnston, and Pattie

(1992) offer empirical evidence that illustrates how voters’ characteristics and regional contexts affect voting behavior in the U.K. Steenbergen and Jones

(2002) argue that citizens’ support for foreign policy is determined by people’s characteristics and by crossnational differences. Duch and Stevenson (2005) state that both citizens’ perceptions and the concentration of the authority in the cabinet are the main determinants of the variation of economic voting.

Jones, Johnston and Pattie (1992) find that local context is a key determinant for voting in the U.K. Regions as the contiguous divisions of the national territory may stimulate some degree of personal identity affecting individual citizens’ political behavior. For example, in regions suffering relative deprivation it is less likely that citizens support the incumbent government that in regions relatively prosperous. How does this finding apply to presidential approval in Mexico? According to the current Constitución de los Estados Unidos

164

Mexicanos (2001), the will of the citizens is to organize Mexico as a representative, democratic, federal republic, integrated by free and sovereign states. Therefore, the main political unit in which citizens are exposed to the influence of other voters is in the states. Although individuals approve of the president, they do it in specific contexts, in the states, which affect people’s preferences and decisions. I expect that in the State of Mexico, with high level of unemployment, citizens’ perceptions of rising unemployment have stronger effects on approval than in Baja California, one state with low level of unemployment.

Both voting behavior and citizens’ support for foreign policy can be affected by aggregate variables. Steenbergen and Jones (2002) argue that people’s support for the European (EU) is influenced by individual characteristics and by crossnational differences such as the date at which a country accedes to the EU and the dependency of the national economy on EU membership. This finding allows me to expect that Mexican voters’ opinion on rising crime, corruption, unemployment and inflation affect approval, but their effects should be easily influenced by subnational contexts. The more unfavorable subnational context for either political or economic problems, the stronger the effects of perceptions of rising crime, corruption, unemployment, and inflation on presidential approval become.

If citizens’ support for foreign policy (Steenbergen & Jones, 2002;) and voting behavior (Jones, Johnston and Pattie, 1992) do not occur in a vacuum,

165

what mechanisms do political institutions develop to influence voters’ behavior when they make political decisions? And how do these mechanisms offer some insights that can be helpful for our understanding of approval? Scholars (Duch and Stevenson 2005; Huber, Kernell and Leoni 2005; Long and Shively, 2005) suggest that institutions offer information resources and affect the use of economic voting in an election (Duch and Stevenson, 2005), the strength of party identification (Huber, Kernell, and Leoni 2005 ), and electoral participation (Long and Shively 2005). These findings are helpful to understand how political institutions impact citizens’ behavior when they evaluate the president.

In a cross national study, Duch and Stevenson (2005) find that the variation of economic voting across the nations is explained by the concentration of the authority. More economic voting exists for chief executive when the authority is more concentrated and parties share power more equally. Voters’ confusion of responsibility is more likely to occur when the cabinet is made up of coalitions of different parties. How does this argument shed light on presidential approval? Theoretically, we cannot expect that the president is approved in the same way in a state controlled by the president’s party that in a state governed by the opposition. When the president and the state governor are member of the same party, citizens easily identify the president as the main actor responsible for negative outcomes. In contrast, when one party controls the federal government and another the subnational government, citizens blame for negative outcomes both the president and the state governor. Divided Government provides

166

ambiguous and conflicting information about which level of government, either the federal or the state, to hold accountable for government performance.

Presidents are less likely to be held accountable for government’s failures when the state governor’s party is different than that of the president’s.

Huber, Kernell, and Leoni (2005) demonstrate that voters are most likely to form party attachments when institutions assist them in retrospectively evaluating parties, such as party systems with few parties in government. As the number of parties in the legislature increases, the consequent power sharing in government impedes an individual’s ability to make retrospective judgments that leads to party attachments. This finding suggests that the effect of party ID on approval may be affected by the institutional context. Therefore, after PRI regime, with the dead of onepartysystem, party competition in Mexico might become stronger and party attachments enjoyed a more powerful impact on presidential approval.

Long and Shively (2005) find that citizens’ level of information is very sensitive to the effective number of parties in political struggles. Citizens’ cognitive resources, their information capacity, and the political institutional context increase a decline in voter turnout. The effective number of parties can alter the information cost. In a multiparty system the information cost may be very high and discourage voter to participate in the electoral process. How does this finding apply for our understanding about presidential approval? People’s cognitive resources such as their perceptions of the president’s policies may alter

167

the way in which citizens evaluate the president. However, these cognitive resources are affected by institutional variables such as the political and economic environment of the subnational context. For example, in a state with a high level of crime, such as Baja California, the information cost about rising crime is low because delinquency has been spread across the majority of the municipalities. In contrast, in a state such as Hidalgo, with low crime rate, the information cost of rising crime may be high, and citizens do no take into account their perceptions of crime when they evaluate the president.

The literature review has showed that the nature of approval is very sensitive and easily affected by factors that are nested. This chapter addresses the next questions: What is the importance of each of these two levels (citizens’ characteristics and subnational contexts) for our understanding of presidential approval in Mexico? How do we account for presidential approval at the different levels? That is, what predictors can explain approval? Do these predictors have the same effects before and after the election? To construct the rationale that allows me to formulate some possible answers to previous questions, I go back to some of the main findings of the first and second chapters of this dissertation. The present analysis translates these questions into the rationale that leads this chapter.

First, what is the importance of each of these two levels (citizens’ characteristics and subnational contexts) for our understanding of presidential approval in Mexico? The first chapter showed that perceptions of crime and corruption impacted voters’ evaluation of the president. The second chapter

168

revealed that subnational contexts affected statelevel presidential approval.

Therefore, I expect that citizens’ perceptions and subnational conditions are key explanatory variables for the analysis of presidential approval in Mexico. The research design of the analysis should include all the possible explanatory variables of presidential approval: citizens’ perceptions, subnational conditions

(aggregate variables), and sociodemographic background. People’s perceptions can work as key explanatory variables. What citizens consider about political and economic issues strongly determines the way in which they evaluate the president, as the first chapter illustrates. However, the second chapter also demonstrates that subnational conditions affect presidential approval. In this chapter, I use citizens’ characteristics as key determinants. The aggregate variables (subnational economic and political conditions) work as the treatments that were not included in the first chapter. Moreover, the sociodemographic background, income, education, age, and gender, is included as control for presidential approval.

Moving to the question about the way necessary to account for presidential approval at the different levels, the findings of the first two chapters suggest that the analysis should include variables such as partisanship and perceptions of rising crime, corruption, inflation, and unemployment for the first level. The research design should incorporate objective political and economic indicators that vary across the states for the second level.

The question about the possible different effects before and after the 2000 presidential election involves analyzing some theoretical implications. The first

169

chapter shows that the political regime works as the market when citizens evaluate the president. From the demand side, citizens have expectations about the government, and from the supply side, the president’s behavior to address these concerns is restricted by the political regime. Since the 2000 transition political issues, crime and corruption, have become salient. Citizens consider combating crime to be the second most important task to be addressed by democracy (Klesner 2001), and view corruption as the main obstacle for a democratic consolidation (Klesner 2001). Thus, after the 2000 presidential election, the turning point of the transition, Mexican citizens have begun to evaluate the president in terms of his anticorruption and anticrime policies, and performance.

The 2000 presidential election significantly affected citizens’ perceptions of both democracy and the president. Camp’s analysis (2004) reveals that only 40 percent of the people thought that Mexico was a democracy in February 2000, four months before the presidential election; in contrast, in July 2000 some days later of the election 63 percent of the citizens believed that Mexico was a democratic regime. Figure 3.1 presents this change in citizens’ perceptions of democracy:

170

70 63 60

50 40 40 Percentage 30 Percentage 20

0 February July Before and After the 2000 Presidential Election

Figure 3.1: Changing Visions of Democracy in 2000 (Camp 2004). Source Camp (2004), Mexico 2000 Panel Study (Domínguez and Lawson 2004). The question asked was: “Do you think Mexico is a democracy?”

This change in citizens’ perceptions affected how people perceived the president. Figure 3.2 shows that Mexican voters began to trust more in the president after the democratic transition. The levels of citizens’ trust in Vicente

Fox are clearly higher than those of Salinas and Zedillo after the first year of tenure of each president.

171

70 58 60

50 40 40 32 Trust 30

(Percentage) 20

Presidential Raiting Trust 10

0 Salinas (1989) Zedillo (1995) Fox (2001) Mexican Presidents

Figure 3.2: Citizens’ Trust in Three Mexican Presidents after Their First Year of Tenure. Figure Done by the Author. The Data Come from the BIIACS and Parametría

The statistical results of the first chapter showed that the effects of citizens’ opinion on corruption and crime in the democratic period were stronger than those during the PRI semiauthoritarian system. Therefore, I expect that in the multilevel analysis perceptions of crime and corruption should have stronger effects after the transition.

The second chapter revealed that some of the subnational conditions affected presidential approval across the federal entities. These are the cases of the state GDP per capita and divided government (the partition between the federal government and the state governments. The state governor’s party is different than that of the president’s). The variations of the number of convicted criminals by state and the number of complaints against members of the civil

172

service by federal entity allow me to expect that state objective indicators of both crime and corruption should affect presidential approval.

Figure 3.3 46 reveals that although crime is a national concern by Mexican citizens, its great variation across the states suggests that voters may have different perceptions about crime at the subnational level.

0.4 0.35

0.3

0.25

0.2 Convicted Criminals 0.15 0.1

100 Inhabitants Inhabitants by 100 State 0.05

Num. Criminals of Convicted per 0 1 3 5 7 9 11 13 15 17 19 21 23 25 27 29 31 States

Figure 3.3: Convicted Criminals by State in Mexico (2005) Source: Justice in Mexico Crime Statistics Database, prepared by the Trans Border Institute at the University of San Diego

The story of the variation of corruption at the subnational level is similar to that of crime. Figure 3.4 shows that the variation of the number of complaints against the civil service across the states is significant. Therefore, I expect that in

46 Aguascalientes (1), Baja California (2), Baja California Sur (3), Campeche (4), Coahuila (5), Colima (6), Chiapas (7), Chihuahua (8), Distrito Federal (9), Durango (10), Guanajuato (11), Guerrero (12), Hidalgo (13), Jalisco (14), Estado de México (15), Michoacán (16), Morelos (17), Nayarit (18), Nuevo León (19), Oaxaca (20). Puebla (21), Querétaro (22), Quintana Roo (23), San Luis Potosí (24), Sinaloa (25), Sonora (26), Tabasco (27), Tamaulipas (28), Tlaxcala (29), Veracruz (30), Yucatán (31), Zacatecas (32).

173

the multilevelanalysis the number of complaints against the civil service by state affects presidential approval in Mexico.

0.05 0.045 0.04 0.035 0.03 0.025 Complaints 0.02 0.015 0.01 0.005 Num. of Complaints the against

Civil Inhabitants Service per 1000 0 1 3 5 7 9 11 13 15 17 19 21 23 25 27 29 31 States

Figure 3.4 47 : Complaints against the Civil Service by State in Mexico (2005) Source: Secretaría de la Función Pública

The introduction of this analysis states that the previous two chapters offer contradictory results. The first chapter reveals that the effects of perceptions of rising crime and corruption on presidential approval during the democratic period are stronger than those of the predemocratic era. In contrast, the second chapter shows that some of the subnational conditions, the number of convicted criminals by state, have stronger effects on approval during the PRI regime. By including citizens’ perceptions and sociodemographic background in the multilevel analysis, I attempt to address this problem. I expect that by aggregating individual

47 Vide footnote 1 to match the numbers of the states with their names.

174

survey data in the analysis the effects of crime and corruption, both perceptions and subnational indicators, should be stronger in the democratic period because the results of chapter one were more reliable than those of chapter two.

3.4 Hypotheses

The general hypothesis of this chapter is that Mexican citizens evaluate the president differently after the 2000 democratic transition. This dissertation has showed that the 2000 presidential election worked as a “critical election” affecting voters’ approval of the president. Citizens started to see the president in a different way after 2000. They had more expectations of a greater competence of the next Mexican president in managing political issues. The current analysis has showed that the political context of Mexico has encouraged citizens to associate democracy with anticrime and anticorruption policies. Mexican people regard combating crime as one of the most important tasks of democracy.

Moreover, citizens consider corruption to be the main obstacle to democracy.

Therefore, my expectations are twofold:

H.1 All else being equal, the effect of both citizens’ perceptions of rising crime

and the number of convicted criminals by state per 100 inhabitants on

presidential approval after the 2000 transition should be stronger than

those during the PRI regime. After the 2000 transition, Mexican citizens

had more expectations about the determination of the new president to

175

address crime. Mexican people believed that one of the main tasks of the

new democratically elected government was combating crime and the new

president could eradicate this problem from Mexican politics.

H.2. All else being equal, the effects of citizens’ perceptions of rising corruption

and the number of complaints against the civil service by state per 1000

inhabitants on presidential approval after the 2000 transition should be

stronger than those during the PRI regime. After the 2000 democratic

transition, Mexican citizens started to have more expectations about the

determination of the new, democratically elected, president to address

corruption in Mexico. They thought that the new president could eradicate

this problem from Mexican politics.

3.5 Data and Method

The fourteen national surveys on which this study is based were organized by the Mexican Office of the Presidency (MOP) In total the surveys contain more than 20,000 household interviews, from November 1994 to November 2005. This period covers three different presidents (Carlos Salinas de Gortari, Ernesto

Zedillo, both Priista presidents, and the Panista Vicente Fox). There is one poll for Salinas’ presidency (in 1994) and six surveys for Zedillo’s presidency (from

1996 to 2000), seven surveys for Fox’s administration (from 2001 to 2005), all

176

conducted by Asesoría Técnica de la Presidencia de la República 48 . Because every survey is an independent representative sample of the population, I pool all the fourteen surveys in a single dataset. Pooling independent cross sections across time has the advantages of increasing sample size, obtaining more precise estimators, and more robust test statistics.

I estimate a series of multilevel models predicting whether citizens approve of the president in order to test my hypothesis about the determinants of presidential approval. With a dichotomous dependent variable, these are logit models. Because the explanatory variables are measured at different levels – perceptions of crime, corruption, unemployment, and inflation, and the sociodemographic background at the individual level, and subnational conditions at the level of the states I use multilevel modeling techniques to acknowledge this data structure explicitly. Citizens’ decisions are nested within the political and economic environment of the subnational context in a specific time. My covariates may not measure all of these forces, creating a correlation between the prediction errors across all the observations introduced.

I incorporate this potential autocorrelation by estimating a model that allows each stateyears (32states across 12 years, from 1994 to 2005, 384 state years) to have its own intercept and thus its own average predictors error to “soak

48 Unfortunately, it is not possible to include observations for every year. In 1995 and 1998, questions about my all explanation variables were not included in the surveys. Although I have survey data of 2006, the database does not provide information about where the interviewees come from. Thus, it was not possible to identify their federal entities in order to aggregate data of sub national conditions in 2006.

177

up” the effects of its distinct and unmeasured political and economic environment.

In these models I control for random effects at the level of stateyear. I am assuming that the intercepts for the 384 stateyears (32 X 12 = 384) that I analyze here are randomly distributed, conditional on the characteristics of the states across the time. The analysis does this because I have unexplained variance across the states and across the time. People are nested at the stateyears (within the states and the time). For example, I have unexplained variance for people who lived in Aguascalientes 1994, 1995 … 2005. To address this problem, this chapter uses random effects at the stateyears. The models also use fixed effects to take into account another hierarchy in the dataset: each observation is nested both in a state and in a year. It is quite likely that unmeasured characteristics at the state level exert common effects on all of the observations. As a consequence, I include both state and year fixed effects in my models, leaving one arbitrarily chose state

(Aguascalientes) and one year (1994) as our reference category. My model specification then is:

Yij = β0j+ β 1 Xij+ β 2 k+ β 3m + εij (1)

where β0j = γ 0+ γ 1Zj+ δj (2)

In this notation, Yij indicates whether a citizen approves or not the president, Xij is the set of citizens’ features and subnational conditions, β1 is the vector of their coefficients, β0j is the vector of unique intercepts for each state

178

year that is a function of zj stateyears characteristics and γ parameters, β2k is the vector of state fixed effects, β3m is the vector of year fixed effects, and εij is the error for each observation. Equation (2) is then substituted into equation (1) in order to obtain the estimated effects of both, citizens’ features and subnational conditions (aggregate variables)

179

Table 3.1: Explaining Presidential Approval (Multi Level): Variables

180

Concept Question Operationalization and Codification Dependent Do you approve the (1 ) Approve Variable : way the president is (0) Disapprove Presidential handling his job? Approval 49

Independent Variables: PRI Regime Dummy variable, 1 during the PRI Regime (9400); 0 after the PRI era (0105) StateLevel Variables: Unemployment at The data come from the state level INEGI Inflation at the state The data come from level the Banco de México. Number of The data come from convicted criminals the Justice in per 100 inhabitants Mexico Crime by state Statistics Database, TransBorder Institute at the University of San Diego Number of The data were complaints against provided by members of the civil Secretaría de la service per 1000 Función Pública inhabitants by state through the Instituto Federal de Acceso a la Informacin

49 Some of the surveys have a wide range of options: people could respond “totally approve”, “partly approve”, “partly disapprove” and “totally disapprove”. Others have a narrow range options: “approve” or “disapprove”. I recoded the former surveys to correspond with the latter ones in order to increase my number of observations

181

Table 3.1, Continued 50

Concept Question Operationalization and Codification Divided Dummy Government variable, 1 when the governor’s party is different than that of the president’s. 0 if this is not the case IndividualLevel Variables Perceptions of rising Since Salinas (1) Decreased unemployment (Zedillo or Fox) is (2) The same president, do you (3) Increased think that unemployment has increased or decreased in Mexico?

Perceptions of Since Salinas (1) Decreased rising inflation (Zedillo or Fox) is (2) The same president, do you (3) Increased think that inflation has increased or decreased in Mexico?

50 In the second chapter I mentioned that states of Mexico are not obligated to report the number of complaints against members of the civil service involved with corruption. Because of this limitation and the modest number of observations, I took out the corruptionvariables in chapter two. However, I trust more in the reliability of the data on corruption in this chapter because I conduct an econometric analysis that includes almost 15,000 observations. In the appendix 3, I present a multilevel analysis excluding the corruptionvariables. The results are similar to these that I present in the next section.

182

Table 3.1, Continued

Concept Question Operationalization Perceptions of rising Since Salinas (1) Decreased crime (Zedillo or Fox) is (2) The same president, do you (3) Increased think that crime has increased or decreased in Mexico? Perceptions of Since Salinas (1) Decreased rising corruption (Zedillo or Fox) is (2) The same president, do you (3) Increased think that corruption has increased or decreased in Mexico? Partisanship Do you consider Dummy variable: 1 (matching the yourself as a if the voter belongs President’s party) Panista , Priista , or to the president’s Perredista? party; 0 if the voter does not. From 1994 to 2000, 1 if the voter was Priista . From 2001 to 2005, 1 if the voter was Panista. Sociodemographic Male (gender) Mark your gender (1) Male (0) Female Age How old are you? Three dummy variables. Young Mature People (from 26 to 40). Mature People (from 41 to 60).The Elderly (from 61 or more) .The base line group is the youngest people in the surveys, from 18 to 25.

183

Table 3.1, Continued

Concept Question Operationalization and Codification Income Which is your Two dummy weekly minimum variables: Lower salary? Middle Class (people who have a salary from 1 minimum salary, MS, to 3 MS). Middle Class and Rich (from 4 MS or more). The base line group is the people who have a salary lower than 1 MS. Education What year of school Four dummy did you reach? variables. Primary, Junior High Scholl, Senior High School, University or more (People who a Bachelor degree or more). The base line group is the people who do not have scholarly education

3.6 Results

The results of models 1 and 2 in Table 3.2 show that subnational conditions affected presidential approval at the individual level before and after

184

the 2000 transition. The analysis reveals that except for the variable Divided

Government the effects of state factors were stronger during the democratic period than those in the predemocratic era. In the predemocratic period, Model

1, when the number of convicted criminals by state per 100 inhabitants increased, presidential approval did not suffer significant changes. However, this variable turned out to have positive significant effects on approval after the 2000 democratic transition, Model 2. From the voters’ perspective, an increase in the number of convicted criminals was a signal about the democratic government’s efforts to address crime in Mexico. This finding reveals that the institutional political context affects the way in which citizens evaluate the president. In a semiauthoritarian regime, objective indicators of crime did not impact voters’ perceptions of the president. In contrast, in an emerging democracy crime became significant for approval. The analysis also demonstrates that the levels of unemployment and inflation by state affected presidential approval. The higher either inflation or unemployment, the lower approval became.

The story of corruption is a little different than that of crime. The number of complaints against the civil service by state was significant at the 1% level during the PRI era and after the predemocratic period. However, the impact of this variable is negative in Model 1, and positive in Model 2. This finding suggests that during the PRI regime citizens thought that an increase in the number of complaints against the civil service in a state was a signal of rising corruption in Mexico. In contrast, after the transition, voters considered that an

185

increase in the number of complaints against the civil service meant that the democratic government attempted to address this problem in Mexican politics.

Divided Government strongly affected presidential approval before and after the transition, but in an opposite direction. During the PRI regime, Model 1, this variable had negative effects on approval. In the predemocratic period when citizens faced adverse conditions in their states, they punished the president with low levels of approval. The PRI controlled most of the state governorships and from the voters’ perspective the president was the main actor responsible. In contrast, in the democratic period, Model 2, Divided Government had positive effects on voters’ perceptions of the president. After the 2000 transition, more members of the opposition stayed in office as state governors. Divided

Government provided ambiguous and conflicting information about which level of government, either federal or state, to hold accountable for government performance. The president was less likely to be held accountable for government’s failures when the state governor’s party was different than that of the president’s.

The results of models 1 and 2 in Table 3.2 also reveal that perceptions of rising unemployment, inflation, crime, and corruption affected presidential approval before and after the 2000 transition. The analysis shows that the effects of perceptions of rising crime are stronger during the democratic period than those in the predemocratic era. Table 3.3 Model 4 shows that all else being equal when perceptions of rising crime increased by one unit, presidential approval decreased

186

by .06. In contrast, Model 5 reveals that when the same variable increased by one unit, approval decreased by .12. Moreover, the interactive variable PRI X

Perceptions of Rising Crime in Model 6 51 reveals that the effects of perceptions of rising crime are significantly different before and after the transition. This study also reveals that the effects of the number of convicted criminals by state per 100 inhabitants during the predemocratic era are significantly different than those of the democratic period. In Model 6 the interactive variable PRI X the Number of

Convicted Criminals by State is significant at the 1 % level. Therefore, these findings offer strong empirical evidence to approve the first hypothesis of this chapter about the stronger effects of rising crime and the number of convicted criminals by state on presidential approval after the 2000 transition.

The econometric analysis also suggests that we can partly accept the second hypothesis of this chapter about the stronger effects of corruption after the

PRI regime. Model 6 in Table 3.3 shows that while the interactive variable PRI X the Number of Complaints against the Civil Service by State is significant, the variable PRI X Perceptions of Rising Corruption does not have significant effect on presidential approval. Thus, while objective indicators of corruption impact presidential approval differently after the transition, the effects of citizens’

51 In Model 6, I test the hypothesized differences between these two periods, the predemocratic period and the emerging Mexican democracy, by including a battery of interaction effects for each of the key independent variables. Every interactive variable is the product of the variable PRI (the nondemocratic age) and the key explanatory variables. Every time the interactive variable is significant the analysis shows that the different impact of the variable tested is significant by itself, rather than by the different number of observations in models 4 and 5.

187

perceptions of corruption are not significantly different after the PRI regime. As the literature has shown (Davis and Langley 1995), these findings reveal that corruption was a relevant issue when citizens evaluated the president even before the 2000 transition. Mexican citizens thought that during the PRI era the president could address this problem even though he was a member of the PRI. Therefore, we can partly accept the second hypothesis of this analysis, the effects of objective indicators of corruption, the number of complaints against the civil service by state, are stronger after the transition, but this is not the case of voters’ perceptions of this problem.

Model 6 also reveals that the effects of subnational conditions are significantly different and stronger after the transition. All the interactive variables of subnational conditions, the variable PRI multiplied by the levels of unemployment, inflation, crime and corruption, respectively, are significant. This finding suggests that Mexican people had a lot of expectations of a greater competence of the new president in managing both the economy and political problems. In contrast, the only perceptionvariable that had significantly different effects after the transition was perceptions of crime. Partisanship had strong effects before and after the transition, but its effects are not significantly different between these two periods.

188

Table 3.2, Multilevel Models Using StateYears Random Effects, (Random Intercepts)

189

Model 1Pre Model 2 Model 3 Interactive Democratic Democratic Era Test Period (0105) (9405) (9400) StateLevel Variables: Objective Sub national Indicators: Unemployment 4.1 7.02** 7.42*** Level by State (3.3) (3.09) (3.04)

Inflation Level by 4.4*** 10.4*** 10.6*** State (.43) (3.25) (3.2)

Num. Convicted .38 2.*** 1.99*** Criminals by State (.42) (.47) (.47) per 100 Inhabitants Num. Complaints 3.35*** 4.8*** 4.6*** against the Civil (.76) (1.5) (1.5) Service by State per 1000 Inhabitants Divided .32*** .28*** .27*** Government (.07) (.08) (.08) IndividualLevel Variables: Citizens’ Perceptions Perceptions of .64*** .52*** .53*** Rising (.05) (.06) (.06) Unemployment

Perceptions of .57*** .56*** .55*** Rising Inflation (.05) (.06) (.06)

Perceptions of .27*** .6*** .61*** Rising Crime (.04) (.06) (.06)

190

Table 3.2, Continued

Model 1 Pre Model 2 Model 3 Democratic Democratic Era Interactive Test Period (9400) (0105) (9405) Perceptions of .41*** .37*** .37*** Rising Corruption (.04) (.06) (.06)

Partisanship 1.6*** 1.6*** 1.7*** (.06) (.08) (.08) Socio demographic Background Lower Middle .23*** .19** .02 Class (.08) (.09) (.06) Middle Class and .17** .24*** .02 Rich (.09) (.09) (.06) YoungMature .04 .03 .01 (.06) (.08) (.05) Mature .01 .12 .03 (.08) (.1) (.06) The Elderly .04 .15 .09 (.12) (.12) (.08) Primary .03 .33* .17 (.15) (.18) (.11) JuniorHigh .01 .37** .16 School (.16) (.18) (.12) SeniorHigh .02 .44** .18 School (.16) (.19) (.12) (.09) University or More .19 .28 .02 (.17) (.2) (.13) Male .01 .08 .02 (.05) (.06) (.04) TestInteractive Variables PRI .83*** (.23) PRI X Level of 11.7*** Unemployment by (3.8) State

191

Table 3.2, Continued

Model 1 Pre Model 2 Model 3 Democratic Democratic Era Interactive Test Period (9400) (0105) (9405) PRI X Level of 6.15* Inflation by State (3.2)

PRI X Num. 1.6*** Convicted (.6) Criminals by State per 100 Inhabitants

PRI X Num. 7.9*** Complaints against (1.65) the Civil Service by State per 1000 Inhabitants

PRI X Divided .6*** Government (.11)

PRI X Perceptions .11 of Rising (.07) Unemployment

PRI X Perceptions .02 of Rising Inflation (.07)

PRI X Perceptions .34*** of Rising Crime (.07)

PRI X Perceptions .05 of Rising (.07) Corruption PRI X Partisanship .06 (.1)

192

Table 3.2, Continued

Model 1 Pre Model Model 3 Democratic 2Democratic Era Interactive Test Period (9400) (0105) (9405) Year Fixed Included Included Included Effects

State Fixed Included Included Included Effects

StateYears Level 3.66e12 (.00001) 3.33e11 (.00002) 6.45e13 (.00001) Variance

2 X Log 7990.3 5512.4 13521.58 Likelihood

8227 5958 14185 *p<.10, **p<.05, ***p<.01 Note: The dependent variable is presidential approval at the individual level. I include year dummies except for one year and state fixed effects in the estimation of the models but not presented in the table. Some of them are significant. Dy/DX is for discrete change of dummy variable from 0 to 1

193

Table 3.3, Multilevel Models Using StateYear Random Effects (Random Intercepts), Results Expressed as Marginal Effects

194

Model 4 Pre Model 5 Model 6 Interactive Democratic Democratic Era Test Period (0105) (9405) (9400) StateLevel Variables: Objective Sub national Indicators: Unemployment .99 1.39** 1.65*** Level by State (.78) (.61) (.67)

Inflation Level by 1.04*** 2.06*** 2.36*** State (.1) (.64) (.72)

Num. Convicted .09 .39*** .44*** Criminals by State (.1) (.09) (.1) per 100 Inhabitants Num. Complaints .79*** .95*** 1.02*** against the Civil (.18) (.29) (.32) Service by State per 1000 Inhabitants Divided .07*** .05*** .06*** Government (.01) (.01) (.01) IndividualLevel Variables: Citizens’ Perceptions Perceptions of .15*** .1*** .11*** Rising (.01) (.01) (.01) Unemployment

Perceptions of .13*** .11*** .12*** Rising Inflation (.01) (.01) (.01)

Perceptions of .06*** .12*** .13*** Rising Crime (.01) (.01) (.01)

195

Table 3.3, Continued

Model 4 Pre Model 5 Model 6 Democratic Democratic Era Interactive Test Period (9400) (0105) (9405) Perceptions of .09*** .07*** .08*** Rising Corruption (.01) (.01) (.01)

Partisanship .34*** .29*** .33*** (.01) (.01) (.01) Socio demographic Background Lower Middle .05*** .04** .006 Class (.02) (.02) (.01) Middle Class and .04** .04*** .005 Rich (.02) (.01) (.01) YoungMature .01 .007 .003 (.01) (.01) (.01) Mature .002 .02 .007 (.01) (.02) (.01) The Elderly .01 .03 .02 (.02) (.02) (.02) Primary .008 .06* .03 (.03) (.03) (.02) JuniorHigh .002 .07** .03 School (.03) (.03) (.02) SeniorHigh .005 .08** .04 School (.04) (.03) (.03) (.09) University or More .04 .05 .004 (.04) (.03) (.02) Male .002 .01 .006 (.01) (.01) (.009) TestInteractive Variables PRI .18*** (.04) PRI X Level of 2.63*** Unemployment by (.86) State

196

Table 3.3, Continued

Model 4 Pre Model 5 Model 6 Democratic Democratic Era Interactive Test Period (9400) (0105) (9405) PRI X Level of 1.37* Inflation by State (.73)

PRI X Num. .35*** Convicted (.14) Criminals by State per 100 Inhabitants

PRI X Num. 1.77*** Complaints against (.37) the Civil Service by State per 1000 Inhabitants

PRI X Divided .12*** Government (.02)

PRI X Perceptions .02 of Rising (.02) Unemployment

PRI X Perceptions .004 of Rising Inflation (.01)

PRI X Perceptions .07*** of Rising Crime (.01)

PRI X Perceptions .01 of Rising (.01) Corruption PRI X Partisanship .01 (.02)

197

Table 3.3, Continued

Model 4 Pre Model 5 Model 6 Democratic Democratic Era Interactive Test Period (9400) (0105) (9405) Year Fixed Included Included Included Effects

State Fixed Included Included Included Effects

The main finding of this chapter is that since the 2000 democratic transition perceptions of rising crime have become more salient and have had stronger effects on presidential approval. One counterhypothesis can question this result: a real increase in crime that started before the transition, the number of convicted criminals by state, encouraged citizens to evaluate the president differently after the 2000 presidential election. To test this counterhypothesis, in

Table 3.4 I conducted an analysis as a robust test in which I only included those most stable states in the number of convicted criminals, with both high and low crime rate. To conduct this analysis I measured the standard deviation of the number of convicted criminals by state per 100 inhabitants from 1994 to 2205. I

198

used the standard deviation to create a new variable that allowed me to divide my data in quartiles. Thus, I ran econometric models that only included the first and second quartiles, the states with the most stable levels in the number of convicted criminals, those with both high and low levels of crime.

The interactive variable PRI X Perceptions of Rising Crime in Model 9 of

Table 3.4 shows that perceptions of crime had different effects after the transition even in the states with the most stable levels of crime. This finding offers strong empirical evidence to support the first hypothesis of this chapter about a real change in citizens’ approval of the president in the democratic period. After the

2000 presidential election, voters’ perceptions of the president’s determination to address crime had stronger effects on presidential approval. Thus, with results of the current analysis we can reject the counter hypothesis that a real increase in crime mainly encouraged citizens to evaluate the president differently after the

2000 transition.

199

Table 3.4, Multilevel Models Using StateYears Random Effects, (Random Intercepts), Only Including States with the Most Stable Levels of Crime

200

Model 7 Pre Model 8 Model 9 Interactive Democratic Democratic Era Test Period (0105) (9405) (9400) StateLevel Variables: Objective Sub national Indicators: Unemployment 1.3 29.4*** 30.2*** Level by State (3.9) (7.1) (7.1)

Inflation Level by 5.2*** 5.3 4.2 State (.65) (6.1) (5.9)

Num. Convicted 2.58 7.43*** 7.32*** Criminals by State (1.85) (2.88) (2.8) per 100 Inhabitants Num. Complaints 14.5*** 42.7*** 42.6*** against the Civil (5.7) (12.5) (12.3) Service by State per 1000 Inhabitants Divided .5*** .1 .04 Government (.2) (.24) (.24) IndividualLevel Variables: Citizens’ Perceptions Perceptions of .51*** .5*** .51*** Rising (.07) (.1) (.1) Unemployment

Perceptions of .61*** .74*** .74*** Rising Inflation (.07) (.1) (.1)

Perceptions of .29*** .54*** .54*** Rising Crime (.06) (.1) (.1)

201

Table 3.4, Continued

Model 7 Pre Model 8 Model 9 Democratic Democratic Era Interactive Test Period (9400) (0105) (9405) Perceptions of .37*** .34*** .34*** Rising Corruption (.06) (.1) (.1)

Partisanship 1.6*** 1.45*** 1.45*** (.09) (.14) (.14) Socio demographic Background Lower Middle .19* .33** .003 Class (.11) (.14) (.08) Middle Class and .19 .12 .07 Rich (.13) (.17) (.1) YoungMature .15 .03 .09 (.1) (.14) (.08) Mature .13 .14 .04 (.12) (.17) (.09) The Elderly .02 .18 .04 (.19) (.23) (.14) Primary .02 .07 .03 (.2) (.3) (.17) JuniorHigh .13 .23 .16 School (.22) (.32) (.18) SeniorHigh .02 .21 .09 School (.23) (.33) (.19) (.09) University or More .15 .01 .09 (.23) (.35) (.19) Male .05 .05 .01 (.08) (.11) (.06) TestInteractive Variables PRI .14 (.51) PRI X Level of 31.7*** Unemployment by (8.07) State

202

Table 3.4, Continued

Model 7 Pre Model 8 Model 9 Democratic Democratic Era Interactive Test Period (9400) (0105) (9405) PRI X Level of 9.45 Inflation by State (6)

PRI X Num. 4.68 Convicted (.3.4) Criminals by State per 100 Inhabitants

PRI X Num. 56.3 Complaints against (13.5) the Civil Service by State per 1000 Inhabitants

PRI X Divided .54* Government (.31)

PRI X Perceptions .004 of Rising (.12) Unemployment

PRI X Perceptions .13 of Rising Inflation (.13)

PRI X Perceptions .25** of Rising Crime (.12)

PRI X Perceptions .02 of Rising (.12) Corruption PRI X Partisanship .11 (.16)

203

Table 3.4, Continued

Model 7 Pre Model 8 Model 9 Democratic Democratic Era Interactive Test Period (9400) (0105) (9405) Year Fixed Included Included Included Effects

State Fixed Included Included Included Effects

StateYears Level 2.38 e09 4.40e15 (.00003) 1.49e10 (.00002) Variance (.000028)

2 X Log 3624.03 1981.4 5617.48 Likelihood

3756 2153 5909 *p<.10, **p<.05, ***p<.01 Note: The dependent variable is presidential approval at the individual level. I include year dummies except for one year and state fixed effects in the estimation of the models but not presented in the table. Some of them are significant. Dy/DX is for discrete change of dummy variable from 0 to 1

204

3.7 Conclusion

By conducting a multilevel analysis, one of the main theoretical goals of this chapter has been to explain how political institutions affect presidential approval. In this dissertation, I have presented empirical evidence to show that the political regime works as the market when citizens evaluate the president. From the demand side, citizens have expectations about the government, and from the supply side, the president’s behavior to address these concerns is restricted by the political regime. In Mexico since the 2000 presidential election, citizens’ main political concerns have become more relevant and have had stronger effects on presidential approval. I have used Mexico as a study case to analyze how democratization affects presidential approval.

In this dissertation, I have argued that in Mexico since the 2000 transition political issues, crime and corruption have become salient. Citizens consider combating crime to be the second most important task to be addressed by democracy, and view corruption as the main obstacle for a democratic consolidation. In this analysis, I strengthen my previous findings and attempt to overcome the contradictory results between the first and the second chapters.

While in chapter one citizens’ perceptions had stronger effects on presidential approval after the democratic transition, in the second chapter subnational conditions revealed stronger explanatory power on approval before the

205

democratic period. By including objective indicators, citizens’ perceptions and background, the econometric analysis provides more reliable and accurate results.

This chapter has showed that subnational conditions affected presidential approval at the individual level before and after the 2000 transition. The analysis reveals that except for the variable Divided Government the effects of state factors, levels of unemployment, inflation, crime and corruption by state, on approval during the democratic period were stronger than those in the pre democratic era. This finding is consistent with the literature. After the 2000 presidential election, voters had more expectations of a greater competence of the next Mexican president in managing not only economic but also political issues

(Magaloni and Poiré, 2004).

The story of citizens’ perceptions is slightly different than that of sub national conditions. In the multilevel analysis, the only variable with significantly different effects after the transition that measures voters’ perceptions is citizens’ opinion on rising crime. As a robust test, I conducted a restricted analysis by including only the states with the most stable levels of crime. Even in the states with the most sable number of convicted criminals, perceptions of crime had stronger and significantly different effects on presidential approval after the

2000 presidential election. This finding shows that the greatest change in voters’ perceptions of presidential approval after the 2000 democratic transition is how he addresses crime in Mexico. Citizens consider unemployment and inflation to be two important issues in Mexican politics. However, they do not strongly associate

206

economic issues with the mains duties of a democratic government. In the case of corruption, this issue has been relevant for Mexican people even before the democratic transition (Davis and Langley, 1995). The results of this analysis suggest that during the PRI era citizens thought that the president could address this problem even though he was a member of the PRI. Thus, voters’ perceptions of the president’ anticorruption policies did not suffer a change as dramatic as that when they evaluated the president’ anticrime policies. Therefore, the greatest change in citizens’ evaluation of the president can be explained by the stronger effects of voters’ perceptions of the president’s determination to address crime after the 2000 transition.

207

3.8 Appendix 3

Table 3.5, Determinants of Presidential Approval (9405). Multilevel Models Using StateYear Random Effects (Random Intercepts), Results Expressed as Marginal Effects, Excluding CorruptionVariables

208

Model 1 Pre Model 2 Model 3 Democratic Democratic Era InteractiveTest Period (0105) (9405) (9400) StateLevel Variables: Objective Sub national Indicators: Unemployment .62 .3 .33 Level by State (.49) (.48) (.5)

Inflation Level by .92*** 1.4*** 1.3*** State (.07) (.47) (.5)

Num. Convicted .005 .44*** .44*** Criminals by State (.08) (.08) (.08) per 100 Inhabitants Divided .08*** .04*** .05*** Government (.01) (.01) (.01)

IndividualLevel Variables: Citizens’ Perceptions Perceptions of .17*** .13*** .13*** Rising (.009) (.009) (.01) Unemployment Perceptions of .15*** .15*** .15*** Rising Inflation (.009) (.01) (.01)

Perceptions of .08*** .16*** .16*** Rising Crime (.007) (.009) (.01)

Partisanship .33*** .32*** .34*** (.009) (.01) (.01)

209

Table 3.5, Continued

Sociodemographic Background Lower Middle Class .04*** .04*** .005 (.01) (.01) (.01)

Middle Class and .02 .05*** .02* Rich (.02) (.01) (.01)

YoungMature .005 .01 .003 (.01) (.01) (.009)

Mature .01 .03** .02** (.01) (.01) (.01)

The Elderly .01 .06*** .04*** (.02) (.02) (.01)

Primary .002 .005 .001 (.02) (.02) (.01)

JuniorHigh School .01 .03 .006 (.02) (.03) (.02)

SeniorHigh School .04 .03 .005 (.03) (.03) (.02)

University or More .07*** .0001 .04* (.03) (.03) (.02)

Male .008 .02** .01*** (.01) (.01) (.007)

210

Table 3.5, Continued

PRI Hegemony Post PRI Era Interactive Variable Model Test TestInteractive Variables PRI .09*** (.03)

PRI X Level of 1.15* Unemployment by (.63) State PRI X Level of .36 Inflation by State (.48)

PRI X Num. .42*** Convicted (.11) Criminals by State per 100 Inhabitants PRI X Divided .11*** Government (.01)

PRI X Perceptions .04*** of Rising (.01) Unemployment PRI X Perceptions .007 of Rising Inflation (.01)

PRI X Perceptions .09*** of Rising Crime (.01) PRI X Partisanship .04*** (.01) 12277 10781 23058 *p<.10, **p<.05, ***p<.01 Note: The dependent variable is presidential approval at the individual level. I include year dummies except for one year and state fixed effects in the estimation of the models but not presented in the table. Some of them are significant. DY/DX is for discrete change of dummy variable from 0 to 1.

211

3.9 Works Cited

Camp, Roderic. 2004. “Citizen Attitudes toward Democracy and Vicente Fox’s Victory in 2000” Mexico’s Pivotal Democratic Election. Ed. Jorge I. Dominguez and Chappell Lawson: Stanford: Stanford University Press: 2546.

Constitución Política de los Estados Unidos Mexicanos. 2001: Ed. Porrúa

Duch, Raymond M. and Randy Stevenson. 2005. “ Context and the Economic Vote: A Multilevel Analysis”. Political Analysis, Vol. 13. Num. 4: 387 409.

Huber, John D., Georgia Kernell, and Eduardo L. Leoni. 2005. “ International Context, Cognitive Resources and Party Attachments”. Political Analysis, Vol. 13. Num. 4: 365386.

Jones, K. R. J. Johnston and C. J. Pattie. 1992.”People, Places and Regions: Exploring the Use of MultiLevel Modeling in the Analysis of Electoral Data”. British Journal of Political Science, Vol. 22, No. 3: 343380.

Long, Karen and W. Phillips Shively. 2005. “Applying a TwoStep Strategy to the Analysis of CrossNational Public Opinion Data”. Political Analysis, Vol. 13. Num. 4: 327346.

Paterson, Lindsay and Harvey Goldstein. “New Statistical Methods for Analyzing Social Structures: An Introduction to Multilevel Models”. British Educational Research Journal, Vol. 17, No. 4: 387393.

212

Snijders, Tom and Roel Bosker. 1999. Multilevel Analysis. London, UK: Library of Congress.

Steenbergen, Marcos and Bradford S. Jones. 2002. “Modeling Multilevel Data Structures”. American Journal of Political Science, Vol. 46, No. 1: 218 237.

3.10 Web Sites

Banco de Información para la Investigación Aplicada en Ciencias Sociales (BIIACS) http://www.biiacs.cide.edu

Banco de México, http:// www.banxico.org.mx

Instituto acional de Geografía y Estadística (IEGI), http:// www.inegi.gob.mx

Parametría, http://www.parametria.com.mx

TransBorder Institute at the University of San Diego, http://www.sandiego.edu/peacestudies/tbi

Secretaría de la Función Pública http:www.funcionpublica.gob.mx

4 PRESIDETIAL APPROVAL AD ROLL CALL VOTIG I MEXICO:

4.1 Introduction

The previous three chapters of this dissertation have showed that in a context different than the U.S. both crime and corruption affect citizens’ approval of the president. I have analyzed presidential approval at the individual and at the state levels. Moreover, I have also conducted a multilevel analysis that strengthens my findings. This chapter examines the other side of the story, the effects of presidential approval on roll call voting before and after the 2000 democratic transition. Using an original dataset that includes 111 presidential bills, more than 1,500 deputies and 64,000 obervations, I estimate the effects of presidential approval on roll call voting in Mexico in the 57 th , the 58 th , and the

59 th legislatures, from 1998 to 2006.

Scholars (Bond and Fleisher 1980; CanesWrone and de Marchi 2002;

Cohen et al., 2000; Edwards 1976; 1980) have analyzed the influence of public opinion of the president on Congress by focusing their attention on legislators’ decisions. Some of them examine the final outcome of the legislative process,

213 214

whether or not the House votes with the president’s position (CanesWrone and de Marchi 2002; Edwards 1976; 1980). However, to measure the impact of presidential approval on congressmen’s voting behavior is an open debate.

Bond, Fleisher, and Northrup (1988), Cohen et al., (2000) use as dependent variable the percentage support for the president’s positions on roll call voting on which he took a public position . In this chapter, the availability of the data allows me to conduct a microunit analysis of roll call voting. This study analyzes each vote or abstention on every presidential bill of each member in the

Chamber of Deputies in Mexico from 1998 to 2006 52 . The microunit analysis provides the opportunity to examine in detail every step of the legislative process by including all the possible observations of my study. Restricting this analysis to the final outcome in the executivelegislative relationships, whether the House supports the presidential bills, may oversight the dynamic of the legislative deliberation process in Congress as a product of the interaction among congressmen. Moreover, I might not take advantage of the availability and power of the data. In this chapter, except for the first five bills discussed during the 58 th

Legislature, the timedifference between the session of the roll call voting of a bill discussed in Congress and the date of the closer survey was fewer than two weeks. In the case of those five bills, the difference was one month.

52 No data of roll call voting in Mexico exist before 1998.

215

Recent studies in American politics reveal that presidential approval affects legislators’ decisions (Bond and Fleisher 1980; CanesWrone and de

Marchi 2002; Cohen et al. 2000; Edwards 1980; 1989; Pritchard 1983; Rivers and

Rose 1985). When the president enjoys high levels of approval, he is in a better position to bargain with Congress. Therefore, in an old and stable democracy such as the U.S. high approval rating can help the president to implement his plans and polices. In an emerging democracy such as Mexico, high levels of approvals are important because both the president is in a better position to bargain with

Congress and presidential approval may be a helpful political instrument for the process of consolidation through the executivelegislative agreements.

Mexico is an ideal case to test the effects of presidential approval on roll call voting in two different institutionalpolitical contexts: the transitional era during the last years of the semiauthoritarian PRI regime and the Mexican new democracy. This fourth chapter is guided by one research question: how different are the effects of presidential approval on roll call voting during the last years of

PRI in office from those after the Mexican democratic transition?

Mexico is a challenging case to investigate. While Vicente Fox, the president of the democratic transition who enjoyed high levels of approval, was not very successful at bargaining with Congress, the Priista president Ernesto

Zedillo, less popular than Fox, enjoyed congressmen’s support in the legislative arena. Although Fox in his last three years in office sent 77 bills to the Mexican

216

Congress and Zedillo only 56 in the same period, the great difference is in the percentage of these bills that had political relevance and were accepted. Out of a total of 56 bills sent by Zedillo to the Chamber of Deputies, 50 were approved by congressmen, 89 percent. Out of a total of 77 bills sent by Fox, only 59 were supported by Mexican deputies, 76 percent. Moreover, 25 of these 50 bills approved by the Chamber of Deputies during the last three years of Zedillo’s tenure were high important bills. This means 50 percent. In contrast, 41 bills out of a total of 59 were rejected by Mexican deputies during the second half of Fox’s administration. These bills might be considered as high important projects and policies. 70 percent of them were rejected (Espinoza and Miguel 2007).

The previous figures challenge the American literature of presidential approval. Most scholars in American politics have, implicitly, assumed that popular presidents should be more successful than unpopular executives when they bargain with Congress. Edwards (1980) finds that the president is in a better position to bargain with Congress about domestic issues when he enjoys high levels of approval. Pritchard (1983) states that a popular president who is responsible for a policy area receives support from his party’s congressional members. Rivers and Rose (1985) estimate that a 1 percent increase in a president’s public support level increases the president’s legislative approval rate by approximately 1 percent. The story is different in Mexico, a popular president was not successful in Congress.

217

To analyze the effects of presidential approval on roll call voting in

Mexico is necessary to examine the political context before and after the transition. Because of his legislative powers, the president has the faculty to send bills to the Congress. According to the Constitución Política de los Estados

Unidos Mexicanos (2001), the Mexican Chamber of Deputies is filled through two different principals of representation 60 percent of the 500 seat House or 300 deputies are elected in single member district (SMD) plurality races, while 40 percent or 200 reach the House through five regional proportional representation

(PR) districts that elect 40 deputies each from a closed list. The no immediate reelection clause of the Mexican Constitution dictates that federal deputies must search out new posts after each period and so cannot rely on incumbency and long term legislative careers. The prohibition of immediate reelection for deputies cut the electoral connections between congressmen and their electorate. Therefore, the no reelection clause allowed the Mexican presidents to enjoy an almost absolute domain over the legislative branch for many decades, regardless if public opinion approved of the president’s policies or not. The legislators’ fate was in the president’s hands. Although the Mexican Constitution has not modified the no immediate reelection clause, the new political context that emerged after the transition allows me to expect that the effects of presidential approval on roll call voting during the PRI regime should be different than those in the democratic period.

218

Although after the democratic transition the president kept some influence over his party members, the president did not hold the power to nominate most candidates to elective office for his party. Mexican presidents are less powerful than they used to be. After the 2000 presidential election, the president has handed over power to political parties for the nomination of candidates (Béjar

2004; 2006; 2009; Béjar and Waldman 2004). Moreover, Langston (2007) finds that in spite of the prohibition of immediate reelection, Mexican deputies can respond to voters’ preferences if legislators can maximize their benefits with this decision. Since the emergence of the democratic period, the president has not enjoyed an absolute domain over the legislative power, and he has bargained with congressmen.

This chapter has two main goals: (1) To analyze the impact of presidential approval on roll call voting in Mexico from 1998 to 2006. (2) To show that the effects of presidential approval on roll call voting in Mexico during the last years of the PRI regime are different from those after the transition. In this chapter, I argue that the effects of presidential approval on congressmen’s decisions are different after the democratic transition. Given the control of the president over the Congress during the PRI regime, presidential approval did not have effects on roll call voting before the 2000 presidential election. After the 2000 presidential election, the politicalinstitutional context changed. The president had less influence on

219

congressmen, even on his party members, and presidential approval became important and began to affect roll call voting.

To support my argument this chapter is divided as follows. The first section of this article provides background information on the effects of presidential approval from the perspective of American politics, and a review of the comparative politics literature. The second section examines the context of

Mexican politics before and after the 2000 democratic transition, focusing my attention on the executivelegislative relationships. Third, I formulate my hypotheses, and explain the data and method. Fourth, I present a logistic analysis in order to test my main argument about the different effects of presidential approval on roll call voting after the 2000 transition. Finally, the conclusion summarizes the main findings of this analysis.

4.2 The Review of the Literature

Recent studies in American politics ( Bond and Fleisher 1980; CanesWrone and de Marchi 2002; Cohen et al. 2000; Edwards 1980; 1989; Pritchard 1983;

Rivers and Rose, 1985 ) have found that presidential approval affects legislators’ decisions. However, the debate is open and divided into two main branches: The first approach states (CanesWrone and de Marchi, 2002; Edwards 1980; 1989;

Pritchard 1983; Rivers and Rose 1985) that presidential approval is a key determinant when the president bargains with Congress. The second perspective

220

claims (Bond and Fleisher 1980; Bond, Fleisher, and Northrup 1988; Cohen et al.

2000) that approval has only marginal effects on legislators’ decisions.

Edwards (1980) was the first scholars that tested the effects of approval on legislators’ support. His analysis reveals strong correlations between presidential popularity and congressional support. Edwards (1989) extends his argument and finds that presidential approval ratings are correlated with legislators’ support in the Senate but not in the House. In the same line of thought, Pritchard (1983) finds that presidents influence congressmen’s voting behavior on domestic issues due to high levels of approval rating. CanesWrone and de Marchi (2002) claim that when a popular president promotes a legislation that is salient and complex, he has a lot of opportunities to influence legislators’ voting.

From the second perspective, in American politics some scholars (Bond and Fleisher 1980; Bond, Fleisher, and Northrup 1988; Cohen et al. 2000) have claimed that approval has only marginal effects on congressmen’s support. Cohen et al., (2000) find that presidential approval itself does not have any effect on senatorial support, but approval has a “positive” indirect effect on opposition party members’ voting behavior through partisanship. Bond and Fleisher (1980) state that partisanship and ideology are the strongest predictors on legislators’ roll call voting. According to these scholars, the effects of approval on congressmen’s support are only marginal

One central point in the debate is how partisan considerations condition the relationship between presidential approval and congressional support.

221

Scholars (Bond and Fleisher 1980; CanesWrone and de Marchi, 2002; Cohen et al. 2000; Edwards 1976; 1977; 1980) find that partisan attachments affect the effects that presidential approval has on roll call voting. The presidents’ popularity may influence congressional decisions to support or oppose his positions between elections.

Edwards (1980) suggests two reasons why public opinion of the president may influence congressional decisions to support the president. First, congressmen who seek reelections can use presidential approval as a signal of citizens’ preferences. High approval rating is a signal of voters’ support for the president’s policies. To advance in their political careers congressmen should approve of the president’s policies in Congress. Research on congressional elections reveals that voters’ evaluations of the president affect the reelections changes of members of his party in Congress (Kernell 1977; Piereson 1975; Tufte

1975). If the president is popular, members of his party benefit from being associated with him. If he is unpopular, voters tend to punish his partisans for his unpopular policies. Second, congressmen use approval as a signal of voters’ preferences to satisfy their constituencies (Edward 1977). Davidson (1969) find that about onethird of House members agree that a representative should work for what his constituents want even though this may not always agree with his personal values.

Bond and Fleisher (1980, 71) provide two main arguments that justify why

“for members of the opposition party in the House, presidential popularity is

222

negatively associated with support”. First, members of opposition are not likely to be impressed by a popular president. Popularity with the public may cause presidents to be less compromising in their dealings with Congress. If they propose partisan policies, then this uncompromising spirit is likely to lead to increased partisan voting, and hence, less support from the opposition (Bond and

Fleisher 1980; Bond, Fleisher, and Northrup 1988). Second, members of the president’s party tend to receive credit for the president’s policies even if they do not support them in Congress, and members of opposition are not likely to receive credit even if they do. Therefore, members of the opposition are likely to follow their basic partisan predisposition and oppose the positions of popular presidents.

They have little to gain from their support ad much to lose if the president succeeds (Bond and Fleisher 1980; Bond, Fleisher, and Northrup 1988).

Thus, favorable public opinion may not influence Congress to approve more of the president’s proposal in all situations (Bond and Fleisher 1980; Bond,

Fleisher, and Northrup 1988). If the president’s party controls Congress, then favorable public opinion may improve his rate of success because he gains support from members of his party at a faster rate than he loses support from the opposition. However, if the opposition party controls Congress, then favorable public opinion may not increase the president’s rate of success.

The vast literature of American politics on the president’s influence on congressmen’s political behavior contrasts with the literature of comparative politics. Studies focused on the executivelegislative relationships in comparative

223

politics are rare. Calvo (2007) finds that presidential approval influenced congressmen’s decisions in Argentina from 1983 to 2001. A popular president is in a better position to bargain with Congress. Mejia examines the effects of approval on the president’s success in Congress in Ecuador, but he did not find effects of approval on legislative support (2004). Although these two studies have significantly improved our understanding of presidential approval in contexts different than that of the U.S., they suffer from two main limitations: (1)

Both of these studies analyze the effect of approval after a democratic transition.

Thus, it is not possible to compare the effects of approval in two different political regimes. (2) Mejia (2004) and Calvo (2007) focus their attention on the final outcome of the legislative procedure, whether a presidential bill was approved by the Congress. In this chapter, I attempt to overcome the limitations of these studies. My analysis examines the effects of presidential approval before and after the democratic transition Moreover, rather than analyzing the final outcome of a presidential bill in Congress, I conduct a microanalysis by paying attention to each legislator’s vote for every bill.

In the Mexican case, scholars have ignored the relationship between presidential approval and the president’s success in Congress because two main reasons: (1) During the PRI era Mexican presidents could hold an almost absolute domain over the legislative power. Therefore, presidential approval did not have any impact on the executivelegislative relationships. (2) The lack of survey data made impossible to test if presidential approval might have any effect on

224

legislator’s decisions. The practice of roll call voting did not exist before 1998 in

Mexico. This chapter is the first study that analyzes the impact of presidential approval on roll call voting in Mexico. I use the literature on American politics to examine the Mexican case.

Bond and Fleisher’s arguments (1980) may shed light on our understanding of the effects of presidential approval on roll call voting in Mexico.

This analysis attempt to solve this puzzle: Why was Vicente Fox, a popular and democratically elected president, less successful in Congress than Ernesto Zedillo, the last president of the predemocratic period? The empirical evidence reveals that a dramatic change of approval occurred in Mexico after the 2000 democratic transition. Figure 4.1 shows that while Zedillo reached a 68 percent of approval rating in November 2000, Fox enjoyed an 85 percent in January 2001. Most of the citizens loved Vicente Fox, the president of the democratic transition, except the

Mexican congressmen.

225

90 85 80 68 70

60 50 Presidential Approval 40 30 20 10

Presidential Approval Rating (Percentage) Zedillo Novemeber 2000 Fox January 2001

Figure 4.1: Presidential Approval in Mexico before and after the 2000 Democratic Transition. Figure Done by the Author. The Data come from the BIIACS, www.biiacs.cide.edu

Figure 4.2 shows that while the president Vicente Fox was not very successful at bargaining with Congress, the Priista Ernesto Zedillo, less popular than Fox, acquired congressmen’s support in the legislative arena. While Fox’s average approval rating during the 58 th Legislature was 66 percent, Zedillo’s reached only 55 percent in the 57 th Legislature. The great difference is that while

50 percent of the bills approved by the Mexican deputies to Zedillo in the 57 th

Legislature might be considered as high important bills, only 15 percent of the bills approved by the Chambers of Deputies to Fox in the Legislature 58 th were important bills. Fox was more popular than Zedillo, but he was less successful to bargain with Congress.

226

70 66 60 55 50 50

40 Percentage 30 20 15 10 0 Bills Passed in House the Zedillo's Approval Important Bills Fox's Approval in Important Bills in the 57th Accepted to the 58th Accepted to Fox Legislature Zedillo in the Legislature in the 58th Percentage of Approval and PresidentialPercentage 57th Legislature Legislature

Figure 4.2: Presidential Approval vs. Presidential Success in the House before and after the 2000 Transition. Figure Done by the Author. The Data Come from Sistema de Información Legislativa, sil.gobernacion.gob.mx and the BIIACS, www.biiacs.cide.edu

Examining the composition of the Chamber of Deputies by party labels in the 57 th , 58 th , and 59 th Legislatures provides insights to understand why popularity was not a key determinant when both Zedillo and Fox bargained with Congress.

According to Bond and Fleisher (1980), the control of the president’s party in

Congress may determine the direction of the effects of presidential approval on roll call voting. If the president’s party controls Congress, presidential approval has positive effects on roll call voting. If this is not the case, the effects should be negative. Figure 4.3 displays the composition of the Chamber of Deputies from the 57 th to the 59 th Legislature. Although in 1998, the president’s party, the PRI, did not control the absolute majority of the seats in the Chamber of Deputies in the 57 th Legislature, they kept almost 50 seat shares. In contrast, the president’s

227

party lost political strength in Congress after the 2000 transition. In the 59 th

Legislature, the president’s party, the PAN, only held 29 seat shares. Therefore, although Vicente Fox enjoyed high approval rating during both the 58th and the 59

Legislatures, he could not take advantage of his popularity because his party did not control the Chamber of Deputies, and the PAN’s presence in Congress became weaker from the 58 th to 59 th Legislature. I analyze the political context in

Mexico before and after the 2000 transition in the next section, focusing my attention on the executivelegislative relationships. This analysis allows me to turn the empirical evidence presented into the hypotheses of this chapter.

80 71 70 61 60 52 48 50 39 40 Percentage 29 30 20 10 0 Members of Members of Memebers Members of Members of Members of the the of the the the the President's Opposition President's Opposition President's Opposition Party L57th L57th Party L58th L58th Party L59th L59th

Figure 4.3: Composition in the Chamber of Deputies by Partisanship. Figure Done by the Author. The Data Come from the Chamber of Deputies in Mexico, www.camaradediputados.gob.mx

228

4.3 The ExecutiveLegislative Relationships in Mexico

During the PRI era, Congress in Mexico played a rather poor role in law making. It was not only that congressmen were seldom initiators of law proposals, but also that they did not play the role of stopping or substantially amending those bills sent by the executive. Congressional checking of presidential powers was not a practice either. Although the legislative branch was endowed with specific and direct powers to check executive acts, during the PRI domain Congress seldom made use of them.

The Mexican Constitution (2001) grants the Mexican Congress the power to provide specific direction to policy through legislation and to oversee the policy implementation and regulatory acts of the executive. While these powers are counterbalanced by the president’s array of legislative powers (which include veto powers, the right to initiate legislative proposals, rule making authority, and decree powers) the Mexican president is not extraordinary strong on paper (Casar

2002, 117). However, before the 2000 transition in Mexico, the Mexican political regime was characterized by a great control of the executive. This is the question to answer: where did the informal powers of the Mexican presidents come from during the PRI domain? They came from three conditions: unified government, the recognition of the president as a party leader, and party discipline.

The first condition, unified government, was related to the formation of the

Revolutionary National Party (PNR), the antecedent of the PRI. After the

229

formation of the PNR in 1929, unified government was secure. The official party controlled the presidency and majorities in both chambers of Congress between

1929 and 1997. In fact, before 1964, all senators, and virtually deputies, were members of the official party, renamed the PRI in 1946 (Weldon 2002, 378).

The second condition was the recognition of the president as a party leader during the PRI era. Between the mid1940s and the mid1990s, the president remained the undisputed leader of his party. He generally held the power to nominate most candidates to elective office for the party, including the power to name his own successor. Without reelection, no one else in the party could build an independent geographical electoral base that might eventually challenge the authority of the national party leadership (Weldon 2002, 382).

The third condition for the maintenance of the metaconstitutional power of the Mexican presidency during the PRI era was high party discipline. This condition is related to one of the most dramatic reforms in the Mexican political regime: the prohibition of immediate reelection for deputies and senators. These constitutional reforms, passed by Congress in late 1932 and ratified in early 1933, also extended the noreelection prohibitions to state legislators and mayors, where the impact was probably greater. The prohibition on consecutive terms cut the electoral connections between the members of Congress and their electorate

(Weldon 2002, 380). The deputies had to be less concerned over local issues and more likely to respond to cues from other political actors. The National Executive

Committee (CEN) of the PNR was the natural resource of these cues. It had

230

already been taking an active role in the legislative process by presenting bills and seeking greater discipline among the deputies. At about the same time, the CEN of the PNR had begun centralizing the nominations process so that it could maintain greater control over who could use the party label (Weldon 2002, 381).

Reforms within the party in the late 1930s made sectorial organization in the

PNR, after 1938 the Mexican Revolutionary Party (PRM), as important as the geographical organizations. Political careers could be made in the PRM through the sectors, without ever having to worry about local politics.

These three factors of metaconstitutional presidentialism (unified government, the recognition of the president as head of the party, and high party discipline) were necessary conditions for the domain of the executive power over the Congress. After the 2000 presidential election and the victory of Vicente Fox, the first non Priista president, the political context changed. The meta constitutional conditions that made the Mexican presidents very powerful were relaxed.

The relaxation of the first condition, unified government , affected significantly the executivelegislative relationships. Since 1997 divided government has been the rule in Mexico. Two presidential elections, 2000 and

2006, and two intermediate 2003 and 2009, have passed and the president’s party has not been able to get the majority in both chambers, the House and the Senate.

Scholars (Béjar 2001; 2004; 2006; 2009; Casar, 2002; Nacif, 2004; Weldon,

2002) have argued that with the emergence of divided government the executive

231

has altered his political behavior. While during the PRI hegemony, the president was the main “legislator”, the principal actor who sent bills to the Congress, since the first divided government in 1997, the president has handed over power to political parties in Congress. Since 1997, divided government has been the rule in

Mexican politics, and at the same time, the strength of the president’s party has been eroded in the last legislatures. While in 1997, during the 57 th Legislature, the

President’s party, the PRI, could keep almost 50 seat shares in the Chamber of

Deputies, after the transition, in the 59 th Legislature, the president’s party, the

PAN, held only 29.

The relaxation of the second condition, the recognition of the president as party leader , has altered legislators’ behavior. Although after the transition the president kept some influence over his party, the president did not hold the power to nominate most candidates to elective office for his party. In fact, Vicente Fox, the presidential candidate of the PAN, did not participate in the nomination process of members of his party to elective office in the 2000 elections. For the

2003 intermediate elections, even though Fox was the president, he did not have a direct intervention in the nomination process of candidates to elective office

(Reveles 2004). The president has handed over power to political parties for the nomination of candidates. According to Béjar (2004) after the transition, political parties have reinforced their internal regulations to guarantee their control over congressmen. If during the PRI hegemony with the control of one party in both branches, the executive and the legislative powers, the main principal who

232

influenced legislators’ political future was the President, for Béjar (2004) after the

2000 presidential election the main political actors who began to control congressmen’s faith were party leaders.

The relaxation of the third condition, party discipline, has affected deputies’ strategies to advance in their political careers. Although this condition has not been strictly relaxed because the prohibition of immediate reelection for deputies and senators is maintained in the Mexican Constitution (2001), the relaxation of the two previous conditions (unified government and the recognition of the president as party leader) has weakened party disciplined. Recent empirical evidence reveals that the three major parties (the PRI, PAN, and PRD) have experienced some subnational divisions, and their party cohesion has been eroded since the 57 th Legislature (Cantú and Desposato 2009).

Moreover, Langston (2007) argues that despite the prohibition of immediate reelection for Mexican deputies, in some conditions they respond to voters’ preferences, ignoring either the president or party leaders as principals.

Even though Mexican deputies cannot be reelected immediately, as rational actors they want to maximize their preferences and advance in their political career. The electoral connection between deputies and their constituencies are not totally broken, it is only adapted to the Mexican political context. Mexican deputies attempt to advance in their political careers. They want to become either a senator or a state governor. Satisfying citizens’ policy preferences at the state level, deputies search for higher positions in politics. Of all the senators who served

233

between 1982 and 1994, almost 68 percent had been federal deputies at least once prior to their arrival in the upper house, and 33 percent became senators immediately after leaving the Chamber Of all the governors who served between

1976 and 1995, approximately 52 percent had been federal deputies at least one time prior to their gubernatorial positions. 21 percent had been deputies just before becoming gubernatorial candidates (Ugalde 2000: 109).

4.4 Hypotheses

Mexico experienced significant changes in the executivelegislative relationships after the 2000 democratic transition: the presence of divided government as a rule in Mexican politics, the weak recognition of the president as the leader of his party, and some “relaxation” of party discipline. Although divided government has been the rule since 1997 in Mexican politics, its presence and effects have been more evident during the democratic period because the strength of the president’s party has been eroded after 2000. While during the 57 th

Legislature the President’s party, the PRI, could keep almost 50 seat shares in the

Chamber of Deputies, after the transition, in the 59 th Legislature, the President’s party, the PAN, held only 29 seat shares. After the 2000 presidential election, the

234

recognition of the president as a party leader became weak. In contrast to the PRI regime, when the president held the power to nominate most candidates to elective office, after 2000, political parties, including the president’s, the PAN, regularly began to use primaries as the main method for nomination. Although the prohibition of immediate reelection for deputies and senators is maintained in the

Mexican Constitution (2001), party discipline has experienced some “relaxation” as a consequence of the political changes in Mexico after the transition, the presence of divided government as a rule in Mexican politics and the weak recognition of the president as a party leader. Therefore, my expectations are twofold:

H. 1 Given the control of the President over the Congress during the PRI regime,

all else being equal, presidential approval does not have effects on roll call

voting before the 2000 transition.

H. 2 Given that the conditions that strengthened the metaconstitutional attributes

of the President were relaxed after the transition, all else being equal,

presidential approval affects roll call voting after the 2000 presidential

election. More precisely, the effects of presidential approval on roll call

voting should be affected by partisanship. Approval should have positive

effects on roll call voting of the president’s party members, and negative

effects on the opposition.

235

4.5 Data and Method

The surveys on which this study is based were organized by the Mexican

Office of the Presidency (MOP). The questionnaires were designed within the

Presidency, and the survey was conducted by Opinion Profesional S.A. de C.V., a private company specializing in policy polling and focus groups. In total the surveys contain more than 20, 000 interviews, from 1998 to 2006, covering two different presidents, Ernesto Zedillo during the 57th Legislature, 19982000, and

Vicente Fox, the 58 th and 59 th legislatures, 20002006. Some of the data come from household national facetoface surveys and other from national telephone surveys.

My dependent variable is roll call voting in Mexico from 1998 to 2006, covering the 57 th (19982000), 58 th (20002003) and 59 th (20032006) legislatures. I have created an original database with the roll call voting of more than 1, 500 Mexican deputes. In this chapter, roll call voting refers to each vote or abstention on every presidential bill of each member in the Chamber of Deputies in Mexico from 1998 to 2006. The analysis includes 111 presidential bills and more than 64,000 observations. I use the Chamber of Deputies rather than the

Senate because of the accessibility and availability of the data. No data of roll call voting in Mexico exist before 1998.

My key independent variable is presidential approval at the national level.

I used in most of the cases the level of approval that the president reached one or

236

two weeks before the roll call voting of the presidential bill. When it was not possible, the minority of the cases, I used the level of approval after roll call voting, one or two weeks later. Except for the first five bills discussed during the

58 th Legislature, the timedifference between the session of the roll call voting of a bill discussed in Congress and the date of the closer survey was fewer than two weeks. In the case of those five bills, the difference was one month.

Consistent with the findings of prior work, I use as control variables the number of presidential bills sent by the executive to the Congress, partisanship, the political importance of the bill, and fiscal incentives. Rivers and Rose (1985) find that when a president offers a large number of policy proposals the likelihood on Congress approving a specific proposal falls. Larger programs, measured as the number of presidential legislative requests, are likely to be ambitious in terms of timeconsuming. The probability that Congress approves of a specific proposal falls because of time constraints. Edwards (1980) claims that partisanship affects the president’s success in Congress. Presidential approval has stronger effects on members of the president’s party. Members of the president’s party who seek reelection should run on his record as well as on their own, they have an incentive to help him succeed. CanesWrone and de Marchi (2002) state that the salience of a bill affects the effects of presidential approval on roll call voting. Congressmen use presidential approval as a signal of voters’ preferences to advance in their political careers. However, president’s approval ratings can signal citizens’ preferences over the policies on the executive agenda only if the policies have

237

some degree of public salience. Langston (2007) finds that orphan deputies, those who do not belong to the governor’s party respond to voters’ preferences. Orphan deputies may use approval as a signal of citizens’ preferences. Orphan deputies support a popular president in order to advance in their political careers. Calvo’s analysis (2007) shows that the fiscal incentives that state governments receive from the federal government affect legislators’ vote. The stronger support that the subnational governments receive from the federal government, the more likely that deputies will approve of the presidential bills. This analysis also uses as control the type of deputy in the legislative arena. Assuming that presidential approval may have stronger effects on Regular deputies’ behavior, those who may be more exposed to public opinion, than Alternate deputies’, those who sporadically participate in the legislative debate. Because no data about the ideological preferences of Mexican legislators exist, this study cannot control for the effects of ideology on roll call voting. In contrast to the U.S., no interest groups in Mexico rate congressmen’s ideological preferences.

In order to test my hypotheses about the effects of presidential approval on roll call voting in the Chamber of Deputies, I conduct a multilevel analysis of congressmen’s voting behavior during the PRI regime, the 57 th Legislature, and after the transition, the 58 th and 59 th legislatures.

With a dichotomous dependent variable, there are logit models. Because the explanatory variables are measured at the different levels, the political context when the roll call voting of a presidential bill takes place (presidential approval,

238

the number of presidential bills yearly, the political importance of the bill, and the

GDP per capita income that each state receives) and individual deputies’ characteristics, I use multilevel modeling techniques to acknowledge this data structure explicitly. Deputies’ voting behavior are nested within the political environment of a legislature in a specific time. My covariates may not measure all of these forces, creating a correlation between the prediction errors across all the observations introduced.

I incorporate this potential autocorrelation by estimating a model that allows each deputy (1747 in three Legislatures: 57 th , 58 th , and 59 th ) to have his/her own intercept and thus his/her own average predictors error to “soak up” the effects of his distinct and unmeasured characteristics. In these models I control for random effects at the level of the deputies. I am assuming that the intercepts for the 1747 deputies that I analyze here are randomly distributed condition on the characteristics of the deputies in their own political environment across the time.

The analysis does this because I have unexplained variance across the political environment of a legislature across the time. For example, I have unexplained variance for deputies who represented Aguascalientes in the 57 th Legislatures in

1998, 1999 and 2000. Moreover, every vote of a deputy should be correlated across the time. Deputies’ voting behavior should be consistent. While some legislatures reject the president bills, others always approve them. The models also use fixed effects to take into account another hierarchy in the dataset: each observation is nested in a legislature, in a state, and in a year. It is quite likely that

239

unmeasured characteristics at the legislature and the state levels exert common effects on all the observations across the time. As a consequence, I include legislature, state, and year fixed effects in my models, leaving one arbitrarily chose legislature (the 57 th ) one state (Aguascalientes), and one year 1998 as our reference category.

240

Table 4.1 Explaining Roll Call Voting in Mexico: Variables

Concept Operationalization and Codification Dependent variable Roll call voting on the President’s Bill in the Chamber of Deputies (1998 2006). It can take two values: 0 or 1. Roll call voting refers to each vote or abstention of every presidential bill of each deputy. Independent variables: Aggregate variables: PRI Regime A dummy variable that takes the value of (1) during the PRI regime, the 57 th Legislature, and (0) if this is not the case (the 58 th and 59 th legislatures). Presidential Approval at the national The percentage of people who support level the President before roll call voting. I aggregate individual survey data at the national level. I use a standard question in all the surveys: “Do you approve the way the president is handling his job?” Number of Bills Yearly The yearly number of bills sent by the president to the Chamber of Deputies Political Importance of the Bill 53 The most reliable scale available of the president’s bills from 1998 to 2006 (Espinoza and Miguel, 2007), Metropolitan Autonomous University. It is an index that takes into account the judicial importance of the bill, the salience of the bill (media coverage), and the debating process of the bill. Ramo 28 The GDP per capita income that each state receives from the federal government. The data come from Anexo 6 Informe de Gobierno. 2006.

53 Few studies analyze the legislative significance of a bill in the Mexican Congress. Puente (2004) proposes an interesting methodology using five main criteria: the number of people affected, media coverage, government publicity, economic consequences, and ideological significance. However, his analysis only covers one legislature in the Mexican Senate (20002003), and it does not provide data for the House. Thus, this chapter cannot use Puente’s methodology.

241

Table 4.1, Continued

Concept Operationalization and Codification Individuallevel variables President’s party Matching the president’s party. (1) If the deputy belongs to the same party to that of the president. (0) If this is not the case. Orphan (1) If the deputy belongs to a different party from that of his/her state governor. (0) If this is not the case Regular (1) For Regular deputies (0) For Alternate deputies PAN (1) If the deputy is Panista. (0) If this is not the case PRI (1) If the deputy is Priista. (0) If this is not the case PRD (1) If the deputy is Perredista. (0) If this is not the case 54

4.6 Results

In Table 4.2, Models 1 shows that presidential approval affects roll call voting before the democratic transition. This is an unexpected finding. Contrary to the hypothesis one, high approval rating of the president influences deputies’

54 For these three categories, PAN, PRD, and PRI the reference groups is Independents and small party deputies.

242

voting behavior. Model 1 analyzes the impact of presidential approval on roll call voting during the last two years of the PRI in office, a semiauthoritarian regime in transition. This chapter has explained that during the 57 th Legislature, when the

PRI still held the power, the metaconstitutional conditions that made Mexican presidents very powerful had changed in some way. Divided government emerged for the first time in 1997, and the Priista president, Ernesto Zedillo, lost control of the Chamber of Deputies. This result suggests that the political institutional context in Mexico two years before the transition did not allow the president to influence the Congress without citizens’ support, and congressmen thought that a president who enjoyed voters’ approval was in a better position to bargain with them.

Model 1 reveals that some of the aggregatevariables significantly affected roll call voting. These are the cases of the political importance of a presidential bill and Ramo 28. When a presidential bill becomes salience, to reach agreements turns difficult because political actors have strong and different political preferences. The result of Ramo 28 indicates that during the predemocratic period those deputies who represented states that received more financial support from the federal government supported the president in the Congress. Model 1 also shows that most of the individuallevel variables influenced deputies’ voting behavior before the democratic transition. While members of the president’s party supported the president, orphan deputies disapproved of the presidential bills. In the first case, members of the president’s party who seek reelection must run on

243

his record as well as on their own, they have an incentive to help him succeed. In the second case, during the PRI regime orphan deputies did not support the presidential bills, when presidential approval increased. This is an unexpected result. Orphan deputies did not use approval as a signal of citizens’ preferences to advance in their political careers before the transition. Most of the orphan deputies before the 2000 transition were not member of the president’s party, the

PRI. This explains why orphan deputies did not support the presidential bills in

Congress, when presidential approval increased.

Model 2 shows that after the transition the effects of presidential approval on roll call voting turned negative. This result is consistent with some findings in

American politics (Bond and Fleisher 1980; Bond, Fleisher, and Northrup 1988).

Favorable public opinion may not influence Congress to approve more of the president’s proposal in all situations (Bond and Fleisher 1980; Bond, Fleisher, and

Northrup 1988). If the president’s party controls Congress, then favorable public opinion may improve his rate of success because he gains support from members of his party at a faster rate than he loses support from the opposition. However, if the opposition party controls Congress, then favorable public opinion may not increase the president’s rate of success. Thus if the president’s party controls

Congress, presidential approval has positive effects on roll call voting. If this is not the case, the effects turn negative.

In the Mexican case, although the president’s party, the PRI, did not control the absolute majority of the seats in the Chamber of Deputies in the 57 th

244

Legislature (19982000), they kept almost 50 seat shares. In contrast, the president’s party lost political strength in Congress after the 2000 transition. In the 59 th Legislature, the president’s party, the PAN, only held 29 seat shares.

Therefore, the results of Model 2 suggest that even though Vicente Fox enjoyed high approval rating during both the 58 th and the 59 th legislatures, he could not take advantage of his popularity because his party did not control the Chamber of

Deputies, and the PAN’s presence in Congress became weaker from the 58 th to

59 th Legislature. After the 2000 democratic transition, high presidential approval rating encouraged deputies to embarrass the president and to frustrate his plans in

Congress, as the best strategy to advance in their political careers.

Model 2 also reveals both similar and different results than those of Model

1. The salience of a presidential bill and partisanship significantly affected roll call voting during the democratic period. However, in contrast to Model 1, in

Model 2 the number of presidential bills sent to the Congress was statistically significant at the 1 % level. This result shows that in an emerging democracy such as Mexico, congressmen have not been as permissive to the President as they were during the semiauthoritarian PRI regime. Moreover, while during the pre democratic period Ramo 28 strongly affected roll call voting, after the transition this variable was not statistically significant. This finding suggests that in an emerging democracy political strategies such as patrimonialism and centralization do not affect congressmen’s voting behavior.

245

In Model 3, I test the hypothesized differences between these two periods, the predemocratic period and the emerging Mexican democracy, by including a battery of interaction effects for each of the key independent variables. Every interactive variable is the product of the variable PRI Regime (the nondemocratic period, during the 57 th Legislature) and the key explanatory variables. Every time the interactive variable is significant the analysis shows that the different impact of the variable tested is significant by itself, rather than by the different number of observations in models 1 and 2. This is the case of presidential approval, the number of bills sent by the President, partisanship, and the salience of the presidential bill.

246

Table 4.2 Effects of Presidential Approval on Roll Call Voting in Mexico (98 06) Multilevel Models Using Deputies Random Effects (Random Intercepts)

247

Model 1 Pre Model 2 Model 3 Model 4 Democratic Democratic Interactive Effects of Period, the Period, the Test, the 57 th , Approval by 57 th 58 th and 59 th 58 th , and 59 th Partisanship, Legislature Legislatures Legislatures Democratic (9800) (0006) Period, the 58 th and 59 th Legislatures Aggregate variables: Presidential 6.25*** 4.4*** 3.97*** 1.3* Approval (2.10) (.56) (.46) (.7) National

Number of .0008 .3*** .09*** .3*** Bills (.03) (.04) (.009) (.04)

Political 1.7*** .85*** .79*** .98*** Importance of (.05) (.05) (.04) (.05) the Bill Ramo 28 .001** .0002 .0002* .0002 (.0005) (.0003) (.0001) (.0003) Individual level variables:

President’s 1.7*** .54*** .56*** .88*** Party Deputy (.09) (.07) (.07) (.14)

Orphan .33*** .07 .12* .12 (.09) (.07) (.07) (.08)

Regular .24 22 .13 .04 (.20) (.16) (.16) (.17)

Test Variables

PRI Regime 8.9*** (1.08)

248

Table 4.2, Continued

Model 1 Pre Model 2 Model 3 Model 4 Democratic Democratic Interactive Effects of Period, the Period, the Test, the Approval by 57 th 58 th and 59 th 57 th , 58 th , Partisanship, Legislature Legislatures and 59 th Democratic (9800) (0006) Legislatures Period, the 58 th and 59 th Legislatures PRI Regime X 11.63*** Presidential (2.1) Approval PRI Regime X .12*** Number of (.02) Bills PRI Regime X .8*** Political (.07) Importance PRI Regime X .00001 Ramo 28 (.00009)

PRI Regime X 1*** President’s .11 Party PRI Regime X .15 Orphan (.11) PRI Regime X .12 Regular (.25)

Disaggregating Presidential Approval by Partisanship President’s 18.3*** Party Deputy X (1.4) Presidential Approval PRI .8*** (.13)

249

Table 4.2, Continued 55

Model 1 Pre Model 2 Model 3 Model 4 Democratic Democratic Interactive Effects of Period, the Period, the Test, the 57 th , Approval by 57 th 58 th and 59 th 58 th , and 59 th Partisanship, Legislature Legislatures Legislatures Democratic (9800) (0006) Period, the 58 th and 59 th Legislatures PRI X 4.8*** Presidential (1.3) Approval PRD 1.65*** (.11)

PRD X 5.8*** Presidential (1.1) Approval Year Fixed Included Included Included Included Effects State Fixed Included Included Included Included Effects Legislature Included Included Included Fixed Effects Deputies 4.63e11 2.41e10 4.2e10 1.04e11 Level (.0001) (.00003) (.00003) (.00002) Variance 2X Log 5314.34 8419.6 13947.3 7273.5 Likelihood 11,517 33,655 45172 33,655 *p<.10, **p<.05, ***p<.01 Note: The dependent variable is roll call voting. I include year dummies except for one year, state and legislature fixed effects in the estimation of the models but not presented in the table. Some of them are significant. DY/DX is for discrete change of dummy variable from 0 to 1.

55 For these three categories, PAN, PRD, and PRI the reference groups is Independents and minor party deputies. In Model 4, to address a multicollinearity problem derived from the inclusion of the interactive variables, I demeaned the variables Presidential Approval, President’s Party, PRI, and PRD to disaggregate the effects of approval by partisanship.

250

In Model 2, I assume that all else being equal the effects of presidential approval on roll call voting were the same for members of the president’s party and for members of the opposition after the 2000 transition. To address this problem the analysis in Model 4 disaggregates roll call voting by partisanship:

Panista, Priísta, Perredista. The reference group is Independent and minorparty deputies. As the following equation shows, to test the hypothesis that the effects of presidential approval on roll call voting are affected by partisanship, I have created three interactive variables: the products of presidential approval and three dummy variables of each of the major parties in Mexico: the PAN, PRD and PRI.

Because during the 58 th and 59 th legislatures, Vicente Fox stayed in office, the

Panista deputies are coded as members of the president’s party. The reference group is Independents and small partydeputies. My model specification then is:

Roll Call Voting= β0 + β1 President’s Party + β2 Presidential Approval + β3 ( β1 X β2 ) + β4 PRI + β5 ( β4 X β2) + β6 PRD + β7 ( β6 X β2 ) + β8 (the rest of the control variables) + U (error)

In Table 4.2, Model 4 reveals that the three interactive variables β3

(President’s Party X Presidential Approval), β5 (PRI X Presidential Approval), and β7 (PRD X Presidential Approval) are significant. The effects of presidential approval on roll call voting are affected by partisanship. As the hypothesis 2 predicted, after the 2000 democratic transition the effects of presidential approval on roll call voting were affected by partisanship. Approval had positive effects on

251

roll call voting for members of the president’s party, and negative effects for the opposition. For an Independent or a minorparty deputy (the reference group) the effect of presidential approval on roll call voting after the democratic transition was β2 (1.3). For a member of the president’s party (Panista) , the effect of presidential approval on roll call voting was β2 + β3 (1.3 + 18.3 = 19.6). For a

Priista deputy , the effect of presidential approval on roll call voting during the democratic period was β2 + β5 [(1.3) + (–4.8) = 3.5]. For a Perredista deputy, the effect of presidential approval on roll call voting is β2 + β7 [(1.3) + (5.8) =

4.5]. These findings coincide with the strategies of the opposition in Mexico (the

PRI and the PRD) when they bargained with a Panista president. Priistas did not always reject Vicente Fox’s policies. They sometimes supported him, and occasionally Priistas rejected Fox’s plans. Priistas behaved strategically, attempting to frustrate Fox’s plans when he enjoyed high levels of presidential approval. The story of the PRD was different. Their policy preferences were different than those of Panistas , and Perredistas behaved less strategically than

Priistas did . This is the reason because the coefficients of the variables PRD and

PRD X Presidential Approval were both negatives

Although the raw coefficients presented in Table 4.2 are essential in determining the accuracy of my hypotheses, they are quite difficult to interpret in nonlinear specifications. To address this problem and highlight substantive findings, in Table 4.3 I present results of the marginal effects of the models specified above.

252

Table 4.3 Effects of Presidential Approval on Roll Call Voting in Mexico (98 06) Multilevel Models Using Deputies Random Effects (Random Intercepts), Results Expressed as Marginal Effects

Model 1 Pre Model 2 Model 3 Model 4 Democratic Democratic Interactive Effects of Period, the Period, the Test, the 57 th , Approval by 57 th 58 th and 59 th 58 th , and 59 th Partisanship, Legislature Legislatures Legislatures Democratic (9800) (0006) Period, the 58 th and 59 th Legislatures Aggregate variables: Presidential .19*** .07*** .08*** .01* Approval (.06) (.009) (.009) (.005) National

Number of .00003 .004*** .002*** .002*** Bills (.001) (.0005) (.0002) (.0002)

Political .05*** .01*** .02*** .007*** Importance of (.002) (.0008) (.0008) (.0005) the Bill Ramo 28 .00004** 3.54e06 4.4e06 1.68e06 (.00002) (.000) (.000) (.0000) Individual level variables:

President’s .06*** .008*** .01*** .006*** Party Deputy (.004) (.001) (.001) (.0009)

Orphan .01*** .001 .002* .001 (.003) (.001) (.001) (6.6e04)

Regular .007 .003 .002 .0002 (.005) (.002) (.003) (.001)

253

Table 4.3, Continued

PRI Regime .94*** (.04)

PRI Regime X .25*** Presidential (.04) Approval PRI Regime X .003*** Number of (.0005) Bills PRI Regime X .01*** Political (.001) Importance PRI Regime X 4.14e07 Ramo 28 (.0000)

PRI Regime X .02*** President’s (.001) Party PRI Regime X .003 Orphan (.002) PRI Regime X .002 Regular (.005)

Disaggregating Presidential Approval by Partisanship President’s .14*** Party Deputy X (.01) Presidential Approval

254

Table 4.3, Continued

PRI X .04*** Presidential (.01) Approval PRD .03*** (.003)

PRD X .045*** Presidential (.009) Approval 11,517 33,655 45,172 33,655 *p<.10, **p<.05, ***p<.01 Note: The dependent variable is roll call voting. I include year dummies except for one year, state and legislature fixed effects in the estimation of the models but not presented in the table. Some of them are significant. DY/DX is for discrete change of dummy variable from 0 to 1.

255

4.6 Conclusion

This chapter has showed that the effects of presidential approval on roll call voting during the PRI regime are different than those in the democratic period. Contrary to my expectations and the hypothesis 1, presidential approval affected roll call voting during the 57 th Legislature, the last two years that the PRI stayed in office. My original thought was that when the PRI held the power, presidential approval did not influence deputies’ voting behavior. I incorrectly assumed that, regardless his popularity, Ernesto Zedillo, the last Priista president was able to control his party members’ behavior in the Chambers of Deputies. I also incorrectly believed that during the last years of the PRI regime, the opposition overlooked the citizens’ support for the president when they bargained with the executive. However, this chapter reveals two important findings: (1)

During the last two years of the PRI era, members of the president’s party took into account the president’s popularity when they bargained with him. Priista deputies used presidential approval as a signal of voters’ preferences to advance in their political careers. This finding is consistent with some scholars’ arguments.

Langston (2007) and Nacif (2004) state that even though legislators in Mexico cannot be reelected immediately, the electoral connection between deputies and their constituencies are not totally broken. It is only adapted to the Mexican political context. Mexican deputies look for office either in the Senate or in a governorship. Satisfying citizens’ policy preferences at the state level, deputies

256

search for higher positions in politics (Langston 2007; Ugalde 2000). (2) During the predemocratic period, presidential approval affected roll call voting of members of the opposition in the Chamber of Deputies. The opposition considered that a president with acceptable levels of approval, like Zedillo whose approval rating was 55 percent during the 57 th Legislature, was not in a bad position to bargain with them.

While presidential approval had positive effects during the 57 th , the effects of presidential approval on roll call voting turned negative after the 2000 transition. This result is consistent with some findings in American politics (Bond and Fleisher 1980; Bond, Fleisher, and Northrup 1988). Favorable public opinion may not influence Congress to approve more of the president’s proposal in all situations (Bond and Fleisher 1980; Bond, Fleisher, and Northrup 1988). If the opposition party controls Congress, then favorable public opinion may not increase the president’s rate of success. That was the case of Mexico after 2000.

In the 59 th Legislature, the president’s party, the PAN, only held 29 seat shares.

Therefore, even though Vicente Fox enjoyed high approval rating during both the

58 th and the 59 th Legislatures, he could not take advantage of his popularity because his party did not control the Chamber of Deputies, and the PAN’s presence in Congress became weaker from the 58 th to 59 th Legislature. In the new

Mexican democracy, opposition deputies considered that the best strategy to advance in their political careers was to embarrass the president and to frustrate his plans. Therefore, the higher the approval of the new president became, the

257

more incentives opposition legislators had to make the president’s life difficult in

Congress.

The normative implication of the fourth chapter is whether a high approval rating is not desirable for the president in a new democracy. I think that the findings of this chapter are not necessarily bad signals for the consolidations of the Mexican democracy. I have analyzed only the first six years after the democratic transition in Mexico. I have examined only one case, Vicente Fox’s tenure from 2000 to 2006. This research should extend its analysis to the Senate and the current administration of Felipe Calderon to offer final conclusions. My expectation is that the effects of approval on roll call voting will turn positive with the eventual consolidation of the Mexican democracy in some years

258

4.8 Works Cited

Bond, Jon R. and Richard Fleisher. 1980. “The Limits of Presidential Popularity As a Source of Influence in the U.S. House”. Legislative Studies Quarterly , Vol. 5, No. 1: 6978.

Bond, Jon R., Richard Fleisher, and Michael Northrup. 1988. “Public Opinion and Presidential Support”. Annals of the American Academy of Political and Social Science. Vol. 499: 4763.

Béjar, Luisa. 2001. “La (Re)Institucionalización del Poder Legislativo en México”. Revista Mexicana de Sociología Vol. 63, No. 3: 99133.

2004. El Poder Legislativo en México. Los Diputados de Partido: el Primer Eslabón del Cambio. México: Universidad Nacional Autónoma de México/Gernika.

2006. Los Partidos en el Congreso de la Unión. La Representación Parlamentaria Después de la Alternancia. México: Universidad Nacional Autónoma de México/Gernika.

2009. “Élites Parlamentarias en México. Los Presidentes de Comisiones”. ¿Qué Hacen los Legisladores en México? El Trabajo en Comisiones. Ed. Luisa Béjar. México: Porrúa: 6598.

and Gilda Waldman. 2004. La Representación Parlamentaria en México. México: Universidad Nacional Autónoma de México/Gernika.

Calvo, Ernesto. 2007. “The Responsive Legislature: Public Opinion and Law Making in a Highly Disciplined Legislature”. British Journal of Political Science (Forthcoming)

CanesWrone, Brandice and Scott de Marchi. 2002. “Presidential Approval and Legislative Success”. The Journal of Politics, Vol. 64, No. 2: 491509.

Casar, Ma. Amparo. 2002. “ExecutiveLegislative Relations: The Case of Mexico (19461997)”. Legislative Politics in Latin America. Ed. Scott Morgenstern and Benito Nacif. United Kingdom: Cambridge University Press:114146.

259

Cohen, Jeffrey E, Jon R. Bond, Richard Fleisher, and John A. Hamman. 2000. “StateLevel Presidential Approval and Senatorial Support”. Legislatives Studies Quarterly, Vol. 25, No. 4: 577590.

Constitución Política de los Estados Unidos Mexicanos. 2001: Ed. Porrúa

Davidson, Roger. 1969. The Role of the Congressman. Indianapolis: Pegasus.

Desposato, Scott W. and Francisco Cantú. 2009. “The Impact of the Federalism on Mexican Legislative Politics”. Presented at the Annual Meeting of the American Political Science Association.

Edwards, George.1976. “Presidential Influence in the House: Presidential Prestige as a Source of Presidential Power”. The American Political Science Review, Vol. 70. No. 1: 101113.

1977. “Presidential Influence in the Senate: Presidential Prestige as a Source of Presidential Power”. American Politics Research, Vol. 5, No. 4: 481499.

1980. Presidential Influence in Congress. United Sates: Freeman and Company.

1989. At the Margins. New Haven and London: Yale University Press.

Espinoza, Ricardo and Mónica Miguel. 2007: “El Poder Ejecutivo y la Producción Legislativa en México (199720060)”. ¿Para Qué Sirve el Poder Legislativo? Ed. Ricardo Espinoza and Jeffrey Weldon. México: Porrúa: 209230.

Kernell, Samuel. 1977. “Presidential Popularity and Negative Voting: An Alternative Explanation of the Midterm Congressional Decline of the President’s Party”. The American Political Science Review, Vol. 71, No. 1: 4466.

Langston, Joy. 2007. “Tracing the Lines of Authority in Mexico’s Congress”, Center for Research and Teaching in Mexico (CIDE), working paper.

Nacif, Benito. 2004. “La Nueva Relación entre el Presidente y el Congreso en México”. Política y Gobierno, Vol. XL, No. 1: 942.

260

Mejía, J. Andrés. 2004. “Ghost Coalitions: Economic Reforms, Fragmented Legislatures and Informal Institutions in Ecuador (19792002)”. Ph.D. diss. University of Notre Dame.

Piereson, James E. 1975. “Presidential Popularity and Midterm Voting at Different Electoral Level”. American Journal of Political Science, Vol. 19, No. 4: 683694.

Presidencia de la República. 2006. Anexo del 6 Informe de Gobierno. México.

Pritchard, Anita. 1983. “Presidents Do Influence Voting in the U.S. Congress: New Definitions and Measurements”. Legislative Studies Quarterly, Vol. 8, No. 4: 691711.

Puente, Khemvirg. 2004. “Viscosity in the Mexican Senate: Legislative Significance and Resisting Patterns under Divided Government (2000 2003).” Master’s thesis. The University of Hull.

Reveles, Francis. 2004. “La Coalición Dominante en el Partido Acción Nacional”. Partidos Polítos. uevos Liderazgos y Relaciones Internas de Autoridad. Ed. Rosa María Mirón Lince and Ricardo Espinoza. México: Universidad Autónoma Metropolitana.

Rivers, Douglas and Nancy L. Rose. 1985. “Passing the President’s Program: Public Opinion and Presidential Influence in Congress”. American Journal of Political Science, Vol. 29, No. 2:183196.

Tufte, Edward R. 1975. “Determinants of the Outcomes of Midterm Congressional Elections”. The American Political Science Review, Vol. 69, No. 3: 812826.

Ugalde, Luis Carlos. 2000. The Mexican Congress: Old Player, ew Power. Washington D.C: Center for Strategic and International Studies.

Weldon, Jeffrey. 2002. “The Legal and Partisan Framework of the Legislative Delegation of the Budget in Mexico”. Legislative Politics in Latin America. Ed. Scott Morgenstern and Benito Nacif. United Kingdom: Cambridge University Press: 377412

261

4.9 Web Sites

Banco de Información para la Investigación Aplicada en Ciencias Sociales (BIIACS) http://www.biiacs.cide.edu

Cámara de Diputados en México, http://www.camaradediputados.gob.mx

COCLUSIO

This dissertation has explored presidential approval before and after a democratic transition. From American politics (Clarke and Stewart 1994; Kenski

1977; MacKuen, Erikson, and Stimson 1992; Markus 1988; Thomas 2003) we know that voters’ perceptions of the economy and economic issues such as unemployment and inflation affect presidential approval. However, few studies have analyzed presidential approval outside the U.S. Most of these studies focus their attention on well consolidated democracies such as France, Germany, and the U.K. (Powell and Whitten 1993; Norpoth 1994; Lafay 1994). While these analyses have contributed a great deal to our understanding of presidential approval, they have neglected the effects that democratization has on citizens’ evaluation of the president. No analysis explaining whether people change their expectations about the president’s responsiveness after a democratic transition exists in the literature. We also lack consistent explanations of the conditions under which political rather than economic issues become relevant for approval after democratization takes place.

I have used Mexico as a case study to improve our understanding of presidential approval in comparative politics. As I have argued, public opinion of the president affected Mexican politics after the 2000 transition. First, since the

262 263

2000 presidential election, crime and corruption have become more salient and citizens have begun to evaluate the president in terms of his determination to address these two problems. Second, public opinion of the president has influenced the executivelegislative relationships differently since the democratic transition. After Vicente Fox’s ascension to power in 2000, opposition deputies began to use as a main political strategy to advance in their political careers embarrassing the president in Congress when he enjoyed high levels of approval ratings.

I have used the theoretical framework of American politics to explain presidential approval before and after a democratic transition. I have conducted four empirical analyses to support my argument about how public opinion of the president affected Mexican politics after the 2000 transition. I summarize the main findings of my dissertation and list some possibilities for future research.

Presidential Approval at the Individual Level

Existing literature on presidential approval in Mexico argues that party ID and citizens’ perceptions of both the economy and economic issues, unemployment and inflation, are the key determinants when citizens evaluate the president. These studies (Buendía 1996; Domínguez and McCann 1995; Magaloni

2006; Villarreal 1999) have significantly improved our understanding of

264

presidential approval in Mexico, yet they ignore the effects of the Mexican political context on presidential approval. These analyses also neglect the impact that democratization has when citizens evaluate the president.

My analysis of presidential approval at the individual level before and after the 2000 democratic transition demonstrates that citizens’ perceptions of the president do not occur in a vacuum. The political regime works as the market when people evaluate the president. From the demand side, citizens have expectations about the government, and from the supply side, the president searches for support by addressing people’s main concerns. The politicalmarket affects how voters evaluate the president. In the Mexican case, citizens regard combating crime to be the second most important task to be addressed by democracy, and view corruption as the main obstacle for a democratic consolidation (Clark 2001, 87; Klesner 2001, 127; Moreno 2001, 43). Thus, since the democratic transition Mexican people have begun to evaluate the president in terms of both his anticorruption and anticrime policies, and performance. The analysis of the first chapter reveals that the effects of citizens’ perceptions of both rising corruption and rising crime in the democratic period were significantly different and stronger than those during the PRI semiauthoritarian system. As a robust test, I conducted an analysis by using the 1997 and 2003 intermediate elections as breaking points, respectively. When I used the 2000 presidential election as the breaking point of the analysis, the effects of both perceptions of rising crime and rising corruption were stronger than those when the breaking

265

points were 1997 and 2003, respectively. This robust test reveals that the greatest change in citizens’ evaluation of the president occurred after the 2000 democratic transition.

Presidential Approval at the State Level

No study explaining statelevel presidential approval in Mexico exists in the literature. My understanding is that the analysis of the second chapter is the first study of statelevel presidential approval outside the U.S. The American literature shows that state factors affect presidential approval (Beyle, Niemi, and

Sigelman 2002; Cohen 2003; Cohen and Powell 2005). The results of the second chapter reveals that as in the U.S. case, in Mexico subnational conditions and perceptions aggregated at the state level affected statelevel presidential approval.

However, in the Mexican case, the empirical evidence suggests that the perceptionvariables had more explanatory power than subnational conditions.

Except for state GDP per capita, the effects of objective indicators did not have generally significant effects on approval. In contrast, perceptions of crime, unemployment, and inflation across the states significantly affected statelevel presidential approval, especially during the predemocratic period. However, in most of the cases the effects of the explanatory variables were not significantly different between these two periods, the semiauthoritarian PRI regime and the

266

democratic term. Except for state crime, the effects of the rest of the explanatory variables were not significantly different before and after the transition.

MultiLevel Analysis of Presidential Approval

The first and the second chapters demonstrate that the factors that affect presidential approval have a different impact after the 2000 presidential election.

The first chapter reveals that the effects of perceptions of rising crime and rising corruption on presidential approval during the democratic period are stronger than those of the predemocratic era. In contrast, the second chapter shows that some of the subnational conditions have stronger effects on approval during the PRI hegemony. By including objective indicators, citizens’ perceptions and background, the multilevel analysis in the third chapter provides more reliable and accurate results. This chapter has shown that subnational conditions affected presidential approval at the individual level before and after the 2000 transition.

The third chapter of my dissertation reveals that the effects of state factors, levels of unemployment, inflation, crime and corruption by state, on approval during the democratic period were significantly different and stronger than those in the pre democratic era.

267

is citizens’ opinion on rising crime. As a robust test, I conducted a restricted analysis by including only the states with the most stable levels of crime. Even in the states with the most stable number of convicted criminals, perceptions of crime had stronger and significantly different effects on presidential approval after the 2000 presidential election. This finding shows that the greatest change in voters’ perceptions of presidential approval after the 2000 democratic transition is how they believe the president addresses crime in Mexico.

Perceptions of economic issues affect presidential approval before and after the transition. However, their effects are not statistically different from one period to the other. The multilevel analysis of this dissertation suggests that citizens considered unemployment and inflation to be two important issues in

Mexican politics before and after the democratic transition. Nevertheless, they did not strongly associate economic issues with the mains duties of a democratic government.

In contrast to the findings of the first chapter, and when I control for sub national conditions, levels of crime, corruption, unemployment, and inflation by state, the multilevel analysis shows that the effects of citizens’ perceptions of rising corruption are not significantly different after the transition. Perceptions of corruption had been a relevant issue for approval of the president even before the

2000 presidential election (Davis and Langley 1995). The results of the multi level analysis suggest that during the PRI era citizens thought that the president could address this problem even though he was a member of the PRI. Thus,

268

voters’ perceptions of the president’s anticorruption policies did not suffer a change as dramatic as when they evaluated the president’s anticrime policies.

Therefore, the greatest change in citizens’ evaluation of the president can be explained by the stronger effects of voters’ perceptions of the president’s determination to address crime after the 2000 transition.

The Effects of Presidential Approval

Most of the existing roll call voting models in American politics shows that popular presidents are more successful than unpopular executives when they bargain with Congress (CanesWrone and de Marchi 2002; Edwards 1980;

Pritchard 1983; Rivers and Rose 1985). However, the Mexican case may challenge these findings in American politics. While Vicente Fox, the president of the democratic transition who enjoyed high levels of approval, was not very successful at bargaining with Congress, the Priista president Ernesto Zedillo, less popular than Fox, enjoyed congressmen’s support in the legislative arena.

The fourth chapter shows that contrary to my expectations, presidential approval affected roll call voting during the 57 th legislature, the last two years of the PRI with Ernesto Zedillo in office. This result suggests that the political institutional context in Mexico two years before the transition did not allow the president to influence the Congress without citizens’ support, and congressmen thought that a president who enjoyed voters’ approval was in a better position to

269

bargain with them. That was the case of Ernesto Zedillo who reached an acceptable approval rating during the last two years of his tenure, before the democratic transition took place.

While presidential approval had positive effects on roll call voting before the democratic transition, after the 2000 presidential election, the effects of presidential approval were only positive for members of the president’s party, but negative for opposition deputies. In the new Mexican democracy, opposition deputies considered that the best strategy to advance in their political careers was to embarrass the president and to frustrate his plans. Therefore, the higher the approval of the new president became, the more incentives that opposition legislators had to make the president’s life difficult in Congress.

The first three chapter of this dissertation also provides opportunities for future research projects, considering my findings about the determinants of

270

presidential approval. First, although in chapter one I tested the mediacoverage hypothesis about the media effects on presidential approval, a more accurate study requires the use of content analysis in order to present a more complete final conclusion. These data do not exist. We do not have analyses of the effects of media coverage on presidential approval in Mexico. Second, an analysis of going public as a political strategy of the president to influence citizens’ opinion in

Mexico should shed light on our understanding of presidential approval in comparative politics. Third, given that corruption and crime have become salient in most of the new Latin American democracies, an analysis of their effects on presidential approval may offer results from a comparative perspective. These are only a few examples of the sorts of research projects that can grow out of the research contained in this dissertation.

271

Works Cited

Buendía, Jorge. 1996. “Economic Reform, Public Opinion and Presidential Approval in Mexico 19881993”. Comparative Political Studies, Vol. 29, No. 5: 566591.

CanesWrone, Brandice and Scott de Marchi. 2002. “Presidential Approval and Legislative Success”. The Journal of Politics, Vol. 64, No. 2: 491509.

Clark, Mary A. 2001. “Costa Rica: Portrait of an Established Democracy”. Citizen Views of Democracy in Latin America. Ed. Roderic Ai Camp. The United States: University of Pittsburgh Press. 7390

Cohen, Jeffrey. 2003. “The Polls: StateLevel Presidential Approval: Results from the Job Approval Project”, Presidential Studies Quarterly, Vol. 33, No. 1 (March): 211220.

, and Richard Powell. 2005. “Building Public Support from the Grassroots Up: The Impact of Presidential Travel on StateLevel Approval” Presidential Studies Quarterly 35, No. 1 (March): 1127.

272

Edwards, George. 1980. Presidential Influence in Congress. United Sates: Freeman and Company.

Kenski, Henry. 1977. “Inflation and Presidential Popularity”. The Public Opinion Quarterly, Vol. 41, No. 1 (Spring): 8690.

Magaloni, Beatriz. 2006. Voting for Autocracy. The Untied States: Cambridge University Press.

Moreno, Alejandro. 2001. “Democracy and Mass Belief Systems in Latin America”. Citizen Views of Democracy in Latin America. Ed. Roderic Ai Camp. The United States: University of Pittsburgh Press. 2750..

273

Pritchard, Anita. 1983. “Presidents Do Influence Voting in the U.S. Congress: New Definitions and Measurements”. Legislative Studies Quarterly, Vol. 8, No. 4: 691711.

Rivers, Douglas and Nancy L. Rose. 1985. “Passing the President’s Program: Public Opinion and Presidential Influence in Congress”. American Journal of Political Science, Vol. 29, No. 2:183196.

Thomas, Dan, Lee Sigelman and Larry Baas. 1984. “Public Evaluations of the President: Policy, Partisan, and ‘Personal’ Determinants”. Political Psychology, Vol. 5, No. 4: 531542.

Villarreal, Andrés. 1999. “Public Opinion of the Economy and the President among Mexico City Residents: The Salinas Sexenio”. Latin American Research Review , Vol. 34, No. 2: 132151.