Online Appendix for Partisan Losers Effects: Perceptions of Electoral Integrity in Mexico Francisco Cantú a and Omar García-Ponce b March 2015 A Survey Information A.1 Pre- and Post-Electoral Surveys Both surveys were conducted by Buendía & Laredo a survey and polling firm based in Mexico City. Mexico s electoral precincts were used as the Primary Sampling Units (PSUs). It is worth mentioning that the combination of geospatial and census data at the electoral precinct level provide the most complete and up-to-date sampling frame available in the country. These data come from the Federal Electoral Institute (IFE, by its Spanish acronym) and are continuously updated, providing a comprehensive assessment of the Mexican electorate approximately 95 percent of Mexicans 18 years old or older are registered at the IFE. As of 2012, about 77.4 million voters were registered at the IFE. These citizens are dispersed across 64,934 electoral precincts. In both surveys, respondents were randomly selected using a stratified multistage cluster sampling design. This drawing procedure was crafted to achieve a nationally representative sample of Mexican citizens, and required the careful implementation of the five methodological steps that we describe below. 1. Stratification by Geographical Region and Type of Electoral Precinct: The sample was allocated to five geographical regions electoral circumscriptions, as defined by the IFE in order to ensure territorial coverage. Strata were then defined based on a precinct s level of urbanization (urban, rural, or mixed) and party affiliation (i.e. the party that received the most votes in the 2006 presidential election). 2. Electoral Precinct Selection: Within each stratum, precincts (PSUs) were randomly selected according to a probability proportional to its size. The size of a electoral precinct is measured as the number of registered voters. In total, 100 precincts were drawn and proportionally distributed in each stratum. A total of 8 and 10 face-to-face interviews were conducted per precinct in the pre- and post-electoral surveys, respectively. a Assistant Professor, University of Houston. E-mail: fcantu10@uh.edu. b Ph.D. Candidate, New York University. E-mail: garcia.ponce@nyu.edu. 1
3. Block Selection within Electoral Precincts: Once electoral precincts in the sample were drawn, the next step was to select two blocks from the precinct using a table with random numbers. For instance, the PSU map shown in Figure 5 has 20 blocks, and the PSU number is 0320. A combination of the number of blocks and the last digit of the PSU number was used to determine which blocks are to be selected. 4. Household Selection within Blocks: Once blocks in the sample were identified, households per block were selected using a systematic random sampling method. Blocks were covered starting by the northeast corner using a systematic random start of 3 households. Blocks were walked clockwise. Once a questionnaire was completed, the interviewer had to move to the next side of the block. 5. Respondent Selection within Households: One respondent per household was selected using a random method (a numbered card). If the selected respondent was not available during the first visit, the interviewer returned up to three times. If the respondent refused the interview, the interviewer moved using a systematic random start of 10 households in order to obtain the interview. The question on voting preferences was administered using a procedure that simulated a secret ballot, which displayed candidates names and official party logos (see Figure 6). In total, 800 and 1,000 face-to-face interviews with Mexicans 18 years old or older were conducted in the pre- and post-electoral surveys, respectively. The response rate was 63% in the former case, and 61% in the latter one, based on AAPOR s Standard Definitions. A.2 Exit Poll The exit poll was based on a similar stratified multistage cluster sampling design. The key difference with respect to the pre- and post-electoral surveys is that the exit poll is representative at the state level. The sample was allocated to Mexico s 31 federated states and its Federal District. Strata were constructed based on a precinct s level of urbanization (urban, rural, or mixed) and party affiliation (i.e. the party that received the most votes in the 2006 presidential election). Within each stratum, precincts were randomly selected according to a probability proportional to its size. In total, 428 precincts were drawn. During the election day, interviewers were located at the exit of the polling stations of each selected precinct. Approximately, one in three voters were systematically interviewed immediately after casting their votes, totaling 7,064 face-to-face interviews. The questionnaire contained only seven questions and was administered using a procedure that simulated a secret ballot that showed candidates names and official party logos (see Figure 6). 2
B Appendix Figures and Tables Figure 5: Example of a PSU Map 3
Figure 6: Simulated Secret Ballot ENCUESTA DE SALIDA PRESIDENTE DE LOS ESTADOS UNIDOS MEXICANOS Por favor, marque con una cruz el partido o candidato por el que votó hoy para Presidente de la República PARTIDO ACCIÓN NACIONAL JOSEFINA EUGENIA VÁZQUEZ MOTA PARTIDO REVOLUCIONARIO INSTITUCIONAL ENRIQUE PEÑA NIETO PARTIDO DE LA REVOLUCIÓN DEMOCRÁTICA PARTIDO VERDE ECOLOGISTA DE MÉXICO ANDRÉS MANUEL LÓPEZ OBRADOR ENRIQUE PEÑA NIETO PARTIDO DEL TRABAJO MOVIMIENTO CIUDADANO ANDRÉS MANUEL LÓPEZ OBRADOR ANDRÉS MANUEL LÓPEZ OBRADOR NUEVA ALIANZA GABRIEL RICARDO QUADRI DE LA TORRE Si desea votar por algún candidato no registrado, escriba en este recuadro el nombre completo ENCUESTA DE SALIDA DIPUTADOS FEDERALES Por favor, marque con una cruz el partido o candidato por el que votó hoy para Diputados Federales PARTIDO ACCIÓN NACIONAL PARTIDO REVOLUCIONARIO INSTITUCIONAL PARTIDO DE LA REVOLUCIÓN DEMOCRÁTICA PARTIDO VERDE ECOLOGISTA DE MÉXICO PARTIDO DEL TRABAJO MOVIMIENTO CIUDADANO NUEVA ALIANZA Si desea votar por algún candidato no registrado, escriba en este recuadro el nombre completo 4
Figure 7: Perceptions of Electoral Fraud in Latin America Notes: The graph shows the proportion of respondents answering a fraud to the question Do you believe that, generally speaking, elections are clean in this country or a fraud?, based on data from the Latinobarómetro survey (2009). 5
Figure 8: Perceptions of Electoral Integrity in Latin America Country Uruguay Surinam Venezuela Chile Belice Argentina Peru Costa Rica Nicaragua El Salvador Mexico Dominican Republic Ecuador Jamaica Paraguay Trinidad and Tobago Guyana Guatemala Panama Brazil Bolivia Colombia Honduras Haiti To what extent do you trust elections in this country? 2 4 6 Mean Score Notes: The graph shows the mean value of respondents answering a fraud to the question To what extent do you trust elections in this country? on a 1-7 scale, where 1 means no trust at all and 7 means total trust. The graph is based on data from the Latin American Public Opinion Project (2012). 6
Table 4: Contingency Tables for Party Identification and Declared Electoral Support PAN PRI PRD Very Somewhat Very Somewhat Very Somewhat Pre-Electoral Survey (June 11-14, 2012) Vázquez Mota 98 (0.852) 122 (0.859) 6 (0.028) 4 (0.020) 0 (0) 0 (0) Penã Nieto 8 (0.097) 10 (0.074) 191 (0.863) 166 (0.843) 1 (0.010) 8 (0.067) López Obrador 3 (0.026) 2 (0.014) 2 (0.009) 6 (0.031) 80 (0.860) 96 (0.835) Total 115 142 213 197 95 115 Post-Electoral Survey (August 2-6, 2012) Vázquez Mota 49 (0.924) 77 (0.895) 1 (0.006) 2 (0.013) 1 (0.010) 0 (0) Penã Nieto 2 (0.038) 5 (0.058) 150 (0.937) 152 (0.956) 4 (0.041) 0 (0) López Obrador 1 (0.019) 4 (0.047) 4 (0.025) 1 (0.006) 89 (0.917) 81 (0.964) Total 53 86 160 159 97 84 Notes: The table shows the frequency values of party identification and declared electoral support for the three most important political parties and candidates in the 2012 Mexican presidential election. The values in parentheses are the proportion of each cell as the total party identification. 7
Table 5: Estimates of the Determinants of Perceived Electoral Integrity (Exit Poll) Dependent variable: Confidence on the Election Multilevel Ordinal logit (1) (2) (3) (4) (5) López Obrador 0.558 0.559 0.564 0.931 0.941 (0.062) (0.058) (0.069) (0.172) (0.213) Vázquez Mota 0.058 0.059 0.072 0.302 0.258 (0.068) (0.064) (0.074) (0.194) (0.197) Female 0.191 0.184 0.185 0.181 0.178 (0.048) (0.046) (0.048) (0.048) (0.048) Age 0.060 0.064 0.063 0.065 0.065 (0.018) (0.018) (0.018) (0.018) (0.018) Education 0.068 0.071 0.070 0.075 0.074 (0.028) (0.027) (0.028) (0.029) (0.029) Decision time 0.298 0.292 0.292 0.293 0.294 (0.045) (0.043) (0.045) (0.045) (0.045) Attention to campaigns 0.937 0.939 0.941 0.949 0.948 (0.069) (0.085) (0.070) (0.069) (0.070) Rural 0.265 0.281 0.279 0.238 0.244 (0.095) (0.098) (0.096) (0.101) (0.102) Observers 0.247 0.155 0.255 0.268 (0.248) (0.302) (0.252) (0.255) Agent Vázquez Mota 0.022 0.019 (0.071) (0.071) Agent López Obrador 0.030 0.025 (0.047) (0.045) PRI 2006 0.351 0.275 (0.696) (0.716) Vázquez Mota PRI 2006 1.059 1.397 (0.806) (0.842) López Obrador PRI 2006 1.714 1.710 (0.672) (0.673) Vázquez Mota Agent Vázquez Mota 0.108 (0.091) López Obrador Agent López Obrador 0.007 (0.053) State Fixed Effects? X X X X X cut1 Constant 2.047 2.054 2.025 1.918 1.981 (0.429) (0.415) (0.428) (0.467) (0.476) cut2 Constant 0.359 0.366 0.338 0.228 0.291 (0.423) (0.401) (0.422) (0.461) (0.473) cut3 Constant 1.380 1.375 1.404 1.516 1.454 (0.429) (0.393) (0.428) (0.466) (0.477) Precinct Constant 0.317 0.319 0.312 0.318 (0.054) (0.054) (0.053) (0.053) Municipality Constant 0.000 (0.000) Observations 6819 6787 6787 6736 6787 AIC 14705.06 14641.86 14642.07 14538.64 14651.11 Notes: Significant at the 0.1 percent level. Significant at the 1 percent level. Significant at the 5 percent level. 8
Table 6: Estimates of the Determinants of Perceived Electoral Integrity (Pre-Electoral Survey) Dependent variable: Confidence on the Election Multilevel Ordinal logit (1) (2) (3) López Obrador 0.544 0.450 1.007 (0.165) (0.166) (0.366) Vázquez Mota 0.009 0.103 0.063 (0.186) (0.202) (0.507) Female 0.094 0.039 0.042 (0.144) (0.138) (0.139) Age 0.171 0.131 0.130 (0.057) (0.060) (0.061) Education 0.127 0.059 0.057 (0.065) (0.064) (0.064) Attention to campaigns 0.934 0.922 (0.162) (0.162) Rural 0.033 0.020 0.003 (0.173) (0.164) (0.167) PRI 2006 0.637 (1.227) AMLO Winner 2006 0.579 0.576 (0.165) (0.164) Vázquez Mota PRI 2006 0.620 (1.988) López Obrador PRI 2006 2.283 (1.271) State Fixed Effects? X X X cut1 Constant 0.915 1.219 1.364 (0.304) (0.346) (0.427) cut2 Constant 0.391 0.137 0.004 (0.274) (0.318) (0.404) cut3 Constant 1.856 1.658 1.520 (0.294) (0.322) (0.426) Precinct Constant 0.028 0.007 0.006 (0.108) (0.101) (0.100) Municipality Constant 0.000 0.000 0.000 (0.000) (0.000) (0.000) Observations 758 758 758 AIC 2009.859 1965.823 1969.216 Notes: Significant at the 0.1 percent level. Significant at the 1 percent level. Significant at the 5 percent level. 9
Table 7: Estimates of the Determinants of Perceived Electoral Integrity (Post-Electoral Survey) Dependent variable: Confidence on the Election Multilevel Ordinal logit (1) (2) (3) López Obrador 1.539 0.956 1.234 (0.156) (0.178) (0.464) Vázquez Mota 0.717 0.928 1.335 (0.235) (0.239) (0.527) Female 0.162 0.150 0.147 (0.106) (0.107) (0.109) Age 0.098 0.109 0.109 (0.056) (0.058) (0.058) Education 0.093 0.081 0.085 (0.063) (0.065) (0.066) Rural 0.174 0.232 0.221 (0.254) (0.275) (0.270) PRI 2006 0.021 (1.057) AMLO Winner 2006 1.452 1.456 (0.195) (0.196) Vázquez Mota PRI 2006 1.736 (1.975) López Obrador PRI 2006 1.131 (1.919) State Fixed Effects? X X X cut1 Constant 2.275 2.488 2.538 (0.621) (0.720) (0.754) cut2 Constant 0.683 0.803 0.852 (0.618) (0.722) (0.758) cut3 Constant 0.991 0.978 0.928 (0.635) (0.739) (0.778) Precinct Constant 0.083 0.077 0.083 (0.134) (0.175) (0.172) Municipality Constant 0.201 0.212 0.199 (0.122) (0.155) (0.153) Observations 973 973 973 AIC 2464.771 2381.133 2386.124 Notes: Significant at the 0.1 percent level. Significant at the 1 percent level. Significant at the 5 percent level. 10
Table 8: Estimates of the Determinants of Perceived Electoral Irregularities Dependent variable: Vote buying Biased Media Vote Miscount Campaign Overspending (1) (2) (3) (4) López Obrador 1.326 1.002 0.642 0.820 (0.257) (0.219) (0.227) (0.238) Vázquez Mota 0.907 0.868 0.413 0.830 (0.246) (0.238) (0.285) (0.253) Female 0.196 0.079 0.051 0.167 (0.144) (0.146) (0.168) (0.145) Age 0.009 0.117 0.041 0.001 (0.060) (0.062) (0.069) (0.061) Education 0.018 0.008 0.009 0.045 (0.055) (0.055) (0.060) (0.044) Unemployment 0.336 0.390 0.167 0.247 (0.240) (0.240) (0.280) (0.247) López Obrador winner 2006 1.139 1.155 1.367 1.232 (0.162) (0.157) (0.176) (0.166) Rural 0.071 0.499 0.232 0.423 (0.158) (0.162) (0.191) (0.158) Observations 944 952 917 926 Log Likelihood 587.631 577.845 457.045 580.641 Notes: Significant at the 0.1 percent level. Significant at the 1 percent level. Significant at the 5 percent level. 11
Figure 9: Change in predicted probabilities for perceptions of electoral integrity among those who declared their vote intention to López Obrador by number of electoral observers (Exit Poll) Not Confident At All Little Confident Observers per polling station Observers per polling station Somewhat Confident Very Confident Observers per polling station Observers per polling station Notes: This graph shows the mean values for the simulated first differences in the predicted probabilities that a voter replies to each of the four categories of the dependent variable as a result of a change in the vote intention and the average number of electoral observers in the polling station. Dashed lines indicate 95% confidence intervals. 12
Figure 10: Change in predicted probabilities for perceptions of electoral integrity among those who declared their vote intention to Vázquez Mota by number of electoral observers (Exit Poll) Not Confident At All Little Confident Observers per polling station Observers per polling station Somewhat Confident Very Confident Observers per polling station Observers per polling station Notes: This graph shows the mean values for the simulated first differences in the predicted probabilities that a voter replies to each of the four categories of the dependent variable as a result of a change in the vote intention and the average number of electoral observers in the polling station. Dashed lines indicate 95% confidence intervals. 13
Figure 11: Change in predicted probabilities for perceptions of electoral integrity among those who declared their vote intention to López Obrador by number of party agents (Exit Poll) Not Confident At All Little Confident 0.2.4.6.8 1 1.2 1.4 1.6 1.8 2 PRD agents per polling station 0.2.4.6.8 1 1.2 1.4 1.6 1.8 2 PRD agents per polling station Somewhat Confident Very Confident 0.2.4.6.8 1 1.2 1.4 1.6 1.8 2 PRD agents per polling station 0.2.4.6.8 1 1.2 1.4 1.6 1.8 2 PRD agents per polling station Notes: This graph shows the mean values for the simulated first differences in the predicted probabilities that a voter replies to each of the four categories of the dependent variable as a result of a change in the vote intention and the average number of party agents in the polling station. Dashed lines indicate 95% confidence intervals. 14
Figure 12: Change in predicted probabilities for perceptions of electoral integrity among those who declared their vote intention to Vázquez Mota by number of party agents (Exit Poll) Not Confident At All Little Confident 0.2.4.6.8 1 1.2 1.4 1.6 1.8 2 PAN agents per polling station 0.2.4.6.8 1 1.2 1.4 1.6 1.8 2 PAN agents per polling station Somewhat Confident Very Confident 0.2.4.6.8 1 1.2 1.4 1.6 1.8 2 PAN agents per polling station 0.2.4.6.8 1 1.2 1.4 1.6 1.8 2 PAN agents per polling station Notes: This graph shows the mean values for the simulated first differences in the predicted probabilities that a voter replies to each of the four categories of the dependent variable as a result of a change in the vote intention and the average number of party agents in the polling station. Dashed lines indicate 95% confidence intervals. 15
Figure 13: Change in predicted probabilities for perceptions of electoral integrity among those who declared their vote intention to López Obrador by by the proportion of PRI votes in the municipality (Exit Poll) Not Confident At All Little Confident Somewhat Confident Very Confident Notes: This graph shows the mean values for the simulated first differences in the predicted probabilities that a voter replies to each of the four categories of the dependent variable as a result of a change in the vote intention and the average number of party agents in the polling station. Dashed lines indicate 95% confidence intervals. 16
Figure 14: Change in predicted probabilities for perceptions of electoral integrity among those who declared their vote intention to Vázquez Mota by the proportion of PRI votes in the municipality (Exit Poll) Not Confident At All Little Confident Somewhat Confident Very Confident Notes: This graph shows the mean values for the simulated first differences in the predicted probabilities that a voter replies to each of the four categories of the dependent variable as a result of a change in the vote intention and the average number of party agents in the polling station. Dashed lines indicate 95% confidence intervals. 17
Figure 15: Change in predicted probabilities for perceptions of electoral integrity among those who declared their vote intention to López Obrador by by the proportion of PRI votes in the municipality (Pre-electoral survey) Not Confident At All Little Confident -.4-.3-.2-.1 0.1.2.3.4.5.6.7 -.4-.3-.2-.1 0.1.2.3.4.5.6.7 Somewhat Confident Very Confident -.4-.3-.2-.1 0.1.2.3.4.5.6.7 -.4-.3-.2-.1 0.1.2.3.4.5.6.7 Notes: This graph shows the mean values for the simulated first differences in the predicted probabilities that a voter replies to each of the four categories of the dependent variable as a result of a change in the vote intention and the average number of party agents in the polling station. Dashed lines indicate 95% confidence intervals. 18
Figure 16: Change in predicted probabilities for perceptions of electoral integrity among those who declared their vote intention to Vázquez Mota by the proportion of PRI votes in the municipality (Pre-electoral survey) Not Confident At All Little Confident -.4-.3-.2-.1 0.1.2.3.4.5.6.7 -.4-.3-.2-.1 0.1.2.3.4.5.6.7 Somewhat Confident Very Confident -.4-.3-.2-.1 0.1.2.3.4.5.6.7 -.4-.3-.2-.1 0.1.2.3.4.5.6.7 Notes: This graph shows the mean values for the simulated first differences in the predicted probabilities that a voter replies to each of the four categories of the dependent variable as a result of a change in the vote intention and the average number of party agents in the polling station. Dashed lines indicate 95% confidence intervals. 19
Figure 17: Change in predicted probabilities for perceptions of electoral integrity among those who declared their vote intention to López Obrador by by the proportion of PRI votes in the municipality (Post-electoral survey) Not Confident At All Little Confident -.6-.5-.4-.3-.2-.1 0.1.2.3.4.5.6 -.6-.5-.4-.3-.2-.1 0.1.2.3.4.5.6 Somewhat Confident Very Confident -.6-.5-.4-.3-.2-.1 0.1.2.3.4.5.6 -.6-.5-.4-.3-.2-.1 0.1.2.3.4.5.6 Notes: This graph shows the mean values for the simulated first differences in the predicted probabilities that a voter replies to each of the four categories of the dependent variable as a result of a change in the vote intention and the average number of party agents in the polling station. Dashed lines indicate 95% confidence intervals. 20
Figure 18: Change in predicted probabilities for perceptions of electoral integrity among those who declared their vote intention to Vázquez Mota by the proportion of PRI votes in the municipality (Post-electoral survey) Not Confident At All Little Confident -.4-.3-.2-.1 0.1.2.3.4.5.6.7.8 -.4-.3-.2-.1 0.1.2.3.4.5.6.7.8 Somewhat Confident Very Confident -.4-.3-.2-.1 0.1.2.3.4.5.6.7.8 -.4-.3-.2-.1 0.1.2.3.4.5.6.7.8 Notes: This graph shows the mean values for the simulated first differences in the predicted probabilities that a voter replies to each of the four categories of the dependent variable as a result of a change in the vote intention and the average number of party agents in the polling station. Dashed lines indicate 95% confidence intervals. 21