Matthew M. Singer and Dr. Jay Goodliffe, Political Science
In the 2000 Mexican Election, the PRI party lost a presidential election for the first time in Mexico’s history to Vincente Fox of the PAN party. There has been a great deal of recent scholarship examining voting behavior in Mexico. Most of it has focused on the calculation that risk-averse voters make in deciding whether to defect from the PRI i. There has been, however, little scholarship on why voters then choose a party other than the PRI. Individuals don’t just all defect; they need someone to defect to. Previous studies have not found evidence of social determinants, such as personal economic interests or social class, influencing individual vote choices.ii However, the changes in the institutional and political environment in Mexico over the last decade have created new opportunities for voters to evaluate the performance of the various parties and their platforms. As uncertainty about party performance decreases, the ability of voters to vote on the basis of the candidates’ issue positions (to vote prospectively) potentially increases. Finding differences between how groups voted would provide evidence of this shift.
The difficulty in doing electoral analysis is that voting is an individual decision and comparative political scientists generally face a lack of individual level data. Unless scholars have access to reliable survey data, their potential data sources are aggregate vote totals and aggregate statistics. Gary King has devised a statistical method and software package called EZI that can use this data while avoiding the ecological fallacy.iii EZI begins with the known bounds of each precinct parameter, which restrict the program’s analysis. EZI then uses a combination of statistical techniques and computer simulations to estimate the aggregate parameters of interest. EZI works in two steps, estimating first the proportion of voters who turned out and then, using that as a baseline, estimating the percentage who voted for each party. Reliability decreases in the second step.
Kings’s EZI method makes it possible to test whether turnout and voting patterns differed between different population segments in the 2000 Mexican elections. I obtained data on turnout and voting in the 2000 election from Mexico’s Federal Election Institute. The data for the various independent variables come from the 2000 Mexican National Census. The analysis focuses on 4 main variable categories, shown below in footnote IV.iv The census data use different populations as bases for counting members of different social groups; hence turnout and voting statistics for each variable category have different bases as well and so the estimated parameters should not be compared between variable categories.
Unfortunately, the only way of combining data from the census and election returns into congruent geographical unities was to aggregate it to the state level. The impact of the reduction in cases is an increased margin of error in EZI’s estimates and somewhat questionable results. A second result of the aggregation of data is a reduction in variance in the independent and dependent variable, thus likely suppressing the value of true relationships.
With those caveats, Tables 1-5 compares the proportion of voters in different social categories that turned out and voted for either the PAN or for the PRI in the 2000 Presidential Election. As Table 1 shows, voters in urban areas were significantly more likely to support the PAN than voters in rural areas; 6.5% of people living in towns with less than 10,000 inhabitants voted for Fox while 21.24% of people living in towns with more than 10,000 inhabitants voted for him. Interestingly, the PRI experienced no such division.
Fox and the PAN also did significantly better with voters who were literate than with those who were not (See Table 2). It also suggests that the PRI did better among the illiterate population than among literates. However, the lack of success Fox had among illiterates is very low, which makes me suspicious of these results. Further exploration of this relationship is needed.
Table 3 provides evidence that people of different economic status also voted differently. The large margins of error make these differences between proportions significantly insignificant, so further exploration is needed. In data not reported here, the left wing PRD did better among the unemployed than the employed.
Table 4 provides additional evidence that prospective evaluations were driving voting decisions. Again, the standard errors are too large to prevent definite conclusions, but these results provide evidence that the PAN did better among workers would benefit most from its economic policies while it did much more poorly among farmers and utility workers and among construction laborers, all of whom are generally members of unions that have traditionally been aligned with the PRI and have relied upon government patronage.
This brief analysis provides evidence of social group influencing vote choice. I propose two possible explanations for this outcome. These group differences might reflect different levels of party organization within different groups or regions. An alternative explanation is that voters are voting prospectively. Both of these forces can also evolve into the foundation of a longstanding partisan identification. I am currently pursuing further research along these lines.
References
- See, for example, Dominguez, Jorge I. And Alejandro Poiré, eds. 1999. Toward Mexico’s democratization. New York: Rutldege; Aldrich, John, Beatriz Magaloni, and Elizabeth Zechmeister. 2001 “Electoral Evaluations and Party Change in Mexico and Taiwan.” Presented at the conference “The Rise of DPP and PAN in Taiwan and Mexico,” Jan 26-7, 2001, Duke University.
- Dominguez, Jorge I. and James A. McCann. 1995. Shaping Mexico’s electoral arena: The construction of party cleavages in the 1988 and 1991 national elections. The American Political Science Review 89 (1): 34-49.
- For details on his method, see King, Gary. A solution to the ecological inference problem. Princeton: Princeton University Press, 1997. EZI Software downloaded from http://gking.harvard.edu/stats.shtml