Social determinants of voting behavior in Mexico, an analysis with longitudinal data and machine learning

Abstract

The purpose of the present study was to evaluate the relationship between social determinants and voting behavior at

the municipal level in the federal elections in Mexico in 2009, 2012, 2015 and 2018. Open data was accessed and

linear interpolation was used to obtain a total of 13 social determinants. Using linear regressions with mixed effects

and the recursive feature elimination technique with the random forest algorithm, the relationships between the

proportion of the general vote, by age and sex groups, and social determinants were explored. Among the main

results, the negative and significant associations between voting behavior and the homicide rate and the social

backwardness index stand out. The population with the highest electoral participation were women aged 30 and over

with 52.69 %, 70.44 %, 60.59 % and 73.41 % in the 2009, 2012, 2015 and 2018 elections, respectively. The random

forest algorithm found that the years of the election, Seguro Popular coverage and education were the most important

variables to predict the proportion of the vote.