On the reliability of voting processes: the Mexican case

Analysis of vote distributions using current tools from statistical physics is of increasing interest. While data considered for physics studies are subject to a careful understanding of error sources, such analysis are almost absent in studies of voting process. As a way to test reliability in electoral processes we choose to investigate the July 2006 presidential election in Mexico. We use records which appeared in the Programa de Resultados Preliminares, PREP, the program which offers electoral results as soon as they are captured. In order to contrast the results we used the congressional election data-set and the July 2000 presidential record. We demonstrate the existence of correlations in the percentage of votes and, we offer evidence of the strong influence of the PRI in the curious results in the presidential case. Distributions of errors in the data-sets are analyzed for all the elections and no large deviations were found. That is, even when the sum of errors is around 45% in all cases, their global behavior is similar. This result gives support to the thesis of no large fraud for the July 2006 presidential election. Distribution of votes in all cases is obtained for all the parties. Parties and candidates with few votes, annulled votes and non-registered candidates follow a power law distribution, and the corporate party follows a daisy model distribution. Parties and candidates with many votes show a mixed behavior.