Detection of Fraud with Agent-Based Models: the 2006 Mexican Election

Authored by Gonzalo Castaneda, Ignacio Ibarra

Sponsors: No sponsors listed

Platforms: NetLogo

Model Documentation: Other Narrative

Model Code URLs: Model code not found

Abstract

In this paper an innovative tool is applied for the detection of electoral fraud by analyzing statistical distributions of the official results. With this aim a computational model is developed to describe the dynamics of an electoral campaign that conditions the individuals' political preferences. The agent-based model is built assuming that in a clean process the electoral outcome depends, essentially, on preferences subjected to the influence of social interaction and global information. In contrast, in order to simulate an electoral fraud the model is modified such that the voting behavior of the citizens is manipulated on election-day. The model is calibrated with data form the 2006 Mexican electoral campaign for the presidency, and the presence or absence of fraud is validated using the Kolmogorov-Smirnov test, a non-parametric procedure. With information aggregated at the electoral district level, the simulation model rejects the existence of a large scale fraud where at least 5-6% of the total vote tally is manipulated in favor of a particular candidate.

Tags

Agent-based model Mexican elections electoral fraud political simulations