Was Mexico's 1988 election truly free? New analysis reveals AI-powered detection using an original database of vote tally sheets uncovered widespread fraud.
Data & Methods
Researchers examined thousands of images from the 1988 presidential election tallies, applying Convolutional Neural Networks (CNN) to identify manipulated returns. This novel approach leveraged optical character recognition challenges posed by altered documents.
Key Findings
Fraud was prevalent in about one-third of all precincts nationwide. The analysis found that tampering occurred most frequently where opposition candidates were absent and in states governed by officials with extensive grassroots election experience.
Implications for Political Science Research
This groundbreaking study demonstrates how autocratic regimes historically manipulated electoral outcomes even after formal reforms. It introduces a revolutionary methodology combining historical archives and machine learning to audit election integrity, offering insights into understanding democratic backsliding.
