Benford’s Law and the Statistical Detection of Election Fraud

Elections worldwide are essential for democratic governance, but election fraud seriously threatens the integrity that underpins democratic societies. This thesis explores the applicability of Benford’s Law as a forensic tool for detecting anomalies in electoral data, a possible manifestation of election fraud, by introducing a novel, reproducible methodology, thoughtfully designed by the author through the integration of multiple established approaches. It combines compliance tests of the leading and last digits distributions according to Benford's law and other indicative measures. These include correlation of vote shares, turnout, and the share of valid votes, and the digital development pattern. The author was able to reach similar conclusions as prominent international organisations, such as the OSCE or the EU, by applying the proposed methodology to presidential elections with different political contexts: the Czech Republic (2023), the United States (2024), and Belarus (2020). While the Czech and U.S. elections showed no significant deviations from expected digit distribution patterns, the Belarusian data revealed multiple irregularities, particularly in the first and last digits of vote counts. This illustrates that statistical analysis of digit patterns can serve as a useful preliminary indicator of electoral manipulation.