Election forensics applies quantitative techniques to officially reported vote totals, turnout figures, and precinct-level data to identify patterns inconsistent with normal voting behavior. Unlike election observation, which focuses on procedures at polling stations, forensics works after the fact on numerical outputs, often when on-the-ground access is restricted.
Common methods include:
- Digit tests, such as examining the distribution of last or second-to-last digits in precinct totals against a uniform distribution, or applying Benford's Law to leading digits (though its applicability to vote counts is contested).
- Turnout-vote share correlations, where unusually strong positive correlations between turnout and a single candidate's share can signal ballot stuffing.
- Cluster analysis of precincts with implausibly round numbers (e.g., suspiciously many precincts reporting exactly 100% turnout or 100% for one candidate).
- Sklar-style 2D histograms popularized by physicists Peter Klimek, Yuri Yegorov, and Stefan Thurner, who used them to flag irregularities in Russian and Ugandan elections.
Prominent academic work includes Walter Mebane's research at the University of Michigan, which developed eforensics models combining fraud indicators, and analyses of the 2009 Iranian presidential election, the 2011 and 2018 Russian Duma and presidential votes, and the 2020 Belarusian election.
Limitations are significant. Anomalies can arise from legitimate causes — ethnic geographic clustering, get-out-the-vote operations, or strategic voting — so forensic findings are typically framed as indicators warranting investigation, not proof of fraud. Methods also require granular precinct or polling-station data, which authoritarian regimes often withhold or aggregate.
Organizations such as the OSCE/ODIHR, the Carter Center, and the National Democratic Institute increasingly incorporate forensic analysis alongside traditional observation. For MUN and policy researchers, election forensics is most useful as one evidentiary stream among many, complementing observer reports, exit polls, and parallel vote tabulations (PVTs).
Example
After the 2020 Belarusian presidential election, independent analysts applied election forensics to leaked precinct data and identified statistical patterns inconsistent with the official result declaring Alexander Lukashenko the winner.
Frequently asked questions
Its use is debated. Scholars including Joseph Deckert, Mikhail Myagkov, and Peter Ordeshook have argued that Benford's Law performs poorly on vote counts because precinct sizes do not span enough orders of magnitude; many practitioners prefer second-digit tests or other methods.
Keep learning