Election Forensics: Identifying Election Fraud

Claims of election fraud get a lot of attention in contemporary political discourse, but how can we tell whether fraudulent votes have actually been counted? University of Michigan political scientist Walter R. Mebane, Jr. has spent much of his career trying to answer this question by developing Election Forensics (eforensics), a statistical model for determining how many votes, if any, are improperly allocated in an election due to fraud. Using Bayesian statistical methods, the eforensics model enables researchers to estimate for each precinct (or any other vote aggregating unit) what voting patterns should look like in each of three conditions: no fraud, incremental fraud, and extreme fraud. Researchers can evaluate a given precinct’s observed voting patterns and determine with which of these three conditions the pattern most closely aligns, thus offering a measure of the extent of realized fraud in an election.

The eforensics model has been applied to test for voter fraud in a number of national elections. Most recently, Mebane examined the 2024 presidential election in Venezuela and found statistical support for the vote tallies provided by the opposition—a finding that strongly suggests that the official numbers provided by President Nicolás Maduro are fraudulent. More specifically, the statistical evidence suggests a miniscule number of fraudulent votes (fewer than 60, if any, out of more than 10 million) were counted among the opposition’s tallies, meaning that it is virtually statistically impossible for those tallies to have been faked. The role Mebane’s analyses have played in coverage of the election has received international attention: “Statistical Analisis of Electoral Fraud Allegations in Venezuela’s 2024 Presidential Election,” Latin American Post (8/29/2024) and “Evidence shows Venezuela’s election was stolen – but will Maduro budge?,” The Guardian (8/6/2024).

If you’re interested in learning more about Prof. Mebane and his work, see this introductory discussion of eforensics. His application of eforensics to this year’s Venezuelan election can be found with this study. Mebane’s working book project, tentatively titled Election Forensics: A Finite Mixture Model that Estimates Realized Election Frauds, will offer a comprehensive approach to eforensics. For more details on Prof. Mebane’s research agenda, check out his website.