Election Forensics: Latent Dimensions of Election Frauds and Strategic Voting∗

Many statistical methods that use low-level election vote count data to detect election frauds have the limitation that they have a hard time distinguishing distortions in vote counts that stem from voters ’ strategic behavior from distortions that originate from election frauds. Identifying latent components that underlie various election forensics statistics (e.g., digit tests, Klimek-related estimates) can help show the extent to which the statistics measure fraudulent as opposed to strategic behavior. We use polling station voting data and postelection complaint data from a recent German election to illustrate the latent variable methods. Geographic ambiguity about the locations at which some complaints occur motivates embedding a geographic mixture structure in the latent variable model. “Fraud ” probabilities that stem from an idea that elections without fraud are unimodal in fact respond strongly to strategic behavior.