Replication Data for: Errors and Calibration in Mail Ballot Signature Rejections

Many U.S. states require election workers to check signatures on ballots returned by mail, causing tens of thousands of ballots to be rejected at each general election. This paper applies Bayes’ theorem, along with findings from prior research, to show how rejection rates are likely to vary under plausible assumptions about signature checker accuracy and the prevalence of invalid signatures. This approach yields a pair of predictions. On the one hand, given findings from research on handwriting forensics, the Bayesian analysis implies election workers are more likely to wrongly reject valid ballots for purported signature mismatch than to correctly reject invalidly signed returns. On the other hand, research on election workers as problem-solvers suggests they may try to minimize the wrongful rejection of ballots. I find empirical support for each prediction, including signs that election workers learn from past errors to calibrate signature rejection rates downwards. This suggests election worker discretion may be an unexpected source of resilience against laws otherwise liable to disenfranchise eligible voters, although I also discuss conditions under which such discretion may conflict with democratic norms.