The Quality of Vote Tallies: Causes and Consequences

Abstract: Virtually all theoretical work on elections and collective choice assumes that votes are either tallied perfectly, or that deviations from the truth stem from partisan fraud. However, in large electorates where votes are tallied by hand, as in most modern democracies, the accuracy of vote totals cannot be guaranteed even in the absence of fraud. We provide the first systematic evidence about the incidence, causes, and consequences of inaccuracies in poll-booth-level vote tallies. Using data for the universe of over seventy million voters in Mexico in five national elections, we exploit various procedural randomizations and rule-based discontinuities to show that the human capital of poll booth officials, tallying difficulty, and workload are major causal drivers of tallying inaccuracies. We find no evidence that tallying inaccuracies reflect partisan manipulation or fraud. Nevertheless, inaccuracies are strongly correlated with subsequent recounts, and in tight races they can change who wins. Even nonpartisan tallying inaccuracies, therefore, threaten public trust in election results and confidence in electoral institutions. Our findings highlight an unrecognized challenge to the practice of democracy and collective choice.