Vote early and vote often? Detecting electoral fraud from the timing of 19th century elections

This paper develops a new approach to detecting electoral fraud. Our context involves repeaters, individuals voting in multiple states in the U.S. during 19th Century Congressional Elections. Given high travel times, and the associated difficulties of voting in multiple states on the same day, we exploit the staggered introduction of holding federal elections on the first Tuesday after the first Monday in November (1T1M). The key finding is that county-level turnout rates fell when the closest neighboring state coordinated on 1T1M. This result is consistent with 1T1M adoption making repeating more difficult. In terms of mechanisms, the pattern is stronger in states that had not yet adopted the secret ballot, consistent with the secret ballot itself reducing voter fraud. The results are also stronger in open-seat elections, which tend to be more competitive. Finally, the pattern is also driven by smaller population counties, consistent with repeaters particularly inflating turnout rates in these places.