Score contribution per author:
α: calibrated so average coauthorship-adjusted count equals average raw count
This paper uses a unique data set and a natural experiment to test if electoral fraud can exist in late-arriving votes. On the night of the 2019 Bolivian elections, the official vote counting system that was expected to publish the results real-time, suddenly stopped. When it resumed, the results had flipped. We estimate several difference-in-differences specifications using a 2016 referendum and the votes of other political parties. We find that the extent of the fraud is 2.51% of the valid votes, sufficient to change the outcome of the election. Our results are robust to polling-station-level shocks common across 2019 and 2016, as well as 2019 specific shocks. This controls for geography, socioeconomic characteristics, unobserved voting preferences, and endogeneity in the arrival of the polling station results. We report evidence of fraud that occurred before the shutdown and document a statistically significant discontinuous jump in the votes during the shutdown. We provide insights on how to apply our different identification strategies to test for fraud in other elections.