Working Paper: CEPR ID: DP15678
Authors: Leticia Abad; Noel Maurer
Abstract: In 2020, many observers were surprised that the Covid-19 outbreak did not appear to have swung the election. Early returns showed little indication that harder-hit areas swung away from the incumbent GOP. In 1918, however, the United States also held an election in the middle of a devastating pandemic. Using county-level epidemiological, electoral, and documentary evidence from 1918-20 we find that flu mortality had a statistically-significant negative effect on the Congressional or gubernatorial vote. The swing, while precise however, was relatively small and not enough to determine the results. We find no effect from flu mortality on turnout rates or on the 1920 presidential election. Our results hold using overall mortality in 1917 and distance to military camps as instruments for 1918 flu deaths. They also withstand tests of coefficient stability and alternative specifications. Considering that the 1918 flu was much more severe than the 2020 Covid pandemic, the historical evidence implies that surprised observers of the 2020 election should not have been so surprised.
Keywords: elections; pandemics
JEL Codes: N0
Edges that are evidenced by causal inference methods are in orange, and the rest are in light blue.
| Cause | Effect |
|---|---|
| flu mortality rate (I12) | voter turnout rates (K16) |
| flu mortality rate (I12) | 1920 presidential election (K16) |
| flu mortality rate (I12) | vote share of incumbent party (D72) |
| flu mortality rate (I12) | electoral outcomes (K16) |