When Drought Travels: Partisan Electoral Responses Through Food Networks in U.S. Presidential Elections

This paper examines whether climate shocks are associated with electoral behavior beyond the counties in which they occur. The analysis covers 3,069 counties in the contiguous United States across five presidential elections from 2008 to 2024. County-level election returns are combined with pre-election drought measured using the Palmer Z Index and with network exposure constructed from estimated county-to-county food flows in 2007. A model in which voters interpret drought through pre-existing partisan preferences predicts that the electoral association of drought becomes more favorable to the Democratic candidate as a county's historical Democratic vote share increases. The estimates are consistent with this prediction. Local drought is associated with lower Democratic two-party vote shares in historically Republican counties and higher shares in historically Democratic counties. Network drought exhibits the same sign-changing pattern, including where measured local drought is absent. The estimated relationships cross zero at historical Democratic vote shares of about 0.40 for local drought and 0.44 for network exposure. Directional exposure through suppliers versus buyers adds little explanatory power conditional on total network exposure. The network relationship is comparatively stable across alternative specifications. Turnout and party-vote results suggest that the pattern reflects changes in party-specific electoral support rather than a common change in aggregate participation across counties.