Association for Public Policy Analysis & Management

We have long known that economics inform voting behavior and decisions (Rees et al. 1962; Feldman 1982; Rosenstone 1982; Bowler and Donovan 1994; Cutler et al. 1999; Vigdor, 2004; Cheung and Cunningham, 2011). Economic circumstances, in particular, can drive political participation through both the capacity to participate and the agitation to rectify an unfair distribution of resources (Brady 2003). Much of the research on economic voting, however, has focused on the economic circumstances of the voters themselves. We intend to fill the “spatial gap” in the economic voting literature to understand how the economic conditions of where voters live influence their likelihood to vote. We consider a scenario, the Great Recession, where economic shocks were quite localized and sudden. Furthermore, we focus on housing-related shocks, which uniquely characterized the Great Recession and were felt particularly hard by middle and working class homeowners in the U.S. We hope to shed light on the ways in which individuals interact with the political process when faced with an economic threat to their biggest asset—their home.

We use a novel dataset on registrants and voters in local, state and national elections for the Tampa, FL, metro area, over the course of two decades. These data were obtained from the state and county boards of elections, and also contain a range of registrant-level characteristics like age, race, gender and party affiliation. We augment these data with information on demographics and housing markets from the U.S. Census and FL Department of Revenue to capture differences in localized socioeconomic conditions and housing-related shocks.

Our data possess four key features. First, we can follow registrants over time and observe whether or not they vote in local, state and national (primary and general) elections. Second, we can identify and differentiate among local offices and measures on the ballot. Third, we can construct very fine-grained geographies to capture localized economic circumstances. Finally, the sample is quite large (over 780,000 unique registrants in total), allowing for enough power in fine-grained analyses to test for heterogeneity. There is meaningful variation across Florida, both geographically and temporally, with respect to baseline economic conditions and to how places weathered the recession.

Using these data we can exploit the geographic variation in (i) how economically vulnerable people and places were leading up to the Great Recession, (ii) how the housing price shock of the Great Recession impacted local communities, and (iii) how individuals responded with respect to voting participation. We estimate a model to determine how the magnitude of a localized housing price shock affects the probability of voting (across various election types and controlling for the composition of the election ballots). The crux of our analytical strategy is individual-level fixed effects, so that we are identifying off of changes in individual voting behavior, controlling for changes in micro-geography conditions.