Social Networks and the Geography of Voting: A Spatial Econometric Approach

点击翻译按钮获取中文摘要

This thesis examines the relationship between the geographic structure of social networks and electoral outcomes. It draws on spatial analysis, econometric methods, and data on the density of social ties between the populations of different places to make empirical contributions to a body of related literatures in the social sciences. The thesis is comprised of an introduction, three independent empirical chapters, and a conclusion. The common theme uniting all three chapters is the observation that non-local, and even geographically distant social ties can play an important role in shaping local voting behaviour. In the first chapter, I investigate the effect of social proximity to trade-related employment shocks on vote choice in the 2016 UK EU membership referendum. Looking at referendum results in International Territorial Level 3 (ITL3) regions and individual-level voting data from the British Election Study, I examine whether regional social proximity to shocks in different local labour markets affected support for the Leave option. Instrumental variable estimates suggest that social spillover effects on voting behaviour are comparable to those of within-region exposure and travel an average distance of 74 to 102 kilometres. In the second chapter, I examine social spillovers on choice of voting method from local rollouts of all-mail voting policy in the United States. Combining individual-level administrative records on the entire electorate of North Carolina and data on the social ties between zip code tabulation areas (ZCTAs) across states, I look at how the share of social ties in policy-switching jurisdictions affected mail voting in the 2020 US presidential election. Difference-in-differences estimates are indicative of positive spillover effects, primarily originating from populous Western jurisdictions located over 2,000 kilometres away. In the third chapter, I explore the geography of partisan homophily in the 2020 US presidential election. I use the local Moran index to identify clusters of politically homophilous ZCTAs and examine their characteristics. The findings suggest that while spatial patterns of partisan homophily broadly track residential segregation along the urban-rural continuum, there are notable partisan differences in the relative density and geographic distance of social ties: homophilous Democratic-leaning areas are likely to have denser and more distant ties elsewhere than homophilous Republican-leaning areas. Overall, the thesis contributes new evidence on the role of social interactions in elections and its links to public policy.