The Veil of Elections

Residential partisan segregation---i.e., the tendency for partisans to live near co-partisans and far from anti-partisans---is an extensive and accelerating trend in the United States. As partisans segregate, opportunities shrink for cross-group contact and dialogue, fueling extremism and distrust. Beyond shaping mass attitudes, partisan segregation may also polarize elites, as both electoral campaigns and news media appeal to fractured constituencies. It is urgent we understand what drives partisan segregation, as such processes can amplify echo chambers and create feedback loops that allow even slight levels of segregation to quickly become extreme.

Why partisans are sorting geographically is not yet understood. Surveys suggest many Americans prefer to live near co-partisans and to avoid anti-partisans. Whether such preferences rank highly in comparison to housing affordability, neighborhood safety, and good school districts is yet unclear. Some scholars argue that Americans generally seek like-minded neighbors, and that partisan segregation is simply a byproduct of the increasingly tight relationship between social and political identity. Others argue instead that sorting is driven entirely by differential preferences for neighborhood, rather than for neighbor---such as by Democrats being more likely than Republicans to accept smaller homes for walkability, as surveys also suggest. These different possible causal mechanisms for why partisans continue to sort suggest different policy countermeasures.

In ongoing research (with Amber Lee, Todd Nief, Panos Toulis, and Kristian Lum), we seek to distinguish between competing causal theories of partisan segregation in the US. Using a novel comprehensive nationwide dataset of millions of housing sales and voter records going back 50 years, our work develops the "veil of elections" methodology to identify the causal effect of partisan homophily on housing purchases. Our approach asks the counterfactual question: would this Democrat home-buyer have bought the same house had her neighbors been instead Republican? To identify the causal effects of co- (or anti-)partisan would-be neighbors on home-buying decisions, we exploit the heightened visibility of neighborhood partisanship during election seasons—e.g., when lawn signs and campaign activity reveal neighbors' political leanings. 

Preliminary results have thus far revealed subtle and inconsistent evidence of partisan homophily influencing home-buying decisions, perhaps suggesting that partisan sorting is being driven by confounding factors. However, restricting attention to contested elections, allowing for realistic shopping-to-purchase delays, and refining the scope and granularity of neighborhoods, are several modeling choices that our ongoing work seeks to better understand. This work generally seeks to provide some of the first large-scale empirical evidence on whether Americans intentionally sort into politically homogeneous communities, with implications for understanding polarization's roots and designing interventions to bridge the partisan divide.