Abstract: We estimate the effect of candidate appearance on vote choice in congressional elections using an original survey instrument. Based on estimates of the facial competence of 972 congressional candidates, we show that in more competitive races the out-party tends to run candidates with higher quality faces. We estimate the direct effect of face on vote choice when controlling for the competitiveness of the contest and for individual partisanship. Combining survey data with our facial quality scores and a measure of contest competitiveness, we find a face quality effect for Senate challengers of about 4 points for independent voters and 1 to 3 points for partisans. While we estimate face effects that could potentially matter in close elections, we find that the challenging candidate's face is never the difference between a challenger and incumbent victory in all 99 Senate elections in our study.

Abstract: I exploit a natural experiment by which a racial threat effect (Key, 1949) can be identified. Between 2000 and 2004, the reconstruction of public housing in Chicago caused the displacement of over 25,000 African Americans, many of whom had previously lived in close proximity to white voters. The removal was a largely systematic process, exogenous to the neighborhood of the public housing and even the city. I apply an original method of Bayesian updating to identify the race of voters. Then using individual level, geocoded data for over 800,000 voters I show that after the removal of their African American neighbors, the voter turnout of white voters dropped by over 5 percentage points.

Abstract: I test for ``racial threat'' using a field experiment. The racial threat hypothesis (Key, 1949), which contends that voters will be politically motivated by the presence of a proximate racial outgroup, has been controversial for over 60 years. While laboratory experiments on racial attitudes are common, political behavior motivated by residential racial context has never been subject to an experimental test.

I take advantage of the geography of Los Angeles County, which brings different racial/ethnic groups into close, yet spatially separated, proximity. This allows for a randomized, controlled field experiment to directly test the effects of stimulating racial threat on individual voter turnout. I conduct a test of African American and Latino voters during the June, 2008 California Primary Election using a direct-mail intervention. Results indicate a large racial threat effect for African Americans, but not for Latinos.

Abstract: The racial threat hypothesis (Key, 1949) proposes that voters of a given racial group will be politicized by the presence of a large, spatially proximate racial outgroup. Racial threat is usually operationalized by looking at variation in white voter turnout in response to the presence of African Americans. A large number of studies of racial threat have found inconsistent support for the hypothesis that white voter turnout should be positively related to the size of the proximate African American population. A number of studies have also extended the hypothesis to vote choice and examined the relationship between white voter support for African American candidates and the size of the proximate African American population. The results of these studies have also been mixed.

One reason for these inconsistent findings could be confusion about the behavioral mechanism behind racial threat. I propose a new theoretical mechanism for racial threat that derives from basic social psychology theories. This theory leads me to propose that racial threat should not only be related to the size of the proximate outgroup, but also the intergroup segregation.

Another reason for the inconsistent findings might also be that different studies use data from a different times and places. Particularities of different locations and elections could affect the relationship between racial threat and white voter behavior. However, until recently, studies of racial threat had to be of a limited geographic scope because there had not been a clearly racialized candidate in a nationwide campaign.

The candidacy of Barack Obama created the first explicitly racialized candidate with nationwide appeal. This allows for the first, truly nationwide measurement of racial threat. I use survey data from two nationwide surveys and aggregate election returns to show that white support for Obama has a clear, negative relationship with both the size of the proximate African American population and with the level of Black/white segregation. The relationship is present using both aggregate and individual-level data and the relationship remains even when other individual and contextual-level variables are controlled using multivariate regression analysis. The strength of racial threat is shown to be conditioned by pre-existing racial attitudes.

These findings provide some of the first nationwide evidence for racial threat and also allow for an understanding of what individual attitudes condition the relationship between racial threat and vote choice.

Abstract: The 2008 Presidential Election allows for new tests of the Racial Threat hypothesis (Key 1949). In Los Angeles County, precincts that are overwhelmingly Latino are in close proximity to precincts that are overwhelmingly African American. These are conditions that may prime racial threat. If racial threat exists, then presumably it would have affected the probability that a non-Black voter cast a ballot for Barack Obama in the 2008 Presidential Primary.

I use Census data and the California voter file to perform a Bayesian analysis of surnames to determine the race/ethnicity of individual voters in Los Angeles County. I also geocode voters and precincts in order to measure the spatial distance between precincts. I argue that an observable implication of racial threat is that proximity to the source of threat should condition its effect. I show that Latino support for Obama was negatively correlated with proximity to African Americans. The unit of analysis is precincts, but hyper-segregated residential patterns in Los Angeles mean there is little concern that an ecological fallacy is driving the results.

Abstract: I experimentally test the effects of spatial segregation on the perception of difference between groups and argue that the experimental phenomenon is related to real-world political behavior. I conduct three original experiments with ten trials and a total of 1439 subjects. In these experiments, subjects are asked to evaluate differences and similarities between objects. I show that when subjects are exposed to objects that are spatially segregated from objects in other groups that subjects tend to judge these objects to be more similar than when the objects are integrated. Additionally, objects that are segregated tend to be more discretely categorized in a perceptual task than integrated objects. I argue that these results demonstrate the cognitive effects of segregation and demonstrate how residential racial segregation directly affects intergroup attitudes and, consequently, voting behavior.

Abstract: Many Americans habitually abstain from the political process, allowing some citizens to achieve better representation than others. Political scientists and pundits assume that greater electoral competition will motivate these under-represented citizens and bring them to the polls. First analyzing observational and survey evidence, we find little support for this claim. Then exploiting the rare opportunity of a tied election for major political office, we conduct a large-scale field experiment. Informing citizens that an upcoming election will be close has little mobilizing effect. To the extent that we do detect an effect of electoral competition on turnout, it is concentrated among frequent voters. Our evidence indicates that increased electoral competition is unlikely to be a solution to inequality in representation.