The predicted effects of distance and population sizeon voting are displayed in Figure 4. These simulations ofvoter turnout in 2004 based on size and distance from thedemolishedprojects,conditionalonhavingvotedin2000,support Hypothesis 2. Between 2000 and 2004, turnoutfor the average white registered voter in Chicago declined(see Figure A.2 in the supporting information). Figure4 reflects voters near the demolished projects decreasingtheir turnout at a faster rate. Figure 4(a) demonstratesthat for a person already inclined to vote, the probabilityof voting increases with distance: by almost 10 percent-age points when moving 500 meters away from the de-molished projects. This indicates that persons living nearprojects were significantly motivated by their proximityto the projects when the projects were still standing. Simi-larly,Figure4(b)showsthatasapersonalreadyinclinedtovote moves from living near projects representing a smallportion of the local population to living near projectsrepresenting a large portion of the local population, herprobability of voting decreases, indicating that, for whitevoters living nearby, the relative size of the local outgrouphad a significant effect on voter turnout.Effects on Vote ChoiceScholars have observed correlations in a variety of set-tingsbetweenproximateoutgroupsandvotingforracially
DEMOLITION OF PUBLIC HOUSING AND RACIAL THREAT13FIGURE4 Effects of Distance and Size of Projects(a) Distance0.600.650.700.75Distance from ProjectPr(vote2004)02004006008001000 1200 1400 1600 1800 2000(b) Percent of Local Black Population0.600.650.700.75Percent of Local Black Population in Demolished Project10%20%30%40%50%60%70%80%90% 100%Note:Predictedeffectsgeneratedfromvote2004=0+1(log(distance))+2(log(localpercent))+vote2000,withwhitevoters.Figure 4(a) is the predicted probability that a person who voted in 2000 will vote in 2004 with increasing distance, while holdingsize at its mean. Figure 4(b) is the predicted probability that a person who voted in 2000 will vote in 2004, with increasing outgrouppopulation size, withdistance=100. Dotted lines represent 95% confidence intervals generated by bootstrapped standard errors.conservativecandidatesoragainstcandidatesper-ceived as representing the outgroup (Carsey 1995; Enos2010; Giles and Buckner 1993; Key 1949; Spence andMcClerking 2010). An observable implication of myclaims about the effect of racial threat on voting is thatthe removal of the outgroup might produce changes invoters’ propensity to vote Republican. With this claim, Iam relying on substantial evidence that more racially con-servative voters prefer Republican candidates (e.g., Teslerand Sears 2010). My prediction in Hypothesis 3 is thatthe demolition of the housing projects should lower theproportion of white voters living near the projects whovote for Republican candidates.