We exploit a natural experiment to study voter taste-based discrimination against nonwhite political candidates. In Illinois Republican presidential primary elections, voters do not vote for presidential candidates directly. Instead, they vote delegate-by-delegate for delegate candidates listed as bound to vote for particular presidential candidates at the Republican nominating convention. To maximize their support for their preferred presidential candidate, voters must vote for all that candidates delegates. However, some delegates names imply they are not white. Incentives for statistical discrimination against nonwhite delegates are negligible, as delegates have effectively no discretion, and taste-based discrimination against them is costly, as it undermines voters preferred presidential candidates. Examining within-presidential-candidate variation in delegate vote totals in primaries from 20002016, we estimate that about 10 percent of voters do not vote for their preferred presidential candidates delegates who have names that indicate the delegates are nonwhite, indicating that a considerable share of voters act upon racially-discriminatory tastes. This finding is robust to multiple methods for measuring delegate race, to controls for voters possible prior information about delegates, to ballot order, and to other possible confounds we consider. Heterogeneity across candidates and geographies is also broadly consistent with taste-based theories.
