A natural experiment on discrimination in elections

Abstract

We exploit a natural experiment to study discrimination in elections. In Illinois Republican presidential primaries, voters vote for delegates bound to presidential candidates, but delegates' names convey information about their race and gender. We identify discrimination from variation in vote totals among delegates bound to the same presidential candidate and who face the same voters. Examining delegate vote totals from 2000 to 2016, we estimate nonwhite delegates receive 9 percent fewer votes. We find essentially no gender discrimination. Negligible incentives for statistical discrimination, costs to preferred presidential candidates, and heterogeneity are consistent with an interpretation of this behavior as taste-based.

Why Did the Democrats Lose the South? Bringing New Data to an Old Debate

“…44 Online Appendix Tables A. 11 to A.14 show that the patterns in Table 3 hold when instead of party ID we use votes in presidential and congressional races as outcome variables.…”

“…Although both large and contentious, the literature on the cause of dealignment has a clear gap: due to the limitations of standard datasets, existing quantitative work is unable to examine racial attitudes before Civil Rights was a key political issue (and often not until several years after that). 11 Due to this limitation, a standard econometric decomposition of the share of dealignment accounted for by those with conservative racial views has not been possible.…”

“…10 For further evidence that enfranchising blacks makes white voters more conservative, see Ang (2017), who uses county-level variation in the requirements of the Voting Rights Act to show that strengthening enforcement of black voting rights increases black turnout, but in fact increases Republican vote share because white voters in these areas shift their political allegiances. 11 Even Shafer and Johnson (2009), the book perhaps most associated with the argument that economic development and not Civil Rights triggered dealignment, write: "Introducing racial attitudes… will prove more diffi-cult… because there is less substantive consistency in the opinion items asked by the [A]NES in the realm of race policy for the full postwar period." While the deficiencies of ANES racial views questions are well documented, we revisited the ANES ourselves, hoping that past scholars had given up too quickly.…”

Do Name-Based Treatments Violate Information Equivalence? Evidence from a Correspondence Audit Experiment

“…Names are common treatments when the goal is to estimate the effect of a characteristic—such as race or ethnic minority status—on a behavioral outcome, and name-based treatments have been used in observational studies (Barth, Mittag, and Park 2019; Broockman and Soltas 2020), survey experiments (Doherty, Dowling, and Miller 2019), and field experiments (Butler and Broockman 2011; Costa 2017). In audit studies, researchers use fictional aliases to manipulate a desired characteristic and measure discrimination based on the different service received (Butler and Homola 2017).…”

The Price of Prejudice

“…(JEL C93, J15, J24, J31, J71) F ield experiments have been used for more than 40 years to investigate the causes of ethnic discrimination in the workplace (see Riach and Rich 2002 for a survey). 1 In so-called correspondence tests (e.g., Bertrand and Mullainathan 2004), pairs of fictitious résumés are submitted to employers by mail. Discrimination is inferred from differential callback or job-offer rates across pairs of workers which are similar in all respects except for ethnicity.…”