Primary Submission Category: Causal Inference and Bias/Discrimination
Parsing Taste-Based from Statistical Discrimination in Audit Experiments
Authors: Viviana Rivera-Burgos, Thomas Leavitt,
Presenting Author: Viviana Rivera-Burgos*
The literature on legislative responsiveness aims to parse racial (taste-based) discrimination from statistical discrimination that is due to legislators’ strategic incentives to appeal to co-partisan constituents. In this paper, we show that extant designs may be unable to do so because of a lack of symmetry in when legislators are exposed to signals of race and signals of party identification. For example, in e-mail audit studies, the putative race of the e-mail sender is signaled by the e-mail address (at which point legislators can choose whether to open the e-mail), but the party of the sender is signaled to legislators only if they open the e-mail. We derive the bias for the effect of race + party treatments that results from this lack of symmetry. We then propose two solutions: (1) We show how to implement sensitivity bounds when scholars can measure whether or not legislators open an e-mail and (2) propose a new design that uses a racially neutral e-mail address and then exposes legislators to race and party cross-cutting treatments within the body of the e-mails. We implement the former solution on an original audit experiment. Both solutions enable scholars to better discern the mechanisms behind – and hence solutions to – racial discrimination in legislators’ responsiveness.