"Does the Size of the Black-White Wage Gap Depend on Who's Asking?"  
(w/ Dan Black and Jeffrey Grogger)

Abstract

We identify a new source of endogenous non-response, and possible measurement error, in survey elicitations of wage and other financial information. By using data on interviewer characteristics, we document a systematic bias in response induced by the social distance—that is the similarity in race, age, and gender—between the interviewer and respondent in the NLSY97 and the SCF. Specifically, we find (i) higher levels of non-response for black respondents when questioned by male interviewers, (ii) higher overall response from female black interviewers, and (iii) suggestive evidence that, conditioned on response, reported wage is mediated by the race-gender of the interviewer and respondent. We illustrate the impact of these findings by re-estimating the gender and racial wage gap using NLSY97 data by interviewer characteristics. We then present a model of incidental selection induced by social context, discuss the relevant cognitive and social psychological principles that may underlie the effect, and consider implications of these findings for broader research in labor economics.