Daniel E. Ho
Does stereotype threat explain part of the black-white test score gap? Recent evidence suggests that the perception of a white test administrator causes black test takers to perform worse, but that race of the test administrator itself has no impact independent of racial perception. If correct, this finding sheds new light on the validity, mechanism, and normative legal implications of stereotype threat. Using a Bayesian principal stratification approach -- a framework that makes explicit and relaxes crucial assumptions implicit in conventional instrumental variables (IV) analyses -- I reanalyze a quasi-experiment in a survey of political knowledge during which respondents only interact orally with the administrator. Examining the race of the interviewer alone, I find more evidence of a stereotype lift for whites than a threat for blacks. Moreover, racial effects do not appear to be mediated by racial perceptions alone. The administrator's race affects performance similarly, regardless of the respondent's perception of administrator's race. This suggests that conscious racial perceptions are not crucial to the operation of stereotype threat. The evidence for stereotype threat from this experiment, however, is a mixed bag. Methodologically, the article illustrates widely-applicable methods for researchers to gain more leverage and assess the credibility and sensitivity of IV studies.