When Performance Trumps Gender Bias: Joint Versus Separate Evaluation

Implicit Preferences Inferred from Choice

“…Then they are "more sensitive" to quality in the choice set {High, Low, P ool} (High is always chosen and Low is never chosen), than in {High, P ool} and {Low, P ool} (High and Low are equally likely to be chosen in each case). We do however show that Bohnet et al (2015)'s subjects exhibit violations of WARP that point to implicit preferences, though we believe are harder to interpret.…”

“…While intuitively the variation in frequency of certain choices points to implicit preferences (as we have argued, considering multiple candidates increases revealingness with respect to their attributes), in fact we show that it is not possible to infer implicit preferences from these data: regular, transitive preferences will generate the patterns of choice that Bohnet et al (2015)'s tests interpret as varying sensitivity. A simple 59 Note that this test is not a valid parallel scissors in general for detecting an implicit preference over work ethic: the target in both scissors is Excellent, while parallel scissors would require one Poor and one Excellent target (see Proposition 5).…”

“…A recent experimental paper by Bohnet et al (2015) can be interpreted as studying implicit preferences. They study whether a decision-maker's choice between candidates for a task becomes more or less sensitive to certain attributes-gender and past performance-when the choice is either between an individual candidate and an unobserved "pool" alternative, or between two candidates and the pool (a paradigm closely related to joint and separate evaluation, although it elicits choices, not evaluations).…”

“…Prior experiments on implicit preferences. A few prior experimental studies have relied on the intuition that we are attempting to formalize: Snyder et al (1979) on implicit discrimination against the handicapped, Exley (2015) on implicit preferences over giving to charity, and Bohnet et al (2015) on implicit gender discrimination. For each of these papers we show that, although they study implicit preferences in our sense, the statistical tests which they use to identify implicit discrimination are imperfect (i.e., they would identify implicit preferences where none exist), and we describe alternative appropriate tests.…”

Does the Gender Mix Among Employers Influence Who Gets Hired? A Labor Market Experiment

“…Figure A.3 in Appendix A shows a sample of the screen employers encountered. We presented information about applicants jointly in pairs following work by Bohnet et al (2012) which shows that joint evaluation reduces gender bias compared with separate evaluation. This enabled us to further isolate individual gender preferences as a source of hiring bias instead of other environmental factors.…”