Using the IAT to predict ethnic and racial discrimination: small effect sizes of unknown societal significance

Greenwald, Banaji, and Nosek (2015) present a reanalysis of the meta-analysis by Oswald, Mitchell, Blanton, Jaccard, and Tetlock (2013) that examined the effect sizes of Implicit Association Tests (IATs) designed to predict racial and ethnic discrimination. We discuss points of agreement and disagreement with respect to methods used to synthesize the IAT studies, and we correct an error by Greenwald et al. that obscures a key contribution of our meta-analysis. In the end, all of the meta-analyses converge on the conclusion that, across diverse methods of coding and analyzing the data, IAT scores are not good predictors of ethnic or racial discrimination, and explain, at most, small fractions of the variance in discriminatory behavior in controlled laboratory settings. The thought experiments presented by Greenwald et al. go well beyond the lab to claim systematic IAT effects in noisy real-world settings, but these hypothetical exercises depend crucially on untested and, arguably, untenable assumptions.