Interpretive Evaluation: E↵ects of Confirmation Bias on the Retribution to Talent

This work studies the e↵ects of confirmation bias on the retribution to unobservable talent in a competitive labor market. Under evaluation with confirmatory-bias, the candidate always exerts a positive level of ef- fort to influence information, but e↵ort decreases over time, converging to zero. While Bayesian beliefs converge at linear rate t to the true tal- ent, confirmatory-biased beliefs converge at an exponential rate 2t to a weighted average between talent and the initial prejudice. If initial beliefs are biased, then for any prior precision h0 > 0, confirmation-biased wages never converge to the talent, so the Retribution Bias is persistent, even with ad infinitum optimal signalling by the candidate. Thus, confirmation bias becomes a new source of market ine ciency, of persistent nature. The Retribution Bias increases in the initial prejudice gap and in the relative prior-to-signal precision h0/h". For the case when the market conditions initial beliefs on observable characteristics (e.g., gender, ethnicity), we analyze di↵erent measures of bias in retribution relative to individual tal- ent. The inter-individual wage gap, commonly used in public discussion, reflects both an inter-group prejudice gap and an inter-individual talent gap, and therefore is a confounding and inconclusive measure of group- based biases. A more appropriate measure of the inter-group prejudice gap is the Retribution Bias Gap, since it is orthogonal to -and therefore controls for- the inter-individual talent gap.