Bias-Motivated Updating in an Online Labor Market

Abstract

In the canonical economics literature on discrimination, it is assumed that statistical discrimination based on inaccurate beliefs will not persist since agents have clear incentives to update as Bayesians based on accurate information. However, if beliefs about group productivity are driven by bias rather than by an agnostic lack of information, agents may be resistant to updating in the face of accurate information that contradicts stereotypes. In an online labor market experiment, I find that employers’ response to information about the labor market productivity of Black and White workers is a function of their implicit biases.

Keywords: discrimination, bias, labor, experiment, beliefs, race

JEL Classification: C91, D83, J15, J71