On Statistical Discrimination as a Failure of Social Learning: A Multiarmed Bandit Approach
Management Science · 2026 · Volume 72 · Issue 1
Abstract
We analyze statistical discrimination in hiring markets using a multiarmed bandit model. Myopic firms face workers arriving with heterogeneous observable characteristics. The association between the worker’s skill and characteristics is unknown ex ante; thus, firms need to learn it. Laissez-faire causes perpetual underestimation: minority workers are rarely hired, and therefore, the underestimation tends to persist. Even a marginal imbalance in the population ratio frequently results in perpetual underestimation. We demonstrate that a subsidy rule that is implemented as temporary affirmative action effectively alleviates discrimination stemming from insufficient data. This paper was accepted by Nicolas Stier-Moses, Management Science Special Issue on The Human-Algorithm Connection. Funding: This work was supported by the Social Sciences and Humanities Research Council of Canada [Grant 430-2020-00088] and JST ERATO [Grant JPMJER2301], Japan. Supplemental Material: The online appendix and data files are available at https://doi.org/10.1287/mnsc.2022.00893 .
Keywords
statistical discrimination, social learning, affirmative action, multiarmed bandit, algorithmic fairness
Enable JavaScript for citation tools and additional paper interactions.