Gender Representation and the Adoption of Hiring Algorithms: Evidence from MBA Students and Executives

Abstract

I examine how job performance and diversity considerations shape recommendations for adopting algorithms in hiring. Between 2019 and 2022, around 400 business managers and executives coded up and evaluated algorithms aimed at improving hiring at a firm. Although these algorithms would lead to large performance improvements for most, managers were unlikely to recommend adoption. Instead, their adoption recommendations were strongly shaped by the demographic impacts of the algorithm, particularly in regard to gender. Algorithms that decreased the number of female hires were half as likely to be adopted as those that had no impact or increased it. These results hold while controlling for the algorithm’s impact on job performance, and are not present for two other protected classes, age and country of origin. I rule out that this behavior was driven by fear of illegal discrimination. Instead, using a regression discontinuity design, I provide causal evidence that even marginal decreases in the number of female hires leads managers to reject hiring algorithms. I conclude by discussing the implications of these results for our understanding of algorithmic aversion, algorithmic bias, and the returns to algorithmic decision-making in business.

Keywords: hiring algorithms, algorithmic aversion, gender, algorithmic decision-making, people analytics

JEL Classification: M5, J7, M15