Algorithmic Bias and the Diversity Paradox in Leadership

We study how algorithmic biases in recruitment affect top management diversity. Individuals differ in talent and invest in human capital before appointment. When appointment thresholds differ systematically across groups, a higher threshold lowers entry-level representation but can raise representation at the top. By motivating the most talented individuals to invest more, higher thresholds generate a thicker upper tail of the ability distribution. Consequently, diversity at senior levels may increase even as entry-level diversity declines. The model challenges the "pipeline argument" and predicts that algorithmic biases may improve top-management diversity in meritocratic firms but reduce it in less meritocratic ones.