Selecting the Best: The Persistent Effects of Luck

Abstract

This article proposes a stylized model of organizational learning where agents’ performance is the sum of time-invariant, unobservable ability, privately-chosen effort, and transitory noise. Our main result shows that, even when performance is almost entirely random, identification of the most able agent requires final selection to be biased in favour of early strong performers. Making luck persistent, e.g. through professional fast-tracks or high-potential programs, is thus rationalized by selective efficiency. Organizational learning propagates disadvantages stemming from an agent’s identity, e.g. race or gender, but identity-dependent biases, such as gender-specific mentoring, create incentives that make selection both more efficient and equitable.