Self-image Bias and Lost Talent

We propose an overlapping-generations model in which established researchers evaluate the research of new researchers. All researchers are differentially endowed with equally desirable research characteristics and belong to two groups, M or F, which have identical ex-ante productivity distributions. Evaluations are group-blind. Yet, when research is evaluated on many characteristics, evaluators' self-image bias and mild between-group heterogeneity lead the initially larger group, say M, to dominate indefinitely. M-researchers only accept the research of young scholars with characteristics close to theirs. Promoted F-researchers are thus few and "similar" to M-researchers, perpetuating the asymmetry. This talent loss is exacerbated by candidates' career concerns and institutions' focus on hiring faculty whose research will be approved by established researchers. Mentorship reduces group imbalance, but it increases the F-group talent loss. Affirmative action reduces both. Our model's predictions are consistent with existing empirical evidence on female participation in academic economics