The Allocation of Talent Under Perfect and Imperfect Information

We develop a general sorting model that incorporates both self-selection by applicants, who assess their own skills, and evaluation by admissions agents, who rely on noisy signals of skills. Building upon the theoretical framework of the Roy model, our analysis examines how the presence of noise in signals of skills influences admission behaviour and the wage distribution. Our analysis reveals two novel behavioural effects on admission procedures. First, institutions may optimally assign positive weight to skill signals that are uncorrelated and unproductive in the relevant sector a phenomenon we call talent hoarding. This talent-hoarding effect disrupts comparative advantages, reallocates talent towards restricted sectors, and diminishes overall efficiency. With the presence of noise in signals, the admissions agent is further incentivised to increase the number of admitted applicants, which lowers wages. This, in turn, reduces reliance on admission rules and promotes more informative self-selection—a behavioural effect we label talent separation. Under relatively lenient assumptions, talent separation improves efficiency. Evidence from Danish administrative data reveals empirical patterns consistent with the predicted talent-hoarding effect, and a structural model of Denmarks education system highlights that the two behavioural effects can have a substantial impact on the distribution of wages.