Efficient Learning, Job Turnover and Wage Dispersion

This paper studies the aggregate consequences of individual learning in the labor market. Specifically, I examine this issue in a model of directed search on the job. Once matched, a firm-worker pair gradually learns the match-specific quality, taking the history of realized production as signals. Heterogeneity in beliefs about the match quality and in the job search behavior of workers naturally occurs, resulting from a variety of individual histories. I describe the efficient learning and searching strategy and implement the efficient allocations through a market mechanism in which the labor contract depends deterministically on tenure. Consistent with the stylized facts, the model successfully predicts the tenure effect on both the job separation rate and the probability of on-the-job search, and when search frictions are small, the model generates a dispersed wage distribution with a at tail, along the lines of observations.