An Equilibrium Theory of Learning , Search and Wages ∗

We propose an equilibrium theory of learning from search in the labor market, which addresses the search behavior of workers, the creation of jobs, and the determination of wages as functions of labor market histories. In the model, each worker has incomplete information about his job-finding ability and learns about it from his search outcomes. The theory formalizes a notion akin to discouragement: over each uninterrupted unemployment spell, unemployed workers update their beliefs about their job-finding abilities downward and reduce their desired wages. One contribution of the paper is to integrate learning from search into an equilibrium framework. By inducing endogenous heterogeneity in workers’ beliefs, learning from search provides a novel explanation for a set of related empirical observations about the labor market, including unemployment duration dependence and wage dispersion. Another contribution is to apply lattice-theoretic techniques to analyze learning from experience, which is useful because learning generates convex value functions and, in principle, multiple solutions to a worker’s optimization problem. JEL classifications: E24, D83, J64.