We introduce an equilibrium labor market model, where workers gradually learn about their unobserved production abilities. While engaged in a productive activity (occupation), workers observe their output and extract information that allows them to make inferences about their unobserved aptitudes. Because workers are learning about themselves, their output informs them not only about their productivity in their current occupation, but about their likely productivity in other occupations as well. As workers acquire more information, they self-select into the occupations in which they expect to perform best, and their wages increase. Returns to experience here capture improved job selection by workers as they sort through occupations and learn about their productive abilities. Our setup can account for the offsetting worker flows across occupations, the within-occupation wage inequality, as well the decline in the probability of occupational switching as workers grow older, that we observe in the data. We use this framework to investigate whether the interaction of learning and search frictions can lead to further decreases in output. Indeed, an increase in the unemployment rate similar to the one experienced by many European countries in the early 1970s is found to reduce the flow of output per employed worker by 1% annually.
