The work laid out in this proposal aims to change our understanding of labor markets by viewing both the mobility as well as the frictions in the market as a consequence of long-term worker heterogeneity. Despite the advances in information technology which substantially reduce the costs of sending information (job advertisements, job applications) extracting the relevant information about worker quality remains hard. Long-term differences in ability coupled with screening frictions are proposed as the main reason for mismatch, for mobility, and for the presence of unemployment.The proposal is based on novel empirical observations on occupational mobility. Both low-paid workers as well as high-paid workers in an occupation tend to leave it. The former tend to move to occupations with lower average pay, while the opposite holds for the latter. This happens even within firms, and after excluding managerial positions.Most work on selection assumes that low-earners leave. This data suggest a novel angle: Workers have a long-term type that affects productivity in their current and in new occupations. They might accumulate human capital, but also their baseline ability is imperfectly known. Unexpectedly low performers (low-wage workers) have to leave towards less demanding tasks, while high performers change to more demanding tasks. This consistently accounts for the observed selection patterns.When workers know more about their ability than new firms, this also explains unemployment: firms spend efforts on screening, and impose costs on workers to induce them to self-select. The latter counteracts exogenous reductions in workers’ search costs. The aim is to develop a tractable model of screening unemployment that can serve as a building block in larger macro-labor models, and to assess the work of the government employment agency through the lens of a mechanism designer that facilitates match-making but relies on firms for additional screening of the unemployed.