Should education become more vocational or more general? We address this question in two steps. We first build and solve a two-sector matching model with generalists and specialists. Generalists pursue jobs in both sectors; however, they come second in job queues. Specialists seek for jobs in a single sector; they come first in job queues. Self-selection in education type vehicles three main externalities: specialists boost job creation in each sector; generalists improve the efficiency of the matching technology; generalists exacerbate firms’ coordination problems. We then calibrate the model on the labor market for upper-secondary graduates in OECD countries. In each country, we match the proportion of specialists and unemployment rates by type of education in 2000. Self-selection is always inefficient: taxing vocational education to reduce the proportion of specialists down to the efficient level could reduce unemployment rates (for upper-secondary graduates) by 1.1 to 1.8 percentage points.