Why do some people become entrepreneurs (and others don't)? Why are firms so heterogeneous, and many firms so small? To start, the paper briefly documents evidence from the empirical literature that the relationship between entrepreneurship and education is U-shaped, that many entrepreneurs start a firm "out of necessity", that most firms are small, remain so, yet persist in the market, and that returns to entrepreneurship have a much larger cross-sectional variance than returns to wage work. Popular models of firm heterogeneity cannot easily account for the U-shape or for the persistence of low-productivity firms. The paper shows that these facts can be explained in a model of occupational choice between wage work and entrepreneurship where agents are heterogeneous in their ability as workers, and starting entrepreneurs face uncertainty about their project's productivity. Then, if agents' expected productivity as entrepreneurs is increasing and not too concave in their ability as workers, the most and the least able individuals choose to become entrepreneurs. This sorting is due to heterogeneous outside options in the labor market. Because of their low opportunity cost, low-ability agents benefit disproportionately from the ability to pursue only good business projects and abandon low-productivity ones. This also makes them more likely to immediately abandon a project for a new one. Data from the NLSY79 gives support to these two predictions. Individuals with relatively high or low wages when employed, or with a high or low degree, are more likely to be entrepreneurs or to become entrepreneurs, and spend more time in entrepreneurship. Among entrepreneurs, more of the firms run by individuals with low wages when employed, or with a low degree, are abandoned after only a year.