These four facts are related. They suggest that: entrepreneurs have very heterogeneous outside options, so some
become entrepreneurs ‘‘out of necessity’’. These may (a conjecture) mainly be people with low levels of education. The
firms they run most likely will remain small, if they manage to survive. Suppose that some variance in returns to
entrepreneurship also arises from heterogeneous quality of projects. Finally suppose that, while any budding entrepreneur
could end up running projects of varying return, those with higher education would on average run their projects better, or
run better projects. Then it is clear that the fact that entrepreneurs come from the extremes of the ability distribution
implies that the observed post-selection cross-sectional variance in returns will be high relative to the variance in returns
any individual might face. Hence, selection from the extremes of the ability distribution, arising from heterogeneous
outside options, increases observed variance in returns to entrepreneurship.
The model developed in the next section shows how selection from the extremes can occur naturally in a simple,
general setting, and also matches the other facts.