Take up of microcredit by the poor for investment in businesses or human capital turned out to be very low. We show that this could be explained by risk aversion, without relying on fixed costs or other forms of non-convexity in the technology, if the investment is aimed at increasing the probability of success. Under this framework, rational risk-averse agents choose corner solutions, unlike in the case of a risky investment with an exogenous probability of success. Our online experiment confirms our theoretical predictions about how agents' choices differ when facing the two types of investments.