Many subjects in lab experiments exhibit small-stakes risk aversion, consistent with loss aversion. Those with greater math skills are less likely to show small-stakes risk aversion. We argue that departures from expected utility maximization may help explain why many firms in developing countries leave high expected return investments unexploited. We show that among a sample of Kenyan shopkeepers, inventories are negatively associated with small-stakes risk aversion and positively associated with math skills.