Confusion in Risk Aversion Experiments in Low-Income Countries

Although early risk aversion experiments were designed to detect subject confusion, many recent studies are not, and pay little attention to the issue. I re-examine four recent studies in low-income countries and find that the empirical distribution of risk aversion in one AER study is statistically indistinguishable from that of confused subjects making choices at random. Eighty percent of subjects in a second AER study answered in a way that is consistent with a heuristic strategy, and responses imply extremely high levels of risk-loving behavior. Researchers should design to detect confusion, and journals should demand the results