AbstractA nonparametric test of the expected utility hypothesis is developed in this paper. The expected utility hypothesis is shown to hold if there exists a feasible solution to a particular system of linear inequalities. Furthermore, when a feasible solution exists, boundaries on the coefficient of absolute risk aversion can be calculated explicitly. The test is applied to data on land allocations modeled as choices over lottery sets. Results, contrary to those obtained in many laboratory experiments, show that in most cases the expected utility hypothesis cannot be rejected.