Farmers’ decisions about how much crop insurance to buy are not generally consistent with expected utility maximization. Taking into account both marginal risk benefits and marginal subsidy effects suggests that most farmers have chosen lower coverage levels than would be predicted by standard models. By modeling financial outcomes as gains and losses, cumulative prospect theory offers an alternative framework to perhaps better understand farmers’ purchase decisions. The role of the reference point that defines outcomes as either a gain or a loss, the degree of loss aversion, curvature of the value function, and the probability weighting function in determining optimal crop insurance coverage levels are explored for three representative farms calibrated to 2009 conditions. Loss aversion and how crop insurance is framed through choice of the reference point are shown to be the key factors that determine whether predictions from prospect theory are consistent with observed crop insurance coverage choices. When crop insurance is framed as a tool to manage farm risk then optimal choices under prospect theory are not consistent with observed choices. If crop insurance is framed as a stand-alone investment where a loss is felt if the indemnity received is less than the premium paid, then prospect theory can generate optimal coverage level choices that are largely consistent with observed decisions. This result is shown to be robust to changes in parameterizations as long as loss aversion is maintained and if curvature of the value function is accompanied by decision weights that overweight low probability outcomes.