The U.S. crop insurance market has several features that set it apart from other insurance markets. These include explicit government subsidies with an average premium subsidy rate of about 60% in recent years, and the legislative requirement that premium rates be set at actuarially fair levels, where the federal government sets rates and pays all costs related to insurance policy sales and services. Bearing these features in mind, we examine the extent to which farmers’ crop insurance choices conform to economic theory. A standard expected utility maximization framework is constructed to analyze trade-offs between higher risk protection and larger subsidy payments. We decompose the effect of coverage level on expected utility into insurance, premium loading, and subsidy transfer effects where the loading effect vanishes if rates are actuarially fair. Given an actuarially fair premium, we infer that a rational farmer should choose either the coverage level with the highest premium subsidy or a higher coverage level. Evidence from a large insurance unit-level dataset contradicts this theoretical inference, and so suggests anomalous insurance decisions. In a novel application of the mixed logit framework, we show that the probability an insurance product is chosen declines as out-of-pocket premium expenditures increase, even though higher values of these expenditures should reflect improved grower welfare. Premium expenditures appear to be more salient than the uncertain future benefits they support. We apply our regression to the recent trend yield adjustment innovation in crop insurance.