Separating Moral Hazard from Adverse Selection: Evidence from the U.S. Federal Crop Insurance Program

We use data from the administrative les of the U.S. Department of Agriculture's Risk Management Agency to examine how the distribution of crop yields changed as individual farmers shifted into and out of the federal crop insurance program. The large panel facilitates use of fixed effects that span each combination of farmer and production practice to account for unobserved differences in farmer abilities, risk preferences and soils, in addition to fixed effects for interactions between all years and all counties to account for geographically-specific technological change, local prices, and weather. We also account for farm-specific yield variances. Conditional on this large set of fixed effects, we estimate the mean shift in yield and non-parametrically estimate the shift in the distribution around the conditional mean associated with enrollment incrop insurance. Because differences between farmer and land types have been accounted for (i.e., controlling for adverse selection), the estimated shifts in yield distributions likely reflect moral hazard. For most crops in most states we find insurance is associated with statistically signi cant but small downward shifts in average yield. The largest shifts occur for cotton and rice, the highest-value of ve crops considered. By integrating the estimated shift in yield distributions over actual indemnities paid, we provide estimates of the total indemnities paid due to moral hazard. Our results indicate moral hazard accounted for an estimated $53.7 million in indemnities between 1992 and 2001, which amounts to 0.9% of indemnities paid to the insured crops and states considered.