Asymmetric Information and Alternate Premium Rating Methods in U.S. Crop Insurance: A Comparison of High and Low Risk Regions

Abstract

Federally subsidized crop insurance has a long history of underwriting losses. These losses may be due to premium rating procedures that do not account, as fully as possible, for differences in yield loss risk across farms. If farmers understand their own yield loss risk in more detail than is reflected in crop insurance premium rates, an information asymmetry may be leading to adverse selection or moral hazard. Regional differences in underwriting losses suggest that the effect of asymmetric information is relatively large where inter-farm yield variability is also relatively large.

An econometric model is used to identify asymmetric information in crop insurance premium rating. A simulation model is used to compare existing crop insurance premium rates to alternative rates calculated using yield loss risk measures based on existing, farm-specific yield history data. Both models are applied to crop/region combinations where inter-farm yield variability is relatively low (non-irrigated corn in the Corn Belt region) and where inter-farm yield variability is relatively high (non-irrigated, continuously cropped wheat in the Northern Plains).

Region-wide asymmetric information effects are identified for both regions, but the asymmetric information effect is found to be larger in the high variability region. This difference explains at least part of the inter-regional difference in underwriting losses. The simulation analysis suggests that, on average, across an entire region, premium rates derived from a farm-specific measure of yield variability are closer to actuarially fair rates than RMA premium rates. At a county- and farm-level, however, it is much more difficult to say, with a high level of statistical confidence, whether these alternate premium rates are closer than RMA rates to the actuarially fair rates.

To provide a foundation for the crop insurance models, an econometric model of crop yields is estimated and used to separate total yield variation into systematic and random components. Random yield variation is tested against several common distributions, including normal, gamma, and beta. The effect of aggregation on the representation of both systematic and random yield variation is also investigated.