The Case of the Missing Data: Short Samples and Multiple Imputation in Rating Crop Insurance

The federal crop insurance program plays a central role in US farm income support but often faces challenges in rating coverage due to missing county-level yield data. Traditional approaches, such as spatial aggregation, may fail to capture true risk by treating proxy values as observed data. This study applies a multiple imputation (MI) framework to address missing yield data using information from related crops, neighboring counties, and temporal indicators. We construct four imputation scenarios using corn and soybean yield data from Adair County, Iowa, varying the type and scope of supplemental inputs. Results show that MI-based approaches outperform direct proxies in capturing uncertainty and producing more stable parameter estimates. While cross-crop data alone introduce variability, combining spatial, cross-crop, and temporal indicators yields the most robust estimates. These findings highlight the importance of accounting for uncertainty in proxy yields and suggest that MI can improve the actuarial soundness of premium rate determination in data-sparse settings.