Purpose – This article seeks to examine the reference yield calculation method used in crop insurance rating and provides recommendations that could potentially improve actuarial performance of the Federal crop insurance program. Design/methodology/approach – Conceptual, numerical, and statistical analysis is utilized to evaluate the reference yield calculation method used in the US Federal crop insurance program. Findings – The results suggest that reference yields, which at the time of this study are calculated using National Agricultural Statistics Service (NASS) data, do not accurately represent the average actual yields of the insured pool of producers in the Federal crop insurance program. In addition, it is found that not regularly updating these NASS-based reference yields exacerbates this problem because these reference yields do not appropriately represent the current state of technological progress. Practical implications – The empirical analysis leads this paper to recommend a reference yield calculation procedure that utilizes county-average yields from the risk management agency (RMA) participation database and an approach that uses spatially aggregated average yields in cases when data for a particular county are sparse. Originality/value – No previous study has investigated the reference yield calculation method in the Federal crop insurance program using both RMA and NASS data sets. Moreover, this study contributes to the small literature that examines various aspects of the actual production history (APH) rating platform and suggests refinements to improve actuarial performance.