This study examines how rainfall behavior relates to local performance in the USDA Pasture, Rangeland, and Forage rainfall index insurance program. Using Arkansas from 2014 through 2023, we construct theoretical loss ratios at the grid-interval level to isolate program performance from producer enrollment behavior. We separate ex post performance from ex ante rating information by comparing rainfall behavior observed during the PRF period with rainfall information available through 2012. Realized loss ratios are strongly associated with PRF-era rainfall levels and variability. Grid-intervals with higher mean rainfall relative to their historical baseline had lower loss ratios, while grid-intervals with larger increases in rainfall variability had higher loss ratios. Pre-period mean rainfall shifts do not reproduce the ex post level relationship. Pre-period changes in rainfall variability, however, carry a consistent within-interval signal for later indemnity rates and rate gaps. The evidence supports a specific variability-channel diagnostic rather than a broad claim of general rating failure.
