Crop insurance is plagued by relatively little historical information but significant spatial information. We investigate the efficacy of using nonparametric Bayesian model averaging (BMA) to incorporate extraneous information into the estimated premium rates. Nonparametric BMA is particularly suited to this application because it does not make any assumptions about parametric form or the extent to which yields are similar. We evaluate the proposed estimator under small-to-medium sample sizes and various geographical restrictions on the distance of spatial smoothing for policy relevance. The nonparametric BMA consistently decreases error and enables statistically significant and economically important rents to be captured.