Combining survey, census, and administrative data improves the precision of survey estimates of mean agricultural land values. A components-of-variance model is developed and applied to cropland value data for the Corn Belt Performance of the model compared with other procedures is tested using cross-validation techniques. Results indicate that use of the proposed estimator would improve upon the USDA estimators at both the State and strata levels. At the strata level particularly, the improvements may be very substantial.