Most data on the market value of farm real estate are presented on the basis of national aggregates. Though continuing to serve many useful purposes, such data have limited use when more exacting economic analysis is required. Data are needed that more accurately reflect 10eal market responses. Thus, an automatic interaction detection model was used to group counties on the basis of their similarity across selected farm and nonfarm factors into optimal farm real estate submarket areas. Through use of the model, factors are identified and examined that help to discriminate between both local markets and variations in land values.