Measures and Forecasts of Agricultural Land Values

Abstract

This thesis is comprised of two articles that examine farmland values and farmland value forecasts for non-irrigated cropland, irrigated cropland, and pasture. The first article considers the collection and reporting procedures each of three data sources: the United States Department of Agriculture's (USDA) annual report, the Federal Reserve's quarterly 10th District Survey of Agricultural Credit Conditions, and sales data. The objective of this essay is to determine if the land value estimates from opinion based surveys are consistent with values observed from land sales. Additionally, Granger causality tests are used to determine if land values published by the Federal Reserve prior to the release of the USDA report are indicators of USDA land values. The second essay examines forecasts from the Federal Reserve's survey to determine if bankers can accurately forecast land values. Various forecasting techniques are used to compare bankers' quarterly forecasts to their own reported changes in land value. In the first essay, land values from each source are found to be highly correlated, but land value averages from sales data is often higher than averages derived from survey data. With respect to the survey data, first quarter Federal Reserve Bank estimates are indicators of USDA land value estimates. The second essay finds that agricultural lenders have some ability to forecast land values for the next quarter. For each of the forecasting methods used, bankers' one-step-ahead forecasts produced had lower root mean squared errors than forecasts produced by nave forecasting models.