An Evaluation of Crop Forecast Accuracy for Corn and Soybeans: USDA and Private Information Agencies

Article excerpt

Using 1971-2000 data, we examine the accuracy of corn and soybean production forecasts provided by the USDA and two private agencies. All agencies improved their forecasts as the harvest progressed, and forecast errors were highly correlated and unbiased. The relative forecast accuracy of the agencies varied by crop and month. For corn, USDA's forecasts ranked as most accurate of the three agencies in all periods except for August during the recent period and improved most markedly as harvest progressed. For soybeans, forecast errors were very similar, with the private agencies ranking as most accurate for August and September and making largest relative improvements for August during the recent period. The USDA forecasts were dominant for October and November. Our findings identify several patterns of relative forecast accuracy that have implications for private and public decision makers.

Key Words: corn, private agencies, production forecasts, soybeans, USDA

JEL Classifications: Q11, Q13, C82, Q18

In industrial production, final output is typically known with a high degree of certainty given a set of inputs. In contrast, agricultural crop production is characterized by large variability in output corresponding to the inputs employed. This variability is often the result of changes in stochastic factors affecting agricultural production (e.g., precipitation, temperature), and makes forecasting of crop proauction a challenge. Uncertainty about the level of final crop production is resolved only as the growing season progresses and more information about crop conditions and crop yields becomes available. It is a well-known theoretical result that crop reports that accurately estimate the size of production before harvest can play an important role in the process of uncertainty resolution and can provide an opportunity for less risky decisions (e.g., Bradford and Kelejian).

For corn and soybeans, crop production forecasts are provided by the National Agricultural Statistics Service (NASS) of the U.S. Department of Agriculture (USDA) and numerous private agencies, of which two, Conrad Leslie and Sparks, are the most prominent. The forecasts produced by these private agencies are initially available to subscribers only but move quickly to the market, typically a few days prior to the release of the NASS forecasts. These private forecasts are widely reported in the popular press, and consequently, the private as well as the USDA forecasts provide potentially valuable information to a large number of market participants. Understanding the forecast accuracy of these agencies can be important to decision makers. Identifying when crop forecasts become accurate reflections of final production-how the uncertainty about final crop production is resolved-can indicate at what point decisions that are influenced by final crop production can be made with relative certainty. Furthermore, information about the relative accuracy of these three agencies' forecasts can provide insights into which forecasts are of most value in a specific decision context. Alternatively, a high degree of correspondence in the three agencies' production forecasts could indicate little difference in their informational value. Finally, these agencies use different procedures to develop their forecasts, and to date, no research has focused on the relationship between these procedural differences and their effect on the relative accuracy of crop forecasts. Differences in the relative forecast accuracy attributable to the methods used by the agencies might point to changes in procedures that could improve overall forecast performance.

The most recent research related to the accuracy of crop production forecasts has analyzed the agencies' accuracy by indirect methods, using aggregate measures, and has not considered the relationship between the procedures used to develop the forecasts and their performance. Indirect methods assess the relative forecast accuracy of public and private agencies by measuring the reactions of corn and soybean cash and futures prices to the release of the respective reports (e. …