Christopher S. McIntosh

Qualitative Forecast Evaluation: A Comparison of Two Performance Measures

Agricultural producers and marketers often have access to several sets of forecasts on the same series. Decision makers must be able to evaluate the relative accuracy of the forecasts. The ability to evaluate whether a series will move up or down is important for series related to futures, options, other asset prices, and situations where processes can be modified according to changing economic conditions. A measure of qualitative forecast accuracy from Naik and Leuthold is compared to a probability-based measure developed by Henriksson and Merton. The Henriksson-Merton measure is shown to provide additional, and more accurate, information concerning a models qualitative forecasting ability.

FORECASTING AGRICULTURAL PRICES USING A BAYESIAN COMPOSITE APPROACH

Forecast users and market analysts need quality forecast information to improve their decision-making abilities. When more than one forecast is available, the analyst can improve forecast accuracy by using a composite forecast. One of several approaches to forming composite forecasts is a Bayesian approach using matrix beta priors. This paper explains the matrix beta approach and applies it to three individual forecasts of U.S. hog prices. The Bayesian composite forecast is evaluated relative to composites made from simple averages, restricted least squares, and an adaptive weighting technique.

Evaluating alternative price expectation models for multiproduct supply analysis

Neoclassical economic theory provides an important conceptual framework for the analysis of agricultural production. Theory provides little guidance, however, in the actual specification of empirical models. This paper applies an integrated approach for choosing between price expectation mechanisms in a multiple-equation model when the alternatives are non-nested. Nine alternative specifications of market price and policy information are developed. Price forecasting accuracy, non-nested tests of hypotheses, and out-of-sample predictive accuracy are examined for agricultural production in Iowa. The results call into question the reliability of using forecasting accuracy as the sole guide to selecting a price expectation proxy.

RATIONAL EXPECTATIONS ESTIMATION OF GEORGIA SOYBEAN ACREAGE RESPONSE

The general method of moments procedure is used for estimating a soybean acreage response function assuming the producers hold rational expectations. Results indicate that soybean, corn, and wheat futures prices, lagged acreage, and government programs are significant factors for determining soybean plantings. Implications of the results are that crop acreage selection by Georgia producers is not very responsive to demand shocks. Thus, producers in other regions are more likely to absorb impacts from these shocks on crop acreage selection.