Improving the accuracy of outlook price forecasts: An application to livestock markets

• Main Author: Colino, Evelyn Del Valle
• Format: Dissertation
• Language: English
• Published: ProQuest Dissertations & Theses 01.01.2009
• Subjects: Accuracy; Agricultural economics; Agriculture; Bias; Business forecasts; Calendars; Cattle; Central banks; Commodities; Decades; Econometrics; Economic forecasting; Economic statistics; Economists; Efficient markets; Essays; Forecasting techniques; Futures market; Futures trading; Hogs; Hypotheses; Livestock; Marketing; Performance evaluation; Prices; Regression analysis; Studies
• ISBN: 1109218699; 9781109218695

The performance and economic value of public outlook forecasts has been of continuing interest to agricultural economists and market participants. This dissertation provide new and powerful evidence on the performance of outlook forecasts relative to futures prices in hog and cattle markets over the last three decades and evaluates numerous time-series models and combinatory procedures as forecasting techniques to improve the predictive accuracy of hog price outlook forecasts. Many of these forecasting techniques have never been applied to livestock markets. Quarterly data from the mid- to late-1970s through 2007 for up to three-quarter ahead is available from four prominent outlook programs: University of Illinois/Purdue University, Iowa State University, University of Missouri, and the Economic Research Service of the U.S. Department of Agriculture (USDA). Overall, results show that in general, futures outperform outlook with some differences statistical significant. However, a combination of futures and outlook forecasts generally provide lower forecast errors than futures alone, and therefore, outlook forecasts of hog and cattle prices provide incremental information relative to futures prices. When compared to numerous time-series models, Iowa's estimates are in general outperformed with statistical insignificant differences. However, even with the use of simple time-series models, findings from the encompassing tests highlight the efficacy of improving Iowa's price forecasting performance via composite procedures. Finally, given the potential benefits of forecast combination, numerous combinatory techniques are evaluated in a true out-of-sample context. A true out-of-sample evaluation of composite forecasts is an issue not always carefully considered in the literature. Results show that significant forecast error reductions can be obtained from the combination of outlook, futures and two simple time-series models under most of the methods considered. More interesting, the simple average composite shows an outstanding performance that tends to increase at longer horizons, a result consistent with previous literature. In addition, evidence says that the accuracy of futures prices is stellar at the first horizon, but weaker at distant horizons, suggesting that the value of market forecasts lies primarily in the short-run.