Commodity Price Model
Xiaoli Etienne, Idaho Wheat Commission endowed chair in commodity risk management, started work on the project while employed at West Virginia University (WVU), where she received a $20,000 grant from USDA’s Economic Research Service to fund the effort. Etienne and her colleagues have continued perfecting the model since she joined U of I.
Etienne’s team included her former graduate student at WVU, Sara Farhangdoost, and USDA ERS economists Linwood Hoffman and Brian Adam. Their paper, “An Alternative Method to Forecast the Season-average Price for U.S. Corn,” is currently in revision for publication in the Journal of Commodity Markets.
The alternative model they created uses only publicly available data, unlike the agency’s go-to forecast, the World Agricultural Supply and Demand Estimates (WASDE), which Etienne describes as using a “black box” model. WASDE is released monthly based on a top-secret equation and private data, including global market factors often unknown by the public.
USDA’s WASDE forecast has major implications, as it can influence markets and produplanting decisions and is used for calculating government payment programs.
The new alternative model slightly outperforms WASDE from January through April.
“Whether it’s economically significant, that’s another paper to write,” Etienne said.
Their model tracks closely with WASDE throughout the rest of the year – with WASDE having a slight edge during the growing season. In most months, their alternative model also outperforms the other model commonly used by USDA ERS – the Hoffman model developed by Etienne’s team member Linwood Hoffman.
The Hoffman model bases its forecasts on publicly available futures price data. Futures contracts involve locking in future delivery of a commodity at a price set today. Etienne’s alternative model also relies heavily on futures prices, but it adds in current cash prices, providing some real-world data while recognizing many commodities aren’t sold on the futures market.
Etienne’s alternative model is slightly more laborious to calculate than the Hoffman model, requiring regression analysis – statistical processes for evaluating the strength of relationships between elements. Both the Hoffman model and the alternative model offer far greater flexibility than WASDE reports, as they can be computed by any economist at any time, while WASDE is secretive and publishes at a set time of each month.
The alternative model has performed especially well when run against actual price data since 2004, despite extreme corn market fluctuations during those years, demonstrating that it’s highly effective during periods of volatility.
“In years of volatility in prices, the model performs much better,” Etienne said. “That’s when the forecasting really matters. When prices are so volatile, no one knows what the price will be.”
The alternative model was also highly accurate when used for predicting soybean prices. Etienne believes each model fills a niche and using them together would be prudent.
“We need to consider a composite approach,” she said.
The research was funded by USDA ERS under cooperative agreement No. 58-3000-7-0014 with West Virginia University.