Multivariate forecasting of a commodity portfolio: application to cattle feeding margins and risk

Traditionally, economists have utilized univariate approaches to forecast prices, even for firms operating in multicommodity environments. This research improves the way in which managerial decision making is analysed by developing a model better representing the price risks and opportunities faced by firms that produce using a portfolio of commodities. Using the situation of cattle feedlot investors and managers as an example, this is accomplished by recognizing the multivariate (live cattle, feeder cattle and corn prices) situation that feedlots operate in and employing corresponding multivariate simulation techniques. Evaluation suggests that properly modelling the cattle feeding margin as a multivariate set of prices significantly improves the accuracy of forecasting future feeding margin values realized in the cash market. The model also suggests incorporating both implied and historical time-varying volatility information when forecasting margin variability. Implications for other multicommodity situations and future research are also provided.