An evaluation of determinants of fed cattle basis and competing forecasting models

Abstract

The objective of this analysis is to develop econometric models for forecasting fed cattle basis as well as compare these models with historic averaging methods of forecasting basis popular in existing literature. The econometric analysis also aims to identify important determinants of fed cattle basis. Both monthly and weekly models were assessed with data provided by the Livestock Marketing Information Center. All models analyzed the three regions of Nebraska, Kansas, and Texas. Monthly historic average approaches utilized historic fed cattle futures and fed cattle cash price series from January of 1995 through December of 2010. Weekly historic average approaches utilized historic fed cattle futures and fed cattle cash prices series from June of 2001 through December 2010. Data collected post mandatory price reporting implementation in 2001 was used in all econometric models. Overall lags of fed cattle basis, the spread between the nearby live cattle futures contract and the next deferred futures contract, and seasonality regularly proved to explain much of the variation in fed cattle basis in the econometric modeling. Multiple historic average based models were examined on both monthly and weekly frequencies. Once all competing models were estimated in-sample, out-of sample testing was conducted. The forecasting errors of all weekly models were compared to determine which methods prove to be dominant forecasters of fed cattle basis. This testing suggests historic averaging methods outperform the alternate econometric models in out-of-sample work. The econometric models helped to reveal some of the important factors determining fed cattle basis, however lags in collecting data on these factors may inhibit the forecaster’s ability to use these techniques in real time. One interesting revelation in regards to historic averages is the potential of Olympic averages as forecasters. These methods have not been explored in previous academic literature but tend to perform quite well in comparison with other methods explored.