Price Determination and Forecasting in the Australian (Non-Storable) Live Cattle Market

Published empirical studies of simultaneous rational expectations models of spot and futures markets for non-storable commodities are extremely rare. Indeed, only two countries, the US and Australia, have produced data sets for the study of such markets. This paper develops, and presents estimates of a simultaneous rational expectations model of the live cattle market in Australia, the world's leading beef exporting country. The model contains functional relationships for short hedgers and speculators combined (there is no disaggregation of hedgers' and speculators' commitments in Australian data), long hedgers and speculators, and consumers, and is completed with a spot price equation and market clearing identity. Unit root tests indicate that all variables in the model are stationary, except for consumption of beef and the price of pork, which are 1(1). Cointegration tests suggest that these two variables are not cointegrated. The model is estimated by the instrumental variables method of McCallum, which provides consistent estimates. The estimates of all 15 structural parameters have the expected sign, and all are significant at the five per cent level. In a 34 month post-sample period, the model forecasts the spot and futures prices with per cent RMSE's of 3.6% and 2.1% respectively, and in forecasting the spot price, the model outperforms conventional benchmarks such as a random walk and an ARIMA model. The model also outperforms a lagged futures price as a predictor of the spot price, thus providing some evidence against the efficient markets hypothesis.