This paper uses Bayesian vector autoregression models to forecast steel consumption in South-East Asia to 2005. This technique uses both historical correlations among the variables in the system and Bayesian priors on the parameters to introduce more flexibility into the forecasting process and align the models closer in nature to more traditional structural commodity market models.The forecasts suggest that steel consumption in South-East Asia will rise from 35 million tonnes in 1997 to between 48 and 57 million tonnes in 2005 under the low and high growth scenarios used in this paper. The difference between these forecasts highlights the sensitivity of steel consumption to GDP growth, particularly in Thailand, Malaysia and Indonesia. Thailand is expected to remain the largest steel consumer in the region, while Malaysia is forecast to experience the most rapid growth in consumption. (C) 1999 Elsevier Science Ltd. All rights reserved.