This paper contributes to the literature on forecasting seaborne trade flow
s by presenting multivariate autoregressive time series models that can be
used to produce long-term forecasts. The models are applied to forecasting
the Grade flows of four commodity markets (crude oil, iron ore, grain and c
oal) on major trade routes. The empirical results indicate that the models
can produce long-term seaborne trade flow estimates that have relatively sm
all forecast errors. (C) 2001 Elsevier Science Ltd. All rights reserved.