In this study, we apply Markov regime-switching models to various foreign trade variables with the aim of characterising the behaviour of the growth of exports and imports in Turkey. The results indicate that the forecasting performance of the proposed Markov regime-switching models is much better than that of the random walk benchmark except for the growth of import values. However, for the growth of export volumes, the proposed Markov regime-switching model always outperforms the random walk benchmark. Furthermore, we investigate whether the growth regimes of exports and imports can be associated with the growth regimes of world trade volumes, world trade prices and industrial production by introducing a probability score and comparing the regime probabilities. The results indicate that world trade prices are more closely related to import values than they are to export values except for low growth regime of export values. In fact, low growth regime of export values is strongly associated with low growth regime of world trade prices. On the contrary, normal growth regime of export volumes is closely related to high growth regime of world trade volumes. Moreover, world trade volumes and industrial production are more closely related to import values and import volumes than they are to export values in all regimes. Actually, low growth regimes of import values and import volumes are strongly associated with low growth regime of industrial production.
