Nonlinear effects of exchange rate changes on the South African bilateral trade balance

In testing for the J-curve, previous studies have shown that the trade balance model is better fitted using cointegration and error correction mechanisms. These mechanisms are able to incorporate the short-term deterioration and the long-term improvement of the trade balance – the definition of the J-curve. However, the drawback of the established cointegration and error correction frameworks is that they assume symmetry in the equilibrium adjustment process. Incidentally, studies which have used the linear frameworks have found little support for the J-curve. Since the adjustment process could be nonlinear, a fresh investigation of the J-curve using nonlinear approaches could provide competing evidence. This paper retested the J-curve by using quarterly data for South Africa and her key trade partners (China, Germany, India, Japan, the UK and the USA) and found the linear specification to support the J-curve phenomenon in only two cases (India and the USA) under relaxed conditions. In contrast, the nonlinear specification supported the J-curve phenomenon in all cases at no cost of serial correlation and functional misspecification. We also found the real exchange rate changes to have significant nonlinear effects on the South African trade balance.