LAUSR

Abstract This article examines the J-curve phenomenon for 16 European transition economies. While previous studies assume a linear relationship between the exchange rate and the trade balance, this paper allows for nonlinearity. Following Bahmani-Oskooee and Fariditavana (2015, 2016), the empirical method used is the nonlinear cointegrating autoregressive distributed lag (NARDL) model of Shin et al. (2013) in which short-run and long-run nonlinearities are introduced via positive (appreciation) and negative (depreciation) partial sum decompositions of the real exchange rate. We argue that the lack of support for the J-curve phenomenon could be due to the linearity assumption. This issue is examined by utilizing the linear and the NARDL models. Using the linear autoregressive distributed lag (ARDL) model, we are unable to find support for the J-curve phenomenon in any case. However, when the NARDL model is used, we are able to find evidence for the J-curve in 12 out of the 16 countries. This suggests that allowing for nonlinearity in the adjustment process is important when studying the J-curve phenomenon.