Testing the J-curve hypothesis for the USA: applications of the nonlinear and linear ARDL models

Abstract

This paper investigates the evidence of the J-curve hypothesis between the United States, a country which has the world’s largest trade deficits and an anchor  currency, and its main 12 trading partner countries over the period 1991M1–2015M2. To this aim, we apply both linear and nonlinear Autoregressive Distributed Lag (ARDL) cointegration approaches and error-correction model (ECM). The nonlinear ARDL approach, recently introduced by Shin et al. (2014), allows us to examine the  separate effects of both the appreciations and depreciations of the USD on the trade  balances of the country. The empirical results indicate that while the linear approach
 supports the evidence of the J-curve for the USA with only 4 trading partners, the nonlinear approach supports such evidence with 8 trading partners. This implies that the nonlinear approach provides more evidence of the hypothesis than the linear approach. Therefore, this study reveals that existing but concealed potential evidence for the J-curve effect may be discovered with the nonlinear  approach, which allows for nonlinearity in the adjustment process. Another empirical finding of this study is that depreciations in the USD seem to have more distinct long-run effects on the US trade balances than appreciations.