Asymmetry cointegration and the J-curve: new evidence from Korean bilateral trade balance models with her 14 partners

ABSTRACT

Introduction of new econometric methods raises interest in assessing the old theories and the J-curve phenomenon is no exception. Like previous research, we first use the linear autoregressive distributed lag (ARDL) approach to investigate the phenomenon between Korea and each of her 14 trading partners. We then employ the recent nonlinear ARDL approach to show that in most cases, exchange-rate changes have short-run and long-run asymmetric effects on bilateral trade balances. Separating depreciations from appreciations, which is the main feature of the nonlinear model, relies upon nonlinear adjustment of the exchange rate and provides relatively more support for the J-curve effect.

KEYWORDS:

• Asymmetry cointegration
• J-curve
• Korea
• 14 partners
• nonlinear ARDL

Acknowledgments

Valuable comments of an anonymous referee as well as the editor are greatly appreciated. Any error, however, is our own responsibility.

Disclosure statement

No potential conflict of interest was reported by the authors.

Notes

1. For a comprehensive review of the literature, see Bahmani-Oskooee and Ratha (2004b) and Bahmani-Oskooee and Hegerty (2010).

2. For review of Korean-related studies, see Bahmani-Oskooee, Xu, and Saha (2017).

3. This section closely follows Bahmani-Oskooee and Fariditavana (2016).

4. They prove that their upper-bound critical values could also be used when we have a combination of I(0) and I(1) variables in a given model. Since most macro-variables are either I(0) or I(1), there is no need for pre-unit root testing under this approach and this is one of the main advantages of this approach.

5. This short-run concept of the J-curve was originally introduced by Magee (1973) and empirically tested by Bahmani-Oskooee (1985) using a traditional approach and not an error-correction modeling technique.

6. Note that, as argued by Bahmani-Oskooee and Fariditavana (2016), the expected sign of normalized coefficient estimates of POS and NEG variables in Equation (4)(4) $\begin{array}{l}\Delta {\text{LnTB}}_{i,t}=a\text{'}+{\displaystyle \sum _{j=1}^{n1}b{\text{'}}_j\Delta {\text{LnTB}}_{i,t-j}}+{\displaystyle \sum _{j=0}^{n2}c{\text{'}}_j\Delta \text{Ln}{Y}_{\text{KOR},t-j}^{}}+{\displaystyle \sum _{j=0}^{n3}d{\text{'}}_j}\Delta \text{Ln}{Y}_{i,t-j}^{}\\ \stackrel{}{{\displaystyle }}\stackrel{}{{\displaystyle }}\underset{}{\overset{}{{\displaystyle }}}\stackrel{}{{\displaystyle }}\stackrel{}{{\displaystyle }}\stackrel{}{{\displaystyle }}\stackrel{}{{\displaystyle }}+{\displaystyle \sum _{j=0}^{n4}e{\text{'}}_j}\Delta {\text{POS}}_{t-j}+{\displaystyle \sum _{j=0}^{n5}f{\text{'}}_j\Delta {\text{NEG}}_{t-j}}+{\theta }_0{\text{LnTB}}_{i,t-1}+{\theta }_1\text{Ln}{Y}_{\text{KOR},t-1}^{}\\ \stackrel{}{{\displaystyle }}\stackrel{}{{\displaystyle }}\stackrel{}{{\displaystyle }}\stackrel{}{{\displaystyle }}\stackrel{}{{\displaystyle }}\stackrel{}{{\displaystyle }}\stackrel{}{{\displaystyle }}+{\theta }_2\text{Ln}{Y}_{i,t-1}^{}+{\theta }_3{\text{POS}}_{i,t-1}+{\theta }_4{\text{NEG}}_{i,t-1}+{\xi }_t\end{array}$(4) is the same as that of REX in Equation (2)(2) $\begin{array}{l}\Delta {\text{LnTB}}_{i,t}=\alpha +{\displaystyle \sum _{j=1}^{n1}{\beta }_j}\Delta {\text{LnTB}}_{i,t-j}+{\displaystyle \sum _{j=0}^{n2}{\delta }_j}\Delta \text{Ln}{Y}_{\text{KOR},t-j}+{\displaystyle \sum _{j=0}^{n3}{\gamma }_j}\Delta \text{Ln}{Y}_{i,t-j}\\ +\underset{}{\overset{}{{\displaystyle }}}\underset{}{\overset{}{{\displaystyle }}}\underset{}{\overset{}{{\displaystyle }}}\underset{}{\overset{}{{\displaystyle }}}\underset{}{\overset{}{{\displaystyle }}}\underset{}{\overset{}{{\displaystyle }}}\underset{}{\overset{}{{\displaystyle }}}\underset{}{\overset{}{{\displaystyle }}}{\displaystyle \sum _{j=0}^{n4}{\pi }_j}\Delta {\text{LnREX}}_{i,t-j}+{\lambda }_1{\text{LnTB}}_{i,t-1}+{\lambda }_2\text{Ln}{Y}_{\text{KOR},t-1}+{\lambda }_3\text{Ln}{Y}_{i,t-1}\\ +\underset{}{\overset{}{{\displaystyle }}}\underset{}{\overset{}{{\displaystyle }}}\underset{}{\overset{}{{\displaystyle }}}\underset{}{\overset{}{{\displaystyle }}}\underset{}{\overset{}{{\displaystyle }}}\underset{}{\overset{}{{\displaystyle }}}\underset{}{\overset{}{{\displaystyle }}}\underset{}{\overset{}{{\displaystyle }}}\underset{}{\overset{}{{\displaystyle }}}{\lambda }_4{\text{LnREX}}_{i,t-1}+{\mu }_t\end{array}$(2) .

7. For some other application of partial sum concept, see Verheyen (2013) on interest-rate pass-through mechanism to deposit rates; and for application of the linear model, see De Vita and Kyaw (2008), Halicioglu (2007), and Hajilee and Al-Nasser (2014).

8. The exception to this rule was the model with China in which the maximum lag order was set at four due to limited number of observations.

9. Note that in order to manage the short-run estimates, the coefficient estimates of the dependent variable are not reported.

10. Pesaran, Shin, and Smith (2001, 303) also tabulate an upper- and a lower-bound critical value for this test but only for large samples.

Additional information

Notes on contributors

Mohsen Bahmani-Oskooee

Mohsen Bahmani-Oskooee is a distinguished professor and Wilmeth professor of economics at the University of Wisconsin-Milwaukee. He is also the director of the Center for Research on International Economics at the same institution. He has published extensively in international finance and open economy macroeconomics and currently serves as the editor of the Journal of Economics Studies.

Jungho Baek

Jungho Baek is a professor and economics program director in School of Management (SOM) at UAF. His main fields include international trade, economics of environment and trade, and energy economics. He has published many articles and prior to SOM, he conducted research for the Korean Institute for International Economic Policy (KIEP) in South Korea and Center for Agricultural Policy and Trade Studies (CAPTS) at North Dakota University.