An empirical analysis of Korea's trade imbalances with the US and Japan

Abstract

This paper investigates the macro-determinants of Korea's persistent bilateral trade deficit with Japan but her trade surplus with the US using Johansen's cointegration-error correction model, which includes bilateral trade balance and real exchange rate, domestic and foreign incomes and relative money supply. The empirical results show that all variables included affect bilateral trade balances and there exists a long-run equilibrium among them. Especially, Korean won depreciation improves Korea–US trade balance according to the Marshall–Lerner condition while the J-curve effect between Korea and Japan exists with a little improvement of deterioration of trade balance followed by Korean won depreciation. In addition, domestic economic growth is found to improve persistent trade deficits against Japan and mitigate trade surplus with the US. The short- and long-run effects of monetary policy on Korea–US trade balance are opposite to those on Korea–Japan. The exogenous US income contributes to increase Korea–US trade balance.

Keywords:

• bilateral trade imbalance
• real exchange rate
• Marshall–Lerner condition
• J-curve effect
• cointegration-error correction model

JEL classifications:

• C32
• F32

Notes

∗∗

∗∗∗ denote 5% and 1% levels, respectively.

^a The Johansen's cointegration test is carried out with the model 4, which allows for linear deterministic trend in data and includes intercept and trend in CE and no trend in VAR.

^b Critical values are adjusted by a scaling factor, T/(T − nk), to the asymptotic values.

∗Rejection of the hypothesis at the 0.05 level.

^a P-values.

^b The LR test statistics, which are calculated as LR = −2∗ (lr − lu), where lr and lu are the maximized values of the log-likelihood function of the restricted (without dummy variable) and unrestricted (with dummy variable) regressions, respectively; the 5% critical value of χ² with one degree of freedom equals to 3.84.

∗

∗∗

∗∗∗

1. About 90% of the Korean imports belong to industrial materials and fuels and capital goods while the rest 10% are consumer goods. About 40% of the Korean imports are used for export production with its 60% for domestic production.

2. Although China is a major trading partner of Korea since 1990, China cannot be included in this study mainly due to the restrictions imposed by data availability.

3. Saudi Arabia's trade share was exceptionally high during the 1970s' oil shocks because of Korea's heavy reliance on imported oil from the Middle East countries.

4. Carrying out the same estimation for 18 industrial trading partners with the US as that for developing countries, Bahmani-Oskooee and Ratha (2004c) got the same results as those with developing countries. That is, they find no J-curve effect in the short run while real depreciation of the dollar has favorable effects on the US trade balance in most cases in which ML condition holds.

5. A time series is stationary if its mean, variance and covariance are time-invariant and is referred to as integrated of order zero, or I(0). However, most series are nonstationary because they usually have a linear or exponential time trend. A nonstationary series can be converted into a stationary one through the process of differencing. If the first differenced series is stationary, the original series is integrated of order one, or I(1).

6. The distribution of a test statistic is known to be sensitive to the lag length used. If the lag order used is less than the true lag, the regression estimates will be biased and the residuals will be serially correlated. If the order of lags used exceeds the true order, the power of the test is likely to be reduced. Moreover, as Johansen's cointegration tests are rather sensitive to under-parameterization in the lag length, though not so to over-parameterization (Cheung and Lai 1993, p. 326), optimal lag selection is based on AIC or the FPE criterion rather than Schwarz information criterion (SIC) and the Hannan and Quinn criterion (HQ), as the former move away from the lowest possible lag order at a slow rate with increase in the sample size.

7. An equivalent way to make finite-sample corrections is to adjust the test statistics and not the critical values. A scaling factor is given by (T − nk)/T (Cheung and Lai 1993, p. 317).

8. To conserve space, the unadjusted test statistics, which have the same results as the adjusted ones, are not reported but are available from the author on request.

9. The correlation coefficient between the won–dollar and won–yen exchange rates is 0.46.

10. This makes sense because the imports of final consumption goods are only 10% of the total imports.