An Analysis of Trade Patterns and the Effects of the Real Exchange Rate Movements in East Asia

ABSTRACT

This article investigates the patterns of vertical specialization in trade among China, Japan and Korea, and the effects of real exchange rate fluctuations under a multistage production process. By extending the models of Yi (2003, 2010), we derive two distinct features of vertical specialization and test them using Time-Varying Parameter (TVP) VAR. We find that a positive shock to China’s final good consumption increases the intermediate goods trade between Korea and China, with expanding magnitude over time. In addition, the positive effect of a real exchange rate depreciation on intermediate goods trade is strengthened through the competitiveness-enhancing channel, with this effect being more pronouncing in Korea-China trade than in Korea-Japan trade.

KEY WORDS:

• multistage production
• real exchange rate
• time varying parameter VAR
• vertical specialization

Acknowledgments

The authors are grateful to Woon Gyu Choi, Hyun-Jeong Kim, Kyungmin Kim, Tae Jeong Kim, Dongyeol Lee, Jin-Su Park and Seryoung Park, as well as seminar participants at the Bank of Korea and at the Korea International Economic Association for their helpful comments.We are grateful to the editor Ali M. Kutan and anonymous referees for valuable comments. We are also grateful to Michael Marking for his careful editing.

Notes

^1. Korea has been running a trade surplus against China since the early 1990s, and the size of which has tended to increase except during the Asian Financial Crisis and the Global Financial Crisis, and has decreased since 2013. The share of intermediate goods in Korea’s total exports to China was approximately 84% in 2000, decreased to 71% in 2010, and remained at around 74% in 2015. In contrast Korea has been always running a trade deficit against Japan, whose size has tended to grow except during the two financial crises and started to decline from 2010. The share of intermediate goods in Korea’s total imports from Japan has ranged between 65% and 74% since the 2000s.

^2. The share of intermediate goods in China’s total exports to Korea increased from 50% in 2000 to 60% in 2010, and remained over 60% until 2015. The share of intermediate goods in Korea’s exports to Japan rose as well, from 63% in 2000 to 70% in 2010, and remained over 70% until 2015. These patterns show the relative importance of the intermediate goods trade, even though the gross value of trade has not been growing significantly of late.

^3. For a summary of these studies on East Asia, see Suh (2008).

^4. The indices of Suh (2008) differ from those of Hummels, Ishii, and Yi (2001), in that the former’s indices are bilateral.

^5. Thorbecke (2008) applied the panel DOLS method to avoid the reverse causality problem.

^6. There is also the estimation of a gravity equation model. For more about that model, see Rose (2000) and Berthou (2008).

^7. Baak, Al-Mahmood, and Vixathep (2007), Chou (2000), Chowdhury (1993), Arize (1995) and Arize, Osang, and Slottje (2000) use the Error Correction Model (ECM). By using DOLS, Arize, Osang, and Slottje (2000) show the robustness of their cointegration results. For exchange rate volatility, Chowdhury (1993), Baak, Al-Mahmood, and Vixathep (2007) and Arize, Osang, and Slottje (2000) utilize standard deviations, while Chou (2000) and Arize (1995) adopt ARCH(1).

^8. Considering the small geographical distances affecting both the Korea-China and the Korea-Japan country pairs, we omit all other distribution costs.

^9. Hummels, Ishii, and Yi (2001) define vertical specialization as occuring “when i) a good is produced in two or more sequential stages, ii) two or more countries provide value-added during the production of the good, and iii) at least one country must use imported inputs in its stage of the production process, and some of the resulting output must be exported.”

^10. Yi (2010) notes that the margins from Case 2 will be only a fraction of goods, and the overall effect will be smaller compared to that in Case 1.

^11. We assume $\sigma =1$, for simplicity in abstracting the internal margin in Yi (2003).

^12. Yi (2010) assumes a Frechét distribution for $A(z)$.

^13. For the detailed derivation, refer to the Supplementary Material, available online.

^14. As in Eaton and Kortum (2002) and Yi (2010), a Frechét distribution is assumed for TFP, and given by $F(A_k)=e^{-T{A}_k^{-n}}$, $k=H,F$. $T$ is a location parameter that increases the mean of $A_k$, while a larger value of n implies lower heterogeneity of A.

^15. The same reasoning applies to home final demand, as in Case 3 in Figure 2.

^16. Think of the first order conditions of consumption and labor. Real wages can either rise or fall depending upon the relative positions of the labor supply and demand curves. Consumption and labor in contrast rise in response to a positive demand shock.

^17. In Equations (15) and (16), an increase in the labor supply, $L^F$, leads to increases in $H^{\ast }$ and $F^{\ast }$. If a multistage production process does not exist, foreign exports to the home country do not increase in response to an increase in foreign demand.

^18. For the detailed derivation, refer to the Supplementary Material, available online.

^19. The result of this article is different from that of Yi (2010). Compared to the effect of tariffs in Yi (2010), the degree of amplification here is lower. In Yi (2010), tariffs have an amplified effect on $\underset{\_}{z}$ in addition. The differing result here is because, unlike in the case with a tariff, real depreciation in one country implies real appreciation in other countries.

^20. We thank an anonymous referee for pointing out these features in a clearer way.

^21. The difference in volume between Korea’s exports to China and China’s exports to Korea may also deepen the gap.

^22. Note that data on the China-Japan pair is not available in the Korean custom data. Due to this data limitation we do not explore the country pair of China-Japan.

^23. China’s (real) retail sales and Korea’s final goods exports to China are also used in turn.

^24. Japan’s (real) retail sales and Korea’s final goods exports to Japan are also used in turn.

^25. In this analysis, we observe that the monthly export and import data is highly volatile even after aggregation, and so we use an HP filter to smooth out such short-term noisy movements. However, the estimation results may be sensitive to the method of extracting the cyclical components, as Ferroni (2011) has shown in the context of Bayesian DSGE estimation.

^26. Considering that the portions of primary goods are very low in all three countries (less than 5% in 2013), we abstract primary goods in our subsequent analysis.

^27. The export price indices of Korea are applied to Korea’s exports to China and Japan, and Korea’s import price indices are used for China’s and Japan’s exports to Korea.

^28. The median responses of these three variables during the six periods are given in the appendices. The responses of the remaining two variables, i.e. 91 day CP and the real exchange rate, are omitted so as to avoid complexity.

^29. The median responses of these three variables during the six periods are given in the appendices. The responses of the remaining two variables, i.e. 91 day CP and the real exchange rate, are omitted so as to avoid complexity.

^30. Changes in China’s (or Japan’s) exports have demand-boosting effects on Korea’s final goods exports as well.

^31. Given that our VAR is constructed with only four variables, there could be some missing explanatory variables that are important in explaining the trade volume movements.