Asymmetric Effects of Exchange Rate Changes on Thailand-China Commodity Trade: Evidence From 45 Industries

Abstract

China is now Thailand’s largest trading partner and 15% of Thailand’s total trade with the world belongs to China. A previous study that assessed the asymmetric effects of the real baht-yuan rate on the bilateral trade balance between the two countries found that a real depreciation of the baht against Yuan has a worsening effect in the long run. We disaggregate the two countries’ trade by industry and consider trade balance of each of the 45 industries that trade between Thailand and China. When we estimated a linear model, we found no favorable effects of baht depreciation in any industry. However, when a nonlinear model was estimated, we found significant short-run cumulative or impact asymmetric effects in 27 industries and significant long-run asymmetric effects in 15 industries. Additional analysis revealed that four industries will benefit from baht depreciation and six will be hurt from baht appreciation, supporting a new definition of the asymmetric J-curve in a total of 10 industries.

Keywords:

• Asymmetry J-curve
• Thailand
• China
• trade balance
• 45 Industries

Notes

1 China’s exports are on the rise not only to Thailand but also to the rest of the world. For more on China’s trade policies see Urdinez et al. (2015), Linbo (2017), Liang (2017), Rasoulinezhad and Wei (2017), and Wen (2018).

2 For more on the analysis of China-ASEAN cooperation see Soong (2016) and Kung et al. (2016).

3 This section closely follows Bahmani-Oskooee and Baek, (2016) who used such methods to estimate bilateral trade balances between the United States and Korea at the commodity level.

4 Note that an estimate of b could also be negative if the increase in Thailand’s economic activity is due to increase in production of import-substitute goods (Bahmani-Oskooee, 1986). Similarly, an estimate of c could be positive if increase in China’s real GDP is due to an increase in import-substitute goods in China.

5 Note that Pesaran et al. (2001) tabulate new critical values for the F-test that account for integrating properties of variables. Indeed, under this approach variables could be combination of I(0) and I(1) which are properties of most macro variables.

6 Note that nonlinearity originates from the method of constructing the POS and NEG variables.

7 For some other application of these methods see Halicioglu (2007, 2008), Verheyen (2013), Gogas and Pragidis (2015), Baghestani and Kherfi (2015), Durmaz (2015), Al-Shayeb and Hatemi-J. (2016), Lima et al. (2016), and Aftab et al. (2017).

8 Reported in Table 3 are a few additional diagnostics. The Lagrange Multiplier statistic is reported as LM and it is insignificant in most cases, supporting autocorrelation free residuals. Ramsey’s RESET statistic is also insignificant in most cases, rejecting misspecification. Application of CUSUM and CUSUMSQ tests for stability of coefficient estimates are reported in the last column as CS and CS². Stables estimates are identified by “S” and unstable ones by “U”. At least by one of the two criteria, estimates are stable. Finally, the size of adjusted R² reflects a reasonable fit in most models.

9 Note that other diagnostic statistics in Table 7 are similar to those in Table 3 for the linear models and need no more analysis, that is, residuals in most models are autocorrelation free, most nonlinear models pass the RESET specification test, estimates are stable, and most models enjoy good fit.

Additional information

Funding

This research was supported by National Institute of Development Administration (NIDA) Business School Research Grants in Collaboration with the Distinguished Visiting Professor.