The export response to exchange rates and product fragmentation: the case of Chinese manufactured exports

Abstract

This paper examines how changes in the Chinese real exchange rate affect China's exports in the context of global production networks. It highlights the misspecification inherent in conventional export models and the importance of distinguishing between the very different impacts of exchange rate changes relative to export destinations and those relative to sources of parts and components. Our empirical estimates cast further doubts on the effectiveness of Chinese exchange rate adjustments for reducing Chinese export volumes.

Keywords:

• export supply function
• real exchange rates
• trade flows
• production networks
• Chinese Renminbi

JEL Classification:

• F14
• F31

Acknowledgements

This paper benefitted from comments and suggestions from Noel Gaston, an anonymous reviewer, and Sandi Suardi, as well as from seminar participants at the Institute of Social and Economic Research, Osaka University, Japan, and participants of the Economic and Policy Developments in East Asia workshop held at Bond University, Gold Coast, Australia, on 13–14 October 2011. We thank the Institute of Social and Economic Research, Osaka University, Japan for providing facilities for undertaking this research.

Notes

1. Many Chinese economists, for example, assert that an appreciation of the Chinese currency will have very little impact on its net exports and the US trade deficit (see, for example, opinion pieces in Financial Times by Yukon Huang, Vice-president of strategic intelligence and a lead China analyst at Stratfor, a Texas-based private intelligence company [Financial Times, 5 October 2011], and Yao Yang, Director of the China Centre for Economic Research at Peking University [Financial Times, 6 October 2011]).

2. For example, Quanta Computer, the largest laptop original design manufacturer (ODM) originating in Taiwan, collects parts and components from around the globe – such as Intel microprocessors and Microsoft operating systems from the United States, graphic tips designed by ATI technologies from Ontario in Canada, hard disc drives from Japan, and liquid crystal display (LCD) screens and memory chips produced from companies in Taiwan and South Korea – and then assembles them at Quanta Shanghai Manufacturing City in China. Quanta Computer was listed as one of the Global Fortune 500 Enterprises in Fortune Magazine in 2006. See http://www.quantatw.com/Quanta/english/Default.aspx

3. See Dixit and Norman (1980) and Woodland (1982).

4. This study uses the GTAP database for estimates of the required parameters and presents export supply and import demand elasticities for a large number of countries, including China, under various scenarios. These calculated elasticities naturally depend on the underlying parameter values of the variables in the expression for elasticities. Many of these have been estimated using data covering varying periods of time and using different econometric approaches.

5. Jongwanich (2010) acknowledges the issue but uses an export-weighted RER in her estimation. For expositions of the traditional approach, see Goldstein and Khan (1978) and Rose (1991).

6. The CCS data contain a harmonised system (HS) eight-digit product level classification of China's trade flows administered by the Customs Office, with information on the type of trade (processing exports using imported intermediate inputs, using locally sourced inputs, normal exports and imported intermediate inputs for the purpose of exports), trading partner countries, the type of trading firms (whether MNEs, pure local firms or international joint ventures), the location of exporters and importers in the regions and cities, the values in US dollars and the quantity in eight different units. Several studies in Feenstra and Wei (2010) also use CCS trade data.

7. These countries are Germany, Japan, Malaysia, the Philippines, Singapore, South Korea, Taiwan, Thailand, and the United States.

8. In the absence of instantaneous purchasing power parity condition holding across all countries, there is no theoretical reason for these to behave in an identical manner. The empirical behaviour of bilateral RERs shows wide fluctuations.

9. To illustrate, consider the case of a country whose import source countries and export destination countries have a 50% share in total trade. Assume that the change in these shares (due to trade volume changes resulting from exchange rate changes) can be ignored, and that there is zero inflation everywhere so that nominal exchange rate changes correspond to RER changes. In this case, a 1% appreciation of the effective (trade-weighted) RER can be the result of a 2% appreciation of the exchange rate relative to import source countries or a 2% appreciation relative to the RER of the export destination countries. Obviously, these would have opposite impacts on export supply. However, the estimation procedure will yield quite different elasticity estimates in the two cases, though the underlying structural value of the export supply parameter had been unchanged.

10. Note that we use the term ‘import source countries’ to denote countries that are suppliers of parts and components (intermediates) for export production, rather than source countries for all imports.

11. Tokarick's (2010) approach to calculating trade elasticities ignores these exchange rate issues by assuming, in effect, a two-country model where the rest of the world is a homogeneous composite of all foreign countries.

12. The share of road vehicles (SITC 78) has remained at around 1% from 1992/93 to 2004/05.

13. During the recent crisis period in 2008/09, ASEAN's share in China's component imports declined to 8%. Similarly, the shares of South Korea and Taiwan also dropped significantly. However, in contrast to component imports, the share of ASEAN in China's final good imports actually went up from 12% in 2005/06 to 17.5% in 2008/09, while the shares of Japan, the United States and the EU15 all went down. The share of Japan in China's final product imports declined from 20% in 2000/01 to 16% in 2008/09.

14. This section is based on Yamashita (2011).

15. Note that ‘bilateral’ in this context refers to the real effective exchange rate between China and the relevant group of countries.

16. These are the same countries chosen by Thorbecke (2011). Australia is not included in this set of countries because, though it is a major source of imported resources, it plays no significant role as a parts and component supplier.

17. It is noted that component exports out of these countries may be by third-country firms, including firms with home bases in the export destination countries. This complex web of trade and investment linkages emphasises the need to go beyond simple measures of gross trade in measurement of bilateral trade and current account imbalances. Kierzkowski and Chen (2010), for example, show that the true China–US trade imbalance falls dramatically when trade in parts and components produced by US firms and affiliates (including those located in third countries) is taken into account.

18. Some countries, such as Japan, are both significant component suppliers and destination markets, but the basic results are robust to changes in the composition of country groupings taking account of such ‘mixed’ cases.

19. The use of CPI in this context is driven by data availability; other measures of price inflation were not available for all countries during the study period.

20. This loss in statistical significance of BER is probably due to the high correlation between BER and RER (see Table 4B).

21. These elasticity estimates are not strictly comparable to elasticity estimates in previous literature because our exchange rate indices, with respect to which elasticities are defined, are quite different; we treat movements of exchange rates relative to supplier countries and destination countries separately, whereas the elasticity measures in previous literature are defined relative to an aggregate index of exchange rates of both groups.