Are Chinese Exports Sensitive to Changes in the Exchange Rate?

Board of Governors of the Federal Reserve System International Finance Discussion Papers Number 987 December 2009 Are Chinese Exports Sensitive to Changes in the Exchange Rate? Shaghil Ahmed NOTE: International Finance Discussion Papers are preliminary materials circulated to stimulate discussion and critical comment. Reference to International Finance Discussion Papers (other than an acknowledgement that the writer has access to unpublished material) should be cleared with the author or authors. Recent IFDPs are available on the Web at www.federalreserve.gov/pubs/ifdp/. This paper can be downloaded without charge from the Social Science Network electronic library at www.ssrn.com.

Are Chinese Exports Sensitive to Changes in the Exchange Rate? Shaghil Ahmed (cid:3) December, 2009 Abstract This paper builds a model of two types of Chinese exports, those processed and assembled laregely from imported inputs ("processed" exports) and "nonprocessed" exports. Based on this model, the sensitivity of Chinese exports to exchange rate changes is empirically examined. Unlike previous work, the estimation period includes the net real appreciation of the renminbi that has occurredoverthepastthreeyears. Theresultsshowthatgreaterexchangerate appreciationdampensexportgrowth,bothfornon-processedandprocessedexports,withtheestimatedcumulativepriceelasticitybeingsubstantiallygreater thanunity. WhenthesourceoftheincreaseintheChineserealexchangerateis appreciations against the currencies of other emerging Asian trading partners, the e⁄ect on processing exports is positive but insign(cid:133)cant, while the e⁄ect on non-processing exports is signi(cid:133)cantly negative. By contrast, when the source oftheincreaseintheChineserealexchangerateisappreciationagainstChina(cid:146)s advanced-economy trading partners, the e⁄ects on both types of exports are negative. These results are consistent with the predictions of the theoretical model. Counterfactualsimulationsbasedontheestimatedmodelstronglysuggestthatifthetrade-weightedrealrenminbihadappreciatedatanannualrate of10percentperquartersincemid-2005,Chineserealexportswouldhavebeen roughly 30 percent lower today. Thus greater exchange rate (cid:135)exibility could contribute to lowering China(cid:146)s huge trade surplus through restraining growth of exports. Keywords:China,exchangerate,exports. JELclassi(cid:133)cation:F31,F32,F41. (cid:3)The author is Chief, Emerging Market Economies Section, Division of International Finance, Board of Governors of the Federal Reserve System, Washington, D.C. 20551, U.S.A. The views expressed in this paper are solely the responsibility of the author and should not be interpreted as re(cid:135)ecting the views of the Board of Governors of the Federal Reserves System or of any other personassociatedwiththeFederalReserveSystem. Anearlierversionofthispaperwaspresented at the 2009 Canadian Economic Association annual meetings in Toronto, Canada. I would like to thank Kavan Kucko and Nikola Kojucharov for excellent research assistance and Jeannine Baillu, Brett Berger, Carlos De Resende, Neil Ericsson, John Rogers, Jaime Marquez, and John Schindler for very helpful comments. 1

1 Introduction China(cid:146)s ballooning current account surplus in recent years (reaching about 10 percent of GDP in 2008) and rapid accumulation of international reserves (to about $2.2 trillion) has raised concerns that Chinese authorities are heavily managing their currency and contributing to global imbalances. At the same time, many also question whether faster currency appreciation would reduce China(cid:146)s trade surplus signi(cid:133)cantly(cid:150)one argument being that, given the high import content of Chinese exports, appreciation of the currency need not make Chinese exports more expensive totherestoftheworldsincethee⁄ectivecostoftheimportedinputswouldalsofall. Despite this tension there is relatively little empirical evidence on how responsive Chinese exports have, in fact, been to currency movements that cover the period since the middle of 2005 when China revalued the renminbi (RMB) and started allowing a moderate appreciation trend, at least until the middle of last year. This paper provides empirical estimates of the sensitivity of Chinese exports to exchange rate changes. It distinguishes between the e⁄ects on "processed" exports (produced using parts and components imported from abroad) and "non-processed" exports (largely sourced from domestic inputs). It also attempts to distinguish between unilateral changes in the Chinese exchange rate and those that are highly correlated with exchange rate changes of other economies in the region from which China imports parts and components, since this distinction is potentially very important when both processed and non-processed exports are being produced. Therearesomeexistingempiricalstudiesthatalsodistinguishbetweenprocessed and non-processed Chinese exports(cid:150)Aziz and Li (2007), Cheung, Chinn and Fujii (2008), Garcia-Herero and Koivu (2009), Marquez and Schindler (2007), and Thorbecke and Smith (2008)(cid:150)and Thorbecke and Smith also consider unilateral versus multilateral(acrossAsia)reale⁄ectiveexchangeratechanges. However,onlytwoof these studies incorporate any part of the period since mid-2005 in their sample period, and none of them consider the period from 2007 to mid-2008, when the pace of appreciation of the RMB apparently was accelerated. All told, the trade-weighted Chinese real exchange rate has appreciated about 13 percent, on net, since the end of 2006. Taking account of the greater recent variability of the exchange rate, this study provides up-to-date estimates and compares these to earlier estimates. Given concerns about possible currency undervaluation it also uses simulations from the empirical model to examine what the behavior of Chinese exports might have been if the RMB had appreciated more in recent years. Another key distinguishing characteristic of this paper is that it develops a theoretical model of Chinese exports that explicitly incorporates the import of inputs for the production of some types of exports goods. This means that the estimated 2

equations for exports are well-grounded in economic theory, including predictions aboutthedi⁄erente⁄ectsofRMBappreciationwhenitssourceismovementsagainst the currencies of other emerging Asian economies and when its source is movements against the currencies of China(cid:146)s other trading partners. The explicit derivation of reduced-form export equations from theory also makes it clear that the estimated relative price elasticity should not be viewed, as it often is, as the slope of the demand curve, and the income elasticity should not be viewed as representing how much the demand curve shifts in response to a change in income, as the equilibrium quantities will incorporate supply-side parameters as well. The main results of the paper can be summarized as follows: First, including the latest period of greater real exchange rate variability reinforces the conclusions of some earlier studies, such as Marquez and Schindler (2007), which found that Chinese exports respond quite strongly to movements in the real exchange rate, and go against studies which (cid:133)nd little e⁄ect of exchange rate changes or e⁄ects that go in the opposite direction to conventional wisdom. Second, considering the components of the real exchange rate, consistent with the theoretical model, when the source of Chinese real exchange rate appreciation is movements of the RMB against other emerging Asian countries, this does not have a signi(cid:133)cant e⁄ect on Chinese processing exports, but it does have a signi(cid:133)cant negative e⁄ect on Chinese non-processing exports. On the other hand, when the source of the renminbi appreciation is movements against the currencies of non-emerging Asian Chinese trading partners, generally both types of exports go down. Moreover, even though processed exports remain very important for China, increases in non-processed exports have recently accounted for more of the overall increase in exports. Finally, model simulations indicate that the path of total Chinese real exports would have been quite a bit lower if the renminbi had appreciated more in recent years. Overall, the results suggest that greater exchange rate (cid:135)exibility could have signi(cid:133)cant impact on China(cid:146)s trade balance by restraining growth of exports, particularly non-processed exports. The remainder of the paper is organized as follows. Section 2 sets the scene by discussing key developments in the Chinese external sector in recent years and Section 3 provides a selective review of the existing empirical work in this area. Section 4 presents a simple theoretical model of the behavior of Chinese exports that forms the basis of the empirical speci(cid:133)cation used. The empirical results on the exchange rate sensitivity of Chinese exports are presented and discussed in Section 5. Section 6 concludes. 3

2 Background: Developments in the Chinese External Sector When discussing global imbalances, one of the key developments often referred to is the phenomenal rise in China(cid:146)s external surpluses since China(cid:146)s entry into the WTO in December 2001. As can be seen from Figure 1, China(cid:146)s current account surplus increased from a relatively modest under 2 percent of GDP ($17 billion) in 2001 to 11 percent of GDP in 2007, before falling a bit to a still-high of just under 10 percent of GDP last year ($430 billion). The ballooning surplus in recent years has raised concerns among some international analysts that China has been following a mercantilist approach and keeping its currency arti(cid:133)cially (and some argue unfairly) undervalued to pursue an export-led growth strategy. In the (cid:133)rst half of 2009, though, the Chinese current account surplus narrowed sharply, to 6.5 percent of GDP, as the global crisis led to Chinese exports falling much more than Chinese imports. (Chinese balance of payments (BOP) data are reported only semi-annually.) Of course, a large trade surplus, just by itself does not necessarily imply intervention in exchange markets. But accompanying the phenomenal rise in the Chinese current account surplus in recent years has been an even more phenomenal rise in China(cid:146)s international reserves, from about $200 billion in 2001 to $2 trillion at the end of 2008(cid:150)this year, reserves have risen further to about $2.2 trillion (see Figure 2), despite the narrowing of the current account surplus. There are many good reasons for emerging market economies (EMEs) to build up a war chest of reserves for insurance purposes against crisis situations, and the relatively high level of international reserves compared to past crises helped the EMEs cope better during the recent global crisis. Nevertheless, the sheer magnitude of China(cid:146)s reserve accumulation has led to questions about the possibility of an undervalued exchange rate. How has the exchange rate itself behaved? This is shown in Figure 3. As can be seen from the thick black line, Chinese authorities maintained a dollar peg until the middle of 2005. After a one-time appreciation of about 2 percent of the RMB against the dollar, Chinese authorities allowed further gradual appreciation until about the middle of last year. Since that time, the RMB appears again to have been de facto pegged against the dollar. On net, the nominal value of the RMB against the dollar has declined roughly 25 percent since mid-2005. The trade-weighted real and nominal e⁄ective exchange rates are shown by the blue dashed and red dotted lines, respectively, in Figure 3. Note that the tradeweightede⁄ectiveexchangeratesvarymorethanthebilateralrateagainstthedollar. EvenovertheperiodwhenthevalueoftheRMBwaspegged,ine⁄ectiveterms(both 4

real and nominal), the exchange rate was varying. At (cid:133)rst, it followed a generally appreciating trend until the end of 2001 and then a generally depreciating one until mid-2005. After the peg was dismantled, there was a generally appreciating trend of the e⁄ective exchange rates until about the end of last year, especially toward the end of that period with the RMB/Dollar rate constant and the dollar appreciating sharply against other major currencies. This year, the real e⁄ective exchange rate has moved down, re(cid:135)ecting dollar weakness against other major currencies and a continued(cid:133)xedRMBagainstthedollar. ThecumulativeappreciationoftheChinese real exchange rate since June 2005 has been about 13 percent. Two other scene-setting developments are useful to note. The (cid:133)rst relates to speculative activity based upon anticipations of the future behavior of the RMB. Figure 4 gives the sources of the change in international reserves, decomposing these changes into what can be accounted for by the trade balance, net foreign direct investment (FDI) in(cid:135)ows, and the remainder(cid:150)what people have often called "hot money" in(cid:135)ows (measured as the residual).1 Note that in 2004, speculation was rife about the abandoning of the peg against the dollar and only less than half of the change in international reserves could be accounted for by the trade surplus (the green bar) and net FDI (the blue bar). Hot money in(cid:135)ows (the red bar) were thus large, given expectations of appreciation of the RMB, which did become partly ful(cid:133)lled in mid-2005. Hot money net (cid:135)ows since then appear to have been more modest, with a small net out(cid:135)ow last year and a somewhat bigger one in the (cid:133)rst half of this year. In recent months, however, although we cannot provide an actual estimate, since Chinese BOP data are only semi-annual, the sharp narrowing of the trade balance together with the acceleration in the pace of reserves accumulation suggests that hot money in(cid:135)ows have picked up again. The second important development is the di⁄erent behavior of processing and non-processing trade. Much of China(cid:146)s trade involves importing inputs and parts and components with very little import duties (imports for processing) and adding value through processing and assembly of these parts and components into goods that are then re-exported (processed exports). The processed exports have a much higher import content than non-processed exports.2 Chinese customs records keep a distinction between "processing" and non-processing trade, labelling the latter as "ordinary" imports and exports. Theseasonally-adjusted(usingX-12)quarterlynominalvaluesofprocessingand non-processing exports and imports are shown in Figures 5 and 6. Each type of export constitutes about half of total exports. Several features are noteworthy. 1MaandSun(2007)alsoprovidesomeestimatesofhotmoney(cid:135)owsintoChinaandtheypresent a model for the behavior of these (cid:135)ows. 2To measure the exact degree of import content in exports is not a trivial matter. Dean et al (2007)dothisforthecaseofChinathroughmeasuringtheverticalspecializationinChinesetrade. 5

First, processing goods exports and imports grow and fall together. Second, over the better part of the last decade, including this year, China ran a de(cid:133)cit on nonprocessing trade, which implies that the surplus on processing trade was larger than China(cid:146)s overall trade surplus. Third, before the global crisis hit, non-processing exports were rising at a higher pace than processing exports and contributing more to export growth. More recently, non-processing imports have picked up again after the collapse of trade during the crisis, but non-processing exports have not yet shown a similar pickup, which has been a big source of the recent narrowing of China(cid:146)soveralltradesurplus. Thedistinctionbetweenprocessedandnon-processed exports(cid:150)and their di⁄erent behaviors(cid:150)is important for the question of looking at the e⁄ects of exchange rate changes, because with a high import content, processed exports may be less responsive to changes in the Chinese exchange rate against the currencies of those countries it imports inputs and parts and components from. 3 Review of Previous Empirical Work The extent of the possible undervaluation of the Chinese exchange rate has been the subject of intense debate in recent years. Cline and Williamson (2008) provide a survey of the range of estimates. Leaving aside a few outliers, the typical range of the degree of undervaluation of the Chinese real e⁄ective exchange rate is 8 percent to55percent. Theaverageundervaluationforstudiesdated2005orafterwardsis26 percent.3 The estimates di⁄er because of the approach taken as well as di⁄erences in assumptions that are made within a given approach. Despite these estimates, some continue to argue, along the lines of Mundell (2004), that China should not be pressured to appreciate its currency substantially, as this, among other things, would undermine its growth miracle. The approaches used to estimate the extent of the undervaluation are also discussedbyClineandWilliamsonaswellasbyDunaway,Leigh,andLi(2009). Oneof thekeyapproachesisthemacroeconomicbalanceapproach, whichtypicallyinvolves (cid:133)rst computing the long-run sustainable equilibrium current account balance and thenthechangeintherealexchangeratethatwouldberequiredtoclosethegapbetween the actual current account balance and this equilibrium value.45 Obviously, 3The costs that might be incurred by the Chinese government in allowing greater (cid:135)exibility of the exchange rate is also a topic of much discussion. Some papers have recently argued that the reasons most often given for why greater exchange rate adjustment would have high costs for China are (cid:135)awed (see, for example, Goldstein (2007), and Goldstein and Lardy (2008)). Some authors have advocated that greater (cid:135)exibility of the RMB would actually be bene(cid:133)cialto China(cid:146)s own economic interests, including allowing appropriate rebalancing of growth and pursuance of a moreindependentmonetarypolciy. ExamplesincludeBernanke(2006),Eichengreen(2007),Lardy (2006), Prasad (2008), and a survey by Lafrance (2008). 4Eamples of the studies which include the use of this approach are Anderson (2006), Cline (2007), Coudart and Couharde (2005), and Goldstein and Lardy (2008). 5The other main approach to computing the extent of undervaluation is to use the PPP or the 6

in computing the change in the real exchange rate required using this approach, assumptions will have to be made about the relative price elasticities of exports and imports. Studies using this approach for China do not always carefully justify the choice of trade elasticities or explicitly lay out the trade models (theoretical or empirical) that are the sources of these choices. It will, therefore, be useful to review what the existing empirical literature says about the sensitivity of Chinese trade to changes in the exchange rate. Since the focus of this paper is on exports, attention is con(cid:133)ned here to estimates of export elasticities. Studies of export e⁄ects of exchange rate changes using earlier data, such as Cerra and Dayal-Gulati (1999), Cerra and Saxena (2003), and Eckaus (2004) do not (cid:133)nd consistent or stable estimates. As Marquez and Schindler (2007) have argued, the estimates of these earlier studies cannot be relied upon, partly because they include a period of transformation from a centrally-planned economy to a marketorientedsystemandbecausetheyincludeperiodsinwhichtherewaslittlemovement in the nominal exchange rate and little movement in the real e⁄ective exchange rate as well. Another important criticism that Marquez and Schindler level against much of the work in the area of estimating trade elasticities for China is that since Chinese tradepricesarenotavailable,studiesuseimperfectproxiesfortheseprices,including using trade price data from China(cid:146)s trading partners, especially Hong Kong. To avoid distorting the results, these authors conduct their own very comprehensive analysis that is based on studying nominal shares of Chinese trade in world trade, modelingthesesharesasdependingoneconomicactivityandtherealexchangerate. Focusing on these nominal shares does not require them to take a stand on which proxy to use for trade prices. Their results are widely-cited. Using monthly data from 1997-2004, they (cid:133)nd thatfornon-processedexports,a10percentappreciationoftheRMBwoulddecrease the world share of Chinese exports by about a half percentage point in the long run, which is a fairly big e⁄ect. The results for processed exports are less clear-cut and sensitive to the lag length used in the estimated equations. ThepointmadebyMarquezandSchindleraboutproxiesforChinesetradeprices extended PPP approach. In the latter, the real exchange rate is related to a small set of variables such as per capita income ratios (the Balassa-Samuelson e⁄ect), and net foreign assset ratios and perhapsothervariables. Thepredictedin(cid:135)uenceofthesefactorsgivesanequilibriumexchangerate which is then compared to the actual to get the misalignment. The equation could be estimated usingdataforasinglecountryorbasedoncross-sectionorpanelestimation. Studiespertainingto ChinausingthisapproachincludeCheungetal(2009),CoudertandCouharde(2005),andFrankel (2006). Cline and Williamson label only studies which incorporate the Balassa-Samuelson e⁄ect as the extended PPP approah. They prefer to label the approach with more variables being used to compute the equilibrium real e⁄ective exchange rate as the behavioural equilibrium exchange rate (BEER) approach. 7

being imperfect is well-taken. However, as they acknowledge, a drawback of their procedure is that the price responsiveness of trade volumes is not identi(cid:133)ed. But someideaofthispriceresponsivenessisveryimportantinoneofthekeyapproaches toobtainingthedegreeofRMBmisalignmentandinformingthedebateinthisarea. Over time, the criticism raised by Marquez and Schindler may have become less important as the U.S. has maintained, since 2003, a Chinese import de(cid:135)ator, which may actually be a reasonably good proxy for the overall price de(cid:135)ator for Chinese exports. Recent studies that use proxies for Chinese trade prices include Aziz and Li (2007) and Cheung et al (2009). Aziz and Li, using quarterly data from 1995-2006 (cid:133)nd an aggregate export price elasticity with respect to RMB real appreciation of about 11, and disaggregated elasticities of about 21 for non-processed exports (cid:0) 2 (cid:0) 4 andabout 1 forprocessedexports.6 Theseelasticitiesarestatisticallysigni(cid:133)cant.7 (cid:0)2 Using rolling regressions, they also (cid:133)nd that, while the elasticity for non-processed exports has stayed relatively constant, the elasticity for processed exports (cid:133)rst decreased and then (in samples that include the period since mid-2005) increased.8 For their latest sub-sample of 1999-2006, the price elasticity of processing exports is about -1.4. Cheung et al use a similar empirical speci(cid:133)cation based on a sample period that generally uses quarterly data over the period 1993:3-2006:2. However, they obtain quite di⁄erent results. Their speci(cid:133)cation for exports includes a foreign activityvariable,arealexchangeratevariable,andasupply-shiftvariable,measured as the capital stock in manufacturing. They (cid:133)nd that, although real exchange rate appreciation lowers exports as expected a priori, the e⁄ect is not statistically signi(cid:133)cant. Theincomee⁄ectsarealsonotgenerallysigni(cid:133)cant,althoughthecapital stock always has a signi(cid:133)cantly positive e⁄ect on exports. They also consider a speci(cid:133)cation that excludes the capital stock, but this results in estimates that are very counterintuitive(cid:150)exchange rate appreciations in this case have a signi(cid:133)cantly positivee⁄ectonbothnon-processingandprocessingexports. Theauthorsconclude that Chinese export price elasticities are not very precisely determined.9 6They compute a Chinese export price index by using a weighted average of U.S. import prices at the SITC 2-digit level, using weights that are proportional to China(cid:146)s share of U.S. imports in each category. 7Their speci(cid:133)cation includes Chinese productivity as an indendent variable in the export equations. This makes it a bit di¢ cult to interpret the estimated elasticity with respect to the real exchange rate, as productivity growth would normally a⁄ect the real exchange rate. However, the authors argue that, in the case of China, the Balassa-Samuelson e⁄ect "breaks down" due to the country(cid:146)s "large excess labor." 8Theytracethechangesovertimeinthepriceelasticityofexportstochangesinthecomposition of trade and variation of sector-speci(cid:133)c elasticities. They also partly attribute the changes to a rising domestic content even in processing trade. 9Cheungetaluseseveraldi⁄erentproxiesfortheChineseexportpricede(cid:135)ator;theirresultsare fairly robust to which proxy is used. 8

Another very recent study is Garcia-Herrero and Koivu (2009). However, the sample period they used in their research only goes up to the end of 2005. The e⁄ects of relative price increases are in the expected direction. For non-processing exports, they (cid:133)nd a long-run relative price elasticity of -2.3 for the full sample period of 1994-2005, but a lower elasticity of -1.6 for the more recent sub-sample 2000-2005.10 For processed exports, the long-run price elasticity is around -1.3 for both sample periods.11 AcarefullydonestudythatemphasizesthecentralimportanceofunilateralRMB changes versus multilateral ones (that are accompanied by similar movements in the exchange rates of other Asian economies against the currencies of their Western trading partners) is Thorbecke and Smith (2008).12 They use an annual panel data set over the period 1992-2005 to model Chinese bilateral real exports (both processing and non-processing) to 33 countries.13 The standard foreign output and bilateral real exchange (rer) rate variables are included in the speci(cid:133)cation for nonprocessing exports. For processing exports, account is taken of how the bilateral real exchange rates of a trading partner with other countries that China imports inputs from would a⁄ect Chinese imported input prices. Speci(cid:133)cally, the relative price variable in the bilateral processing exports equation for trading partner i is an integrated real exchange rate (irer), which is a weighted average of the rer and the average real exchange rate of the other countries in Asia (from which China imports the bulk of parts and components) with trading partner i (wrer). The weight attached to rer in the computation of irer is the share of valued added by China in total processing exports and the weight of wrer is 1 minus that. This is a very interesting approach that also yields interesting empirical results. The authors results imply that if only the RMB appreciated, say a 10 percent rise in rer with wrer unchanged, processing exports would fall by 4 percent only (a price elasticity of -0.4). If, however, all of the currencies in emerging Asia appreciated against other currencies along with the RMB (a 10 percent rise in both rer and wrer), then processing exports fall by much more (10 percent, or an elasticity of -1). Intuitively, this holds because when wrer is unchanged, this means the RMB is appreciating against its other emerging Asian trading partners also, which means imported inputs become cheaper, partially o⁄setting the negative e⁄ect on 10These authors use the Chinese CPI as a proxy for Chinese export prices. 11Garcia-HerreroandKoivustudyalsoestimateimportpriceelasticities. They(cid:133)ndthatexchange rate appreciation actually decreases rather than increases imports. Thus, real exchange rate adjustment would not help the trade balance as much as their export elasticity results suggest. 12A subsequent paper, Thorbecke and Zhang (2009) focuses on disaggregated Chinese laborintensivemanufacturingexports,usingthesameapproach. WhenlookingatbilateralChinese-U.S. trade, Cheung et al (2009) also included a real e⁄ective exchange rate relative to China(cid:146)s other trading partners, in addition to the U.S.-China bilateral real exchange rate. 13Several di⁄erent de(cid:135)ators are used as proxies fot the Chinese export price de(cid:135)ator. 9

processing exports.14 Since processing exports, being more sophisticated, are the ones that European and U.S. goods could be potential substitutes for, the authors argue that only general exchange rate adjustment throughout emerging Asia would do something for global imbalance adjustment. While this approach is interesting and informative, it does not address some issues. First it implicitly assumes that the share of value-added in total processing exports is exogenously given and not itself a⁄ected by the relative prices. Second, the speci(cid:133)cations using bilateral trade and bilateral real exchange rates do not allow for third-party competition e⁄ects so that the equations might be misspeci(cid:133)ed. Third, from the viewpoint of the demanders of the (cid:133)nal processed export goods, e.g. U.S. consumers, only the relative price of the (cid:133)nal good should matter; the distinction between what is happening to the RMB versus other Asian currencies is relevant for the supply-side of processing exports only, and should focus directly on RMB movements against these currencies rather than indirect movements of these currencies against particular trading partners, such as the U.S. The approach taken inthispaperisdi⁄erent, butcomplementary, andisgroundedinatheoreticalmodel that derives the demand and supply functions of processing and non-processing exports explicitly. As mentioned before, none of the empirical studies discussed above includes the period since the end of 2006(cid:150)a period over which interesting and new developments have taken place in the Chinese external sector(cid:150)as part of the sample period. In particular, the greater variability in the real exchange rate observed over this period should give more precise estimates.15 4 Modeling Chinese Exports: Theory In this section, the demand and supply of Chinese exports are derived theoretically and market-clearing conditions used to obtain equations for the equilibrium growth rates of Chinese exports that form the basis of the empirical work. The estimated equations will allow more dynamics than is embedded in the simple theoretical model. 14In an earlier paper, Rahman and Thorbecke (2007), rer and wrer were used separately in the bilateral proceesing exports equations, rather than an integrated irer, and di⁄erent statistica techniques were also used. That resulted in estimates that were in many ways counterintuitive. 15Before leaving this section reviewing previous empirical work, we should note that another completelydi⁄erentapproachtakentodeterminehowmuchChinesetradewouldadjustforagiven size change in the RMB is to do simulation analysis from a calibrated general-equilibrium model. Wedonotdiscussthatliteraturehere,althoughanexampleofthisapproachisWangandWhalley (2007). 10

4.1 Demand for Chinese Exports Importers of Chinese goods from the rest of the world are assumed to consume three types of goods: a Chinese good that is produced largely from inputs and components that are imported into China and then assembled into (cid:133)nal products in China for export (M )(cid:150)what we have been calling processed exports; a Chinese A export good that relies more heavily on inputs produced in China, i.e. domestically sourced (M )(cid:150)what we have been calling non-processed exports; and an aggregate D of all other goods consumed by the rest of the world, including goods they produce domestically and import from other countries (C ). The preferences of the rest O of the world consumers for these three types of goods are given by a CES utility function: (cid:27) u(M A ;M D ;C O ) = (cid:30) 1 A =(cid:27) M A (cid:27) (cid:0)(cid:27) 1 + (cid:30) 1 D =(cid:27) M D (cid:27) (cid:0)(cid:27) 1 +(1 (cid:0) (cid:30) A (cid:0) (cid:30) D )(cid:27) 1 C O (cid:27) (cid:0)(cid:27) 1 1 (cid:0) (cid:27) (1) (cid:20) (cid:21) where(cid:27) istheelasticityofsubstitution, andthe(cid:30)(cid:146)sarepreferenceparameters. The division of aggregate consumption expenditure by the rest of the world on the three goods they consume can be written as: P M +P M +P C = P C (2) A(cid:3) A D(cid:3) D O(cid:3) O C(cid:3) (cid:3) where the P (cid:146)s represent foreign prices and P is the aggregate foreign consumer (cid:3) C(cid:3) price (CPI) and C is aggregate real foreign consumption. The (cid:133)rst order neces- (cid:3) sary conditions of maximizing (1) subject to (2) give rise to the following standard demand functions: P (cid:27) M = (cid:30) A(cid:3) (cid:0) C (3) A A P (cid:3) (cid:18) C(cid:3)(cid:19) P (cid:27) M = (cid:30) D(cid:3) (cid:0) C (4) D D P (cid:3) (cid:18) C(cid:3)(cid:19) P (cid:27) C = (1 (cid:30) (cid:30) ) O(cid:3) (cid:0) C (5) O (cid:0) A (cid:0) D P (cid:3) (cid:18) C(cid:3)(cid:19) where the aggregate foreign CPI is de(cid:133)ned as: 1 P C(cid:3) = (cid:30) A P A(cid:3) 1 (cid:0) (cid:27) + (cid:30) A P D(cid:3) 1 (cid:0) (cid:27) +(1 (cid:30) A (cid:30) D )P O(cid:3) 1 (cid:0) (cid:27) 1 (cid:0) (cid:27) (6) (cid:0) (cid:0) h i Looking a bit ahead, the real exchange rate is going to be assumed to be a policy variable that Chinese authorities will target. Accordingly, it will be useful to rewrite the demand functions (3) and (4) in terms of the real trade-weighted CPI-based Chinese real exchange rate and the local currency prices of the Chinese 11

export goods. De(cid:133)ne the real exchange rate to be: EP Q = C = (Q )$(Q )1 $ = Q Q (7) P 0A 0B (cid:0) A B C(cid:3) where E is the nominal trade-weighted Chinese exchange rate, expressed as foreign currency per unit of the RMB, and P is the Chinese aggregate CPI. Note that C a rise in Q represents a real exchange rate appreciation for China. The aggregate real exchange rate index, Q, is decomposed into components attributable to China(cid:146)s trading partners in the rest of emerging Asia, Q , and China(cid:146)s other trading part- A ners, Q , with Q ; Q being the component real exchange rates and $ and 1 $ B 0A 0B (cid:0) being the weights attached to the two sets of countries, respectively, in the aggregate real exchange rate index. This distinction will be useful when we consider the supply of Chinese exports. Equation (7) can be used to rewrite (3) and (4) as: M = (cid:30) Q (cid:27) P A (cid:0) (cid:27) C (8) A A (cid:0) P (cid:3) C (cid:18) (cid:19) M = (cid:30) Q (cid:27) P D (cid:0) (cid:27) C (9) D D (cid:0) P (cid:3) C (cid:18) (cid:19) where P ;P represent the domestic currency (RMB) prices of the goods, and A D purchasing power parity for traded goods is being assumed. In the empirical work, we will use growth rates of the variables to ensure stationarity. Equations (8) and (9) in growth-rate form can be approximated by taking logs and (cid:133)rst-di⁄erencing to yield: (cid:1)m = (cid:27)[(cid:1)q+(cid:1)(p p )]+(cid:1)c (10) A A C (cid:3) (cid:0) (cid:0) (cid:1)m = (cid:27)[(cid:1)q+(cid:1)(p p )]+(cid:1)c (11) D D C (cid:3) (cid:0) (cid:0) where lower case letters represent the natural logarithm of a variable, and (cid:1) is the (cid:133)rst di⁄erence operator. Note that the elasticity with respect to foreign consumptionisunitybecauseofthechoiceofCESutilityfunction. However,intheempirical work we will be estimating reduced forms that allow for this elasticity to di⁄er from unity if the data so dictate. 4.2 Supply of Chinese Exports ThreegoodsareproducedinChina:anon-tradedaggregategoodandthetwoexport goods whose demand was discussed above. The aggregate non-traded good is assumed to be the only consumption good in the economy. The supply of the three 12

goods is subject to Cobb-Douglas technology: X = A L(cid:11)CK1 (cid:11)C exp (cid:17) (12) C C C C(cid:0) f C g X = A L(cid:11)DK1 (cid:11)D exp (cid:17) (13) D D D D(cid:0) f D g X A = A A L(cid:11) A AZ(cid:11)ZK A 1 (cid:0) (cid:11)A (cid:0) (cid:11)Z exp f (cid:17) A g (14) where X is the supply of the Chinese non-traded (consumption) good, X is the C D supply of the Chinese export good using only domestic inputs, X is the supply of A the Chinese "assembled" export good, the A(cid:146)s represent the state of technology, the L(cid:146)s and the K(cid:146)s represent labor and capital, respectively, allocated to the various sectors, Z represents the imports of inputs, parts, and components from abroad, and the (cid:17)(cid:146)s represent the technological innovations to the di⁄erent sectors.16 Firms are assumed to choose their demand for labor and their demand for the imported input based on pro(cid:133)t maximization. However, in the case of China, it is highly debatable to what extent the quantity of capital and its allocation across sectors is determined by pro(cid:133)t-maximizing behavior in response to changes in interest rates and their e⁄ects on the user cost of capital. The government has quite a substantial degree of in(cid:135)uence through directed lending and informal assigned lending quotas (the so-called "window guidance") as well as through state approval of investment projects. In light of this, we assume the stock of capital and its allocation across sectors to be determined outside the model. Pro(cid:133)t maximization yields the following standard labor demand functions and theimportedinputdemandfunctionthatfollowfromtheCobb-Douglastechnology: X W i (cid:11) = (15) i L P i i for i = C;D;A; respectively. X P A Z (cid:11) = (16) Z Zd P A where W is the economy-wide wage rate, Zd is the demand for the imported input, P is the local currency (RMB) price of the imported input, and L ;L ;L are Z C D A the quantities of labor demand in the three sectors. For simplicity, labor is assumed to be elastically supplied at a (cid:133)xed economy-wide real wage rate, measured 16Puttingtheimportedinputintotheproductionfunctionhasalsobeentheapproachtakenina few otheroptimizingmodelsoftheopeneconomy,suchasMcCallum andNelson(1999)andErceg et al (2008), although our Cobb-Douglas functional form is simpler. 13

in consumption units.17 Speci(cid:133)cally, W = (cid:22)P (17) C Other countries in emerging Asia are assumed to supply the input Z to China. The supply of this input is assumed to be price elastic and given by: P (cid:13) Q P (cid:13) Zs = Z0 B = 0A Z B (18) P P (cid:18) C0 (cid:19) (cid:18) C (cid:19) where B is the scale factor for the input supply, P is the foreign (Asian) currency z0 price of the input exported by rest of emerging Asia to China, P is the domestic C0 price of the aggregate consumption good in the input-exporting countries, and purchasing power parity for traded goods is assumed to hold, so that E P = P where 0 Z Z0 E is the nominal exchange rate expressed as units of foreign Asian currency per 0 RMB. Recall that Q is the CPI-based real exchange rate of China vis-a-vis other 0A emerging Asian countries, which are the countries China is assumed to imports its inputs and parts and components from:18 E P 0 c Q = (19) 0A P C0 ForagivenChinesecurrencypriceoftheinputP , ariseinQ meansforeigners Z 0A supplying the input would receive more for it, increasing the supply of it, as can be seen from (18). An alternative equivalent way to characterize the situation is that for a given foreign currency price of the input P , a rise in Q means less Z0 0A would have to be paid for a given amount of the input by the Chinese in RMB, thus increasing the demand for the input. Viewed either way, whether we think of the demand and supply curves for the imported input being drawn in P or P Z0 space, the equilibrium quantity of the imported input would rise following a real RMB appreciation against the other emerging Asian currencies. To derive, in growth rate terms, the supply functions of the two goods that China exports, the growth rate of the equilibrium quantity of the imported input for any given level of production of the export good (X ) that uses this input is (cid:133)rst A 17In the past, this assumption has been defended by appealing to China(cid:146)s limitless supply of laborthatcaneasilybemovedfromruraltourbanareas. However,althoughthismightbetrueof non-skilledworkers,thereisanecdotalevidenceatleast,ofasubstantialshortageofskilledoreven semi-skilledworkers. Therefore,theassumptionofanelasticlaborsupplyshouldberegardedasa simplying one, rather than a realistic one. The qualitative results derived here should not change with the introduction of an upward sloping supply curve for labor, although the exact solutions for the equilibrium quantities of labor allocated to each sector and the equilibrium quantity of the imported input would di⁄er. 18In practice, China imports parts and components from outside the region as well, but most of them come from other emerging Asian countries as documented, for example, in Haltmaier et al (2009). 14

derived. This is done by taking logs of equations (16) and (18), (cid:133)rst di⁄erencing and then equating growth of the demand for the input to the supply of the input to solve for the price level change (cid:1)p , which is then substituted back into either the Z growth of demand or the growth of supply of the input to yield: 1+(cid:13)(1 (cid:22) ) 1 1 (cid:13) (cid:1)z = 1+ (cid:0) (cid:13) Z + 1+(cid:13) (cid:1)q A0 + 1+(cid:13) (cid:1)(p A (cid:0) p C )+ 1+(cid:13) (cid:1)x A (20) where recall that lower case letters represent logs, (cid:1) represents the (cid:133)rst-di⁄erence, (cid:22) is the long-term growth of the supply shift factor B in (18). As discussed above, Z (cid:1)q > 0wouldincreasetheequilibriumquantityoftheimportedinput. Risesinx A0 A and p represent positive shifts to the demand for the input and raise equilibrium A imports of the input. As to the e⁄ect of p , for a given real exchange rate q , a c A0 rise in p will mean a higher p and/or a lower e. In either case, the relative price C 0c 0 facing local suppliers would go down, decreasing the supply of the input and its equilibrium quantity. To solve for the equilibrium supply of processed exports ((cid:1)x ), the equilibrium A quantity of the input (cid:1)z; shown in equation (20), along with the standard labordemand function based on Cobb-Douglas technology, is substituted into the logdi⁄erenced version of the production function in (14) to yield: (cid:22) 1 (cid:10) (cid:11) (cid:1)x A = (cid:10) A + (cid:0) (cid:10) 1 (cid:1)(p A (cid:0) p C )+ (cid:10) (1 Z +(cid:13)) (cid:1)q A0 1 1 1 (1 (cid:11) (cid:11) )(cid:21) (cid:24) + (cid:0) A (cid:0) Z A (cid:1)k + A (21) (cid:10) (cid:10) 1 1 where (cid:22) = a +(cid:11) [1+(cid:13)(1 (cid:22) )]=(1+(cid:13)) with a representing the growth rate A A A z A (cid:0) of the state of the technology in sector A, i.e. the long-term growth rate of A , A (cid:10) = 1 (cid:11) (cid:13)(cid:11)Z; and (cid:21) is the share of capital allocated to sector A. 1 (cid:0) A (cid:0) 1+(cid:13) A The choice to specify the variables in growth rate form anticipates that di⁄erencing will be required to render stationarity to the variables used in the empirical work. Implicitly, this translates into an assumption that the (cid:17) shocks in the production functions (12)-(14) are random walk forcing variables. The white noise innovations to these random walk shocks are labeled the (cid:24)s, such as (cid:24) in the above A equation. Note that the supply function (21) is upward sloping with respect to relative prices, and that a greater availability of the capital stock or a positive productivity shockre(cid:135)ectsanoutwardshifttothissupply. Thee⁄ectsofchangesinq ,operating A0 throughtheequilibriumquantityoftheimportedinput, z, in(20)havealreadybeen discussed. The supply function of exports that are domestically sourced only does not involvetheinputZ. Tosolveforit,takethelog-di⁄erencedversionsoftheproduction 15

function (13), labor demand represented by (15) for i = D, and the wage equation (17) to obtain: (cid:11) (cid:11) (cid:24) (cid:1)x = D + D (cid:1)(p p )+(cid:21) (cid:1)k+ D (22) D D C D 1 (cid:11) 1 (cid:11) (cid:0) 1 (cid:11) D D D (cid:0) (cid:0) (cid:0) where (cid:21) is the share of capital allocated to sector D. As with the other supply D function it is upward slowing in the relative price and shifts out with more capital as well as with a productivity shock. 4.3 Market-Clearing The speci(cid:133)cation assumes that relative prices of the Chinese exported goods adjust to clear the market. The real exchange rate is regarded as a variable that the authorities target to try to engineer shifts in the demand function for these goods. In light of this, equating the growth rate of demand given by (11) to the growth rate of supply given by (22) gives a solution to the equilibrium growth rate of the relative price of the domestically sourced export, (cid:1)(p p ), which can then be D C (cid:0) substituted back into either of the two equations to yield the equilibrium growth rate of domestically-sourced exports: (cid:27)a (cid:27)(cid:11) (cid:11) (cid:1)me = (cid:1)xe = D D ((cid:1)q +(cid:1)q )+ D (cid:1)c + D D (cid:10) (cid:0) (cid:10) A B (cid:10) (cid:3) 3 3 3 (cid:27)(cid:21) (1 (cid:11) ) (cid:27) D D (cid:0) (cid:1)k+ (cid:24) (23) (cid:10) (cid:10) D 3 3 wherea isthegrowthrateofthestateoftechnologyinsectorD, i.e. thelong-term D growth rate of A and (cid:10) = (cid:11) +(cid:27)(1 (cid:11) ): Note that we have split the change D 3 D D (cid:0) in the real exchange rate into changes in its two components. Similarly, equating the right hand sides of (10) and (21) gives a solution to the equilibrium growth rate of the relative price of the processed export, (cid:1)(p p ), A C (cid:0) whichcanbeusedtoderivethefollowingequilibriumgrowthrateofChineseexports that use imported inputs: (cid:27)(cid:22) (cid:27)(cid:11) (cid:27)(1 (cid:10) ) (cid:27)(1 (cid:10) ) (cid:1)me = (cid:1)xe = A + A (cid:0) 2 (cid:1)q (cid:0) 2 (cid:1)q A A (cid:10) (cid:10) (1+(cid:13))$ (cid:0) (cid:10) A (cid:0) (cid:10) B 4 4 4 4 (cid:18) (cid:19) (1 (cid:10) ) (cid:27)(cid:21) (1 (cid:11) (cid:11) ) (cid:27) 2 A A Z + (cid:0) (cid:10) (cid:1)c (cid:3) + (cid:0) (cid:10) (cid:0) (cid:1)k+ (cid:10) (cid:24) A (24) 4 4 4 where 1 (cid:10) = (cid:11) (cid:13)(cid:11) =(1+(cid:13)) and (cid:10) = (cid:27)(cid:10) +(1 (cid:10) );and we have used the 2 A z 4 2 2 (cid:0) (cid:0) (cid:0) accounting relationship (cid:1)q = 1(cid:1)q . Of course, equations (23) and (24) describe A0 $ A growth rates of exports along equilibrium paths only if initially there was an equilibriuminlevelstobeginwith. Beingequilibriumgrowthrates, thesetwoequations 16

can be interpreted in terms of shifts in demands and supplies. An income e⁄ect that raises aggregate consumption demand in the countries that are the destination for Chinese (cid:133)nal goods exports, (cid:1)c > 0, shifts out the demand for both types of (cid:3) Chinese exports, leading to an increase in their equilibrium quantities. Similarly permanent productivity shocks, (cid:24)s > 0, as well as a higher capital stock, (cid:1)k > 0, shift out the supply functions, also leading to increases in the equilibrium growth rates of both types of exports. With respect to the e⁄ects of exchange rate changes, consider (cid:133)rst when the source of Chinese real exchange rate appreciation is appreciation of the Chinese currency against the currencies of countries other than those it imports parts and components from, i.e. (cid:1)q = 0; (cid:1)q > 0. For example, this might happen if all A B emerging Asian exchange rates appreciate simultaneously against the currencies of advanced economies. This appreciation makes Chinese (cid:133)nished goods more expensive to other countries, leading to lower demand for them and thus lower growth rates for both types of exports in equilibrium. It is not clear, which of the two export goods growth rates will fall by more(cid:150)this depends on various parameters of the model, including labor demand elasticities and the demand and supply elasticities of the imported input. Now consider an appreciation of the Chinese real exchange rate resulting from an appreciation against the rest of its emerging Asian trading partners only, i.e. (cid:1)q > 0; (cid:1)q = 0: Because the other Asians also buy (cid:133)nal goods from China, A B there is a lower demand as before tending to decrease the equilibrium quantity of each type of export good for China. However, in the case of the export good produced with imported inputs there is also an e⁄ect in the opposite direction. With the inputs now e⁄ectively cheaper, the supply of the export good also shifts out tending to increase the equilibrium quantity of exports. Thus for the processed exports, there is an ambiguous e⁄ect on exports from a change in Q (cid:150)what we can A say is that if the e⁄ect is negative, it will be less negative than from a change in Q . B 5 Modeling Chinese Exports: Empirics 5.1 Empirical Speci(cid:133)cation The empirical speci(cid:133)cations used are motivated by the above theoretical model but allow for richer dynamics. They also allow for a structural break in export growth after China(cid:146)s entry into the World Trade Organization (WTO) in December 2001andconsidersomealternativemodelsthatuseforeignoutputinsteadofforeign consumption, as wellas those that justuse the aggregate Chinesereal exchange rate 17

instead of considering movements against emerging Asian and non-emerging Asian currenciesseparately. Inaddition, althoughanattempttoconstructaproxyforthe Chinese capital stock was made from available investment and depreciation data, it did notproveverysuccessfulin termsof explainingexportbehavior. Therefore, the proxy for supply-side factors used is cumulative foreign direct investment (FDI), as in Marquez and Schindler (2007), which does have a signi(cid:133)cant e⁄ect on exports.19 Before getting to the actual export equations estimated, there is one other important issue to be clari(cid:133)ed. The theoretical model presented in the previous section began in levels and may imply some cointegrating relationships between the levelsofvariablesthatgetlostwhenwetakethegrowthrates. Forexample, thereal exchange rate, level of exports, foreign consumption, and domestic capital stock are likelytobecointegrated in themodelunlessmorepermanentshocksareintroduced. However, such a cointegrating relationship would presume that the real exchange rate was tending toward its long run path over the sample period, which in the case of China may not be appropriate. One could try to test for cointegration, but thereisunlikelytobemuchpowerinsuchtestsforasamplebasedonquarterlydata fromthemid-1990sonward. Accordingly, wechosetoestimatethemodelingrowth rates, rather than an error-correction speci(cid:133)cation that would imply cointegration. Speci(cid:133)cally,thefollowingequationswereestimatedfornon-processedandprocessed exports: DREXPN = a +a DUM +a (L)DREXPN + a (L)DCF t 01 11 t 21 t 1 31 t (cid:0) +a (L)DQ +a (L)DFDIK +" (25) 41 t 51 t Dt DREXPP = b +b DUM +b (L)DREXPP + b (L)DCF t 01 11 t 21 t 1 31 t (cid:0) +b (L)DQ +b (L)DFDIK +" (26) 41 t 51 t At DREXPN = a +a DUM +a (L)DREXPN + a (L)DCF t 02 12 t 22 t 1 32 t (cid:0) +a (L)DQA +a (L)DQB +a (L)DFDIK +" (27) 42 t 52 t 62 t Dt DREXPP = b +b DUM +b (L)DREXPP + b (L)DCF t 02 12 t 22 t 1 32 t (cid:0) +b (L)DQA +b (L)DQB +b (L)DFDIK +" (28) 42 t 52 t 62 t At 19Some earlier studies use a series for the capital stock in manufacturing constructed in Bai et al(2006) to capture export supply e⁄ects. However, because this series does not exist for much of our sample period, it could not be used in this paper. 18

where the a(L)s and the b(L)s are polynomials in the lag operator; DREXPN; DREXPP; DCF; DQ; DQA;DQB;DFDIK are the empirical proxies for (cid:1)x ; D (cid:1)x ; (cid:1)c ;(cid:1)q; (cid:1)q ;(cid:1)q , and (cid:1)k, respectively; and DUM is the dummy variable A (cid:3) A B for China joining WTO, which takes on the value of 1 for 2002:1 onwards and 0 otherwise. Wealsoestimatedsimilarequationswithaggregaterealexportgrowth(DREXP) as the dependent variable as well as speci(cid:133)cations that used foreign output growth (DYF) instead of foreign consumption growth. These alternatives facilitate comparison with results of previous studies. The estimated models were then used to obtain the statistical "long-run" e⁄ects (by setting L = 1 in the estimates of the above lag polynomials), which essentially sums the dynamic responses over time to give the cumulative response. Even though it is a cumulative response, since the variables are in growth rates with no error-correction terms, this cannot be interpreted as a structural long run elasticity, consistent with the idea that China(cid:146)s real exchange rate is probably not yet settled at its very long-run path. The cumulative e⁄ects should be interpreted as quasi-medium-term ones(cid:150)that is what the cumulative e⁄ect on export growth over time would be of a change in one of the explanatory variables, before the explanatory variables and export growth all slowly go back to their mean-reverting values. The estimated equations are then interpreted in light of the theoretical model and also used to simulate the path of Chinese exports under the assumption of greater exchange rate appreciation. 5.2 Data Issues To classify exports into processed and non-processed categories, we used Chinese o¢ cial statistics. Chinese customs data have categories of trade that are labelled "processingandassembly." Forexports,thisconsistsprimarilyofexportsofcomponents for assembly outside of China and exports of goods assembled using imported inputs. Both of these types of exports are generally believed to have a high import contentofexports. Weusedtheaggregateofthesecategoriesforobtainingthenominal value of processed exports. Exports that are not in these categories(cid:150)labeled "ordinary" exports in the Chinese statistics(cid:150)are non-processed exports. Data on Chinese trade prices are not available to be able to turn these nominal quantities into real quantities. As was noted earlier, the usual procedure has been to use a proxy for the Chinese export price de(cid:135)ator, such as the Hong Kong export price de(cid:135)ator or the U.S. price de(cid:135)ator for imports from non-industrial countries. We have also followed this approach.20 Speci(cid:133)cally, from 2003 onwards, when it 20We have already acknowledged and disscussed Marquez and Schindler(cid:146)s objections to this approach. 19

is available, we used the U.S. import price de(cid:135)ator for imports from China as a proxy for the Chinese export price de(cid:135)ator; for earlier periods we backcasted this series by using the growth rates of the U.S. import price de(cid:135)ator for non-industrial countries. In constructing foreign consumption/output growth, we used data on growth of real personal consumption expenditures/real GDP for the top ten destinations of Chinese exports. These growth rates were then weighted in proportion to the share of Chinese exports going to these destinations to obtain the aggregate foreign variables. For the Chinese trade-weighted real exchange rate, we used a Federal Reserve sta⁄estimate that is constructed as a weighted geometric average of bilateral CPIbased real exchanges rates with China(cid:146)s important trading partners. The weights take into account both import shares and export shares, as well as third party competition e⁄ects, as discussed in Loretan (2005).21 The idea of third-party competition is that if tradeable goods in Europe, say, become less expensive relative to those in China, not only would Chinese exports to Europe fall and Chinese imports from Europe rise due to the usual relative price e⁄ects, but there would also be a decline in Chinese exports to third parties as they switch imports toward Europe and away from China. Twenty six trading partners of China were used in the construction of the Chinese real exchange rate. Motivated by the theoretical model, we also split the Chinese real exchange rate into two components. The (cid:133)rst component, Q , is one in which the move- A ments in the real exchange rate are due to changes in the bilateral real exchange rates with respect to the other major emerging Asian economies which are the main sources of China(cid:146)s imports of parts and components. These are the newly industrializedeconomies(NIEs)ofHongKongSouthKorea, Singapore, andTaiwanandthe ASEAN-4 economies of Indonesia, Malaysia, the Philippines, and Thailand. The second component, Q , is one in which the movements in the real exchange rate B are due to changes in the bilateral real exchange rates with respect to the rest of the 26 economies in China(cid:146)s full index, including its main advanced-economy trading partners, such as the United States, the euro area, and Japan. The quarterly growth rate of the aggregate real exchange rate index and the contribution to this of each of its two components is presented in Figure 7. Note that the two components generally move in the same direction indicating that when the RMB appreciates or depreciates against the non-emerging Asian currencies, it also moves in the same direction against the emerging Asian currencies. Also, generally, the contribution of the non-emerging Asian currencies (the blue dotted line) to movements in the 21Note that Loretan is describing the construction of U.S. exchange rate indexes, but the same approach has been followed by Federal Reserve sta⁄in constructing the exchange rate indexes for other countries as well, including China. 20

overall index is greater than that of the emerging Asian currencies (the dashed red line), although there are some exceptions such as during the Asian Crisis years when the overall real appreciation of the Chinese currency was largely driven by RMB appreciation against other Asian currencies. The proxy for supply-side factors for Chinese exports that seemed to work best is based on FDI. Speci(cid:133)cally, starting in 1995, a cumulative stock series was constructed from FDI (cid:135)ows in each period and the rate of growth this stock series, DFDIK, was used to roughly capture supply-side in(cid:135)uences on Chinese exports. 5.3 Results Equations (25)-(28) and other variants of them described in the text were estimated using OLS applied to quarterly data. We started with a lag length of four for all the variables in the estimated equations.22 From these initial estimates, more parsimonious empirical models were obtained by successively removing insigni(cid:133)cant lags of some of the variables. The model reduction was guided by two rules. First, the reduced model had to satisfy a battery of statistical tests for model adequacy, including no autocorrelation in the error terms. Second, within the models that satis(cid:133)edthestatisticalcriteria,choiceofthe(cid:133)nalmodelusedwasguidedbytheminimization of the Hannan-Quinn (HQ) information criterion, the Schwartz criterion (SC), and the Akaike information criterion (AIC).23 The goal was to use as many observations as possible to get more precise estimates. However, a lot of structural changes were taking place in the Chinese economy in the early 1990s, and the economy also su⁄ered from quite high rates of in(cid:135)ation in that period. The 12-month CPI in(cid:135)ation rate peaked in late 1994 at nearly 30 percent and had not come down into single digits until the beginning of 1996. Thus we begin our estimation period in 1996:1, and the sample extends to the latest available data point for quarterly data of 2009:2. China(cid:146)s entry into the WTO in December 2001 was also a structural break, and a case might be made for starting the sample after 2001. However, this would not leave us with enough quarterly observations. Instead, an attempt was made to partly address the WTO-related structural break problem by including a dummy variable for China(cid:146)s WTO membership years since 2001. 22The one exception is DFDIK, where we started with a lag length of 3 to be able to start our regression estimation periods at the beginning of 1996. However, if we start the estimation one quarter later and do begin with a lag length of 4 for DFDIK, the fourth lag is never signi(cid:133)cant. 23Generally, the same model minimized all three of HQ, SC and AIC. When this was not the case, at least two of the three were minimized for the same model, which is the model that was chosen for those cases. 21

5.3.1 Model Estimates and Exchange Rate E⁄ects First, export equations were estimated using total exports and the aggregate real exchange rate index to see what results are obtained if we ignore the distinction between non-processing and processing exports and also ignore which trading partners are the source of the movements in the Chinese real exchange rate. The results for the model using foreign consumption are presented in table 1. The R2 of 0.56 does not seem too bad for the variant of the model that is estimated in (cid:133)rst di⁄erences rather than levels. The reported test statistics show that model adequacy criteria are satis(cid:133)ed. These tests include: a Lagrange-multiplier test for fourth order residual autocorrelation (AR 1-4 test); a test for autoregressive conditional heteroscedasticity (ARCH 1-4 test); a Normality test for the distribution of the error term; two tests for heteroscedasticity (based on a regression of squared residuals on the original regressors and their squares (Hetero test) and on all squares and products of the original regressors if the number of observations permit this (Hetero-X test); and a regression speci(cid:133)cation test (RESET ) that tests whether the linear functional form is adequate. Theresultsindicatethatrealexchangerateappreciationshavecontemporaneous andlaggednegativee⁄ectsonrealexportgrowth, whileforeignconsumptiongrowth has positive e⁄ects. The growth of the FDI capital stock has (cid:133)rst a positive e⁄ect and then a small, but signi(cid:133)cant, negative one later on export growth. The longrun solution of the statistical model, also presented in table 1, shows that a one percentage point increase in the annual rate of appreciation of the real exchange ratewouldhaveacumulativenegativee⁄ecton realexportgrowth of 1.8percentage points, which is statistically signi(cid:133)cant. A one percentage point increase in foreign consumption growth would increase export growth by 5.9 percentage points, which is also statistically signi(cid:133)cant, and appears to be an implausibly large e⁄ect. Also, a 1 percentage point increase in the growth rate of the FDI capital stock raises export growth by a cumulative and statistically signi(cid:133)cant 0.3 percentage points. This suggests signi(cid:133)cant supply-side factors at work in the determination of the equilibriumgrowthrateofexports. Alltheestimatede⁄ectsareinlinewiththeory. The estimated model also indicates a large and signi(cid:133)cant e⁄ect on export growth associated with China(cid:146)s entry into WTO. Table 2 presents the results for the total exports model in which the foreign consumption growth variable is replaced by a foreign real GDP growth variable. This model also passes the statistical adequacy tests. Qualitatively, very similar results are obtained, except that the cumulative e⁄ect of a rise in the rate of appreciation of the real exchange rate on export growth is smaller in magnitude, at about -1.1. percentage points. Theresultswhenseparateequationsareestimatedfornon-processedandprocessed 22

exports but still using an aggregate real exchange rate index are reported in tables 3 and 4 for the model with foreign consumption and tables 5 and 6 for the model with foreign output. The battery of statistical tests are satis(cid:133)ed for all these models. The cumulative e⁄ect of a 1 percentage point appreciation in the real exchange rate on growth of non-processing exports is -1.9 percentage points (table 3) while that on growth of processing exports is a bit less, at -1.5 percentage points (table 4), with both these e⁄ects being statistically signi(cid:133)cant. However, the long-run elasticity of exports with respect to a rise in foreign consumption is much higher at 10.7 for non-processing exports (table 3) than for processing exports at 2.0 (table 4); it is puzzling why these e⁄ects should be so di⁄erent. Also, note that WTO appears to have a signi(cid:133)cant e⁄ect on growth of non-processing exports only and not processing exports, where the WTO dummy was dropped because of lack of statistical signi(cid:133)cance. The (cid:133)t of the non-processing exports equation, with an R2 of 0.52 is about the same as for the aggregate exports equation but the (cid:133)t of the processing exports equation is lower, with an R2 of 0.35. The results from using foreign output growth instead of foreign consumption growth are qualitatively similar but again the magnitudes are a bit di⁄erent, as can be seen from tables 5 and 6. Speci(cid:133)cally, the cumulative e⁄ect of foreign output growth on growth of non-processing exports is somewhat smaller at 6.1 percentage points (instead of the 10.7 percentage points with foreign consumption growth) and ongrowthofprocessingexportsissomewhathigherat4.6percentagepoints(instead of 2 percentage points). The real exchange rate elasticities are somewhat lower for bothprocessingandnon-processingexportsandroughlyequaltoeachotherinthese models at -1.4. In sum, incorporating the most up to date recent data on real exchange rate movements gives us price e⁄ects on real exports that are statistically signi(cid:133)cant and consistently toward the upper end of the range that has been found in earlier studies. In particular, we do not get the insigni(cid:133)cant or wrong-signed e⁄ects that some in the literature have found. OneimportantfocusofourpaperinlightoftheimportanceofChina(cid:146)sprocessing trade was stated to be a distinction between Chinese real exchange rate movements against other emerging Asian economies versus Chinese real exchange rate movements against its other important trading partners. We now turn to results which examine whether the two components of the real exchange rate have di⁄erent e⁄ects as predicted by the theoretical model. The results for non-processing exports and processing exports are presented in tables 7 and 8, respectively, for the model using foreign consumption.24 Once again, the models pass all the standard statistical 24Equations were also estimated for aggregate exports with the two di⁄erent components of the realexchangerate. Theresultswereinbetweenthoseobtainedforthenon-processingexportsand processing exports reported here and have been excluded in the interest of brevity. 23

tests of speci(cid:133)cation. As can be seen from table 7, the (cid:133)t of the non-processing exports equation is quite good with an R2 of 0.64. The e⁄ects of foreign consumption growthislarge,asbefore,withacumulativee⁄ectof7.9percentagepointsongrowth of non-processing exports. In addition, real appreciation of the RMB against the other emerging Asian currencies consistently has a negative e⁄ect on growth of nonprocessing exports, whereas real appreciation against other currencies has dynamic e⁄ectsthatvaryinsignovertime. Intermsofthecumulativee⁄ects,a1percentage point increase in the rate of appreciation of the RMB against other emerging Asian currencies (DQ > 0) has a statistically signi(cid:133)cant, cumulative negative e⁄ect of A 3.9 percentage points on non-processing export growth. A same-sized appreciation against the other currencies (DQ > 0) lowers non-processing export growth by B less, about 1 percentage point, which is statistically insigni(cid:133)cant. The e⁄ect of the 2 FDI capital stock on non-processing exports is positive and statistically signi(cid:133)cant. As in the earlier speci(cid:133)cation with the overall real exchange rate index, the WTO e⁄ect on non-processing export growth is large and highly signi(cid:133)cant. As can be seen from table 8, for processing exports, the e⁄ects of foreign consumptiongrowtharesimilartothoseofnon-processingexportspresentedabove,but the e⁄ects of the real exchange rate are quite di⁄erent. The long-run elasticity with respect to foreign consumption is still more than 7 percentage points. However, real exchange rate appreciation of the RMB against the other emerging Asian currencies (DQ > 0) has a positive and insigni(cid:133)cant cumulative e⁄ect on processing A exports growth, whereas a real exchange rate appreciation against other currencies (DQ > 0) has a cumulative negative, and statistically signi(cid:133)cant, e⁄ect of 1.7 B percentage points. These results imply that if there was unilateral appreciation of the RMB (DQ = DQ > 0), the fall in processing exports would be much less A B than if all of the emerging Asian regions(cid:146)s exchange rates appreciated against other currencies (DQ = 0;DQ > 0). Although we have followed a totally di⁄erent ap- A B proach, these results are quite consistentwith those of Thorbecke and Smith (2008). Going back to our analysis of processing exports, the supply side variable was not signi(cid:133)cant and was dropped from the processing exports equation, according to the statistical criteria used. The R2 of the regression was 0.39. How do these results hold up to the predictions of the theoretical model presented earlier? The positive supply side e⁄ects and the positive e⁄ects of foreign consumption are in line with the theory, although throughout our analysis we (cid:133)nd the estimated income e⁄ects (whether using foreign consumption or foreign output) to be implausibly large in magnitude, as have some others such as Cheung et al (2008). The insigni(cid:133)cant e⁄ects on processing exports of a real RMB appreciation against other emerging Asian currencies, from which China imports much of its parts and components, and clearly signi(cid:133)cant negative e⁄ects of real RMB 24

appreciation against non-emerging Asian currencies are quite consistent with the predicted model. The negative e⁄ects of each type of real exchange appreciation on non-processing exports is also consistent with the theory. However, the exact theoretical model presented implies an equal elasticity of non-processing exports with respect each type of real exchange rate movement, which does not hold up. The results instead suggest that the price elasticity of (cid:133)nal goods imported from China is higher in the case of other emerging Asian economies than in the case of China(cid:146)s advanced-economy trading partners. Now consider what happens when foreign consumption growth is replaced by foreign output growth. The results are presented in tables 9 and 10. The (cid:133)ts of the models are similar, and most of the results are qualitatively similar to the model with foreign consumption. The long-run e⁄ect of foreign output on both types of exports is still large, although not as large as the foreign consumption e⁄ect was in earlier speci(cid:133)cations. Once again, real appreciation of the Chinese exchange rate against other emerging Asian currencies has a positive and insigni(cid:133)cant e⁄ect on processing exports, whereas the e⁄ect of Chinese real appreciation against non-emerging Asian currencies is signi(cid:133)cantly negative although a bit less large in magnitude than with the foreign consumption model. The one big di⁄erence in the results is that the e⁄ects of changes in the two components of real exchange rate on non-processing exports are now about equal in magnitude, both being negative e⁄ects and statistically signi(cid:133)cant, which is more in line with the original theoretical speci(cid:133)cation. The supply-side and WTO e⁄ects are very similar to the foreign consumption models.25 AlternativeRelativePriceMeasure Thomas,Marquez,andFahle(2009)have argued that typically used aggregate real exchange rate indexes may not always accurately capture some movements in aggregate relative prices. This is because of thewaytherealexchangerateindexiscomputed(cid:150)itisanindexwhosegrowthrateis derived from a weighted average of the growth rates of bilateral real exchange rates, and the level of the index itself does not have a relative price interpretation. If the underlying bilateral real exchange rates (relative prices) do not change, the index will not change even if the weights do. Of course, if the underlying bilateral relative pricesdochange, thentheweightswillberelevantforhowmuchtheaggregateindex changes. But the aggregate real exchange rate index will not capture any changes in relative prices that occur purely because of a change in weights. If, say, China(cid:146)s 25In this paper, we have not studied the implications of the rise of Chinese trade for the trade of other emerging Asian economies. Such implications have been a focus of a number of papers, including Ahearne et al (2009), Eichengreen et al (2004), Haltmaier et al (2009), and Hanson (2007). Ito (2008) discusses the in(cid:135)uence of the RMB on the exchange rate policies of the other Asian economies. Cui and Syed (2007) focus on the shifting structure of China(cid:146)s trade over time, including the switch to more sophisticated exports, and its implications. 25

tradepatternshiftsfromtradingwithhigh-priceindustrialcountries, suchasJapan, to low-price countries, such as other countries in emerging Asia, but the underlying bilateral relative prices have not changed, the real exchange rate index will not capture this, but the true relative price of China with respect to its trading partners would have increased. The authors propose an alternative measure, the weighted average relative price (WARP), that does not su⁄er from this problem. It is computed as a geometric weighted-averageofbilateralrelativepricelevels,wheretheweightschangeovertime and the bilateral price levels are computed based on World Development Indicator(cid:146)s estimates of PPP. They illustrate their measure by computing WARP measures for the United States and China and comparing those to the more conventional real e⁄ective exchange rate indexes, (cid:133)nding signi(cid:133)cant di⁄erences between the two measures. Inthispapersofar, wehaveusedtheconventionalrealexchangeratemeasureto facilitate comparison of our results with the considerable existing literature on the topic. However,giventhedi⁄erencesThomasetalfoundbetweenthetwomeasures, we also re-estimated our total exports equations, with the aggregate real exchange rate replaced by the WARP. Qualitatively the same results were obtained(cid:150)an increase in the rate of change of Chinese relative prices had a statistically signi(cid:133)cant negative e⁄ect on Chinese real export growth. The magnitudes, however, were quite di⁄erent, with the WARP measure yielding signi(cid:133)cantly higher estimated relative price elasticity of exports, which would just reinforce the conclusions from our earlier results. 5.3.2 In-Sample Fits Next we examine in more detail the in-sample (cid:133)ts of our models with the conventional real exchange rate measure. The top panels of (cid:133)gure 8 plots the actual (the dashed red lines) and (cid:133)tted (the solid blue lines) growth rates for the models that use foreign consumption growth and the aggregate real exchange rate index. The (cid:133)tsarefarfromperfect,especiallyfortheequationmodelingprocessingexports,but as was suggested by the R2 values, they appear fairly good; although the models do not always capture the quarter-to-quarter movements in export growth, they do appear to capture trend movements in the growth rates over time. To see this a bit more clearly, four-quarter growth rates implied by the model (shown as the solid blue lines in the bottom panels) were also compared to actual four-quarter growth rates of exports (the dashed red lines in the bottom panels). The upward trend in the four-quarter growth rates of exports between about 2001 to 2004 and the sharp downward trend during the recent global crisis are very wellcaptured by the models. 26

Figure 9 presents the in-sample (cid:133)ts for the speci(cid:133)cations with foreign output growth but still the overall real exchange rate index used. For the total exports equation and the non-processing exports equation, the (cid:133)ts are quite similar to those shown in Figure 8. In these two cases, statistically there appears to be little to choose between the models with foreign consumption growth and with foreign output growth. However, the statistical (cid:133)t of the processing exports equation is signi(cid:133)cantly better with the foreign output growth speci(cid:133)cation. Figure10givestheactualand(cid:133)ttedvaluesforthemodelswithforeignconsumption growth that distinguish between the source of the changes in the Chinese real exchange rate. The (cid:133)ts are somewhat better than those presented in the plots in (cid:133)gure 8; allowing di⁄erent responses to the two components of the real exchange rate does appear to make some di⁄erence. However, as a comparison of (cid:133)gure 11 with(cid:133)gure9shows, inthemodelswithforeignoutputgrowth, thegainfrommoving to the two components of the real exchange rate is only marginal. 5.3.3 Counterfactual Analysis: E⁄ects of alternative real exchange rate path Counterfactual analysis is conducted to gauge how di⁄erent would the level of Chinese exports expected to be today if the RMB had appreciated in real terms at a faster rate since mid-2005 than the average rate of appreciation that was actually observed. The counterfactual simulation assumes that the structure of the economy (the demand and supply equations) remain the same as the path of the real exchange rate is altered. In doing the counterfactual simulations, we must determine which of the estimated empirical models to use. Going with the models that distinguish between the two components of the real exchange rate would require taking a stand on how other emerging Asian economies would react to greater exchange rate appreciation in China. Rather than make arbitrary assumptions about this, for the counterfactual analysis, we used the model with only the overall real exchange rate but with di⁄erent equations for non-processing and processing exports. In addition, even though the statistical (cid:133)t for processing exports is better with the model in which foreign output growth is used, we selected the foreign consumption growth models (tables 3 and 4) since those are more directly related to the theoretical model which motivated the empirical analysis.26 Speci(cid:133)cally, we ask what the path of Chinese total exports have been if since breakingthedollarpeginmid-2005, theChineserealexchangeratehadappreciated per quarter at an annual rate of 10 percent instead of an average annual apprecia- 26Note, however, that we did also conduct simulations with the foreign output growth models as well, with quite similar results. 27

tion of 51 percent until mid-2009.27 The (cid:133)gure of 10 percent was chosen because 2 this would put the real exchange rate today at about 20 percent more appreciated than its actual value, which is in the rough neighborhood of the average degree of undervaluation of the Chinese RMB that is estimated by analysts. To present the counterfactual results, level series for predicted real exports (in 2004 U.S. dollars) are (cid:133)rst constructed from the (cid:133)tted values of the growth rate of exportsfortheselectedmodels. Thenacounterfactuallevelseriesforexportsissimulatedusingthesamemodelestimatesbutbyinputtinganalternativerealexchange rate path with the path of other explanatory variables the same. The constructed series work o⁄of the actual level of exports in the quarter preceding when the counterfactual path of the real exchange rate starts. Since separate non-processing and processing exports equations were estimated, the levels of exports constructed from each equation are added together to get the predicted and counterfactual paths of total real exports. Figure 12 presents the results of the counterfactual. Looking at the top panel, (cid:133)rst note that the model(cid:146)s predicted path of real exports (the blue dotted line) using the actual path of the real exchange rate matches quite well the actual path of exports until about 2008:3. Thereafter, the model does predict a downturn in real exports, but not to the extent that was actually observed (the black thick line) during the crisis. The counterfactual path of exports (the dashed red line) is well below the model prediction (as well below the actual path of exports), and the gap between the predicted and the counterfactual paths widens until about 2008:3, at whichpointthegapisabout$150billionin2004pricesataquarterlyrate. Towards the end of 2008, the gap between the predicted and counterfactual simulated path starts to narrow signi(cid:133)cantly. This narrowing continues, but the gap is still about $100 billion at the end of the sample period in 2009:2. The narrowing of the gap occurs because of the lagged e⁄ects of the substantial appreciation of the real e⁄ective RMB seen in the second half of 2008, which was more than the assumed appreciation of an annual rate of 10 percent. This e⁄ective appreciation occurred because the RMB was de facto pegged again against the dollar over this period, but the dollar appreciated substantially against other major currencies due to its safe haven role. With the dollar having moved down substantially against other major currencies in 2009:2 and 2000:3, and with the RMB still unchanged against the dollar, the e⁄ective RMB has fallen sharply during these two quarters, which should lead to a widening again of the gap, going forward. The percent deviation of the counterfactual path from the predicted path is shown in the bottom panel. Speci(cid:133)cally the bottom panel plots the values shown 27Note that the actual average annual appreciation per quarter would go down to about 4.5 percent if 2009:3 was included, given a real depreciation at an annual rate of about 12 percent in this quarter. However, our estimated equations use the sample period that ends in 2009:2. 28

by the dashed red line in the top panel less the values shown by the dotted blue line, expressed as a percent of the blue dotted-line values. The results indicate that if theRMBhadappreciatedinreale⁄ectivetermsatanannualrateof10percent, real exports from China would have been 40 percent less in 2008:3 than what the model predicts under the actual observed path of the real exchange rate. Because of the greater than assumed appreciation actually observed in the second half of 2008, this (cid:133)gure falls a bit to roughly 30 percent of observed real exports by 2009:2. Thus the assumed alternative path of the exchange rate implies a very large adjustment to Chinese external balances from the exports side, according to the estimated model. This does not even factor in the adjustment that would take place from the imports side. These results suggest strong scope for the Chinese current account to adjust in the face of greater exchange rate appreciation. 6 Conclusion Analysis and discussions of China(cid:146)s external sector and its implications for global imbalances(cid:150)both in academic circles and among policy makers(cid:150)appear to have an underlying tension embedded in them. On the one hand, many analysts argue that China(cid:146)s exchange rate is considerably undervalued and that it needs to appreciate for the current account balance to adjust to a path that can be sustained permanently. On the other hand, a number of analysts also question whether exchange rate appreciation would really have that much of an e⁄ect on Chinese exports because of the high content of imports in Chinese exports. There have been some but not too many empirical studies that have directly estimated the sensitivity of Chinese exports to real exchange rate changes in the actual experience of the Chinese economy. A few of these studies have argued that the distinction between non-processed and processed exports is important in this regard. However, these studies have come up with di⁄erent answers and, generally, because of the time when they were conducted, they do not incorporate the experience over much of the period since the revaluation of the RMB-dollar peg in mid-2005. ThispaperhasprovidedestimatesofthesensitivityofChineseexportstochanges in the exchange rate, distinguishing between processed and non-processed exports, as in some of the other studies, and using data up to mid-2009. The results show thattherehasbeensigni(cid:133)cantvariationinthetrade-weightedrealexchangerateover time and movements in the exchange rate substantially a⁄ect export growth in the direction predicted by theory; that is a greater exchange rate appreciation dampens export growth, and our estimated price elasticity is generally greater than unity, and towards the high end of elasticities found in previous work. Both processing and non-processing exports are found to be sensitive to real exchange rate changes. 29

We also obtained some results which strongly suggest that it matters which trading partners of China are the source of the real exchange rate changes. Specifically, since China imports most of its inputs and parts and components from other emergingAsianeconomies,appreciationsagainsttheircurrencieshaveanambiguous predicted e⁄ect on Chinese processing exports. On the one hand, these countries being direct consumers of Chinese exports, should be buying less of Chinese exports if they become more expensive to them. On the other hand, with a more appreciated RMB against these other Asian currencies, the imported inputs also become cheaper to China, which shifts out the supply of processing exports and increases their equilibrium quantity. By contrast, the e⁄ect of an appreciation of the Chinese currency against the currencies of its advanced-economy trading partners should unambiguously reduce both Chinese processing and non-processing exports. The empiricalresultsaregenerallyconsistentwiththesepredictions(cid:150)thereisasigni(cid:133)cant negative e⁄ect on exports of Chinese RMB appreciation against the currencies of China(cid:146)s advanced-economy trading partners, and there is a positive but statistically insigni(cid:133)cante⁄ectonprocessingexportsofChineseRMBappreciationagainstother emerging Asian currencies. The counterfactual simulations that were undertaken suggest that if the rate of real appreciation of the trade-weighted renminbi had been 10 percent at an annual ratefrom2005:3to2009:2insteadoftheannualrateof51 percentactuallyobserved, 2 on average, over this period, Chinese real exports in the middle of 2009 would have been roughly 30 percent less than they actually were. The implications of the results for global imbalances depend on what is exactly meant by global imbalances, which is not always clear-cut. If China(cid:146)s large current account surplus or its bilateral current account surplus with the United States by itself contributes to global imbalances, along with the U.S. bilateral current account de(cid:133)citwithChina,thenourresultssuggestthatgreaterdegreeofappreciationofthe Chinese currency would substantially help mitigate global imbalances. If, however, the big part of global imbalances is the U.S. overall current account de(cid:133)cit and the current account surplus of the emerging market world taken together, then it is less clear that greater appreciation of the Chinese currency would make a signi(cid:133)cant dent to global imbalances. For example, following an adjustment of the Chinese real exchange rate one scenario could well be that the fall in exports by China is largely matched by a rise in exports by other emerging market economies, including in emerging Asia, leaving aggregate current account balances of the United States and of emerging market economies more broadly unchanged. But the results do seem to imply that greater (cid:135)exibility of the exchange rate would help China toward its stated desired goal of shifting the sources of growth more toward domestic demand with less dependence on external demand. 30

References [1] Ahearne, Alan, John G. Fernald, Prakash Loungani, and John W. Schindler (2009). "Flying Geese or Sitting Ducks: China(cid:146)s Impact on the Trading Fortunes of Other Asian Economies," in Y. Cheung and K. Wong eds., China and Asia: Economic and Financial Interactions, Routledge, London. [2] Anderson, Jonathan (2006). "The Complete RMB Handbook" (4th edition), UBS Investment Research Asian Focus, September. [3] Aziz, Jahangir and Xiangming Li (2007). "China(cid:146)s Changing Trade Elasticities," IMF Working Paper WP/07/266, November. [4] Bai,Chong-En,Chang-TaiHsieh,andQingyiQian(2006). "ReturnstoCapital in China," Brookings Papers on Economic Activity, 2006:2. [5] Bernanke, Ben, (2006). "The Chinese Economy: Progress and Challenges," RemarksattheChineseAcademyofSocialSciences,Beijing, China, December. [6] Cerra, V. and A. Dayal-Gulati (1999). "China(cid:146)s Trade Flows: Changing Price Sensitivities and the Reform Process," International Monetary Fund (IMF) Working Paper WP/05/99, Washington DC. [7] Cerra, V. and S. Saxena (2003). "How Responsive is China(cid:146)s Export Supply to Price Signals?" China Economic Review, 14, pp.240-270. [8] Cheung, Yin-Wong, Menzie D. Chinn, and Eiji Fujii (2009). "China(cid:146)s Current Account and Exchange Rate," NBER Working Paper No. 14673, Cambridge, MA. [9] Cline, William R. and John Williamson (2008). "Estimates of the Equilibrium Exchange Rate of the Renminbi: Is There a Consensus and If Not, Why Not?" Chapter4inM.GoldsteinandN.Lardy, eds., Debating China(cid:146)s Exchange Rate Policy, Peterson Institute for International Economics, Washington, DC. [10] Couderte, Virginie and Cecile Couharde (2005). "Real Exchange Rates in China," CEPII Working Paper 2005-01, Paris, France. [11] Cui, L. and M. Syed (2007). "The Shifting Structure of China(cid:146)s External Trade and Its Implications," IMF Working Paper WP/07/214, Washington DC, September. [12] Dean, Judith M., K. C. Fung, and Zhi Wang (2007). "Measuring the Vertical Specialization In Chinese Trade," U.S. International Trade Commission Working Paper 2007-01-A,Washington DC, April. 31

[13] Dunaway, Steven, Lamin Leigh and Xiangming Li (2009). "How Robust Are Estimates of Equilibrium Real Exchange Rates: The Case of China," Paci(cid:133)c Economic Review, 14:3. pp.361-375. (Also, IMF Working Paper WP/06/220.) [14] Eckaus, R. (2004). "Should China Appreciate the Yuan?" MIT Working Paper 04-16. [15] Eichengreen, Barry (2006). "Is a Change in the Renminbi Exchange Rate in China(cid:146)s Interest?" Asian Economic Papers, 4:1, pp.40-75. [16] Eichengreen, B., Y. Rhee, and H. Tong (2004). "The Impact of China on the Exports of Other Asian Countries, National Bureau of Economic Research (NBER) Working Paper 10768. [17] Erceg, Christopher, Christopher Gust, and David Lopez-Salido (2008). "The Transmission of Domestic Shocks in Open Economies," NBER Conference volume International Dimensions of Monetary Policy, eds. Jordi Gali and Mark Gertler, University of Chicago Press, forthcoming. [18] Frankel, Je⁄rey A. (2006). "On the Yuan: The Choice Between Adjustment Under a Fixed Exchange Rate and Adjustment Under a Flexible Exchange Rate," in Understanding the Chinese Economy, ed., Gerhard Illing, Munich. [19] Frankel, Je⁄rey A. and Shang-Jin Wei (2007). "Assessing China(cid:146)s Exchange Rate Regime," manuscript, Paper prepared for the 44th Panel Meeting of Economic Policy, 2007. [20] Garcia-Herrero, Alicia and Tuuli Koivu (2009). "China(cid:146)s Exchange Rate Policy and Asian Trade," Bank for International Settlements Working Paper 282, April. [21] Goldstein, Morris (2007). "A (Lack of) Progress Report on China(cid:146)s Exchange Rate Policies," Peterson Institute for International Economics Working Paper 07-5, June. [22] Goldstein, Morris and Nicholas Lardy (2008). "China(cid:146)s Exchange Rate Policy: An Overview of Some Key Issues," Chapter 1 in M. Goldstein and N. Lardy, eds., Debating China(cid:146)s Exchange Rate Policy, Peterson Institute for International Economics, Washington, DC. [23] Haltmaier, Jane T., Shaghil Ahmed, Brahima Coulibaly, Ross Knippenberg, Sylvain Leduc, Mario Marazzi, and Beth Anne Wilson (2009). "The Role of China in Asia: Engine, Conduit, or Steamroller?" in Yueh, Linda, ed., The Future of Asian Trade and Growth: Economic Development with the Emergence 32

of China, London: Routledge. (See also, International Finance Discussion Paper 904, Board of Governors of the Federal Reserve System, Washington, DC, September 2007.) [24] Hanson, Gordon H. (2007). "China and the Manufacturing Exports of Other Developing Countries," manuscript, University of California at San Diego and NBER, July. [25] Ito, Takatoshi (2008). "In(cid:135)uence of the Renminbi on Exchange Rate Policies of Other Asian Currencies," Chapter 7 in M. Goldstein and N. Lardy, eds., Debating China(cid:146)s Exchange Rate Policy, Peterson Institute for International Economics, Washington, DC. [26] Lafrance, Robert (2008). "China(cid:146)s Exchange Rate Policy: A Survey of the Literature," Discussion Paper 2008-5, Bank of Canada. [27] Lardy, Nicholas (2007). "China: Rebalancing Economic Growth," in The China Balance Sheet in 2007 and Beyond, eds., C. Fred Bergsten et al, Peterson Institute for International Economics, Washington, DC. [28] Lardy,Nicholas(2006). "China:TowardaConsumption-DrivenGrowthPath," Peterson Institute for International Economics Policy Brief 06-6, October. [29] Lau, F., Y. Mo, and K. Li (2004). "The Impact of a Renminbi Appreciation on Global Imbalances and Intra-Regional Trade," Hong Kong Monetary Authority Quarterly Bulletin, March, pp.16-26. [30] Loretan, Mico (2005). "Indexes of the Foreign Exchange Value of the Dollar," Federal Reserve Bulletin, Winter 2005, pp.1-8. [31] Ma, Yue and Huayu Sun (2007). "Hot Money In(cid:135)ows and Renminbi Revaluation Pressure," Journal of Chinese Economic and Business Studies, 5:1, pp.19- 36, February. [32] Makin, Anthony (2007). "Does China(cid:146)s Huge External Surplus Imply an Undervalued Renminbi?" China and the World Economy, 15:3, pp.89-102. [33] Marquez, Jaime and John W. Schindler (2007). "Exchange Rate E⁄ects on China(cid:146)s Trade," Review of International Economics, 15:5, pp.837-853. (See also, International Finance Discussion Paper 887, Board of Governors of the Federal Reserve System, May 2006.) [34] McCallum, Bennet T. and E. Nelson (1999). "Nominal Income Targeting in an Open Economy Optimizing Model," Journal of Monetary Economics, 43:3, pp.553-579. 33

[35] Mundell, Robert (2004). "China(cid:146)s Exchange Rate: The Case for the Status Quo," Paper presented at IMF seminar on the Foreign Exchange System, Dalian, China, May. [36] Prasad, Eswar (2008). "Monetary Policy Independence, the Currency Regime, and the Capital Account in China," Chapter 2 in M. Goldstein and N. Lardy, eds., Debating China(cid:146)s Exchange Rate Policy, Peterson Institute for International Economics, Washington, DC. [37] Rahman, Mizanur and Willem Thorbecke (2007). "How Would China(cid:146)s Exports Be A⁄ected By a Unilateral Appreciation of the RMB and by a Joint Appreciation of Countries Supplying Intermediate Imports?" manuscript, George MasonUniversity, FairfaxCAandNationalGraduateInstituteforPolicyStudies, Tokyo, Japan. [38] Rodrik, D. (2006). "What(cid:146)s So Special About Chinese Exports? NBER Working Paper 11947. [39] Thomas, Charles P., Jaime Marquez, and Sean Fahle (2009). "Measures of International Relative Prices for China and the USA," Paci(cid:133)c Economic Review, 14:3, pp.376-397. [40] Thorbecke, Willem and Gordon Smith (2008). "How Would An Appreciation of the RMB and Other East Asian Currencies A⁄ect China(cid:146)s Exports?" Review of International Economics, forthcoming. [41] Thorbecke, Willem (2006). "How Would An Appreciation of the Renminbi Affect the U.S. Trade De(cid:133)cit with China? The B.E. Journal in Macroeconomics, 6:3. [42] Thorbecke, Willem and Hanjiang Zhang (2009). "The E⁄ect of Exchange Rate Changes on China(cid:146)s Labor-Intensive Manufacturing Exports," Paci(cid:133)c Economic Review, 14:3, pp.398-409. [43] Wang,LiandJohnWhalley(2007)."TheImpactsofRenminbiAppreciationon Trades Flows and Reserve Accumulation in a Monetary Trade Model," NBER Working Paper 13586, Cambridge, MA, November. [44] Yu, Yongding (2007). "Global Imbalances and China," Australian Economic Review, 40:1, pp.1-21. [45] Zhang, Z. (2001). "Real Exchange Rate Misalignment in China: An Empirical Investigation," Journal of Comparative Economics, 29, pp.80-94. 34

TABLE 1: Total Exports Model with Foreign Consumption and Aggregate Real Exchange Rate Index Modeling DREXP by OLS Sample: 1996:1 to 2009:2 Coefficient Std.Error t-value t-prob Part. R2 Constant -11.2484 7.448 -1.51 0.138 0.0482 DUM 18.1291 5.263 3.44 0.001 0.2087 DCF_1 3.07279 1.306 2.35 0.023 0.1095 DCF_4 2.82694 1.474 1.92 0.061 0.0756 DQ -0.472920 0.1991 -2.38 0.022 0.1114 DQ_1 -0.520574 0.2345 -2.22 0.031 0.0987 DQ_3 -0.775973 0.2542 -3.05 0.004 0.1716 DFDIK_1 0.370752 0.1225 3.03 0.004 0.1692 DFDIK_3 -0.0445431 0.01344 -3.31 0.002 0.1963 Sigma 14.6184 RSS 9616.38226 R2 0.564136 F(8,45) = 7.28 [0.000]** Log-likelihood -216.543 DW 2.17 No. of observations 54 No. of parameters 9 Mean(DREXP) 18.5355 Var(DREXP) 408.571 AR 1-4 test: F(4,41) = 0.23062 [0.9196] ARCH 1-4 test: F(4,37) = 0.74188 [0.5696] Normality test: Chi^2(2) = 1.6276 [0.4432] Hetero test: F(15,29) = 1.7757 [0.0900] Hetero-X test: not enough observations RESET test: F(1,44) = 2.2335 [0.1422] Solved static long-run equation for DREXP Coefficient Std.Error t-value t-prob Constant -11.2484 7.448 -1.51 0.137 DUM 18.1291 5.263 3.44 0.001 DCF 5.89972 1.550 3.81 0.000 DQ -1.76947 0.4315 -4.10 0.000 DFDIK 0.326208 0.1098 2.97 0.005

TABLE 2: Total Exports Model with Foreign Output and Aggregate Real Exchange Rate Index Modeling DREXP by OLS Sample: 1996:1 to 2009:2 Coefficient Std.Error t-value t-prob Part. R2 Constant -8.55763 6.957 -1.23 0.225 0.0325 DUM 15.3868 5.065 3.04 0.004 0.1702 DYF 2.43240 0.9771 2.49 0.017 0.1210 DYF_1 0.840562 1.029 0.817 0.418 0.0146 DYF_4 3.13271 1.282 2.44 0.019 0.1171 DQ_1 -0.503546 0.2677 -1.88 0.066 0.0729 DQ_3 -0.632178 0.2691 -2.35 0.023 0.1093 DFDIK_1 0.269907 0.1299 2.08 0.043 0.0876 DFDIK_3 -0.0369404 0.01416 -2.61 0.012 0.1314 Sigma 14.6573 RSS 9667.61739 R2 0.561814 F(8,45) = 7.212 [0.000]** Log-likelihood -216.687 DW 2.35 No. of observations 54 No. of parameters 9 Mean(DREXP) 18.5355 Var(DREXP) 408.571 AR 1-4 test: F(4,41) = 1.6475 [0.1808] ARCH 1-4 test: F(4,37) = 2.3147 [0.0755] Normality test: Chi^2(2) = 2.3165 [0.3140] Hetero test: F(15,29) = 2.0202 [0.0509] Hetero-X test: not enough observations RESET test: F(1,44) = 2.6928 [0.1079] Solved static long-run equation for DREXP Coefficient Std.Error t-value t-prob Constant -8.55763 6.957 -1.23 0.225 DUM 15.3868 5.065 3.04 0.004 DYF 6.40566 1.555 4.12 0.000 DQ -1.13572 0.4635 -2.45 0.018 DFDIK 0.232967 0.1164 2.00 0.051

TABLE 3: Non-Processing Exports Model with Foreign Consumption and Aggregate Real Exchange Rate Index Modeling DREXPN by OLS Sample: 1996:1 to 2009:2 Coefficient Std.Error t-value t-prob Part. R2 Constant -28.7267 12.55 -2.29 0.027 0.1064 DUM 24.7134 7.656 3.23 0.002 0.1915 DCF_1 3.61228 1.848 1.95 0.057 0.0799 DCF_3 2.76520 1.947 1.42 0.163 0.0438 DCF_4 4.35587 2.044 2.13 0.039 0.0936 DQ -0.420892 0.2796 -1.51 0.139 0.0490 DQ_1 -0.532913 0.3504 -1.52 0.135 0.0499 DQ_3 -0.926765 0.3831 -2.42 0.020 0.1174 DFDIK 0.821115 0.2844 2.89 0.006 0.1593 DFDIK_2 -0.282275 0.08974 -3.15 0.003 0.1836 Sigma 20.5309 RSS 18546.8741 R2 0.523986 F(9,44) = 5.382 [0.000]** Log-likelihood -234.278 DW 2.36 No. of observations 54 No. of parameters 10 Mean(DREXPN) 20.5157 Var(DREXPN) 721.534 AR 1-4 test: F(4,40) = 2.5594 [0.0532] ARCH 1-4 test: F(4,36) = 1.0572 [0.3917] Normality test: Chi^2(2) = 1.4354 [0.4879] Hetero test: F(17,26) = 0.76381 [0.7144] Hetero-X test: not enough observations RESET test: F(1,43) = 0.13080 [0.7194] Solved static long-run equation for DREXPN Coefficient Std.Error t-value t-prob Constant -28.7267 12.55 -2.29 0.026 DUM 24.7134 7.656 3.23 0.002 DCF 10.7334 2.906 3.69 0.001 DQ -1.88057 0.6323 -2.97 0.005 DFDIK 0.538840 0.2081 2.59 0.013

TABLE 4: Processing Exports Model with Foreign Consumption and Aggregate Real Exchange Rate Index Modeling DREXPP by OLS Sample: 1996:1 to 2009:2 Coefficient Std.Error t-value t-prob Part. R2 DREXPP_3 0.340745 0.1440 2.37 0.022 0.1045 Constant 10.2899 5.480 1.88 0.066 0.0684 DCF_1 3.98883 1.294 3.08 0.003 0.1653 DCF_2 -2.65937 1.402 -1.90 0.064 0.0698 DQ -0.540404 0.2233 -2.42 0.019 0.1087 DQ_3 -0.459832 0.2636 -1.74 0.087 0.0596 Sigma 17.2638 RSS 14305.8804 R2 0.351644 F(5,48) = 5.207 [0.001]** Log-likelihood -227.268 DW 2.19 No. of observations 54 No. of parameters 6 Mean(DREXPP) 17.8406 Var(DREXPP) 408.608 AR 1-4 test: F(4,44) = 0.57494 [0.6823] ARCH 1-4 test: F(4,40) = 0.36431 [0.8325] Normality test: Chi^2(2) = 0.61914 [0.7338] Hetero test: F(10,37) = 1.3003 [0.2662] Hetero-X test: F(20,27) = 0.88804 [0.6025] RESET test: F(1,47) = 1.7066 [0.1978] Solved static long-run equation for DREXP Coefficient Std.Error t-value t-prob Constant 15.6084 7.237 2.16 0.036 DCF 2.01661 2.334 0.864 0.392 DQ -1.51722 0.5905 -2.57 0.013

TABLE 5: Non-Processing Exports Model with Foreign Output and Aggregate Real Exchange Rate Index Modeling DREXPN by OLS Sample: 1996:1 to 2009:2 Coefficient Std.Error t-value t-prob Part. R2 DREXPN_4 -0.333686 0.1094 -3.05 0.004 0.1815 Constant -7.48718 9.023 -0.830 0.411 0.0161 DUM 22.6201 6.509 3.48 0.001 0.2234 DYF_1 4.47758 1.502 2.98 0.005 0.1746 DYF_2 -3.61862 1.892 -1.91 0.063 0.0802 DYF_3 4.02534 2.190 1.84 0.073 0.0745 DYF_4 3.29777 1.938 1.70 0.096 0.0645 DQ -0.342788 0.2738 -1.25 0.217 0.0360 DQ_1 -0.638371 0.3348 -1.91 0.063 0.0797 DQ_3 -0.878803 0.3592 -2.45 0.019 0.1247 DFDIK 0.507421 0.2739 1.85 0.071 0.0755 DFDIK_2 -0.180497 0.08418 -2.14 0.038 0.0987 Sigma 18.5026 RSS 14378.5596 R2 0.630967 F(11,42) = 6.528 [0.000]** Log-likelihood -227.404 DW 2.42 No. of observations 54 No. of parameters 12 Mean(DREXPN) 20.5157 Var(DREXPN) 721.534 AR 1-4 test: F(4,38) = 1.3428 [0.2720] ARCH 1-4 test: F(4,34) = 0.28832 [0.8835] Normality test: Chi^2(2) = 3.5026 [0.1735] Hetero test: F(21,20) = 0.62605 [0.8527] Hetero-X test: not enough observations RESET test: F(1,41) =0.0079106 [0.9296] Solved static long-run equation for DREXPN Coefficient Std.Error t-value t-prob Constant -5.61390 6.877 -0.816 0.418 DUM 16.9606 4.747 3.57 0.001 DYF 6.13493 1.471 4.17 0.000 DQ -1.39460 0.4633 -3.01 0.004 DFDIK 0.245128 0.1566 1.57 0.124

TABLE 6: Processing Exports Model with Foreign Output and Aggregate Real Exchange Rate Index Modeling DREXPP by OLS Sample: 1996:1 to 2009:2 Coefficient Std.Error t-value t-prob Part. R2 DREXPP_3 0.377162 0.1382 2.73 0.009 0.1448 Constant 5.74402 5.139 1.12 0.270 0.0276 DYF 3.07464 1.218 2.52 0.015 0.1266 DYF_1 1.62290 1.552 1.05 0.301 0.0242 DYF_2 -4.83961 1.690 -2.86 0.006 0.1572 DYF_3 2.98542 1.615 1.85 0.071 0.0720 DQ -0.440072 0.2799 -1.57 0.123 0.0532 DQ_2 -0.433449 0.2712 -1.60 0.117 0.0549 DFDIK_1 0.129601 0.07449 1.74 0.089 0.0644 DFDIK_2 -0.0958437 0.05421 -1.77 0.084 0.0663 Sigma 15.6045 RSS 10713.9975 R2 0.514432 F(9,44) = 5.179 [0.000]** Log-likelihood -219.461 DW 2.43 No. of observations 54 No. of parameters 10 Mean(DREXPP) 17.8406 Var(DREXPP) 408.608 AR 1-4 test: F(4,40) = 1.4287 [0.2422] ARCH 1-4 test: F(4,36) = 0.37417 [0.8255] Normality test: Chi^2(2) = 0.045851 [0.9773] Hetero test: F(18,25) = 0.27752 [0.9965] Hetero-X test: not enough observations RESET test: F(1,43) = 3.2582 [0.0781] Solved static long-run equation for DREXPP Coefficient Std.Error t-value t-prob Constant 9.22233 7.405 1.25 0.219 DYF 4.56515 2.483 1.84 0.072 DQ -1.40249 0.7794 -1.80 0.078 DFDIK 0.0541994 0.07126 0.761 0.450

TABLE 7: Non-Processing Exports Model with Foreign Consumption and Components of Real Exchange Rate Modeling DREXPN by OLS Sample: 1996:1 to 2009:2 Coefficient Std.Error t-value t-prob Part. R2 Constant -17.1326 11.94 -1.44 0.159 0.0478 DUM 24.5741 7.262 3.38 0.002 0.2183 DCF_1 4.65579 1.674 2.78 0.008 0.1588 DCF_3 3.24708 2.022 1.61 0.116 0.0592 DQA_2 -1.41476 0.9025 -1.57 0.125 0.0565 DQA_4 -2.46026 0.9167 -2.68 0.010 0.1494 DQB -0.603464 0.5016 -1.20 0.236 0.0341 DQB_1 -1.27142 0.4887 -2.60 0.013 0.1417 DQB_2 1.18251 0.6428 1.84 0.073 0.0762 DQB_3 -1.29887 0.5138 -2.53 0.015 0.1349 DQB_4 1.50164 0.6187 2.43 0.020 0.1256 DFDIK_1 0.444159 0.1721 2.58 0.014 0.1398 DFDIK_3 -0.0633066 0.01830 -3.46 0.001 0.2260 Sigma 18.5286 RSS 14075.6738 R2 0.638741 F(12,41) = 6.041 [0.000]** Log-likelihood -226.83 DW 2.35 No. of observations 54 No. of parameters 13 Mean(DREXPN) 20.5157 Var(DREXPN) 721.534 AR 1-4 test: F(4,37) = 1.9975 [0.1151] ARCH 1-4 test: F(4,33) = 0.13729 [0.9673] Normality test: Chi^2(2) = 0.91921 [0.6315] Hetero test: F(23,17) = 0.40892 [0.9765] Hetero-X test: not enough observations RESET test: F(1,40) = 0.16450 [0.6872] Solved static long-run equation for DREXPN Coefficient Std.Error t-value t-prob Constant -17.1326 11.94 -1.44 0.158 DUM 24.5741 7.262 3.38 0.001 DCF 7.90287 2.598 3.04 0.004 DQA -3.87502 1.477 -2.62 0.012 DQB -0.489593 1.100 -0.445 0.658 DFDIK 0.380853 0.1549 2.46 0.018

TABLE 8: Processing Exports Model with Foreign Consumption and Components of Real Exchange Rate Modeling DREXPP by OLS Sample: 1996:1 to 2009:2 Coefficient Std.Error t-value t-prob Part. R2 DREXPP_3 0.284791 0.1549 1.84 0.073 0.0699 Constant -5.32935 8.579 -0.621 0.538 0.0085 DUM 9.00861 5.882 1.53 0.133 0.0495 DCF 2.47386 1.470 1.68 0.099 0.0592 DCF_1 4.54602 1.499 3.03 0.004 0.1697 DCF_2 -1.93421 1.455 -1.33 0.190 0.0378 DQA_1 1.09498 0.8170 1.34 0.187 0.0384 DQB -0.779904 0.3687 -2.12 0.040 0.0904 DQB_1 -0.403640 0.5147 -0.784 0.437 0.0135 Sigma 17.3523 RSS 13549.5887 R2 0.38592 F(8,45) = 3.535 [0.003]** Log-likelihood -225.801 DW 2.14 No. of observations 54 No. of parameters 9 Mean(DREXPP) 17.8406 Var(DREXPP) 408.608 AR 1-4 test: F(4,41) = 0.39830 [0.8087] ARCH 1-4 test: F(4,37) = 0.67013 [0.6169] Normality test: Chi^2(2) = 0.32613 [0.8495] Hetero test: F(15,29) = 0.89438 [0.5776] Hetero-X test: not enough observations RESET test: F(1,44) = 3.6713 [0.0619] Solved static long-run equation for DREXPP Coefficient Std.Error t-value t-prob Constant -7.45146 12.44 -0.599 0.552 DUM 12.5958 7.966 1.58 0.120 DCF 7.11076 3.230 2.20 0.032 DQA 1.53100 1.193 1.28 0.205 DQB -1.65482 0.8825 -1.88 0.067

TABLE 9: Non-Processing Exports Model with Foreign Output and Components of Real Exchange Rate Modeling DREXPN by OLS Sample: 1996:1 to 2009:2 Coefficient Std.Error t-value t-prob Part. R2 DREXPN_4 -0.403497 0.1081 -3.73 0.001 0.2536 Constant -3.52734 8.486 -0.416 0.680 0.0042 DUM 22.7974 6.354 3.59 0.001 0.2389 DYF_1 4.29667 1.453 2.96 0.005 0.1757 DYF_2 -4.33453 1.905 -2.28 0.028 0.1121 DYF_3 7.65291 1.954 3.92 0.000 0.2722 DQA_2 -1.59898 0.8529 -1.87 0.068 0.0790 DQB -0.770140 0.4806 -1.60 0.117 0.0589 DQB_1 -0.832029 0.4742 -1.75 0.087 0.0699 DQB_2 1.21766 0.6073 2.00 0.052 0.0893 DQB_3 -1.28807 0.5434 -2.37 0.023 0.1205 DFDIK 0.461978 0.2731 1.69 0.098 0.0653 DFDIK_2 -0.183870 0.08342 -2.20 0.033 0.1059 Sigma 18.0625 RSS 13376.3657 R2 0.656689 F(12,41) = 6.535 [0.000]** Log-likelihood -225.454 DW 2.36 No. of observations 54 No. of parameters 13 Mean(DREXPN) 20.5157 Var(DREXPN) 721.534 AR 1-4 test: F(4,37) = 1.3654 [0.2646] ARCH 1-4 test: F(4,33) = 0.75832 [0.5599] Normality test: Chi^2(2) = 2.3477 [0.3092] Hetero test: F(23,17) = 0.28800 [0.9969] Hetero-X test: not enough observations RESET test: F(1,40) = 0.24778 [0.6214] Solved static long-run equation for DREXPN Coefficient Std.Error t-value t-prob Constant -2.51325 6.088 -0.413 0.682 DUM 16.2433 4.395 3.70 0.001 DYF 5.42576 1.156 4.69 0.000 DQA -1.13928 0.6059 -1.88 0.066 DQB -1.19172 0.6960 -1.71 0.093 DFDIK 0.198154 0.1478 1.34 0.186

TABLE 10: Processing Exports Model with Foreign Output and Components of Real Exchange Rate Modeling DREXPP by OLS Sample: 1996:1 to 2009:2 Coefficient Std.Error t-value t-prob Part. R2 DREXPP_3 0.349783 0.1429 2.45 0.019 0.1276 Constant -3.88970 7.943 -0.490 0.627 0.0058 DUM 5.39876 5.546 0.973 0.336 0.0226 DYF 3.50596 1.147 3.06 0.004 0.1856 DYF_1 2.69077 1.614 1.67 0.103 0.0635 DYF_2 -6.30495 1.830 -3.45 0.001 0.2245 DYF_3 4.20255 1.686 2.49 0.017 0.1316 DQA_1 1.04518 0.5558 1.88 0.067 0.0794 DQA_2 -0.978664 0.5726 -1.71 0.095 0.0665 DQB -0.832348 0.4054 -2.05 0.046 0.0932 DFDIK 0.420619 0.2649 1.59 0.120 0.0579 DFDIK_2 -0.253848 0.1173 -2.16 0.036 0.1026 DFDIK_3 0.0219210 0.009646 2.27 0.028 0.1119 Sigma 15.236 RSS 9517.5085 R2 0.568658 F(12,41) = 4.504 [0.000]** Log-likelihood -216.264 DW 2.44 No. of observations 54 No. of parameters 13 Mean(DREXPP) 17.8406 Var(DREXPP) 408.608 AR 1-4 test: F(4,37) = 1.4160 [0.2478] ARCH 1-4 test: F(4,33) = 0.73177 [0.5768] Normality test: Chi^2(2) = 0.027575 [0.9863] Hetero test: F(23,17) = 0.48888 [0.9448] Hetero-X test: not enough observations RESET test: F(1,40) = 0.78027 [0.3823] Solved static long-run equation for DREXPP Coefficient Std.Error t-value t-prob Constant -5.98216 12.52 -0.478 0.635 DUM 8.30301 8.206 1.01 0.317 DYF 6.29686 2.472 2.55 0.014 DQA 0.102295 1.180 0.0867 0.931 DQB -1.28011 0.6470 -1.98 0.054 DFDIK 0.290199 0.2673 1.09 0.283

Figure 1 Chinese External Sector Percent $ Billion 12 450 Annual 400 350 9 = Current Account/GDP Current Account (right axis) 300 250 6 200 150 3 100 50 0 0 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009* *2009 is annualized 2009H1 data. Figure 2 Chinese International Reserves Percent $ Billion 200 2400 Annual Reserves 2200 175 (monthly) 2000 150 = Reserves/GDP 1800 1600 125 1400 100 1200 1000 75 800 50 600 400 25 200 0 0 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 Figure 3 Chinese Exchange Rate RMB per dollar, inverted scale Index, Jun. 2005=100 6.5 130 Monthly 125 RMB appreciation 120 7.0 115 REER* 110 7.5 NEER* 105 100 8.0 95 Nominal 90 8.5 85 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 * Trade-weighted, CPI-based (Staff estimates).

Figure 4 Change in Reserves Billions of U.S. dollars 500 Change in reserves Trade balance (BOP) Net FDI 400 Hot money inflows 300 200 100 0 -100 2001 2002 2003 2004 2005 2006 2007 2008 2009:H1* *At an annual rate. Figure 5 Processing Trade Billions of U.S. dollars* 800 Quarterly 700 600 500 400 Processing goods exports 300 200 Processing goods imports 100 0 2001 2002 2003 2004 2005 2006 2007 2008 2009 *Customs data; seasonally adjusted, annual rate. Figure 6 Non-Processing Trade Billions of U.S. dollars* 900 Quarterly 800 700 600 500 Non-processing goods imports 400 300 200 Non-processing goods exports 100 0 2001 2002 2003 2004 2005 2006 2007 2008 2009 *Customs data; seasonally adjusted, annual rate.

Figure 7. Comparing Exchange Rates Quarter-over-Quarter Annualized Percent Change 40 Q* Q ** A Q *** B 30 20 10 0 -10 -20 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 *Trade-wheighted, CPI-based real export exchange rate (Staff estimates). **Component due to movements against other emerging Asian currencies. ***Component due to movements against currencies of trading partners outside of emerging Asia.

Figure 8. Actual and Fitted Export Growth: Models with Foreign Consumption and Aggregate Real Exchange Rate Total Exports Total Non-Processing Exports Total Processing Exports Percent (a.r.) Percent (a.r.) Percent (a.r.) 75 75 75 2 2 2 R =.56 R =.52 R =.35 50 50 50 25 25 25 0 0 0 DREXP -25 DREXPN -25 DREXPP -25 DREXP_fit1 DREXPN_fit1 DREXPP_fit1 -50 -50 -50 1995 2000 2005 2010 1995 2000 2005 2010 1995 2000 2005 2010 Total Exports Total Non-Processing Exports Total Processing Exports 4-quarter percent change 4-quarter percent change 4-quarter percent change 75 75 75 50 50 50 25 25 25 0 0 0 D4REXP -25 D4REXPN -25 D4REXPP -25 D4REXP_fit1 D4REXPN_fit1 D4REXPP_fit1 -50 -50 -50 1995 2000 2005 2010 1995 2000 2005 2010 1995 2000 2005 2010

Figure 9. Actual and Fitted Export Growth: Models with Foreign Output and Aggregate Real Exchange Rate Total Exports Total Non-Processing Exports Total Processing Exports Percent (a.r.) Percent (a.r.) Percent (a.r.) 75 75 75 2 2 2 R =.56 R =.63 R =.51 50 50 50 25 25 25 0 0 0 DREXP -25 DREXPN -25 DREXPP -25 DREXP_fit2 DREXPN_fit2 DREXPP_fit2 -50 -50 -50 1995 2000 2005 2010 1995 2000 2005 2010 1995 2000 2005 2010 Total Exports Total Non-Processing Exports Total Processing Exports 4-quarter percent change 4-quarter percent change 4-quarter percent change 75 75 75 50 50 50 25 25 25 0 0 0 D4REXP -25 D4REXPN -25 D4REXPP -25 D4REXP_fit2 D4REXPN_fit2 D4REXPP_fit2 -50 -50 -50 1995 2000 2005 2010 1995 2000 2005 2010 1995 2000 2005 2010

Figure 10. Actual and Fitted Export Growth: Models with Foreign Consumption and Components of Real Exchange Rate Total Exports Total Non-Processing Exports Total Processing Exports Percent (a.r.) Percent (a.r.) Percent (a.r.) 75 75 75 2 2 2 R =.63 R =.64 R =.39 50 50 50 25 25 25 0 0 0 DREXP -25 DREXPN -25 DREXPP -25 DREXP_fit3 DREXPN_fit3 DREXPP_fit3 -50 -50 -50 1995 2000 2005 2010 1995 2000 2005 2010 1995 2000 2005 2010 Total Exports Total Non-Processing Exports Total Processing Exports 4-quarter percent change 4-quarter percent change 4-quarter percent change 75 75 75 50 50 50 25 25 25 0 0 0 D4REXP -25 D4REXPN -25 D4REXPP -25 D4REXP_fit3 D4REXPN_fit3 D4REXPP_fit3 -50 -50 -50 1995 2000 2005 2010 1995 2000 2005 2010 1995 2000 2005 2010

Figure 11. Actual and Fitted Export Growth: Models with Foreign Output and Components of Real Exchange Rate Total Exports Total Non-Processing Exports Total Processing Exports Percent (a.r.) Percent (a.r.) Percent (a.r.) 75 75 75 2 2 2 R =.59 R =.65 R =.50 50 50 50 25 25 25 0 0 0 DREXP -25 DREXPN -25 DREXPP -25 DREXP_fit4 DREXPN_fit4 DREXPP_fit4 -50 -50 -50 1995 2000 2005 2010 1995 2000 2005 2010 1995 2000 2005 2010 Total Exports Total Non-Processing Exports Total Processing Exports 4-quarter percent change 4-quarter percent change 4-quarter percent change 75 75 75 50 50 50 25 25 25 0 0 0 D4REXP -25 D4REXPN -25 D4REXPP -25 D4REXP_fit4 D4REXPN_fit4 D4REXPP_fit4 -50 -50 -50 1995 2000 2005 2010 1995 2000 2005 2010 1995 2000 2005 2010

Figure 12. Counterfactual Simulation: 10 Percent Annual Real Appreciation of RMB beginning 2005:3 Level of Total Real Exports Billions of 2004 US Dollars 400 Predicted 350 Acutal 300 250 Counterfactual 200 150 2005 2006 2007 2008 2009 Deviation of Counterfactual Path from Model Prediction Percent 10 0 -10 -20 -30 -40 -50 2005 2006 2007 2008 2009