Product Sophistication and the Slowdown in Chinese Export Growth

Transcription

1 Saff Discussion Paper/Documen d analyse du personnel Produc Sophisicaion and he Slowdown in Chinese Expor Growh by Mark Kruger, Waler Seingress and Sri Thanabalasingam Bank of Canada saff discussion papers are compleed saff research sudies on a wide variey of subjecs relevan o cenral bank policy, produced independenly from he Bank s Governing Council. This research may suppor or challenge prevailing policy orhodoxy. Therefore, he views expressed in his paper are solely hose of he auhors and may differ from official Bank of Canada views. No responsibiliy for hem should be aribued o he Bank.

2 Bank of Canada Saff Discussion Paper November 2017 Produc Sophisicaion and he Slowdown in Chinese Expor Growh by Mark Kruger, Waler Seingress and Sri Thanabalasingam Inernaional Economic Analysis Bank of Canada Oawa, Onario, Canada K1A 0G9 ISSN Bank of Canada 2

3 Acknowledgemens We would like o hank Gino Caeau, Michael Francis, Chrisopher Hajzler, Daniel Trefler, Xiaodong Zhu and seminar paricipans a he Bank of Canada and he Banque de France for helpful commens and suggesions. We also hank Alexander Lam for excellen research assisance. Any remaining errors and omissions are our own. i

4 Absrac Chinese real expor growh deceleraed considerably during he las decade. This paper argues ha he slowdown largely resuled from China moving o a more sophisicaed mix of expors: China produced more sophisicaed goods over which i had pricing power insead of producing greaer volumes of less sophisicaed producs. Indeed, we show ha he share of highly sophisicaed producs in Chinese expors increased seadily over ime and ha Chinese expors became less price sensiive, suggesing increased pricing power. Furher, a decomposiion of China s marke share gains shows ha China coninues o gain marke share despie exporing producs wih higher-han-average world prices. China s coninuous gain in global expor marke share suggess ha is expor machine is far from broken. Bank opics: Inernaional opics; Exchange raes; Developmen economics JEL codes: F14, F17, O10 Résumé La croissance des exporaions réelles de la Chine a décéléré de façon significaive au cours des dix dernières années. Nore analyse démonre que ce ralenissemen s explique en grande parie par la présence de biens plus sophisiqués dans les exporaions chinoises. En effe, la Chine a produi des biens plus sophisiqués pour lesquels elle es en mesure d imposer ses prix au lieu de fabriquer de grands volumes de biens moins sophisiqués. En fai, nous monrons que la par des produis rès sophisiqués parmi les exporaions chinoises a augmené coninuellemen au fil du emps, e que les exporaions de la Chine son devenues moins sensibles aux mouvemens des prix, phénomène qui indiquerai une capacié accrue du pays d imposer ses prix. Lorsqu on décompose les gains de pars de marché de la Chine, on consae qu ils se poursuiven même si les prix des produis exporés son plus élevés que la moyenne inernaionale. La hausse consane des pars de marché de la Chine end à monrer que la machine à exporer du pays es loin d êre en panne. Sujes : Quesions inernaionales ; Taux de change ; Économie du développemen Codes JEL : F14, F17, O10 ii

5 1 Inroducion When China joined he World Trade Organizaion (WTO) in December 2001, he world economy opened o Chinese expors as rade barriers were relaxed. As a resul, Chinese expor volumes grew rapidly more han 25 per cen per year, on average, from 2002 o However, since 2011, China s real expor growh has slowed o abou 6 per cen, on average. While he slowdown in global demand in he afermah of he financial crisis explains par of China s weak expor performance, i does no accoun for all of he moderaion. Indeed, a sandard expor regression conrolling for compeiiveness and foreign demand significanly overesimaes Chinese expor growh in he pos-crisis period. 1 In his paper, we explore wheher China s move o a more sophisicaed expor mix has played a role. 2 In paricular, following Kwan (2002) and Hausmann, Hwang and Rodrik (2007), we consruc produc and expor sophisicaion indices and calculae China s sophisicaion gap relaive o he fronier, i.e., he counry wih he mos-sophisicaed expors. When we accoun for expor sophisicaion in our regression, he under-predicion disappears, and China s move oward he fronier can explain much of he expor slowdown. The inuiion here is sraighforward and is derived from he convergence hypohesis. In he same way ha he growh lieraure predics ha a counry s rae of gross domesic produc (GDP) growh will be faser he furher i is from he seady sae given by he neoclassical growh model (Barro and Sala-i-Marin 1992), our resuls sugges ha as a counry s sophisicaion gap diminishes, here is less scope for cach-up, and expor growh will decelerae. Even hough increased sophisicaion resuls in slower growh in expor volume, counries moving up he value chain benefi from producing more finely differeniaed goods over which hey have more pricing power. This, o some exen, compensaes for he loss in expor volume. Indeed, our resuls show ha Chinese expors became less price sensiive, suggesing ha pricing power increased over ime. To srenghen he argumen ha China indeed upgraded is expor mix, we look for indusrylevel evidence of he changing naure of he expor baske. We examine wheher Chinese expors become less price sensiive, boh wihin and across indusries. Based on indusry-level US impor daa, our resuls show ha changes in he expor baske of goods are driven by boh 1 The pos-crisis period refers o he period afer he global financial crisis of Suon and Trefler (2016) sugges ha endogenous facors higher domesic wages and producion coss provide he impeus for a counry o upgrade is expor mix. 1

6 cross-indusry specializaion and wihin-indusry upgrading. The share of highly sophisicaed producs in Chinese expors does increase over ime, and we find ha more-sophisicaed goods are generally less price sensiive. Moreover, we also find some evidence of wihinindusry upgrading as price elasiciies decreased for 11 of 19 manufacuring indusries over ime. This poenially provides he scope for greaer markups. The changes in he mix of expor producs have implicaions for Chinese compeiors in global markes. When looking a he evoluion of China s share of global expors, we find ha China coninues o ou-compee oher counries and gain share a a fairly consan rae. Upgrading he produc mix oward more differeniaed producs enables China o gain marke share despie charging higher prices. A decomposiion of marke share gains shows ha more han half of he increase in he recen pas comes from prices raher han quaniies. A second decomposiion provides evidence ha China coninues o gain marke share despie exporing producs wih higher-han-average world prices. China s coninuous gain in global expor marke share suggess ha is expor machine is far from broken. The res of he paper is organized as follows: Secion 2 presens he sylized facs based on he aggregae expor daa. In secion 3, we use indusry-level, US impor daa o shed ligh on changes ha may have conribued o he slowdown in Chinese expor growh. Secion 4 focuses on changes in Chinese expor marke share, and secion 5 concludes. 2 Assessing Chinese Expor Dynamics To analyze Chinese real expor growh over ime, we esimae he following sandard regression: α d REER d F CNY e, (1) d log X Chn, = + β1 log Chn, + β2 log Chn, + + Chn, where X Chn REER Chn, is Chinese expors in real erms;, is China s broad real effecive exchange rae (REER) index as compiled by he Bank for Inernaional Selemens; F Chn, is foreign demand for Chinese producs, defined as he average of China s main rading parners indusrial producion, weighed by heir share of Chinese expors a ime ; 2

7 CNY represens Chinese New Year dummy variables each dummy conrols for he monh in which he New Year ook place; and e Chn, is he error erm. Expors, he real effecive exchange rae and foreign demand are expressed as monhly yearover-year growh raes. The resuls in column 1 of Table 1 show ha he coefficiens in our equaion have he expeced signs and magniudes. A one-percenage-poin appreciaion of he REER reduces Chinese expor growh by 0.83 percenage poins, and a one-percenage-poin increase in foreign demand increases expors by 1.45 percenage poins. While foreign demand and he REER capure he dynamics of Chinese real expor growh quie well, we find ha he consan explains a large par of he expor growh. Our esimae implies ha, absen changes in foreign demand and compeiiveness, Chinese expors would grow, on average, a 12.6 per cen over he sample period. This large consan is an arifac of China s accession o he WTO and capures he one-ime increase in expors associaed wih increased marke access. Figure 1a compares acual and prediced expor growh. We observe ha a he beginning of he period, our regression esimaes underesimae expor growh, while we overesimae oward he end of he sample period. Indeed, from 2015 onward, he model over-prediced Chinese expor growh by around 10 percenage poins. These resuls sugges ha he errors are rended. To invesigae his poin furher, Figure 1b plos he error erm of equaion 1 ogeher wih a linear rend and 90 per cen confidence inerval. Noe ha he linear ime rend is downward sloping. 3

8 Figure 1a: Acual and prediced real expor growh Year-over-year growh raes, monhly esimaes % 50 Figure 1b: Chinese expor growh regression residuals Esimaed linear rend and 90% confidence bands Percenage poins Real expor growh Aggregae residuals Sources: China Cusoms via Haver Analyics, Bank of Canada calculaions Model fi Las observaion: Dec Trend esimaion Aggregae residuals Noe: The confidence bands are a he 90% level. Source: Bank of Canada calculaions Las observaion: Dec A poenial cause of he rend in he esimaion errors migh be ime-varian price elasiciies. Our base model imposes a ime-consan relaionship beween relaive prices and expor growh. However, changes in he ype of producs expored may change his elasiciy. For example, if China upgrades he qualiy of is producs and produces more-sophisicaed goods, expor volumes may reac significanly less o price changes. A high-level look a China s expor mix suggess ha here has, indeed, been a shif from lower-ech producs, such as clohing, foowear and oys, o high-ech goods, such as compuers, elecommunicaions equipmen and elecronics. The share of high-ech producs in China s global expors was roughly a fifh in 2001, and i rose o almos a hird by 2006 (Figure 2). 3 Figure 2: Composiion of Chinese expors Share of oal expors, 2001 and % High-ech producs 19% 18% 18% 45% 11% 58% 29% Mechanical and elecrical producs Low-ech producs Oher Source: China Cusoms via Haver Analyics Las observaion: High-ech goods are as defined by Chinese Cusoms. They consis of elecronics, compuers and elecommunicaions, and aerospace, among oher producs. Low-ech goods consis of foowear, oys, garmens and clohing accessories. 4

9 This rapid ransformaion in China s expor baske suggess he aggregae price elasiciy migh be quie differen from a decade ago. To es his predicion, we run he base model specificaion in rolling 10-year window regressions over a 15-year sample period ( ). The price elasiciy (he parameer β1 in our base model) for each window wih is 95 per cen confidence band is ploed in Figure 3, which clearly shows a rend of increasing inelasiciy. Figure 3: China's price elasiciy over ime, oal expors Change in oal expors in response o a 1% change in prices, calculaed using rolling 10-year window regressions % % confidence inerval Price elasiciy Source: Bank of Canada calculaions Las observaion: 2016 To shed furher ligh on he changes in he Chinese expor produc mix, we use monhly, disaggregaed, indusry-level daa from he Unied Saes impor regisry. The key advanage of he US impor daa is ha hey provide a maching of monhly, indusry-specific rade and producion informaion, which is useful for disaggregaed analysis. To check wheher he 20 per cen of Chinese expors ha go o he Unied Saes are represenaive of he overall rend in Chinese expors, we es wheher he observed paerns in he aggregae Chinese expor daa are also presen in he US impor daa. The US daa do indeed exhibi he same rends in he errors and he price elasiciy as we found in he aggregae daa (Figure 4). 5

10 Figure 4a:China o Unied Saes expor growh regression residuals Percenage Esimaed linear rend and poins 90% confidence bands Residuals Source: Bank of Canada calculaions Trend esimaion Las observaion: Dec Figure 4b: China's price elasiciy over ime, US expors Change in expors o he Unied % Saes in response o a 1% change in prices, calculaed using rolling year window regressions % confidence inerval Price elasiciy Source: Bank of Canada calculaions 0-1 Las observaion: 2016 Nex, we exploi he US indusry-level daa and plo he Chinese marke share of eigh manufacuring indusries over he period from 2001 o Figure 5 shows ha since 2010 China has been broadly gaining marke share in he following indusries: machinery, compuer and elecronics, elecrical equipmen and, o a lesser exen, ransporaion equipmen. These are generally considered o be higher-ech indusries. On he oher hand, since 2010 China has been losing marke share in lower-ech indusries such as exiles, apparel and accessories, leaher and furniure (see Figure 6). These indusry-specific resuls sugges ha changes in he ype of producs China expors may be a he roo of he decline in Chinese expor growh. In he nex secion, we es his idea more rigorously. Figure 5: China's share of he US marke in high-ech goods 1-year moving averages, monhly daa Sources: US Census Bureau, Bank of Canada calculaions Compuer and elecronics Elecrical equipmen Machinery Transporaion equipmen % Las observaion: Dec Figure 6: China's share of he US marke in low-ech goods 1-year moving averages, monhly daa Leaher and allied Apparel and accessories Toal exiles Furniure Sources: US Census Bureau, Bank of Canada calculaions Las observaion: Dec % For each indusry, we calculae China s share of he US marke as impors from China divided by US absorpion. Absorpion is defined as US sales minus expors plus impors. 6

11 3 Analyzing Srucural Changes in Chinese Expors As discussed above, he observed decline in he price elasiciy of Chinese expors could be due o a change in he ype of producs ha China expors. One meric o capure such changes is he expor sophisicaion index developed by Kwan (2002) and uilized by Francis, Painchaud and Morin (2005). This index ries o measure he income level of a counry's expors so as o assess heir produciviy based on he hypohesis ha richer counries produce moresophisicaed producs. Thus, he higher he impor expendiure share on goods from rich counries, he more sophisicaed he produc or indusry is. Inspired by Kwan, we firs consruc a produc sophisicaion index for each hree-digi Norh American Indusry Classificaion Sysem (NAICS) code manufacured good (see Table 2) as follows: PS k = i= 1 w ik ( GDPpc i / GDPpc US ), (2) where PS k is he sophisicaion of good k, = IMP IMP is he raio of impors of good k from counry i o oal US impors w ik i, k / US, k of ha produc in 2002, GDPpc is purchasing power pariy (PPP) adjused GDP per capia of counry i in 2002, i and GDPpc is PPP-adjused GDP per capia of he Unied Saes in US Our analysis differs from Kwan (2002) in a couple of ways. Firs, he looked a sophisicaion a a global level, while we confine our analysis o he US marke. Second, we fix he produc sophisicaion index o he iniial year (i.e., 2002) o capure poenial endogeneiy concerns ha arise because as China eners more sophisicaed producs, he produc sophisicaion index will naurally decline. Having calculaed he sophisicaion of each produc, we follow Kwan (2002) in calculaing China s expor sophisicaion index as follows: 7

12 Chn, K ES = w k = 1 Chn, k, PS k, (3) where ES Chn w Chn k,, is China s expor sophisicaion index, and, is good k s share of oal US impors from China a ime. Figure 7 shows China s sophisicaion index along wih hose of oher op exporers o he Unied Saes. The mos-sophisicaed exporers are Germany, Canada and, increasingly, Japan. Mexico and Souh Korea have boh made gains in expor sophisicaion since 2001, wih Souh Korea oupacing Mexico in recen years. China has seadily become more sophisicaed over he sample period bu coninues o lag he oher exporers. Figure 7: Expor sophisicaion indices Annual esimaes Index Germany Canada Japan Souh Korea Mexico China Source: Bank of Canada calculaions Las observaion: 2016 Nex, we compare China s expor sophisicaion index o ha of he counry ha sends he mos-sophisicaed producs, on average over he sample period, o he Unied Saes: Gap = ES / max( ES ). (4) Chn, Chn, j The idea here is ha he counry wih he highes expor sophisicaion index represens he echnological fronier. The gap gives an indicaion of he disance beween China and he fronier. Of he major exporers o he Unied Saes, Germany has he mos-sophisicaed expors, on average, over he sample period. Figure 8 plos he evoluion of he sophisicaion gap beween China and Germany over ime. We observe ha during he las 16 years China upgraded is produc mix by exporing more-sophisicaed producs. While a he beginning of 8

13 he period China s average produc sophisicaion was around 70 per cen of Germany s sophisicaion, his raio coninuously increases over ime o above 75 per cen in China has been converging wih Germany a a somewha slower rae han Korea and abou as fas as Mexico. Figure 8: China's expor sophisicaion gap wih Germany Per cen of Germany's expor sophisicaion index level, annual esimaes % China expor sophisicaion gap Source: Bank of Canada calculaions Las observaion: 2016 We include he gap in our aggregae regression o represen he change in he mix of Chinese expors: d log X Chn, = α + β1 log REER + β2 log F + β3gap + CNY + e. (5) Chn, Chn, Chn, Chn, Column 2 of Table 1 presens he resuls. Including he gap improves he fi of he equaion. Moreover, he coefficien on he real effecive exchange rae falls from in he baseline model o This implies ha he gap measure is negaively correlaed wih he compeiiveness measure (dlogreerchn, ) and hence capures par of he ongoing changes in China s supply of expors. On he oher hand, he coefficien on he demand elasiciy (β2) does no change significanly. Furhermore, unlike he base model, he downward rend in residuals disappears under his specificaion. Nex, we include he square of he gap erm in equaion 5 o capure any non-linear relaionship beween he gap and expor growh. The gap-squared variable is significan a he 1 per cen level, and we find he conribuion of he gap, is squared erm and he consan o be broadly similar o ha from equaion 5. The resuls are presened in column 3 of Table 1 and Figure 9. 9

14 Figure 9: Chinese expor growh model resuls Year-over-year growh raes, monhly esimaes % Real expor growh Model fi Model residuals Sources: China Cusoms via Haver Analyics, Bank of Canada calculaions Las observaion: Dec We hen calculae he conribuion of each variable in equaion 5 o overall real expor growh. Given ha he heory of convergence suggess ha expor growh depends on he disance o he echnological fronier, we combine he conribuion of he consan wih he level of he gap and is square ino a single facor. Figure 10 shows he resuls. This equaion suggess ha changes in he Chinese expor produc mix he gap conribue he mos o he slowdown in Chinese expor growh. The gap was a missing variable in our baseline regression; excluding i led o he large consan and he declining rend in he errors. As Chinese producs become more sophisicaed and approach he echnological fronier, expor growh of 20 per cen is increasingly difficul o mainain. Convergence oward he fronier implies ha he reurns o producion facors diminish, and overall produciviy growh will ineviably slow down. A he end of he sample period, he conribuion of he gap o expor growh is negaive bu relaively small. Thus, going forward, a he curren level of expor sophisicaion, changes in compeiiveness and foreign demand will be he main deerminans of Chinese expor growh. A prominen explanaion for he rapid growh in Chinese expors since China s accession o he WTO in 2002 is he seady decline in Chinese rade coss. For his reason, we include he change in average rade coss of manufacuring ha China faces in is expor markes as an addiional conrol variable. 5 Column 4 of Table 1 presens he resul. Noe ha, conrary o our expecaions, average rade coss do no have a significan impac on Chinese expor growh. One poenial explanaion is ha he consruced rade coss from he World Bank cover only 5 Daa on rade coss comes from he Trade Coss Daase provided by he World Bank. The daa se conains esimaes of bilaeral rade coss in agriculure and manufacured goods for he period. I is buil on rade and producion daa colleced in 178 counries. Symmeric bilaeral rade coss are compued using he Inverse Graviy Framework (Novy 2013), which esimaes rade coss for each counry pair using bilaeral rade and gross naional oupu. 10

15 one par of oal rade coss. An addiional source of rade coss no covered by he daa is nonariff barriers, which are difficul o quanify. Figure 10: Conribuions o Chinese real expor growh Monhly esimaes Percenage poins Expor sophisicaion gap Exchange rae Foreign demand Real expors -50 Source: Bank of Canada calculaions Las observaion: Dec Why would China move up he value chain? The resuls in he previous secion creae a puzzle. Why would China produce moresophisicaed producs if i means weaker real expor growh? According o he heory oulined by Suon and Trefler (2016), early on in he developmen process, counries produce relaively unsophisicaed goods and compee mainly on price. Since relaively unsophisicaed producs are fairly homogeneous, he abiliy o produce a a low price can resul in large marke share gains. However, as real incomes and by exension uni coss of producion sar o rise, coninuing o compee on prices in relaively unsophisicaed goods becomes more difficul as margins are squeezed. One sraegy for mainaining margins in he face of increasing producion coss is o improve expor mix, since more sophisicaed goods provide increased produc differeniaion and greaer pricing power. However, his expor-upgrading process is cosly, as a counry needs o acquire he new echnologies and disribuion channels. Because of hese high ransiion coss, i may no be opimal for every counry o move up he value chain. As a resul, he number of counries capable of producing highly sophisicaed goods decreases wih he level of echnology. Having fewer compeiors reduces compeiive pressures, improves pricing power and increases markups. The more sophisicaed produc mix will be less sensiive o price changes and resul in a lower price elasiciy of expors. Overall, his heory provides a coheren explanaion of he sylized facs oulined in secion 2. 11

16 Nex, we provide furher empirical evidence consisen wih he heory oulined by Suon and Trefler (2016). More specifically, we es wheher China s increasingly sophisicaed produc mix has been accompanied by a lower price elasiciy of expors. There are wo poenial channels hrough which his effec migh work. Firs, over ime, he mix of expors could shif o goods ha are more sophisicaed and less price elasic. Second, wihin a given indusry, producs become more differeniaed, and heir price elasiciy may decrease over ime. To es he across-indusry implicaions, we run he following specificaion for each of he 19 manufacured goods a he NAICS hree-digi level: d log X α realabs + e, (6) Chn, US, = + β1d log relprichn, US, k, + β2 log US, k, Chn,, where d X Chn, US, log is real growh of Chinese expors o he Unied Saes of produc k in year-overyear erms; log relpri Chn US k is he year-over-year change of Chinese impor prices, relaive o changes d,,, in he US prices, for each hree-digi NAICS produc k a ime ; d log realabs US, k, capures he year-over-year changes in US demand for produc k and is defined as US absorpion deflaed by he producer price index (PPI) of produc k; 6 and e Chn, is he error erm. Before discussing he resuls of regression 4, we wan o poin ou ha he US Bureau of Labor Saisics does no provide bilaeral, indusry-specific price indexes. For his reason, we consruc a uni-value-based, indusry-specific US impor price index for Chinese goods using deailed monhly impor daa. More precisely, we firs calculae year-over-year changes in uni values for impored producs from China a he 10-digi Harmonized Sysem (HS) level by monh beginning in We follow Gaulier e al. (2008) and aggregae he 10-digi HS uni value changes ino weighed hree-digi NAICS chained Tornqvis price indexes. We hen divide hese Chinese impor price series by heir US PPI equivalen, giving us a relaive China-Unied Saes price index in domesic currency for each indusry k a ime. Resuls from he regressions across producs are shown in he firs hree columns of Table 3. Apar from chemical producs and primary meal producs, all he regressions show a negaive sign for β1, and, excep for peroleum and coal producs, all of he regressions show a posiive sign for β2. 6 US absorpion is US producion plus impors minus expors. 12

17 In general, hese resuls show ha as China s relaive price of produc k increases, expor growh of ha good o he Unied Saes falls, and as he US demand for good k rises, expors increase. Figure 11 plos he esimaed β1 coefficien for each hree-digi NAICS indusry agains he produc sophisicaion index defined above. I shows ha as expor sophisicaion increases across producs, he price elasiciy of he produc falls. Thus, o some exen, China can gain pricing power even as expor growh slows. Figure 11: Expor sophisicaion and price elasiciies Comparison by 3-digi NAICS manufacuring indusry Price elasiciy (%) Produc sophisicaion based on US impor shares -2.5 Sources: China Cusoms via Haver Analyics, Bank of Canada calculaions In addiion o producing a more sophisicaed mix of goods, China migh also be upgrading is producs wihin given indusries. This would also have he effec of increasing pricing power. In he following regression, we es wheher he price elasiciy has declined in he pos-financial crisis period for each indusry k: d log X Chn, US, = α + β log d log relpri + β d 3 1 crisis Chn, US, k, *log d log relpri Chn, US, k, + β log realabs 2 + β d 4 crisis + e US, k, Chn, US,, (7) where dcrisis is a dummy o idenify he pos-crisis period, which we have defined as he period beginning in We find ha 11 of he 19 manufacuring producs experienced a decline in price elasiciies over he sample period. Column 5 in Table 3 presens hese resuls; a posiive sign on β3 indicaes a decline in he price elasiciy. This provides some evidence ha along wih a shif o 13

18 more-sophisicaed producs, China has also increased qualiy wihin producs, hereby compeing more on qualiy raher han on price. Overall, his secion relaes changes in he mix of Chinese expors o heir price elasiciy. Consisen wih he heory of upgrading oward more-sophisicaed producs, we documen changes wihin produc groups as well as shifs across produc groups oward less price elasic goods. In he nex secion, we show ha despie he lower real expor growh, China is sill rapidly gaining global expor marke share. However, he source of hese gains recenly shifed from volumes oward prices. 4 China s Expor Machine Coninues o Hum In addiion o an analysis of expor growh, an assessmen of Chinese expor performance can also be made from considering he evoluion of China s global expor marke share. Using informaion on bilaeral quarerly rade flows from he Expor Compeiiveness Daabase, we plo he share of oal Chinese expor sales relaive o global expor sales, i.e., is expor marke share (Figure 12). Apar from he sligh jump in 2015 and he dip in 2016, China has gained global expor marke share a a fairly consan rae. This suggess ha while Chinese expor growh slowed, China has coninued o ouperform is compeiors. In he following paragraphs, we show ha his increase in marke share is consisen wih our previous finding ha China upgraded is expor model. Figure 12: China's share of world expors Annual daa % China's expor share Source: Inernaional Moneary Fund via Haver Analyics Las observaion: 2016 We firs employ he shif-share analysis developed by Chepea, Gaulier and Zignago (2005) o shed ligh on he origins of Chinese marke share gains. This mehod allows us o disinguish beween secoral and geographical composiion of expors as well as price and quaniy effecs and rack heir evoluion over ime. To sar wih, we decompose he quarerly change in 14

19 Chinese expor marke share ino price (rade prices, P, are approximaed by changes in uni values) and quaniy (Q) effecs o capure heir respecive roles during he period from 2005 o The decomposiion akes he following form: d log X Chn, / 12 d log X / 12 = d logqchn, / 12 d logq / 12 + d log PChn, / 12 d log P / 12, (8) where he differences capure year-o-year Chinese expor growh (dlogxchn, /-12) relaive o global expor growh (dlogx /-12) on a quarerly basis. Figure 13 plos he log growh raes for all hree series. The analysis confirms ha before he 2008 crisis, changes in quaniies largely drove China s expors expansion, while Chinese prices generally declined relaive o hose of is compeiors. Excep for he crisis period in 2009, his rend coninued unil he end of Since 2012, hese dynamics have changed. The expansion of quaniies deceleraed and subsequenly urned negaive, and prices became he main conribuor o he marke share growh of Chinese expors. This resul is in line wih our previous finding ha China upgraded is produc mix oward less price sensiive producs and compeed on qualiy raher han price. Figure 13: Breakdown of changes in Chinese expor marke shares By price and quaniy, quarerly esimaes Percenage poins Change in quaniies Change in prices Change in expor marke shares -15 Source: Bank of Canada calculaions Las observaion: 2016Q2 To provide furher evidence ha he Chinese expor growh model shifed oward more sophisicaed goods, we apply a second shif-share analysis (Chepea, Gaulier and Zignago 2005) and decompose he changes in China s global expor marke share ino a China-specific effec, a desinaion-marke-specific effec and a produc/secor-specific effec. For noaional convenience, we denoe he exporing counry by i, he imporer by c, he produc by k and he ime by. Nex, we calculae he mid-poin growh rae 7 (gick) of he produc-specific bilaeral 7 The mid-erm growh rae is defined as g ick = (X ick - X ick-1) / (½(X ick + X ick-1)) where X ick is he value of expors of produc k beween i and c. 15

20 rade flow and pariion he variaion of his growh rae ino hree componens using he analysis of variance (ANOVA) mehodology. Wihin his mehod, we regress he mid-poin growh rae on hree ses of ime-varying fixed effecs. The firs se consiss of ime-varyingexporer fixed effecs (fi). The dummy variable, fi, equals one for every ime-exporer pair and zero oherwise. The second se consiss of a ime-varying-imporer fixed effec (fc), and he hird se consiss of a ime-varying secor/produc-fixed effec (fk). The las wo ses of dummy variables are defined in a similar manner o fi. We esimae he resuling regression by means of a weighed ordinary leas squares (OLS) esimaion: 8 g = α + ϕ f + β f + γ f + e. (9) ick i i c c k k ick The weighs are equal o he share of he expor value relaive o counry i s oal expor a ime, (sick). Nex, we normalize he esimaed fixed effecs (yi, βc, gk) o quanify hem as deviaions from he average growh raes of expors for he overall sample in he daa hrough a leas squares esimaion. 9 Based on his normalizaion, we can decompose he growh of counry i s expor marke share as follows: k sick gick = i + s ˆ ickβc + c s ˆ, ickγ k c, k c, k ϕ. (10) The firs erm of he righ-hand side of equaion 10 is equal o he exporer-specific fixed effec ( ϕ i ) and capures any effecs ha are specific o he exporer; i is orhogonal o imporing counries and secoral issues. This is our measure of Chinese-specific expor performance. Examples of such exporer-specific effecs are hose ha are common across all secors and independen of imporers, such as changes in he exporer s exchange rae, is aggregae produciviy shocks and he exporer s overall compeiiveness. Changes in his facor improve or worsen he compeiiveness of all goods expored by a counry China, for example. The second erm represens he imporer-specific fixed effec ( β ) weighed by he expor shares over all imporing counries (sick). This marke-specific facor is independen of he exporing counry and no secor specific. I capures, for example, aggregae demand shocks in he imporing counry. The marke-specific effec also provides a measuremen of he capaciy o direcly expor o desinaion markes wih an increasing impor demand. For example, if impor demand shifs from he euro area o he Unied Saes and China expors a large par of is expors o he Unied Saes, hen he marke-specific componen will play a bigger role in explaining Chinese expor marke share growh. c 8 A complee descripion of he daa se can be found a hps://mec.worldbank.org/. 9 See Chepea, Gaulier and Zignago (2005) for more deails. 16

21 Finally, he hird erm ( γ ) in equaion 10 capures he produc-specific effecs of expors. The k measure absracs from changes in overall impor demand and focuses on changes in he relaive imporance of producs in a counry s expor baske, for example, he share of less sophisicaed producs (exiles) relaive o more sophisicaed producs (compuers). The produc-specific componen will have a srong predicive power if China realigns is expor produc mix in line wih producs for which here is increased global demand or producs whose global prices are rising fases. If prices or demand rise srongly for highly sophisicaed producs, and China adjuss is expor baske s produc mix accordingly, hen he explanaory power of his componen will be high. Figure 14 plos his decomposiion of he changes in Chinese global expor marke shares. The average growh rae from he firs quarer of 2006 o he second quarer of 2016 was 4.6 per cen wih significan cyclical variaion afer he 2008 crisis as well as in he firs quarer of In addiion o he evoluion of China s global expor marke shares, he figure shows he decomposiion in he hree underlying facors as defined in equaion 10. Over he enire sample period, he bigges conribuor o Chinese growh in expor marke shares was he China-specific effec. Poenial facors responsible for ha srong expor growh were a reducion in he overall rade coss and srong overall produciviy growh. However, more recenly, he conribuion of he China-specific effec dropped markedly, while producspecific facors sared o play a much more prominen role. On he oher hand, he markespecific effec played only a minor role in he growh of Chinese expors. The large posiive conribuion of he produc-specific facor is in sark conras o he previous periods when he secorial composiion of Chinese expors was mainly a drag on expor marke share gains. To exclude he possibiliy ha his effec is enirely driven by demand, we furher decompose he produc facor ino a price and a quaniy componen. The decomposiion (see Figure 15) shows ha in he case of China he large posiive conribuion of produc-specific facors is enirely driven by higher prices, while he quaniy componen conribues negaively o expor marke share growh over he whole sample period. This suggess ha China gained expor marke shares because i shifed is expors oward producs wih fas-rising world prices and presens furher evidence ha China s realignmen of is expor produc composiion helped o suppor gains in global marke share. Absen hese changes, Chinese expor growh migh have been much slower. 17

22 Figure 14: Breakdown of changes in Chinese expor marke shares By specific facor, quarerly esimaes Percenage poins China-specific facor Produc-specific facor Marke-specific facor Change in expor marke shares Source: Bank of Canada calculaions Las observaion: 2016Q2 Figure 15: Breakdown of changes in Chinese produc-specific facors By price and quaniy, quarerly esimaes Percenage poins Change in quaniies Change in prices Change in produc-specific facor -10 Source: Bank of Canada calculaions Las observaion: 2016Q2 Overall, China coninues o gain global expor marke share a a fairly consan rae. The shifshare analysis highlighed wo imporan changes in he underlying sources of his growh. Firs, changes in expor prices now conribue more, compared o changes in quaniies. Second, produc specializaion is no longer a drag on expors and suppors expor growh. These wo developmens are consisen wih our argumen ha China s expor model shifed oward more-sophisicaed producs and ha his swich allowed China o gain marke share despie slower volume growh. 18

23 5 Conclusion We began his paper by using a sandard expor equaion o assess he dynamics of Chinese expor volumes. The equaion fis he daa raher well, and he coefficiens on he foreign demand and he real exchange rae boh had he expeced signs. However, we noe wo anomalies. Firs, he consan explains a large par of he expor growh. We believe ha his is an arifac of China s accession o he WTO and ha such srong rend growh is unlikely o persis. Second, he residuals are rended, wih he equaion over-predicing expors a he beginning of he period and under-predicing hem a he end. This suggess ha here have been srucural changes ha are no well capured by he sandard expor equaion. These srucural changes could be relaed o China s moving up he value chain, as China s expor mix has become less relian on lower-ech producs, such as clohing, foowear and oys, and more relian on higher-ech producs, such as compuers, elecommunicaions equipmen and elecronics. To assess how srucural changes affeced he growh of Chinese expor volumes, we calculae expor sophisicaion indices. We hen measure how far China has been from he sophisicaion fronier. We pu his gap (and is square) ino he sandard expor equaion. These appear o have been missing variables in our baseline regression. Excluding hem led o he large consan and he declining rend in he errors. I does appear puzzling ha a counry would choose o expor more sophisicaed goods if his led o slower expor volume growh. However, his phenomenon is consisen wih he insighs from he growh lieraure ha posis ha GDP growh will slow as a counry approaches he echnological fronier and he scope for cach-up is diminished. Moreover, he rade lieraure suggess ha he choice o expor more sophisicaed producs is an endogenous one ha depends on a counry s coss rising faser han is compeiors as i grows relaively more rapidly. One of he benefis of a counry s expors becoming more sophisicaed is ha i will produce more differeniaed producs over which i can exercise more pricing power and benefi from higher markups. To es wheher his occurred in China s case, we examine wo channels: changes in China s expor mix and changes wihin produc caegories. Here we focus on he US marke, where we can make use of disaggregaed impor, demand and price daa. Looking across 19 manufacuring indusries, we find ha increased sophisicaion is, indeed, associaed wih lower price elasiciies and increased pricing power. Looking wihin indusries, we find ha here is also some evidence of decreasing price inelasiciy since he crisis, suggesing ha China has made wihin-produc-line improvemens. 19

24 We find supporing evidence for he effec of srucural change on expor growh from wo decomposiions of China s global marke share. Decomposing changes in marke share ino price and volume effecs, we find ha, prior o he crisis, China s expor growh had been driven by volumes, bu depended more on increased prices hereafer. We hen go a sep furher and decompose he marke share changes ino a China-specific effec, a geographic marke effec and a produc effec. We find ha, over he enire sample, he China-specific effec was he bigges conribuor o expor growh. More recenly, however, he produc effec has become more imporan, providing evidence ha changes in expor composiion have suppored expor growh. In 2010, China became he world s bigges exporer of goods. In 2016, is share of global goods expors was 13 per cen, compared wih 9 per cen for he Unied Saes and 8 per cen for Germany, which were in second and hird place, respecively. I is unrealisic o suppose ha from his base, Chinese expors will again grow a he rapid pace recorded afer WTO accession. However, wha heir expors have los in growh dynamism has been, o some exen, offse by composiional dynamism in paricular, growing sophisicaion. Given China s disance from he fronier, i is reasonable o assume ha his dynamic has no ye fully played ou. 20

25 References: Barro, R. J. and X. Sala-i-Marin Convergence. Journal of Poliical Economy 100 (2): Chepea, A., G. Gaulier and S. Zignago World Trade Compeiiveness: A Disaggregaed View by Shif-Share Analysis. CEPII Working Paper No Francis, M., F. Painchaud and S. Morin Undersanding China's Long-Run Growh Process and Is Implicaions for Canada. Bank of Canada Review (Spring): Gaulier, G., J. Marin, I. Méjean and S. Zignago Inernaional Trade Price Indices. CEPII Working Paper No Hausmann, R., J. Hwang and D. Rodrik Wha You Expor Maers. Journal of Economic Growh 12: Kwan, C. H The Rise of China and Asia s Flying-Geese Paern of Economic Developmen: An Empirical Analysis Based on US Impor Saisics. The Research Insiue of Economy, Trade and Indusry Discussion Paper Series 02-E-009. Novy, D. 2013, Graviy Redux: Measuring Inernaional Trade Coss wih Panel Daa. Economic Inquiry 51 (1): Suon, J. and D. Trefler Capabiliies, Wealh, and Trade. Journal of Poliical Economy 124 (3):

26 Tables Table 1: Real expor growh regressions Variable Foreign demand REER Gap Gap-squared Trade coss Consan 1.45*** (0.153) -0.83*** (0.126) 12.58*** (0.741) 1.55*** (0.121) -0.21* (0.118) -4.15*** (0.419) *** (30.65) 1.45*** (0.114) -0.24** (0.109) *** (31.10) -1.11*** (0.215) *** ( ) 1.50*** (0.120) (0.162) *** (41.57) -1.18*** (0.285) (3.02) *** ( ) Observaions R-squared Adjused R-squared Noe: Chinese New Year dummies included in each regression. ***,**, and * denoe saisical significance a he 1%, 5% and 10% level, respecively. Sandard errors are shown in parenheses. 22

27 Table 2: Sophisicaion by indusry Indusry Sophisicaion Leaher & Allied Produc Mfg (316) 0.30 Texile Produc Mills (314) 0.37 Apparel Mfg (315) 0.40 Misc. Mfg (339) 0.51 Furniure (337) 0.53 Elec Eqp, Appliance & Componen Mfg (335) 0.57 Compuer & Elecronic Produc Mfg (334) 0.63 Non-Meallic Mineral Produc Mfg (327) 0.64 Peroleum & Coal Producs Mfg (324) 0.68 Texile Mills (313) 0.69 Fabricaed Meal Produc Mfg (332) 0.71 Plasics & Rubber Producs Mfg (326) 0.74 Primary Meal Mfg (331) 0.76 Prining & Relaed Suppor Aciviies (323) 0.78 Wood Produc Mfg (321) 0.86 Machinery Mfg (333) 0.86 Transporaion Equipmen Mfg (336) 0.87 Paper Mfg (322) 0.96 Chemical Mfg (325) 1.00 Noe: Indusries and hree-digi codes are from he Norh American Indusry Classificaion Sysem. 23

28 Table 3: Indusry regression coefficiens Indusry Consan (Eq.4) Relaive Prices (Eq.4) Foreign Demand (Eq.4) β3 in Eq.5 Texile Mills (313) 13.12*** -1.10*** 0.96*** Texile Produc Mills (314) 8.26*** -1.00*** 1.00*** Apparel Mfg (315) 7.12*** -2.02*** 1.23*** 0.91*** Leaher & Allied Produc Mfg (316) -0.85** -0.55*** 1.19*** Wood Produc Mfg (321) 7.58*** -0.84*** 0.70*** 0.27 Paper Mfg (322) 6.28*** -0.97*** 0.86*** 0.05 Prining & Relaed Suppor Aciviies (323) 7.97** -0.80*** 1.04*** 0.79*** Peroleum & Coal Producs Mfg (324) ** Chemical Mfg (325) 14.67*** 0.33** 0.82*** -1.02*** Plasics & Rubber Producs Mfg (326) 7.86*** -0.62*** 0.87*** -0.52*** Non-Meallic Mineral Produc Mfg (327) 4.38*** -0.63*** 1.07*** 0.36 Primary Meal Mfg (331) 7.75** 0.62* 1.98*** Fabricaed Meal Produc Mfg (332) 6.34*** -0.88*** 1.25*** 0.32 Machinery Mfg (333) 7.48*** -0.63*** 1.06*** -0.86** Compuer & Elecronic Produc Mfg (334) 6.87*** -0.63*** 1.39*** 0.18 Elec Eqp, Appliance & Comp. Mfg (335) 6.15*** -1.46*** 1.24*** 0.39 Transporaion Equipmen Mfg (336) 11.41*** -0.81*** 0.83*** 0.34 Furniure (337) 5.92*** -1.20*** 1.07*** 0.33 Misc. Mfg (339) 2.19** -1.00*** 1.35*** -0.85** Noe: ***,**, and * denoe saisical significance a he 1%, 5% and 10% level, respecively. Chemical Manufacuring and Primary Meal Manufacuring experienced a decline from a posiive coefficien on heir relaive prices in he pre-crisis period. 24