ECONOMICS CHINA AND GLOBAL MACROECONOMIC INTERDEPENDENCE

Transcription

1 ECONOMICS CHINA AND GLOBAL MACROECONOMIC INTERDEPENDENCE by Rod Tyers Busness School Unversty of Western Australa and Research School of Economcs Australan Natonal Unversty DISCUSSION PAPER 15.05

2 CHINA AND GLOBAL MACROECONOMIC INTERDEPENDENCE Rod TYERS * Revsed March 2015 DISCUSSION PAPER ABSTRACT Chna s transtonng toward more nward-focussed growth, causng adverse changes n the product and fnancal terms of trade n the advanced economes. At the same tme, nternatonal fnancal markets tussle between tghtenng forces assocated wth the US recovery on the one hand and unconventonal monetary expanson n Europe and Japan on the other. The way these shocks nteract s examned n ths paper usng a global macro model wth natonal portfolo rebalancng and asset dfferentaton and a representaton of unconventonal monetary polcy. Results are found to be senstve to the contrbutons of productvty and captal accumulaton to Chna s growth. When these are offered n realstc combnaton, the combned shocks are deflatonary n the US and Chna, mplyng that contractonary US monetary polcy s not mmnent. Monetary responses n the US and Chna then combne wth prce targetng regmes n the EU and Japan to expand lqudty globally, amplfyng mpacts on fnancal markets and the global dstrbuton of real nvestment. Key words: Global performance; US recovery; Chna s transton, macro nterdependence, fnancal ntegraton * Busness School, Unversty of Western Australa, and Research School of Economcs, Australan Natonal Unversty. Fundng for the research descrbed n ths paper s from Australan Research Councl Dscovery Grant No. DP Useful dscussons on the topc wth Dong He, Bert Hoffman, Song Lgang, Paul Luk, Da M, Peter Robertson, Wenlang Zhang and Peter Warr are acknowledged, along wth comments receved at semnars at the Chna Center for Economc Research, Pekng Unversty, the Hong Kong Insttute of Monetary Research and the Australan Natonal Unversty. Thanks for assstance wth data gatherng for ths research are due to Yng Zhang Tsun Se Cheong and Kazuk Tomoka.

3 1 Introducton Central to understandng the behavour of the global economy s the nteracton between the macroeconomc polcy regmes of the major economc regons, the US, the Western Europe and Japan, recently joned by Chna. These regons are all large n that the polces of each affect the others as a group as well as the world s many smaller economes. Ther behavour s therefore hghly nter-actve and strategc. The rse of Chna and other Asan, heretofore developng, economes snce the 1980s has not only underwrtten global economc performance but hgh East Asan savng rates have contrbuted to what became known as the Asan savngs glut. 1 Global real nterest rates peaked n the md-1980s and have fallen snce, n part because of ths relatve ncrease n global savngs supply, though more recently because of unconventonal monetary polcy, prncpally n the US and Japan (He and McCauley 2013, Arora et al. 2014). Ths pattern of nteracton wth Asa began to change, frst wth the slowdown n Japan n the 1990s but more recently wth the poor performance of the other large regons. It has become clear that the export-led growth model s unsustanable for Chna, for three man reasons. Frst ts lght manufacturng exports now loom large n global trade and there s nsuffcent market growth for ther expanson to contnue. Second, a demographc contracton s mmnent that wll eventually reduce the avalablty of surplus agrcultural workers, rasng labour costs and slowng productvty growth. And thrd, reforms that wll help contnue Chna s hgh growth rate must now venture nto the heavy manufacturng and servces sectors, whch are olgopolstc, and therefore not the vanguard of future growth (Aghon et al. 2013), and whch have, n any case, not been export orented. A turn nward s requred that s therefore poltcally dffcult. For these reasons slower growth s lkely, not only n Chna but also n those Asan economes that depend on ts market for manufacturng components. 2 Moreover, Asa s contrbuton to global savng wll also declne as reforms ensure that Chnese households are offered the choce to consume from more of ther corporate ncome and as populatons age, partcularly n Chna and Japan. 3 Ths change presents both opportuntes and dangers n the advanced economes. On the one hand, rsng Chnese consumpton could be one of the Keynesan stmul requred to rase demand and reduce unemployment queues n the advanced economes. On the other, t wll 1 See Bernanke (2005), Chnn and Ito (2007), Cho et al. (2008), Ito (2009) and Tyers (2015b). 2 For detals on the extent of producton fragmentaton and networkng n Asa, see Athukorala (2011). 3 For a comprehensve analyss of Asan household savng, see Horoka and Terada-Hagwara (2012), and for dscusson of Chna s hgh corporate savng see Kujs (2006), Tyers and Lu (2008). 1

4 place upward pressure on global nterest rates. Prvate portfolos that have tended to hoard money durng deflatonary tmes wll eventually rebalance and central banks wll need to soak up lqudty, ncludng by sheddng the non-tradtonal assets acqured va unconventonal monetary polcy (UMP). Ths wll rase nterest rates n both long and short maturty bond markets. Because the long nstruments are extensvely traded nternatonally, and held by Asan savers whose excess demand for them wll declne, the rse n fnancng costs could be very sharp. Ths comes at a tme when a recovery n the US economy has led to the declared abandonment of US UMP and when an expected (and actual) apprecaton of the US$ rases nvestment demand there. Meanwhle Japan and Europe are mantanng, and Europe wll bolster, the global lqudty flood. Low nterest rates n Europe and Japan, combned wth the US apprecaton, would suggest stll lower rates n the US, mltatng aganst a US tghtenng. In ths context t s possble that the change n Chna s growth regme, whch s unleashng of ts consumpton, could yeld the offsettng fnancal tghtenng force requred. Clearly, these many shocks pont n dfferent drectons, cloudng vson of the eventual trends n economc performance and fnancal behavour. To explore the nteracton of shocks emergng n the major economc regons, a parsmonous global macroeconomc model s ntroduced that ncorporates blateral lnkages across sx regons va both trade and fnancal flows. It ncludes a number of nnovatve elements. Frst, although t s determnstc, by allowng for asset dfferentaton t ncorporates optmsng fnancal portfolo management n each regon that serves to drect savng from each nto nvestments across all regons. Second, the degree of asset dfferentaton s quantfed to reflect fnancal ntegraton that dffers by regon. Thrd, long maturty assets are focal and unconventonal monetary polcy (UMP) places drect demands on the global markets for these assets that are endogenous to chosen targets. Ths tends to enhance the spll-over effects of monetary polcy (Chen et al. 2014), whch proves mportant because Chna s growth surge was deflatonary abroad, necesstatng monetary expanson despte slow growth, whle ts transton shock s, by contrast, nflatonary abroad. Smulaton results are found to be senstve to the contrbutons of productvty and captal accumulaton to Chna s growth n ways that end up consstent wth old assertons about the sources of Chnese growth (Krugman 1994) and about the effects of t on US employment (Krugman 2010, Rodrk 2010). When productvty and captal growth are offered n realstc combnaton, the collectve shocks are deflatonary n the US and Chna, mplyng that 2

5 contractonary US monetary polcy s not mmnent. Monetary responses n the US and Chna then combne wth prce targetng regmes n the EU and Japan to expand lqudty globally, amplfyng mpacts on fnancal markets and the global dstrbuton of real nvestment. In the end, however, current year real effects of the change to Chna s growth regme prove to be modest. In the secton to follow a revew s offered of changes n Chna s relatve sze and the assocated domestc transtons that presage ts reduced excess savng. Secton 3 then brefly dscusses recent macroeconomc developments n the advanced economes. The model used for quanttatve analyss s then presented n Secton 4 and smulaton results are descrbed n Secton 5. Conclusons are offered n Secton Chna s Transton and ts Fnancal Effects Abroad Chna s unadjusted share of global GDP s now greater than Japan s and ts shares of global exports, savng and nvestment are larger than those of the US and smlar to those of the EU (Table 1). The mplcatons of ths rapd expanson n output and trade for the large economc regons have been the subject of consderable debate. On the postve sde are global benefts n the form of a terms-of-trade mprovement (cheaper lght manufactured mports relatve to exported luxury and captal goods). On the negatve sde, s the relocaton of lght manufacturng from the advanced economes to Chna and an assocated contracton n the demand for low-skll labour. On macroeconomc polcy, there has been crtcsm of Chna, focusng on ts current account surplus and perceved exchange rate dstortons. Ths debate has centred on the growth surge perod between Chna s WTO accesson and the GFC. 4 But that controversal perod s now passed and Chna s n transton toward more nwardsourced growth that s tendng to unpck the growth surge shocks. The growth surge Durng ths perod consumpton fell short of output growth, causng substantal surplus producton and the consequent current account surplus lense llustrated n Fgure 1. As ncomes and wealth ncreased, outward captal controls prevented dversfcaton of prvate portfolos abroad and so the excess savng went abroad n the form of accumulatng offcal foregn reserves. The majorty of ths accumulaton was n long maturty US Treasury 4 The Amercan perspectve on the growth surge perod s offered by Bergsten et al. (2008) and Lardy (2006, 2012), whle the case aganst the currency manpulator charge s made by, amongst others, Tyers and Zhang (2011). The advanced resource economes lke Australa and Canada all enjoyed undsputed gans from Chna s rse, va ther raw materal exports. Asan neghbors were hurt by Chna s export competton n the early stages but helped n the later stages as Chna became a major destnaton for component exports (Athukorala 2011). 3

6 nstruments (Beltran et al. 2012)). Ths bolstered the contnung effects of the global refocussng of ncome generaton nto the hgh-savng Asan economes, whch had commenced n the 1980s and whch had already accelerated the rate at whch the global savngs supply curve was advancng. Wth slower growth n the advanced regons, global nvestment demand was comparatvely slow, causng a declne n the Wcksellan (1898) natural rate of nterest at the global level. 5 Fnancal mplcatons Consstent wth the segmentaton theory of the yeld curve (Johnson et al. 2010), long rates are not merely the commonly clamed (Boro and Dsyatat 2011) expectatonal extensons of short polcy rates. The transacton cost of fnancng long term nvestments va a successon of short contracts s prohbtve, allowng short and long maturty nstruments to trade at prces and yelds that dffer beyond what would be expected from tme preference and expectatons forces (Shller et al. 1983, He and McCauley 2013). Short bonds prmarly serve domestc fnancal sectors and are nstruments of conventonal domestc monetary polcy. Movements n ther yelds show ther clear lnks to regon-specfc busness cycles. Long bonds, by contrast, arbtrage wth the major nstruments of prvate savng and nvestment and are extensvely traded nternatonally. Ther yelds tend therefore to be more stable through tme than busness-cycle drven short yelds and reflect movements n the equlbrum between global savng and nvestment. He and McCauley (2013) fnd evdence of mperfect substtutablty along the yeld curve and use t to explore monetary polcy spll-over effects, whch they see as enlarged by the global ntegraton of long bond markets. Ito (2013: 8) offers support for ths vew. He argues that as a result of fnancal globalsaton, domestc fnancal markets are more susceptble to nternatonal factors and that tends to decouple short-term and long-term rates. Consstent wth Bernanke (2005), he concludes that the long-term nterest rate s ted down by global savng mbalances and hence reflects the natural rate of nterest. Ths reasonng, and that of Rey (2013), both mply that, when there s free captal moblty, there s nter-regonal arbtrage at the long end of the yeld curve whereas the short end of the yeld curve s conventonally controlled by monetary authortes. Support for ths nternatonal arbtrage at the long end, and for the weght of Chnese and other Asan savng, s suggested by the tme paths of the 5 Ex ante shfts n savng supply and nvestment demand cannot be observed. See Tyers (2015b) for a dscusson of the avalable data. 4

7 advanced regon long yelds n Fgure 2, whch are smoother through tme than short yelds and have trended downward. 6 The transton Chna s excess producton and savng durng the growth surge perod conferred on the advanced economes not only gans n ther product terms of trade but also n ther fnancal terms of trade. As the recent trends n Fgure 1 attest, however, these surpluses are now n contnuous declne. Indeed, the domestc focus on growth va lght manufacturng exports has stalled as Chna has come to domnate these ndustres at the global level and as the domestc supply of excess rural labour has begun to dry up. 7 There s therefore wde agreement that ts growth wll, and should, be ncreasngly underpnned by rsng home consumpton rather than exports. 8 Superfcally, t would seem that a swtch to domestcally drven growth should be possble, just by consumng more and exportng less. But ths smple dea s problematc because the export led growth strategy focusses producton on lght manufacturng, whle Chna s growng mddle class demands sophstcated manufactures and hgh qualty servces that nclude transport, telecommuncatons, health and educaton, all of whch are suppled by olgopolstc ndustres that accumulated consderable rents durng the growth surge and drected these nto corporate savng, bolsterng the current account surplus. 9 To dversfy Chna s output toward these products requres poltcally dffcult polcy reforms, not to menton contnued nvestment n human captal and nfrastructure. Whle the potental s there (Tyers 2014), and the anecdotal evdence suggests the reforms are occurrng, overall growth has slowed. Chna s new transtonal growth regme can therefore be charactersed as offerng reduced rates of productvty and captal accumulaton and contnung declnes n excess savng. 6 The separaton of the seres for Japan s assocated wth ts long term current account surplus and the major Yen apprecaton shocks of the late 1980s and early 1990s, whch establshed a negatve rsk premum amongst Japanese savers. In all regons, nflaton rates were low throughout the perod shown n the fgure and so the trend of nomnal long rates reflects that of correspondng real rates. For formal analyss of the ntegraton of these long bond markets see Arora et al. (2014). 7 The tmng of Chna s Lews turnng pont s a subject of controversy, as suggested by the contrasts between the vews expressed by: Ca (2010), Garnaut (2010) and Golley and Meng (2011), whch offer just a samplng of a substantal lterature. There s, however, lttle doubt that the turnng pont s on ts way, even f there s lttle agreement as to whether recent real wage rses suggest ts presence. 8 For the Chnese offcal lne on the rebalancng of ts economy, ncludng ts external accounts, see Wen (2007, 2011). 9 For analyss of the greater growth potental n compettve relatve to olgopolstc markets, see Aghon et al. (2013) and Aghon et al. (2014). Olgopoly rents n Chna are lnked to corporate savng (Kujs 2006, Song et al. 2011), last measured n the regon of a ffth of GDP. 5

8 3. Developments n the Advanced Regons Global fnancal markets contnue to be domnated by the US, whch has sustaned a structural current account defct snce the 1990s. Though much that has been unpalatable snce 2007 has been blamed on the GFC, the broad pattern of nternatonal fnances dd not appear to be permanently changed by t. Crtcally, t brought about a reverson by the prvate sectors n the US, the EU and Japan to net savng postons whle all three governments assumed net borrowng postons. 10 It therefore replaced prvate debt, some of whch had been unsustanable, wth soveregn debt, some of whch s also unsustanable, leavng heghtened global uncertanty as to soveregn fnancng. 11 Bond yelds and UMP A key change took place around 2005, before whch the large US defct had been fnanced by surpluses n Japan and the ol producng countres. Thereafter, however, the burden of ths fnancng rested ncreasngly wth Chna, whose dollar contrbuton became the sngle largest n 2005 and peaked n 2008 (Tyers et al. 2013). After 2010, however, Chna s relatve role as surplus fnancer began to dmnsh wth the shfts n ts domestc savng-nvestment balance already dscussed. It remans the world s largest sngle natonal buyer of US debt and equtes, however, hghlghtng the potental for dsrupton n US fnancal markets should Chna s excess savng contnue to declne (McKnnon 2013). As Fgure 2 ndcates, the downward trend n long yelds n all three advanced economes perssted beyond the peak of Chna s bond purchases, as central bank buyng took over wth expansonary monetary polces, ncludng the mplementaton of UMP n Japan, the US and the UK. For economes that have been comparatvely stagnant n real terms snce 2007, central bank asset holdngs rose mmensely, quadruplng n the US, almost quadruplng n the UK and doublng wth more to come n the Eurosystem and Japan (He and McCauley 2013). 12 These purchases rased the prces of long bonds and related nstruments and suppressed ther yelds. Unlke more conventonal monetary polcy, the hgh spll-over rate of UMP forces the domestc monetary cycle beyond natonal borders wth mmedacy. When mplemented at home t causes prvate fnancal outflows as nvestors seek out better yelds abroad and these 10 For analyss of ths see Tyers et al. (2013). 11 Accordng to the OECD Economc Outlook 89 Database, the soveregn debt stock problems of contnental European economes and Japan predated the GFC though they were exacerbated by t. The largest proportonal mpacts of the GFC were to the US and the UK. Even so, though these countres debt stock to GDP ratos became not sgnfcantly larger than that of France and are dwarfed by those of Portugal, Italy, Greece and Japan. 12 It s notable that Chna s central bank assets are large compared wth the others, due to reduced money creaton by Chna s commercal banks n response to hgh reserve to depost ratos. 6

9 apprecate foregn currences, threatenng deflaton and nducng expansonst responses. In part for ths reason, expansonary monetary polcy has spread across the three large blocs. 13 The US recovery and the lqudty flood Durng 2014, US-trggered tghtenng was commonly antcpated n global fnancal markets. Ths n turn was seen as beng lkely to place downward pressure on prces of stocks and other assets. The Fed resorted to UMP 14 on the expectaton that ths would rase the monetary base suffcently to tame deflatonary forces, to nduce a more healthy demand for productve assets and, ultmately, to reduce the rate of unemployment. In the thrd phase, enttled QE3, ths took the form of open-ended purchases at a rate of $85 bn per month delmted only by the prospect of the US unemployment rate eventually fallng below sx per cent. In md-2013 t was announced that ths buyng program would be tapered over the year to come. The effect of ths was to stmulate a prvate scramble out of long bond holdngs that saw ther yelds rse from below two per cent to three per cent n a few months. Ths was referred to as the taper tantrum. But the Fed bond purchases contnued and yelds stablsed agan. The announced taper actually began n December 2013 and, as the unemployment rate fell below sx per cent, t ended wth a fnal $15 bn purchase n October To the tme of wrtng there has been no further tantrum. The path of long term nterest rates has shown no tendency toward a lastng ncrease. How can purchases at such a hgh rate cease wthout causng the prce of these globally sgnfcant bonds to fall and hence ther yelds to rse? There are many possble explanatons. 15 Central are those relatng to the US recovery and the comparatvely poor performances of the European and Japanese economes. Hgher growth n the US, combned wth the declared end of US UMP, generates greater expected returns and ths attracts prvate fnancal nflows. These, n turn, tend to apprecate the US$, nducng central banks n other regons (especally n Chna and Japan) to rebalance ther portfolos toward US Treasures. The nomnal effectve US exchange rate has ndeed apprecated, by almost sx per cent n the second half of Whle concrete data on very recent foregn 13 UMP on the part of the US Federal Reserve was that the substtuton would be away from US bonds to US equtes. And ths has happened too. It s ths arbtrage that has sustaned contnued rses n market captalzaton snce the GFC. 14 Ths comprsed acqustons of a combnaton of long term Treasury bonds and mortgage backed securtes, the former domnatng n the early phases and the latter more recently 15 One clam s that the credblty of the US commtment to endng QE s doubtful, but the December 17 Federal Reserve Statement on Monetary Polcy ndcates that dssentng board members preferrng not to rule out a return to QE were out-voted by a clear majorty. Another s that, wth the gradual narrowng of the US fscal defct, the supply of long term Treasures has declned. Ths s a possble contrbutor, though US government fnancng requrements reman consderable. 7

10 purchases of US Treasures s not yet avalable, there are further reasons to expect such purchases to have ncreased recently. Chna s the largest sngle foregn holder of US Treasures and t has a clear nterest n sustanng ther value. In the case of Japan, also a large foregn holder, ts own QE has been proceedng strongly under Abenomcs, but t has caused Japanese short and long bond yelds to fall near the zero lower bound. Further monetary expanson requres the acquston by the Bank of Japan of other assets. Purchases of US Treasures meet ths requrement, wth the complementary beneft that they further deprecate the Yen. Fnally, there s the effect of the US real recovery on prvate US demand for Treasury bonds, and those assets that arbtrage closely wth them. The volume of US prvate savng s ncreasng and some f t s chasng such assets. Thus, whle the US has, at least temporarly, ceased contrbutng less to the global lqudty flood, the European Central Bank s commtted to UMP n 2015 and the Bank of Japan s showng no sgns of taperng ts correspondng program. These central banks are therefore takng up the slack and acqurng US assets n the process. Whle ever nterest rates outsde the US reman low for ths reason, a US recovery that causes nvestors to expect an apprecatng US$ mples that US nterest rates wll tend to declne rather than to ncrease. The US real recovery and the struggle n Europe and Japan to break out of stagnaton are therefore key global macroeconomc challenges. In what follows the changes n Chna s growth regme are consdered n ths context. The approach adopted s through the use of a global general equlbrum model. Ths approach also allows counterfactual analyss to separate out the roles of Chnese from other shocks. The most parsmonous model that wll allow such an analyss s presented n the secton to follow. 3. Modellng Macro Interdependence A mult-regon general equlbrum structure s used that centres on the global fnancal captal market. 16 It s assumed that the fnancal products of each regon are dfferentated and that portfolo managers assgn new net savng across regons so as to maxmse expected portfolo returns gven ths dfferentaton. Ths retans Feldsten-Horoka (1980) home bas whle allowng sgnfcant redrectons n fnancal flows at the margn. It also allows the level of 16 The model used s a more advanced varant on that used n Tyers (2015a). That model assumed a perfectly ntegrated global bond market and so tended to generate unrealstcally large spll-over effects. Here, all fnancal products, ncludng government bonds, are represented as regonally dfferentated and so there s no perfectly ntegrated global market for any asset class. Also, ths model ntroduces UMP explctly. 8

11 global fnancal market ntegraton to be parametersed by varyng the degree of dfferentaton. The scale of short run spll-over effects assocated wth growth performance, excess savng and monetary polcy therefore depend on t. Although there s a tendency for fnancal flows to move the global economy toward nterest party, n the length of run consdered asset dfferentaton leaves ths process ncomplete. At the same tme, regonal rates of return on equty nvestments depart from regonal bond yelds, the former reflectng expected rates of return on nstalled captal and the latter short run equlbrum n regonal fnancal markets between savers, ndebted governments and nvestors. Wthn each regon the demand for money s drven by a cash n advance constrant applyng across the whole of GDP. For any one household, home money s held n a portfolo wth long maturty bonds, whch are clams over physcal captal and government debt across the regons. 17 On the supply sde of the money market, n regons mplementng UMP, expansons rase demand for long maturty bonds, reducng ther yelds and hence reducng the opportunty cost of holdng money. 18 Sx regons are dentfed: the US, the EU, Japan, Chna, Australa and the Rest of the World, though the focus of ths paper s on the frst four. 19 Each regon supples a sngle product that s also dfferentated from the products of the other regons. On the supply sde, there are three prmary factors wth producton labour (L) a partally unemployed varable factor whle the stocks of physcal captal (K) and skll (S) are fxed and fully employed. Collectve households are net savers wth reduced form consumpton dependng on current and expected future dsposable ncome and the home nterest rate. Aggregate consumpton s subdvded va a sngle CES structure between the products of all the regons. The followng offers detal on the aspects of the model central to ths analyss. 20 Fnancal markets Here the modellng departs from conventon by ncorporatng explct portfolos of assets from all regons. Data on regonal savng and nvestment for 2011 s frst combned wth that on nternatonal fnancal flows to construct an ntal matrx to allocate total domestc savng n each regon to nvestment across all the regons. From ths s derved a correspondng matrx 17 Expectatons are exogenous n the model and are formed over future values of home nomnal dsposable ncome, the rate of nflaton, the real rate of return on home assets and blateral real exchange rates. 18 By contrast, conventonal monetary polcy nvolves trade n short term nstruments whch has no drect mpact on the market for long term bonds. 19 The EU s modeled as the full 26 and t s assumed that ths collectve has a sngle central bank. 20 The more routne components of the model are shared wth an earler verson that uses a somewhat less developed fnancal structure. See Tyers (2015a, Appendx 1). 9

12 of ntal shares of regon s net (prvate and government) savng that are allocated to the local savngs supply that fnances nvestment n regon j, S0 j. When the model s shocked, the new shares are calculated so as to favour nvestment n regons, j, wth comparatvely hgh expected after tax yelds, mplyng hgh expected real gross rates of return, r ce. (2) P MP ϕ r r e e P K 0 ce c e e = + ˆ = ˆ K + P ϕ, where P P s the prce of captal goods, whch n ths model s γ P, where γ s the prce of K captal goods relatve to that of output, P P, the producer prce of the regon s generc good. 21 The (exogenous) expected proportonal change n the real exchange rate s e ˆe. A further adjustment s made usng an nterest premum factor, ϕ, that s defned relatve to the US ( ϕ US = 1). Ths permts consderaton of the effects of changes n soveregn rsk n assocaton wth the fscal balance. Increments to regonal soveregn rsk cause nvestments n regons wth ncreasng fscal defcts to be less attractve. (3) 0 G GUS ϕ = ϕ, " "US", T TUS φ where φ s an elastcty ndcatng senstvty to soveregn rsk. In regon, then, the demand for nvestment fnancng depends on the rato of the expected real rate of return on nstalled captal, ce r and a domestc market clearng real bond yeld or fnancng rate, r. It s real nvestment that responds to these real rates but the model tracks nomnal nvestment expendture. For ths reason and adjustment s also needed for change n the prce of captal goods. D ce I P K r (4) = 0 0 I PK r I e, I where ε s a postve elastcty enablng the relatonshp to reflect Tobn s Q-lke behavour. Ths nvestment demand s then matched n each regon by a supply of savng that ncorporates contrbutons from all regonal households. 21 The producer prce level s the factory door prce of the regonal good, whch dffers n ths model from the GDP prce level, P Y, due to ndrect taxaton and from the consumer prce, P C, whch ncludes mported products. See Tyers (2015a: Appendx 1) for further explanaton of ths. 10

13 Regon s portfolo manager allocates the proporton S j of ts annual (prvate plus government) savng to new nvestments n regons j, such that S j = Because the newly ssued equty s dfferentated across regons based on un-modelled and unobserved regonspecfc propertes, ther servces are combned va a constant elastcty of substtuton (CES) functon specfc to each regonal portfolo manager. Thus, regon s household portfolo management problem s to choose the shares, S j, of ts prvate savng net of any government defct, S = S + T + T G, whch are to be allocated to the assets of regon j so as to D P D I maxmse a CES composte representng the value of the servces yelded by these assets: j (5) 1 ρ F D S ρ max U ( ) S = S aj j such that S = 1. j j j j Here αj s a parameter that ndcates the beneft to flow from regon s nvestment n regon j. The CES parameter, ρ, reflects the preparedness of regon s household to substtute between the assets t holds. To nduce rebalancng n response to changes n rates of return the α j are made dependent on ratos of after-tax yelds n destnaton regons, j, and the home regon,, va: 23 K rj τ j (6) αj = βj, j, λ > 0 K r τ λ. Here, K τ s the power of the capta ncome tax rate n regon. Ths relatonshp ndcates the responsveness of portfolo preferences to yelds, va the (return chasng) elastcty λ. The allocaton problem, thus augmented, s: (7) K λ rj τ F D j S r max U ( ) S = S βj K j such that j j r τ S = 1. j j 1 r Solvng for the frst order condtons we have, for regon s nvestments n regons j and k: 22 The manager does not re-optmse over total holdngs every year. Ths s because the model s determnstc and rsk s ncorporated only va exogenous prema. The motvatons for contnuous short run rebalancng, other than the arrval of new savng, are therefore not represented. 23 Note that regon s market bond yeld, r, s determned concurrently and ndcates the replacement cost of captal n regon and therefore the opportunty cost for regon s household of nvestment n regon j. 11

14 (8) 1 S 1 K j βj r rj τ j S K k βk rk τk + = λ 1+ r. Ths reveals that regon s elastcty of substtuton between the bonds of dfferent regons s ( ) I σ = λ 1+ ρ > 0, whch has two elements. The return-chasng behavour of regon s household ( λ ) and the mperfect substtutablty of regonal bonds, and therefore the sluggshness of portfolo rebalancng ( ρ ). For the purposes of ths analyss the values of are seen as ndcatng the extent of each regon s ntegraton wth global fnancal markets. The optmal share of the net domestc savng of regon that s allocated to assets n regon j then follows from (8) and the normalsaton condton, that S k = 1: k I σ (9) S j = 1 I σ λ K βk rk τk K k βj rj τ j σ I. The key matrx for calbraton s β j. These elements are readly avalable, frst, by notng that only relatve values are requred and hence, for each regon of orgn,, one value can be set to unty, and second, by makng the assumpton that the ntal database has the steady state property that the net rates of return n regons j are ntally the same as the market bond yeld, r j. Then, snce n the base data e0 0 e0 0 rj = rj, rk = rk, the βj s are avalable from a modfed (6): Regonal money market equlbrum A cash-n-advance constrant s assumed to generate transactons demand for home money across all components of GDP. The opportunty cost of holdng home money s set at the nomnal after-tax yeld on home long term bonds. 24 Real money balances are measured n terms of purchasng power over home products. (10) m a ( y ) ( + ) MR e e S B MY D MD e r 1 π M ηm = = = K Y Y τ P P. 24 Thus, t s assumed here that the opportunty cost of holdng money s measured by the long bond yeld, whch s the domnant determnant of non-money portfolo yelds. Short rates, at least as they have a role n conventonal monetary polcy, are here embedded n the determnaton of the monetary base. Whle housng nvestment can be senstve to short rates n economes where most mortgage contracts have varable rates, the assumpton that nvestment fnancng depends on the long maturty market s accurate n a comparatve sense and t s a useful abstracton. 12

15 Here y s real regonal GDP, P Y e s the GDP prce and π s the expected nflaton rate of the consumer prce level, P C, defned as a CES aggregate of home and mported consumer prces. 25 The money multpler s η and B M s the monetary base. 26 The monetary base, B M, can be set as an exogenous polcy varable or as endogenous to a prce level or exchange rate target. Regonal fnancal market clearance The home fnancal market n each regon clears separately. For regon, the nomnal value of domestc nvestment, D I, represents the sum total of all domestc long bond ssues. Ths s then equated wth demand for those bonds from home and foregn (net prvate and government) savngs, along wth demands for home long bonds that arse from the quanttatve easng components of monetary expansons by both home and foregn central banks. Total nvestment spendng n regon, n s local currency, s then: (11) E = + θ D, D S D QE QE B j I S j j j sj M j, j E where E s the nomnal exchange rate of regon relatve to the US$, whch s the numerare n the model (E US =1). The quanttatve easng component of the current perod s expanson of QE the monetary base by regon j s central bank, s j, and the share of ths expanson that takes the QE form of acqustons of regon s long bonds, θ j, both determne central bank demand. These flows are orgnally n foregn currency and are therefore converted at the approprate cross rates. The regonal real bond yelds (nterest rates, r j ) emerge from ths equalty. Ther I convergence across regons s larger the larger are the elastctes of asset substtuton, σ j. Balance of payments The sum of net nflows of payments on the current account and net nflows on the captal and fnancal accounts, measured n a sngle (home) currency s zero: E X M + S + s M S + s M = 0, " "US" j E j (12) S D QE QE B j ( S D QE QE B j j θ j j D j j θj D ) 25 For detals of the specfcaton of consumpton and prce aggregates, see Tyers (2015a, Appendx 1). 26 In ths study the money multpler s held constant. In applcatons to fnancal shocks, however, t falls as confdence n fnancal nsttutons declnes and those nsttutons hold more lqud reserves. 13

16 The second two terms n (12) are fnancal nflows and outflows. The frst parenthessed term represents acqustons of regon s home-ssued long bonds by foregn savers and by foregn central banks, the latter assocated, as above, wth the quanttatve easng component of the current perod s expanson of the monetary base n each regon. These net savng and central bank flows are orgnally n foregn currency and so are converted at the approprate cross rates. The second represents acqustons of foregn-ssued long bonds by regon s home savers and ts own central bank. 27 A balance of payments n the US s mpled by balance n all the other regons. These equatons determne the nomnal exchange rates and, snce these are defned relatve to the US$, that for the US s always unty( E = 1). Model database, parameters and operaton The model database s bult on natonal accounts as well as nternatonal trade and fnancal data for the global economy n As ndcated prevously, the scale of the Chnese and advanced economes s ndcated n Table 1. Of partcular nterest here are the fnancal flows between these regons, the pattern of whch s suggested n Table 2. It s noteworthy that, whle the share of Chna s trade n goods s larger wth Europe than the US, the share of the US n ts blateral fnancal flows s comparatvely hgh whle that of Europe s low. 28 These dfferences prove mportant n the results that emerge from the modellng. US 5. Analyss The model s frst appled to the nternatonal sgnfcance of the sources of Chna s growth. Ths ssue stems from the debates over whether Chna s growth has prmarly depended on captal accumulaton on the one hand or productvty on the other and the assocated debate on the extent to whch ts growth has dsplaced employment n the advanced economes. It s confrmed that the sources of growth matter for nternatonal mpacts but that the results are not as smple as some of the rhetorc. The second applcaton s to the magntudes of the nternatonal mpacts of Chna s growth regmes n the context of other landmark changes to the global economy, namely the US real recovery and the contnung European and Japanese monetary expansons. 27 The nomnal values of mports and exports, M and X, are formulated n Tyers (2015a, Appendx 1). 28 Further detals as to the sources and constructon of the database can be obtaned from Tyers (2015a, Appendx 2). 14

17 5.1 Implcatons of the source of Chna s growth Krugman (1994) s amongst those who have, at varous tmes, clamed that Chna s growth has been drven by captal accumulaton rather than productvty growth. At the same tme, many have clamed that the global economc ntegraton that has facltated Chna s growth has facltated job-stealng from the advanced economes. 29 Superfcally at least, job stealng would seem more lkely to stem from productvty shfts n Chna because t adds effectve workers to the global economy, drvng down the global wage or the employment level outsde Chna. Alternatvely, t can be seen as suckng n nvestment and thereby reducng growthenhancng nvestment expendture n the advanced regons. A growth path drven by captal accumulaton only, however, would reduce Chna s rate of return and drve nvestment back to the US, rasng demand there and so beneftng US workers. The smulaton results dffer from these smple scenaros because of the dfferent real exchange rate and monetary effects of these shocks. The labour market and fscal polcy closures adopted for ths purpose are gven n Table 3. To make t clear whether the effects of Chnese growth are nflatonary or deflatonary abroad, the monetary target s assumed to be the monetary base n all regons. The results are summarsed n Table 4. Productvty drven growth: effects on Chna Ths does drect nvestment to Chna, where the new nvestment demand drves up real nterest rates globally. By tself, the reallocaton of nvestment globally would cause a real apprecaton n Chna but here t does not because of the prmary effect of the productvty shock n reducng home costs. The result s a modest real deprecaton. In Chna s money market, where no new prce level changes are antcpated, the hgher nterest rate tends to reduce money demand but ths s more than offset by growth n transactons demand from the productvty gan. The net effect, whle the monetary base s held fxed, s a deflaton. The combnaton of a real apprecaton wth the domestc deflaton requres a nomnal apprecaton. Whle the deflaton creates a drag on employment growth, the productvty change and the new nvestment ensure that there s a net real expanson, both n terms of real GDP and the real purchasng power of ncome. 29 See, for example, Rodrk (2010), Helpmann et al (2010), Haskell et al. (2012), Autor et al. (2013) and D Govann et al. (2013). Krugman (2010) declares, n a blog, that Chna s makng all of us poorer. 15

18 Chnese productvty drven growth: effects on the US The outflow of nvestment nto Chna from the US s larger than that from ether the EU or Japan and so the US experences a real apprecaton relatve to Chna but real deprecatons relatve to the EU and Japan. Global fnancal ntegraton ensures that ts real nterest rate also rses and, wthout any real shock to drve transactons demand for money, ths causes a modest nflaton. The correspondngly modest real exchange rate change, combned wth the domestc nflaton, mples a nomnal deprecaton. Wth the nflaton comes ncreased employment and rses n both real GDP and the home purchasng power of US ncome. 30 Nonetheless, the loss of US nvestment must eventually slow growth there. Captal drven growth: effects on Chna Consstent wth the smple scenaro outlned earler, more Chnese captal reduces ts rate of return and ths drects nvestment out of Chna, tendng to deprecate ts real exchange rate. Importantly, however, the ncreased captal stock reduces Chnese producton costs. These two effects here renforce one another and so the Chnese real deprecaton s larger than n the case of productvty drven growth. Because Chnese savng s reallocated abroad where expected rates of return are now hgher, real nterest rates tend to fall both at home and abroad. 31 The home real nterest rate declne boosts money demand and ths s renforced by the rse n transactons demand due to the captal-drven rse n output and ncome. Wth fxed money supply ths causes a larger deflaton than occurs n the productvty shock case. But the real deprecaton s larger too, suffcent to accommodate both the deflaton and a nomnal deprecaton. The deflaton reduces labour demand, placng a drag on real output, whch stll rses because of the captal stock expanson. The real purchasng power of Chnese ncome falls, however. Ths s because Chna s consumer prces deflate by less than ts GDP prce snce, at the reduced nomnal exchange rate, the cost of purchasng foregn goods holds up. Chnese captal drven growth: effects on the US In ths case there s an nflow of nvestment from Chna, whch causes a real apprecaton relatve to the rest of the world. The global nterest rate falls wth the excess supply of Chnese savng. 32 In the US money market there s no major shft n transactons demand and so, wth a fxed money stock, ths causes a deflaton. Ths, combned wth the real apprecaton, mples a 30 These effects suggest a monetary tghtenng response, whch would elmnate the nomnal deprecaton and the domestc nflaton, though ths would also elmnate the employment and real GDP gans. 31 Ths s wth the sngle excepton of the EU, whch behaves dfferently because t has less captal market nteracton wth Chna than ether the US or Japan. 32 Agan, the EU s an excepton here because of ts smaller fnancal exchanges wth Chna. 16

19 nomnal apprecaton whch s moderated by the fact that the US deflaton tends to be smaller than n other regons. The deflaton, n turn, reduces labour demand and contracts real GDP. 33 At the same tme the real apprecaton ensures that the real purchasng power of US ncome rses and, n any case, US growth s a benefcary of ncreased nvestment. Reprse on growth sources The results suggest that the US would gan employment n the short run, though lose nvestment, n the event that Chna s growth s productvty drven. On the other hand, t would lose employment n the short run, but gan nvestment, were Chna s growth to be drven by captal accumulaton. These conclusons cut aganst the smple logc that follows the flow of nvestment but gnores monetary and exchange rate effects. But they te together the early assertons that Chna s growth was captal drven and that t caused labour dsplacement n the US and other advanced regons. 5.2 The effects of Chnese growth n the context of other global macroeconomc change At the tme of wrtng the most mportant change n the global macro-economy s the real recovery n the US. Long awated snce the GFC, ths recovery leads the older advanced economes and t has brought an uptake n employment combned wth GDP growth at ts strongest level for a decade. Here ths s represented by three shocks: to total factor productvty, to the relatve cost of captal goods and to the scale of UMP. These shocks, lsted n Table 5, are not founded on emprcs but they are suffcent to yeld a representatve real US growth rate of between three and four per cent. In Europe and Japan the most sgnfcant macro-economc developments are upscale monetary expansons va UMP. These are seen to address deflatonary expectatons and are desgned to yeld two per cent consumer prce nflaton. Agan, these shocks, whch are also lsted n Table 5, are not emprcally based but they are structured to delver modest but postve real growth. In ths context are placed the two alternatve Chnese growth regmes dscussed earler: the growth surge regme that charactersed the decade followng Chna s accesson to the WTO and the transton. The growth surge accompaned a combnaton of captal and skll accumulaton along wth productvty mprovements brought about by the relocaton of vast numbers of rural workers nto urban employment n manufacturng and servces. Of course, the most sgnfcant characterstc of ths regme for the rest of the world was ts comparatvely 33 Of course, ths would be readly offset by a modest monetary expanson. 17

20 modest growth n consumpton and the rse n Chna s current account surplus. Ths combnaton of shocks s constructed from the hstorcal sources descrbed n Tyers (2014) and s represented as the thrd bundle n Table The transton represents the post-surge pattern of Chna s growth, whch combnes slower expanson of captal, skll and productvty, combned wth some upward nomnal wage pressure assocated wth the slowdown n rural to urban mgraton. Most sgnfcantly, however, t ncludes a recovery n consumpton growth. 35 In each case the closures chosen for labour markets and fscal polcy are as ndcated n Table 3 and those for monetary polcy are lsted at the base of Table 5. In the frst nstance at least, the US, EU and Japan are assumed to be constraned as to monetary expansons by the conventonal lower bound. Chna s monetary polcy s enslaved to a US$ peg n the growth surge regme but t s assumed to be s more expansonary durng the transton, targetng the level of employment. In general, although Chna has become macro-economcally large, the US s stll almost twce as large and t has an open and well ntegrated fnancal market. Its recovery shocks have substantal real mpacts that are strongest at home. Yet t emerges that the Chnese shocks are roughly as strong at home and almost as strong abroad. By comparson, the monetary expansons n the EU and Japan are nomnal shocks offsettng deflatonary expectatons and so have comparatvely small mpacts both at home and abroad. Ths can be seen, ntally, from the unlateral mplementaton of these shocks, one regon at a tme, wth the monetary closures also lsted n Table 5. Unlateral shocks US recovery: These stmulate US nvestment, shftng out the global nvestment demand curve and so tghtenng global fnancal markets. Wth no monetary accommodaton n the US these shocks are deflatonary there because they rase the supply of goods relatve to the stock of money. In the EU and Japan, however, they are nflatonary, because hgher nterest rates release lqudty and because the nflow of fnance to the US causes ther exchange rates to deprecate aganst the US$. In Chna the effect s deflatonary and contractonary because the assumed US$ peg forces a monetary contracton. EU Japan monetary expanson shocks: the scale of these depends on the deflatonary expectatons they are mplemented to offset and the szes ndcated n Table 5 are merely 34 The constructon of the shocks depends mportantly on the meta-analyss by Wu (2011). 35 Agan, ths cocktal of shocks s as compled n Tyers (2014). 18

21 llustratve. Importantly, the monetary expansons are actng aganst deflatonary expectaton shocks, and so the global effects are more modest than otherwse. 36 Indeed, the nternatonal effects of the expansons are not large n scale compared wth the effects of the US recovery. They tend to ease nternatonal fnancal markets and, whle they yeld the ntended nflaton at home, they cause deflaton n the other markets because the lower rates they lead to rase money demand and ther exchange rates deprecate wth the greater abundance of ther currences. In Chna, the US$ peg, combned wth the effectve apprecaton of the US$, forces a monetary contracton that s deflatonary. Overall, the EU-Japan expanson causes a sheddng of nvestable funds from those regons, retardng future growth, but the nflaton t causes at home does rase employment and GDP n the short run. Chna surge: So far as ts nternatonal effects are concerned the growth surge s domnated by the surge n excess savng. Ths causes an excess supply of product to enter global markets and a compressng effect on real global bond yelds. 37 In combnaton, these are deflatonary n the advanced economes, whch mantan fxed money supples n ths smulaton, prmarly because the lower rates soak up lqudty. The excess savng fnances ncreased real nvestment n all regons but the deflaton n the advanced economes contracts employment and real GDP. Substantal real apprecatons aganst Chna, however, ensure that the purchasng power of home nomnal ncomes rses n all the advanced regons. Chna transton: Here the Chnese government s assumed to dscard the US$ peg and to adopt monetary polcy that targets home employment, durng what would otherwse be a contractonary transton, notwthstandng contnung growth n productvty, captal and skll. Increased Chnese consumpton and decreased excess savng tghtens global fnancal markets. As the source of ths change, Chna s yeld rses by more than those n the advance regons. Whle nvestment falls n the advanced regons as a consequence, t rses n Chna because of the more than offsettng effects of the productvty and skll supply shocks (Table 5), that rase Chna s expected rate of return. The hgher yelds n the advanced economes release lqudty and are nflatonary, rasng employment and real GDP. The real purchasng power of ncome n these regons s mxed n drecton, wth ncreases n the EU and Japan but a margnal declne n the US, prmarly because the EU and Japan experence larger real apprecatons aganst Chna. 36 In ths comparatve statc analyss these expectatonal shocks are not subsequently muted. 37 The effect on the EU s bond yeld s smaller because t has lttle fnancal exchange wth Chna, at least as suggested by data on fnancal flows n