Explaining the Slowdown in Global Trade

Transcription

1 Explaining he Slowdown in Global Trade Jarkko Jääskelä and Thomas Mahews* Following he global financial crisis, global rade conraced sharply and, afer an iniial recovery, grew a an unusually slow pace relaive o global GDP. This aricle reviews cyclical and srucural explanaions for his phenomenon, and finds economeric evidence ha cyclical facors namely shifs in he composiion of aggregae demand oward less impor-inensive componens and heighened economic uncerainy can explain mos of he slowdown in rade in a panel of advanced economies. Alhough he slowdown in aggregae global rade is well explained by he model used, he resuls vary by counry. For Ausralia, he model performed well unil 212, afer which i over-prediced impor growh, mos likely because i did no adequaely capure he impor inensiy of mining invesmen and resource expors. Inroducion Before he global financial crisis, inernaional rade had been growing a abou 1½ imes he rae of global GDP for around four decades, resuling in a subsanial increase in he global rade-o-gdp raio (Graph 1). The increase in rade over his period is widely aribued o srucural facors, including subsanial declines in he cos of rade, expansions of global supply chains and he increased imporance of emerging economies in inernaional rade (WTO 213). Trade as a share of GDP increased dramaically over his period for emerging economies, like hose in eas Asia. In conras, rade as a share of GDP in he Unied Saes and oher advanced economies increased a a slower pace and heir share of global merchandise expors has declined since he early 199s (Graph 2). Since he financial crisis, however, growh in global rade has no resumed is earlier pace. Growh rebounded srongly in 21, following widespread recession, bu real global rade has subsequenly grown a an average annual rae of around 2 per cen, well below he pre-crisis average of 6½ per cen. Susained weakness in economic growh, especially in advanced economies, can help o explain he * The auhors are from Economic Group and Inernaional Deparmen Graph 1 Global Trade* PercenofglobalGDP * Average of impors and expors; dashed line shows increase a pre-crisis average rae Sources: OECD; RBA; World Bank slowdown in rade. Bu over he pas four years rade has grown a an even slower rae han GDP, resuling in he raio of rade o GDP falling a lile over his period. This has led o some suggesions ha he slowdown in rade migh be of a more permanen, or srucural, naure. This aricle reviews possible srucural and cyclical facors behind he recen slowdown in global rade and finds evidence ha a large par of he slowdown in rade can be explained by cyclical facors. This suggess ha if growh in global business BULLETIN SEPTEMBER QUA R T E R

2 Graph 2 OECD Counries Share of Global Expors Source: World Bank invesmen, which is he mos impor-inensive componen of demand, were o pick up o pre-crisis levels and he general economic uncerainy were o dissipae, growh in rade would also be likely o increase again. Srucural Explanaions for he Slowdown One possible explanaion for he recen slowdown in global rade is ha srucural policies and facors ha had boosed rade, such as rade liberalisaion and supply-chain expansion, have largely run heir course. In suppor of his explanaion, Ferranino and Taglioni (214) noe ha rade in complex goods ypically associaed wih supply chains has grown paricularly slowly afer he financial crisis. Consaninescu, Maoo and Rua (215) argue ha rade has become less responsive o changes in income since he early 2s, perhaps because of a slowing in he pace a which supply chains are inernaionalising. Neverheless, i may be oo early o conclude ha srucural expansion in rade is over. Global supply chains are sill a relaively recen phenomenon and increasing demand for consumer durables from emerging economies is likely o boos he imporance of he large-scale, efficien producion ha hey afford. There is also poenial for furher rade liberalisaion. According o he World Trade Organizaion, he average good impored ino he Unied Saes or he euro area sill faces around a 1½ per cen ariff and he average global impor faces a ariff of around 3 per cen (Graph 3). Furhermore, oher economically significan non-ariff barriers, such as ani-dumping and buy local legislaion on governmen purchases, remain in place in many counries (Evene 214) Graph 3 Impor Tariffs* Per cen of impor values European Union * Weighed mean applied rae Source: World Bank US World Cyclical Explanaions for he Slowdown Cyclical facors, such as he composiion of he recovery in global GDP growh and lingering economic uncerainy following he crisis, could also help o explain he slowdown in rade growh relaive o GDP growh. Domesic demand is an imporan driver of impors in mos economies. Business invesmen is usually he mos rade-inensive componen of domesic demand because firms ofen require specialised capial goods no available locally and ypically have more direc inernaional access o purchase goods from overseas han do households (see, for example, Bems, Johnson and Yi (213); Bussière e al (213)). As a resul, he coninuing weakness in business invesmen, relaive o oher componens of demand, is likely o have slowed growh in global rade in he pos-crisis period (Graph 4) RESERVE BANK OF AUSTRALIA

3 15 Graph 4 Advanced Economies GDP componens* sd dev Governmen consumpion Invesmen** March25= GDP * Aggregaed a fixed 211 PPP exchange raes Privae consumpion ** Aggregae privae and public gross fixed capial formaion Sources: OECD; RBA Graph 5 Economic Uncerainy Trend, sandard deviaion from average Measures of economic uncerainy have been elevaed since he financial crisis, reflecing, in par, difficulies in forecasing he iming and pace of economic recovery, and his may also have weighed on rade (Graph 5). Novy and Taylor (214) argue ha a large par of he collapse in rade during he financial crisis can be aribued o uncerainy. They showed ha an increase in uncerainy alered firms invenory policies, and if firms reduced foreign orders by more han domesic orders, his could lead o a bigger conracion in inernaional rade flows han in domesic economic aciviy. Uncerainy can also affec rade by resricing credi growh, o which Japan Sources: Economic Policy Uncerainy; RBA US Europe 95 sd dev rade appears o be paricularly sensiive (Chor and Manova 212). Esimaing he Conribuion of Cyclical Facors o he Slowdown An economeric model can be used o quanify he conribuion of boh he weak recovery in he pos-crisis period in he componens of aggregae demand ha have high impor inensiy and he lingering economic uncerainy during his period. The impor demand model of Bussière e al (213) is augmened wih a measure of economic uncerainy from Baker e al (215) for his purpose. 1 Bussière e al s model relaes impors o relaive prices and a measure of aggregae demand ha weighs he expendiure componens according o heir impor inensiies. This impor-adjused demand measure (IAD), has he form: IAD = c g i x c, g, i, x, j, j, j, j, j, (1) where c, g, i and x represen he expendiure componens of GDP (privae consumpion, governmen consumpion, invesmen and expors, respecively) of counry j while ω represens he impor inensiy of each componen. 2 For example, in 25, he impor conen of Ausralian invesmen was 26 per cen, a bi higher han ha of privae consumpion a 18 per cen and expors 1 Where available, he uncerainy measures consruced by Baker e al (215) (available a < are used. For European economies wihou a direc uncerainy measure, he aggregae European measure consruced by Baker e al is used. For non-european counries wihou uncerainy indices, uncerainy is proxied by a GDP-weighed average of he available uncerainy measures, excep for Ausralia where he measure consruced in Moore (forhcoming, closely following Baker e al (215)) is used. The counry-specific uncerainy measures are relaively highly correlaed, so he weighed average is likely o be a reasonable approximaion for he counries ha have no individual measure available. However, he resuls are robus o including only hose counries for which a direc uncerainy measure is available. 2 Derived from he OECD inpu-oupu (I/O) ables; he weighs are ime-varying and scaled o sum o one in each period. The I/O ables have hree vinages for mos counries: 1995, 2 and 25. The weighs are inerpolaed linearly beween hese poins. BULLETIN SEPTEMBER QUA R T E R

4 a 14 per cen. The impor conen of governmen consumpion was relaively low a 1 per cen. Bussière e al find ha he impor-adjused demand measure explains he fall in rade volumes during he crisis beer han oher, more radiional, demand measures, such as domesic demand or GDP. This aricle re-esimaes Bussière e al s model, updaing he sample period o he firs quarer of 215 and adding economic uncerainy as an explanaory variable, and examines is performance during he pos-crisis slowdown. Specifically, he following equaions for impors are esimaed for a panel of 18 advanced economies, using daa from he OECD s harmonised naional accouns daabase. 3 In he models, he quarerly growh of real impors ( ln m j, ) depends on he quarerly growh raes of aggregae demand ( ln D j, ); relaive impor prices ( ln RPM j, ); uncerainy ( ln UNC j, ); and ime-invarian counry-specific facors (FE j ). IAD, domesic demand and GDP are alernaively used as he measure of aggregae demand ( D j, ). lnm lnd j, 1 2 j, lnrpm lnunc 3 j, 4 j, FE,. j j (2) The economeric resuls sugges ha he relaive slowdown in aggregae global rade growh can be largely explained by cyclical facors. The parameers are saisically significan (wih one excepion) and ake heir expeced signs: higher demand has a posiive effec on impor growh, while higher impor prices and an increase in uncerainy end o have a negaive effec on impors (see Table A1). 4 3 Only OECD counries are included in he sample, hereby excluding several rade-inensive emerging economies, mos noably China. Quarerly naional accouns daa by componen are unavailable for China. Including oher emerging Asian economies in he sample was no found o maerially affec he resuls. 4 Model specificaions wih lags were also esimaed; hese were found o improve he fi marginally bu did no significanly change he poin esimaes of he coefficiens of ineres. Including ime-varying fixed effecs (assumed o average zero over he sample period) is anoher way o esimae how cyclical facors affec rade. These parameers are esimaed o be negaive over mos of he pos-crisis period, consisen wih he resuls presened below ha he rade slowdown is largely cyclical. The models using he IAD measure of domesic demand perform noiceably beer over he full sample (1985:Q2 215:Q1) han he models using he convenional demand variables. The IAD model ha includes a measure of economic uncerainy has he mos explanaory power. 5 To analyse he performance of hese models over he recen period, hey are re-esimaed using daa up o he end of 21, so ha he models observe he rade collapse during he financial crisis and he subsequen rebound, bu no he recen slowdown in global rade growh relaive o GDP growh. The esimaed model coefficiens are hen used o projec wha he models would have forecas growh o be over he pas four years, given he acual movemens in he various demand and uncerainy indicaors. The resuling projecions for he level of impor volumes from he IAD model, wih and wihou uncerainy, are shown in Graph 6, along wih he acual impor volumes. The fied values from he models prior o 211 are also shown. Noably, he exen of he decline in rade in 29 is no enirely explained by he models (alhough hey do much beer han he models wih radiional domesic demand measures, which are no shown). This is hardly surprising, given he exraordinary circumsances of he financial crisis. The difference beween he dashed line and he blue line on Graph 6 shows ha around half of he slowdown in rade growh from is average pre-crisis rae can be explained by he imporadjused demand model on is own; his resul is consisen wih oher work looking a demand-side facors such as Boz, Bussière and Marsilli (214) and Morel (215). Almos all of he remaining slowdown, paricularly over he earlier sages of he pos-crisis period slowdown, is explained when measures of uncerainy are included in he model. The IAD model wih uncerainy suggess ha uncerainy was associaed wih a reducion in 5 Using he pos-crisis sample period only, he relaive performance of differen specificaions does no change, and he poin esimaes for he coefficiens are similar. However, he parameer esimaes are less precise owing o he shorer sample period. 42 RESERVE BANK OF AUSTRALIA

5 Graph 6 Graph 7 Advanced Economies Real Impors* March211= Pre-crisis average growh Seleced Economies Real Impors* March211= Germany France 9 8 Impor-adjused demand(iad) wih uncerainy Acual IAD * Aggregaed a fixed 211 PPP exchange raes Sources: OECD; RBA aggregae impor growh in boh 211 and 212 of around 1½ percenage poins, and a slighly smaller reducion in he pas couple of years. During 211 and 212, he uncerainy measures were paricularly high, wih concerns abou a Greek exi from he euro area, deb-ceiling debaes in he Unied Saes and he upcoming elecion of he Abe Governmen in Japan all conribuing. Over his period, business invesmen was paricularly weak bu, as he blue IAD model line in Graph 6 shows, his alone would no have been enough o cause rade growh o slow as much as i did. As he uncerainy eased and business invesmen growh picked up in 213, boh models correcly prediced he pick-up in rade. More recenly however, hrough 214 and early 215, impor growh was slighly slower han he models predicion. Alhough he aggregae slowdown in rade is well explained by he model, he resuls vary by counry. Graph 7 shows he predicions for real impors for seleced economies from he IAD model wih uncerainy, compared wih he acual growh (he solid grey line) and he pre-crisis rend growh (he dashed line) in real impors. The resuls for he enire sample of counries can be found in Table A2. For he large euro area economies, such as France and Germany, he model performs fairly well. For Canada, he model under-predics impors growh slighly, 9 Ausralia Unied Saes Canada Japan Impor-adjused demand model wih uncerainy Acual * Dashed lines represen growh a pre-crisis average rae Sources: OECD; RBA alhough mos of he error occurs in early 211 and he model-implied growh raes over he res of he period are broadly in line wih he observed daa. For he Unied Saes, impor growh has been below wha he model would predic, bu demand composiion and uncerainy help explain some of he weakness. In Japan, he model under-predics impor growh, which was quie close o is long-run average despie pressure from slowing GDP growh and a subsanially lower exchange rae. In Ausralia, he model over-predics impor growh from 212 onwards. Given Ausralia s recen resource boom, he 25 inpu-oupu ables by he OECD used o weigh domesic demand by impor inensiy are likely o undersae he impor conen of invesmen and oversae he impor conen of expors. This resuls from mining invesmen being more impor-inensive han oher ypes of invesmen, and resource expors being significanly less impor-inensive han oher expors. 6 Subsequen o he esimaion of he 25 inpu-oupu ables, mining invesmen and resource expors increased subsanially (Graph 8). Mining invesmen peaked in 212 and has fallen subsanially since hen, while 6 ABS inpu-oupu ables alhough no necessarily harmonised o he OECD ones are available up o he 29/1 period, which could provide an avenue for fuure research o es his explanaion for he residual. 8 BULLETIN SEPTEMBER QUA R T E R

6 Graph 8 Graph 9 Ausralia Measures of Mining Aciviy Nominal Laes OECD I/O able Ausralia Seleced GDP Componens Chainvolume,March211= Resource expors (per cen of oal expors) 4 Expors 2 Mining invesmen (percenof oal invesmen) Sources: ABS; RBA 2 Invesmen Impors Source: ABS he volume of resources expors has coninued o increase. The model s over-predicion of growh in impor volumes since 212 could herefore be explained by he implici undersaemen of he impor inensiy of resources invesmen, which is in decline, and oversaemen of he impor inensiy of resources expors, which are increasing (Graph 9). One poenially imporan reason for he model fiing he aggregae daa beer han daa on a counry level is ha he relaive impor price variables which measure he naional accouns impor price deflaor relaive o he GDP deflaor may no be fully capuring he effecs of real exchange rae movemens and associaed changes in compeiiveness. The GDP deflaor is a broad measure of domesic prices, which is no necessarily he relevan cos comparison in making decisions abou wheher o buy domesically or impor. These measuremen errors are, however, likely o ne ou in aggregae, bu could maerially affec counry-level resuls; Morel (215) finds some enaive evidence for his by comparing he European core o he periphery in a similar model. Conclusion Growh in global rade has no recovered o he same exen as global growh in GDP in he period since he financial crisis, perhaps because of a lesser role for srucural facors in supporing increased rade of lae. However, he relaive weakness in global rade growh is well explained by he composiion of he recovery in global growh, wih invesmen remaining subdued, and lingering economic uncerainy. This suggess ha, should invesmen recover o pre-crisis growh and he general economic uncerainy dissipae, growh of global rade is likely o srenghen more in line wih global GDP growh. Of course, if he slowdown in GDP growh, and invesmen in paricular, are longer-erm phenomena, hen rade growh is likely o be subdued for an exended period. I should also be noed ha his work does no conclusively rule ou a moderaion in he srucural expansion of rade associaed wih globalisaion and indusrialisaion, bu here remain reasons o expec ha he srucural expansion in rade has furher o run. R 44 RESERVE BANK OF AUSTRALIA

7 Appendix A The baseline regression akes he form: lnmj 1 2 lndj 3 lnrpmj 4 lnunc j FEj j,,,,, (A1) where ln m j, is he log change in naional accouns impors for counry j a ime, ln D j, is he log change in he demand measure, UNC uses he Baker e al (213) and Moore (forhcoming) policy uncerainy measures (sandard deviaions from average level) and ln RPM j, is he log change in he relaive price of impors o domesic goods, measured as he raio of he impor deflaor o he GDP deflaor. Uncerainy measures are available for he G7 economies and Ausralia; for ohers a GDP-weighed average is used as a proxy. The uncerainy measures are no enirely harmonised; for insance, for he Unied Saes he measure includes a wider range of facors. Counry fixed effecs are also included in all specificaions (denoed by FE j ). GDP daa are obained from he OECD harmonised daabases; where he harmonised ime series are oo shor, paricularly for he relaive deflaors, he longer series available in he OECD s Economic Oulook daabase are used. The resuls are robus o he exclusion of hese longer ime series in favour of he harmonised daa. Daa for all counries cover he period 1985:Q2 215:Q1, excep for Germany, which is esimaed from 1991:Q1 onwards. For he uncerainy model, he sample period is shorer due o he availabiliy of he uncerainy measures. Table A1: Esimaion Resuls (a) Dependen variable: quarerly growh of real impors, 1985:Q2 215:Q1 GDP model Domesic Demand model Impor-adjused demand model Impor-adjused demand wih uncerainy 1.48***.45***.3**.21* Y 1.2*** DD 1.38*** IAD 1.14*** 1.26*** UNC.19*** RPM.15**.4.16**.15* Observaions R 2 (wihin) (a) ***,** and * denoe esimaes ha are significan a 99, 95 and 9 per cen level, respecively. Sandard errors are esimaed using a clusered sandwich esimaor BULLETIN SEPTEMBER QUA R T E R

8 Table A2: Acual and Prediced Annual Average Real Impor Growh Per cen Pre-crisis ( ) Acual Values Model Fied Values in Pos-crisis Period (a) Pos-crisis ( :Q1) Impor-adjused demand model Impor-adjused demand model wih uncerainy Ausralia Belgium Canada Denmark Finland France Germany Ialy Japan Korea Neherlands Norway New Zealand Porugal Spain Sweden Unied Kingdom Unied Saes Advanced Economies Toal (b) (a) Using coefficiens esimaed over he period , fied o he period 211:Q1 215:Q1 (b) Aggregaed a fixed 211 PPP exchange raes References Baker SR, N Bloom and SJ Davis (215), Measuring Economic Policy Uncerainy, Economic Policy Uncerainy sie. Available a <hp:// media/bakerbloomdavis.pdf>. Bems R, RC Johnson and K-M Yi (213), The Grea Trade Collapse, Annual Review of Economics, 5(1), pp Boz E, M Bussière and C Marsilli (214), Recen Slowdown in Global Trade: Cyclical or Srucural, VoxEU sie. Available a <hp:// Bussière M, G Callegari, F Ghironi, G Sesieri and N Yamano (213), Esimaing Trade Elasiciies: Demand Composiion and he Trade Collapse of 28-29, American Economic Journal: Macroeconomics, 5(3), pp Chor D and K Manova (212), Off he Cliff and Back? Credi Condiions and Inernaional Trade during he Global Financial Crisis, Journal of Inernaional Economics, 87(1), pp Consaninescu C, A Maoo and M Rua (215), The Global Trade Slowdown: Cyclical or Srucural?, IMF Working Papers 15/6. Evene SJ (214), The Global Trade Disorder: The 16h GTA Repor, Cenre for Economic Policy Research, London. Ferranino M and D Taglioni (214), Global Value Chains in he Curren Trade Slowdown, World Bank - Economic Premise, pp 1 6. Moore A (forhcoming), Measuring Economic Uncerainy and is Effecs, RBA Research Discussion Paper. Morel L (215), Sluggish Expors in Advanced Economies: How Much is Due o Demand?, Bank of Canada Discussion Paper Novy D and A Taylor (214), Trade and Uncerainy, NBER Working Paper WTO (213), World Trade Repor 213: Facors Shaping he Fuure of World Trade, WTO, Geneva. 46 RESERVE BANK OF AUSTRALIA