Economic Compuaion and Economic Cyberneics Sudies and Research, Issue 3/2016, Vol. 50 Assisan Professor Murad A. BEIN, PhD E-mail: mbein@ciu.edu.r Deparmen of Accouning and Finance Faculy of Economics and Adminisraive Sciences Cyprus Inernaional Universiy Lefkosa, Norh Cyprus, Turkey Associae Professor Gulcay TUNA, PhD Deparmen of Economics Faculy of Business and Economics Easern Medierranean Universiy Famagusa, Turkey E-mail: gulcay.una@emu.edu.r COMPARING SPILLOVER EFFECTS AMONG EMERGING MARKETS WITH A HIGHER (LOWER) SHARE OF COMMODITY EXPORTS: EVIDENCE FROM THE TWO MAJOR CRISES Absrac. The paper empirically analyses he spillover ino emerging markes wih a higher (lower) share of commodiy expors during he Global Financial Crisis (GFC) and he European Sovereign Deb Crisis (ESDC). To invesigae such spillover effecs, a group of rapidly growing emerging economies collecively known as BRICS (Brazil, Russia, India, China, and Souh Africa) is seleced. The findings of he paper are as follows. Firs, a subsanial increase in he average condiional correlaion is noiced wihin all BRICS sock markes during he GFC. When considering he ESDC period, we also observed an increase in all markes, excep for Brazil. Furhermore, he dynamic evaluaion significanly increased from 2007 and i remained high during he ESDC. Second, rade profiles can help in explaining he spillover and correlaion levels beween emerging and developed markes. Among he BRICS counries, Brazil, Russia and Souh Africa heavily depend on commodiy expors and he resuls show ha hese economies have a higher correlaion wih he developed economies. Furher, Brazil and Russia are he mos volaile when compared o he oher BRICS counries, since hese counries commodiies are dominaed by food and agriculural expors and fuel and agriculural expors, respecively. Keywords: Commodiy expors, sock marke co-movemens, volailiy ransmission, DCC-GARCH, Crisis. JEL Classificaion: G01, G15, F30, C22 265
Murad A. Bein, Gulcay Tuna 1. Inroducion Uncerainy abou he fuure course of acion on he par of he US Federal Reserve Bank (he Fed) in erms of quaniaive easing (QE) as well as he recen inroducion of QE by he European Cenral Bank (ECB) coninue o dominae he headlines. I is worh remembering ha he Fed inroduced QE o simulae he economy afer i was severely affeced by he Global Financial Crisis (GFC). If he Fed chooses o end QE, i may boos capial inflow o he USA, which has already begun due o he high expecaions of invesors ha ineres raes will increase from heir currenly level of zero. 1 Alongside his, he ECB launched QE o ry and fully recover from he impac of he European Sovereign Deb Crisis (ESDC), which resuled from high deb and budge deficis in peripheral economies and quickly spread o oher member saes. Therefore, he consequences of such QE policies in developed economies (USA and he EU) increase he vulnerabiliy and risk of emerging markes, especially hose economies ha have srong rade and financial linkages wih he developed economies. BRICS is he acronym used o describe cerain emerging economies (Brazil, Russia, India, China and Souh Africa) ha have close economic ies wih he developed economies. Over he las few decades, he BRICS economies have been growing faser when compared o many developing and developed economies (see Table 1). They also seem o arac considerable inflows of capial ino heir growing financial markes. On he oher hand, he BRICS counries face some challenges ha canno be ignored. For example, hey are highly dependen on high-income economies for heir expors. Table 1 demonsraes ha merchandise expors o high-income economies are well above 50% for all he BRICS counries, wih he highes expor level being ha of China. I can also be observed from Table 1 ha during boh he GFC and he ESDC, expors o high-income economies declined. In addiion, mos of he BRICS economies have greaer volailiy in exchange markes ha respond quickly o policy changes in developed economies. There have been several sudies ha suppor dynamic inerrelaionships beween rade and he exchange rae wih he sock marke. For example, on he linkage beween rade and he sock marke, see Forbes (2002), who argued ha inernaional rade linkages ransmi counry-specific crises hrough sock markes o oher counries worldwide. There have also been sudies ha examine he ineracion beween he sock price and he exchange rae (Ning, 2010; Ulku and Demirci, 2012). Given hese relaionships, i is necessiy o invesigae equiy markes behaviour during boh 1 However, some financial analyss and media commenaors such as Peer Schiff (CEO and Chief Global Sraegis for Euro Pacific Capial Inc.) have argued ha if he Fed decides o raise he ineres rae, i will cause an even greaer financial collapse since he USA now holds huge deb. Schiff did in fac predicae he global financial crisis and is he auhor of several books as well as regularly appearing on several TV channels. 266
Comparing Spillover Effecs among Emerging Markes wih A Higher (Lower) Share of Commodiy Expors: Evidence from he Two Major Crises ranquil and urbulen periods. Indeed, i is criical o undersand wheher here is greaer inerdependence and higher correlaion among sock markes for individual invesors and corporae managers for he purpose of porfolio diversificaion, since he benefis of diversificaion can only be achieved by invesing in markes wih lower correlaion (Wason, 1978). Based on he above, his paper aims o address hree imporan issues. Firs, o wha degree have he BRICS counries been affeced by he GFC and he ESDC? Did he correlaion change from ha of he pre-crisis period? Which of he wo crises had a greaer impac on he emerging markes? Second, do economies wih a higher (lower) share of commodiy expors o oal merchandise have a higher (lower) level of correlaion and more (less) volaile markes? In oher words, does he rade profile help o explain he level of correlaion and spillover? Third, do he BRICS counries provide he opporuniy for inernaional diversificaion during periods of urmoil in developed markes? The sudy conribues o he exisen lieraure in several ways. Firs, we show ha a counry s rade profile is significan in explaining he levels of correlaion and volailiy beween emerging and developed economies. Second, previous sudies ha invesigaed he spillover o emerging economies mosly failed o ake ino accoun he role of he EU in he emerging economies. We believe ha he EU s role should be considered because some of he member saes fel he GFC severely and experienced a crisis. Third, unlike previous sudies ha assumed he whole sample in invesigaing he impac of he GFC on emerging economies, his sudy divides he sample ino hree subsamples (pre-crises, global crisis and Europe crisis) and compares he spillover. Fourh, he sudy employs a DCC model ha has several advanages as compared o convenional mehodologies such as long run coinegraion and he error correcion model. The key findings of his paper are as follows. Firs, a subsanial increase in condiional correlaion is noiced wihin all BRICS sock markes during he GFC and, considering he ESDC period, we also observed an increase in all markes excep for Brazil. Furhermore, he dynamic evaluaion of mos of he BRICS counries significanly increased from 2007 and remained high during he ESDC. Second, rade profiles can help in explaining he spillover and correlaion levels beween emerging and developed markes. Among he BRICS counries, Brazil, Russia and Souh Africa are highly dependen on commodiy expors and he resuls show ha hese economies have a higher level of correlaion wih he developed economies. Furher, Brazil and Russia are he mos volaile when compared o he oher BRICS economies, since 267
Murad A. Bein, Gulcay Tuna hese counries commodiies are dominaed by food and agriculural expors and fuel and agriculural expors, respecively. The remainder of his paper is organised as follows. The nex secion will summarise he empirical lieraure on spillovers. Secion hree discusses he daa and he mehodology of he sudy. The empirical findings are presened in secion four and, finally, secion five concludes he sudy. Table 1. Macroeconomic profile of BRICS GDP growh (annual %) Period Brazil China India Russia SA 2002-2004 3.17 9.73 6.53 6.41 3.72 2005-2007 4.40 12.72 9.45 7.69 5.48 2008-2010 4.12 9.77 7.54 0.64 1.75 2011-2013 3.90 10.74 7.84 4.91 3.65 Merchandise expors o high-income economies (% of oal merchandise expors) 2002-2004 63.67 85.55 70.56 65.67 61.47 2005-2007 59.96 82.45 67.48 66.59 70.67 2008-2010 53.11 77.11 65.33 60.58 62.36 2011-2013 49.84 74.23 63.05 61.52 49.71 Souh Africa=SA 2. Lieraure Review A reasonable number of sudies have been conduced ha examine he spillover and volailiy ransmission ha resuled from he GFC, paricularly in he conex from developed economies (USA, UK, Germany, France, and Japan) o emerging markes ( Dajčman and Alenka, 2011; Syllignakis and Koureas, 2011). There have also been sudies ha invesigaed he specific impac of he GFC on he BRICS economies. For example, Alou e al. (2011) as well as Dimiriou e al. (2013) observed subsanial spillover o he BRICS economies from he USA during he GFC, especially afer he collapse of Lehman Brohers in 2008. However, hese sudies did no invesigae he impac of he ESDC, and mos of hem failed o ake ino accoun he role of he EU. Given ha mos of he BRICS economies are regionally close and have srong economic (rade and financial) ies o Europe, i is necessary o consider he role of EU index in he BRICS counries. Regarding he ESDC, here have been 268
Comparing Spillover Effecs among Emerging Markes wih A Higher (Lower) Share of Commodiy Expors: Evidence from he Two Major Crises sudies ha invesigaed spillover wihin he European sock markes. For example, Dajčman (2013) noed conagion during he Greek deb crisis from he Irish, Ialian and Spanish sock markes o he sock markes of France and Germany. Along he same lines, Tamakoshi and Hamori (2013) argued ha he sock reurns of he five major European financial insiuions under sudy were highly affeced by he Greek deb crisis. Furhermore, Harmann (2014) examined conagion from wesern European counries o eigh emerging European economies and observed an increase in correlaion during he ESDC. Employing daily daa ha covered he ESDC period, Popa el al. (2015) invesigaed spillover among emerging European sock markes (including Russia) and wo developed sock markes, namely he USA and Germany. In heir analysis, Popa el al. (2015) documened ha here is no significan relaionship beween Russian sock reurns and any oher equiy marke. There have also been sudies ha examine spillover effecs from peripheral counries (Greece, Ireland, Porugal, Spain and Ialy; GIPSI) o emerging European markes. For insance, Bein and Tuna (2015) argued ha he dynamic condiion correlaions beween Poland and Hungary and he GIPSI counries increased subsanially during he sovereign deb crisis. Furher, hey demonsraed ha he Czech Republic remained he mos volaile sock marke, alhough hey did no observe an increase in he dynamic evaluaion correlaion. However, o bes of our knowledge, sudies ha examine he spillover effecs ino he BRICS counries and oher non-european emerging economies are rare. Only Ahmed el al. (2013) have invesigaed he spillover from he GIPSI counries o he BRIICKS (BRICS plus Indonesia and Souh Korea) markes. Neverheless, Ahmed e al. (2013) did no ake ino accoun he impacs of he GFC on he BRICS economies, nor did hey compare he wo crises. In addiion, heir daa sops in January 2012, which does no adequaely cover he enire ESDC since he crisis persised for a longer period. 3. Daa and Mehodology 3.1 Daa The weekly sock indices for five emerging and wo developed sock markes are used from 6 January 2003 o 23 March 2015. We used weekly price indexes in order o minimise boh he cross-counry differences and he end-of-week effec. The emerging marke indexes are he BOVESPA for Brazil, he SSE Composie Index for China, he 269
Murad A. Bein, Gulcay Tuna CNX for India, he MICEX for Russia, and he FTSE/JSE All Share for Souh Africa. The developed sock price indexes are he S&P500 for he USA and he EUROSTOXX50 (from now on EU index) sock price index for he Eurozone. We prefer he EU (EUROSTOXX50) index as a proxy for he Eurozone because is represens 50 blue-chip companies ha operaes in welve Eurozone counries. 2 All of he sock price indexes are obained from DaaSream and hey are all US dollardenominaed. The reason for choosing a common currency is o accoun for he local inflaion rae. In he curren lieraure, here is no precise dae given for when he GFC sared. In deciding on he sar of he crisis, researchers generally follow eiher an economeric or an economic approach, alhough here are also sudies ha consider boh approaches. In his sudy, we follow he economics approach, so he saring dae for he global financial crisis is deermined as 6 Augus 2007, which is in line wih he approach of he Federal Reserve Bank S. Louis (2009). In deermining he sar of he ESDC, we consider he dae when he Greek governmen firs officially requesed a bailou from an inernaional organisaion, which is 23 April 2010. However, since he sudy makes use of weekly daa, he sar dae for he crisis is deermined as 26 April 2010, which is he closes pracical opion o he requesed dae. Finally, yearly rade profiles from 2002-2013 are all obained from he World Bank. 3.2 Mehodology Following he work of Forbes and Rigobon (2002), researchers have been using more advanced echniques, including regime-swiching models, dynamic copulas wih and wihou regime-swiching, dynamic condiional correlaion (DCC), and nonparameric approaches. To avoid several resricions, such as he heeroskedasiciy problem, he conagion mus involve evidence of a dynamic incremen in he regressions, affecing a leas he second momen correlaions and covariances. In his sudy, o overcome several problems involved in measuring correlaion and volailiy, a mulivariae DCC-GARCH model of (Engle 2002) is used. Engle s (2002) model has many advanages over oher models, for example, unlike consan correlaion dynamic condiional correlaions (DCC) allow he deecion of possible changes in condiional correlaions over ime, which is very imporan since sock reurns are negaive during urbulen periods and posiive during ranquil periods. In addiion, he model esimaes 2 Including Ausria, Belgium, Finland, France, Germany, Greece, Ireland, Ialy, Luxembourg, he Neherlands, Porugal and Spain. 270
Comparing Spillover Effecs among Emerging Markes wih A Higher (Lower) Share of Commodiy Expors: Evidence from he Two Major Crises he correlaion coefficiens of he sandardised residuals and accouns for heeroscedasiciy direcly (Chiang e al. 2007). The esimaion of Engle s (2002) DCC-GARCH model comprises wo seps: firs, he esimaion of he univariae GARCH model for he sock reurns and second, he esimaion of he condiional correlaions ha vary over ime. The DDC model of Engle (2002) can be expressed as H D R D (1) where H is he condiional covariance marix ha is decomposed ino condiional 2 2 sandard deviaions, D ( 1/,..., 1/ diag h1,1, hn, N, ) in which i, i is any univariae GARCH process and R is he ime dependen condiional correlaions marix, which defined as: ( 1/ 2 2,..., 1/ 2 2 ) ( 1/ R diag q q Q q,..., q 1/ ) (2) 11, NN, where Q is a symmerical posiive definie marix ha defines he dynamic ' correlaion srucure as Q ( 1 a b) Q au 1u 1 bq 1 (3) where u is a vecor of he sandardised residuals, Q is an uncondiional variance marix of u, and a and b are non-negaive one-period lagged auoregressive and correlaion coefficiens saisfying a+b<1. Therefore, he condiional correlaion beween he wo sock reurns (1 and 2) can be expressed as (1 a b) q12 au1, 1u 2, 1 b12, 1 (4) 12, 11, NN, 2 2 (1 a b) q11 u1, 1 bq 11, 1 (1 a b) q22 au2, 1 bq 22, 1 Where ρ 12 is he elemen on he 1 h line and 2 h column of he marix Q. The quasimaximum likelihood mehod (QMLE) is used o esimae he parameers. Disribuion used is he Suden s -disribuion. 4. Empirical Resuls Table 2 panels A-D show descripive saisics for he whole sample (6 January 2003-23 March 2015), pre-crisis (6 January 2003-30 July 2007), global financial crisis/pos (6 Augus 2007-19 April 2010) and European deb crisis/pos (26 April 2010-23 March 2015) periods, respecively. In general, he emerging economies have higher reurns as measured by mean and also have more volaile sock markes as measured h, 271
Murad A. Bein, Gulcay Tuna by sandard deviaion when compared o he developed economies (USA and EU). Among he emerging economies, he highes reurn is observed for India (0.30), followed by Brazil (0.23) (in panel A), Brazil (0.87) and Russia (0.79) (in panel B), and Brazil (0.24) and Souh Africa (0.01) (in panel C). However, in panel D (European deb crisis/pos period), he highes reurn is observed in he USA (0.21), which can be inerpreed as meaning ha he USA sock marke became less volaile during he ESDC. Considering he volailiy of he sock markes shown in Table 2, he emerging economies display higher volailiy in all of he subsamples (panel A-D), wih Russia and Brazil being he mos volaile. For example, for he whole sample Russia was 5.68 and Brazil was 5.53; for he pre-crisis period Brazil was 5.00 and Russia was 4.56; for he global financial crisis/pos period Russia was 8.52 and Brazil was 7.70; and during he European deb crisis/pos period Russia was 4.54 and Brazil was 4.44. Table 2 also shows ha he reurns are negaively skewed for all he markes, wih he excepion of Souh Africa and USA in panel C and China in panel D. All of he reurns are also lepokuric disribuions and hey confirm he financial series characerisics. The Jarque-Bera es saisics indicae he non-normaliy of he reurn series. An auoregressive condiional heeroskedasiciy (ARCH) es a lag (5) on he reurn series reveals ha he generalised auoregressive condiional heeroskedasiciy (GARCH) is consisen and appropriae for modelling he reurn. The Ljung-Box (LB) Q-saisics are also presened on he reurn, and he squared reurns (Q 2 ) a lag (20) indicae he presence of auocorrelaion on he reurn. Finally, he Augmened Dickey- Fuller (ADF) es on he level series failed o rejec he null hypohesis ha he series uni roo agains he alernaive hypohesis series is saionary (no repored in he able). However, he ADF es on he reurn series rejecs he null of a uni roo. The reurn series is obained as follow: r= [log (P ) log (P -1)]*100, where P is he sock marke index on day. Table 3 panels A-B show he uncondiional correlaion for he hree subsamples (pre-crisis, global financial crisis, and European sovereign deb crisis) beween he BRICS economies and he developed economies (USA and EU). Looking a panel A, he uncondiional correlaion wih he USA, a higher correlaion is noiced wih Brazil and Souh Africa in he hree subsamples (pre-crisis, GFC, and EDSC). Similarly, in he panel B correlaions wih he EU, he same counries (Brazil and Souh Africa) have a higher correlaion. In addiion, Russia also has a higher correlaion wih he USA and he EU during boh he GFC and he EDSC, which means ha he sock marke became vulnerable o he crises. Ineresingly, China, he larges emerging economy, is he counry wih he lowes uncondiional correlaion wih he EU and he USA in all hree subsamples. Comparing he uncondiional 272
Comparing Spillover Effecs among Emerging Markes wih A Higher (Lower) Share of Commodiy Expors: Evidence from he Two Major Crises correlaion increase during he GFC and he EDSC, i is observed ha here is a greaer increase during he GFC. Table 2. Descripive saisics of weekly sock reurns Full sample Panel A BRAZIL CHINA INDIA RUSSIA SA EU USA Mean 0.2387 0.2050 0.3052 0.1572 0.2184 0.0662 0.1284 Sd. Dev. 5.5358 3.7962 4.4413 5.6811 4.1082 3.608 2.5372 Skewness -0.6420-0.2891-0.7108-0.507-0.2772-0.519-0.362 Kurosis 7.76 4.52 8.26 11.34 7.20 6.25 8.76 J-Bera 645 *** 70 *** 788 *** 1874 *** 477 *** 309 *** 895 *** ARCH 31.09 *** 8.40 *** 12.29 *** 31.47 *** 29.49 *** 27.76 *** 33.99 *** Q(20) 35.76 ** 53.12 *** 33.60 ** 41.86 *** 43.38 *** 42.90 *** 48.29 *** Q )2 (20) 157.2 *** 225.1 *** 127.5 *** 267.8 *** 381.6 *** 368.3 *** 521.3 *** ADF(5) -9.29 *** -9.62 *** -9.42 *** -9.00 *** -9.76 *** -9.50 *** -9.96 *** Panel B Pre crisis (sable period ) Mean 0.8771 0.5444 0.7305 0.7958 0.5368 0.3289 0.1939 Sd. Dev. 5.0074 3.6118 4.0890 4.5609 3.3066 2.4882 1.6609 Skewness -0.7744-0.0437-1.680-0.906-0.905-0.595-0.364 Kurosis 3.94 5.33 12.62 5.92 5.22 5.41 4.46 Panel C Global Financial crisis/pos Mean 0.2452-0.2402 0.0443-0.2003 0.0128-0.2705-0.144 Sd. Dev. 7.7023 5.2350 6.5063 8.5275 6.2185 5.0048 3.8582 Skewness -0.6683-0.4157-0.214-0.0976 0.0185-0.3485 0.0343 Kurosis 7.36 2.79 4.61 8.48 5.00 5.19 5.12 Panel D European deb crisis/pos Mean -0.3427 0.1212 0.0505-0.2443 0.0400 0.0113 0.2155 Sd. Dev. 4.4461 2.9126 3.178 4.5401 3.2307 3.582 2.2937 Skewness -0.2819 0.1928-0.320-0.9492-0.321-0.368-0.789 Kurosis 4.637 3.788 3.325 5.953 4.358 3.863 9.271 Noe: The Jarque-Bera (J-Bera), ARCH, and Ljung-Box saisics for serial correlaion in he sandardised reurn a lag (20) and he squared sandardised reurn a lag (20) and he ADF es a lag (5) for he hree subsamples are no repored o save space bu are available on reques. 273
Murad A. Bein, Gulcay Tuna Table 3. Uncondiional correlaions wih he US Counry Pre-crisis GFC/pos EU crisis/pos BRAZIL 0.684513 0.812074 0.671212 CHINA 0.057292 0.323736 0.318361 INDIA 0.410207 0.680898 0.474739 RUSSIA 0.390017 0.71233 0.6138 SA 0.546411 0.793966 0.759505 Uncondiional correlaion wih he EU BRAZIL 0.664732 0.824872 0.629378 CHINA 0.098462 0.345104 0.264842 INDIA 0.450755 0.733294 0.507579 RUSSIA 0.446033 0.810486 0.608009 SA 0.664635 0.908525 0.76529 Auhor s calculaion. Table 4 panels A-C show he univariae esimaion for each counry index as well as he generaed dynamic condiional correlaion (DCC) wih he US and EU sock markes, respecively. The univariae esimaion in Table 1 panel A shows ha he ARCH and GARCH coefficiens are saisically significan a 1% for all he counries and so confirm ha GARCH (1;1) is appropriae for modelling he sock markes. In oher words, a significan ARCH coefficien means ha he previous day s informaion on he reurn reflecs in oday s volailiy, whereas a significan GARCH means ha he previous day s reurn volailiy reflecs on oday s volailiy. The significance of he wo coefficiens means ha he sock reurn volailiy is influenced by is own shock. Considering he derived mulivariae DCC equaion, he condiion a+b<1 is saisfied beween he developed markes (EU and USA). In addiion, he coefficiens (a and b) are non-negaive. Furhermore, Table 3 shows ha he -suden disribuions for all of he sock markes are saisically significan a 1%, confirming ha he -suden is an appropriae disribuion. The pormaneau mulivariae saisics repored as mulivariae Q(20), and Q 2 (20) are due o Li and McLeod s (1981) esing of serial correlaion in he mean and variance equaions, respecively. The resuls in panels A and B confirm he successful eliminaion of serial correlaion in he mean and variance equaions. 274
Comparing Spillover Effecs among Emerging Markes wih A Higher (Lower) Share of Commodiy Expors: Evidence from he Two Major Crises Table 4. Esimaion resuls from GARCH-DCC using weekly reurn daa Panel A Condiional mean and variance equaions for each marke Counries Mean equaion Variance equaion µ ω α β Brazil 0.2664 1.5847 ** 0.1366 *** 0.8131 *** (0.1837) (0.6906) (0.0367) (0.0385) China 0.1335 0.3331 0.0863 *** 0.8922 *** (0.1388) (0.2336) (0.0328) (0.0434) India 0.4059 ** 0.6922 0.1610 *** 0.8160 *** (0.1434) (0.4611) (0.0566) (0.0626) Russia 0.2845 2.4587 ** 0.1359 *** 0.7782 *** (0.1871) (1.078) (0.0487) (0.0686) SA 0.2543 *** 0.5529 0.1084 *** 0.8589 *** (0.1343) (0.3746) (0.0367) (0.0551) US 0.2588 *** 0.3143 *** 0.2049 *** 0.7415 *** (0.0734) (0.1181) (0.0763) (0.07) EU 0.1644 0.4706 ** 0.1274 *** 0.8346 *** (0.1127) (0.2257) (0.0382) (0.0446) Panel B Mulivariae DCC wih he US Brazil China India Russia SA a 0.1328 ** 0.0066 0.0456 0.0165 ** 0.0579 *** (0.0589) (0.0077) (0.0408) (0.0069) (0.0156) b 0.7497 *** 0.9798 *** 0.7498 *** 0.9820 *** 0.9184 *** (0.1612) (0.015) (0.279) (0.0085) (0.0239) df 11.311 *** 9.142 *** 7.426 *** 6.3062 *** 7.0161 *** (3.007) (1.683) (1.183) (0.772) (1.031) Diagnosic checking Log-L -3074-3031 -3008-3122 -2858 MQ(20) 185.5 225.8 201.8 205.7 179.9 275
Murad A. Bein, Gulcay Tuna [0.4790] [0.1012] [0.4501] [0.3763] [0.8421] MQ 2 (20) 149.7 155.5 131.8 192.2 200.4 [0.9956] [0.9886] [0.9999] [0.6020] [0.4375] Panel C Mulivariae DCC wih he EU Brazil China India Russia SA a 0.0374 ** 0.0058 0.0629 0.0487 0.0653 *** (0.0186) (0.0072) (0.0485) (0.037) (0.0181) b 0.9402 *** 0.9775 *** 0.3044 0.9126 *** 0.9045 *** (0.0425) (0.017) (0.3053) (0.0954) (0.0267) df 8.500 *** 8.020 *** 7.112 *** 5.236 *** 6.867 *** (1.634) (1.352) (1.129) (0.5328) (1.077) Diagnosic checking Log-L -3351-3289 -3248-3339 -3037 MQ (20) 190.1 231.3 205.5 203.3 175.8 [0.6799] [0.0641] [0.3798] [0.4221] [0.8901] MQ 2 20) 190 171.2 169.1 194.7 255.9 [0.6454] [0.9159] [0.9324] [0.5520] [0.9948] Noe: Log-L (Log-likelihood),he numbers given in ( ) are sandard error while he numbers given in [ ] are he p-values. ***, **, and * donae saisical significance a 1%, 5%, and 10% respecively. Table 5 panels A-B show he weighed condiional correlaion beween he BRICS economies and he developed markes (US and EU) for he hree subsamples (pre-crisis period, GFC/pos period, and ESDC/pos period). The weighed condiionals are all derived using DCC-GARCH (1;1). As can be observed from he able, mos of he weighed correlaions are higher wih he EU han wih he US, which is also he case in Table 3 (uncondiional correlaions). In addiion, he weighed correlaions beween he BRICS economies and he developed marke subsanially increased during he GFC (column 2) and during he ESDC, expec wih Brazil in column 3. Therefore, he Brazilian sock marke is he leas affeced by he ESDC. Comparing he level (magniude) of he weighed correlaions in penal A, Brazil has he highes correlaion in a sable period while Russia and Souh Africa have he highes during he GFC/pos and ESDC/pos period, respecively. In panel B (correlaion wih he EU), he highes correlaion is observed wih Souh Africa during he hree subsamples. Ineresingly, he Chinese sock marke has he lowes correlaion wih boh he EU and he USA during ranquil and urbulen periods (in he 276
Comparing Spillover Effecs among Emerging Markes wih A Higher (Lower) Share of Commodiy Expors: Evidence from he Two Major Crises hree subsamples). In general, economies ha have a higher commodiy expor share have a higher correlaion wih he developed markes (see Table 6). Table 5. Condiional correlaion during for hree subsamples using GARCH (1;1) Panel A wih he USA Counry Pre-crisis GFC/pos EU crisis/pos Brazil 0.6831 *** 0.7935 *** 0.6471 *** (0.0451) (0.0674) (0.0497) China 0.0884 0.3673 *** 0.2322 ** (0.0720) (0.0762) (0.1046) India 0.4505 *** 0.6914 *** 0.4984 *** (0.0723) (0.0484) (0.0440) Russia 0.4070 *** 0.8957 *** 0.5938 *** (0.0723) (0.2877) (0.0708) SA 0.5411 *** 0.7180 *** 0.7052 *** (0.0826) (0.1230) (0.0627) Panel B wih he EU Brazil 0.6534 *** 0.7944 *** 0.6005 *** (0.0415) (0.0365) (0.0774) China 0.1843 ** 0.4184 *** 0.2239 ** (0.0807) (0.0616) (0.0941) India 0.5058 *** 0.7233 *** 0.5149 *** (0.0545) (0.0463) (0.0425) Russia 0.4694 *** 0.7884 *** 0.6240 ** (0.0755) (0.0532) (0.0411) SA 0.6601 *** 0.9002 *** 0.7295 *** (0.0513) (0.025731) (0.036807) The numbers given in ( ) are sandard errors. ***, **, and * donae saisical significance a 1%, 5%, and 10% respecively. 277
Murad A. Bein, Gulcay Tuna Figure 1 shows he evoluion of he condiional correlaion beween he USA and he BRICS economies. A sudden sharp increase in he condiional correlaion is noiced saring from 2007 and 2008, especially for China, Russia, and Souh Africa. These markes sayed higher during boh he GFC and he ESDC. Saring from 2014, he correlaions fall gradually for Russia and Souh Africa, whereas China experienced a sharp decline during he same year. A sharp increase in he correlaion is regarded as a change in invesors appeie for risk and heir herding behaviours. Invesors appeie for risky invesmens falls during he crisis, since hey experience loss in some markes. Therefore, o offse heir losses hey may decide o sell heir shares in anoher marke, which will lead o a decline in he sock price. Regarding he volailiy of he correlaion, i is observable ha India remains he mos volaile as compared o he oher emerging economies. The highes correlaion is observed wih Brazil, varying beween 60-80%. In general, he condiional correlaions are higher wih Brazil, Souh Africa and Russia and he lowes wih he Chinese sock marke, which reached is highes poin in 2011 a around 30%. Figure 2 shows he evaluaion correlaion beween he BRICS economies and he EU for he full sample. The correlaion shows almos he same rend as wih he USA. The highes correlaion is noiced wih Souh Africa, reaching approximaely 90%, and he lowes wih China. Similarly, he correlaion wih India remains very volaile hroughou he sample, unlike he oher emerging sock markes, which display herding behaviours and changes in he risk appeies of invesors. BRAZIL CHINA INDIA RUSSIA 278
Comparing Spillover Effecs among Emerging Markes wih A Higher (Lower) Share of Commodiy Expors: Evidence from he Two Major Crises SOUTH AFRICA Figure 1. Condiional correlaion beween he USA and he BRICS economies BRAZIL CHINA INDIA RUSSIA SOUTH AFRICE Figure 2. Condiional correlaions beween he EU and he BRICS economies 279
Murad A. Bein, Gulcay Tuna Table 6 deails he rade profiles for he BRICS economies as consruced using yearly daa from 2002-2013. We presen i by using hree-year averaging o accoun for he business cycle and he disorion due o he GFC and he EDSC. We define he share of commodiy expors as he sum of agriculural raw maerials, food expors, fuel expors, and ores and meals expors o he oal merchandise expors (all in percenages). Higher commodiy expors are observed in Brazil, Russian, and Souh Africa. I should be recalled ha Brazil and Russia have he mos volaile markes (as measured by sandard deviaion; see Table 2) as compared o he res of he emerging markes, which could be due o he fac ha commodiy expors for Brazil are dominaed by food and agriculural, while for Russia fuel expors and agriculural expors are dominan. In inernaional markes, here is normally greaer flucuaion in he price of hose wo producs, and his is expeced o reflec in he domesic sock markes hrough he ineracion of he exchange rae and he price of commodiies and he sock marke. A higher share of manufacuring expors as a percenage of merchandise expors is noed for boh China and India. Those wo economies do have lower marke volailiy (see Table 2) and a lower correlaion wih he developed economies. Table 6. Trade profile for he BRICS economies Agriculural raw maerials expors (% of merchandise expors) YEAR BRAZIL CHINA INDIA RUSSIA SA 2002-2004 4.13 0.65 1.12 3.21 2.67 2005-2007 3.77 0.49 1.66 2.75 1.82 2008-2010 3.75 0.45 1.64 2.15 1.82 2011-2013 3.62 0.48 1.95 1.84 1.86 Food expors (% of merchandise expors) 2002-2004 28.19 4.28 11.39 1.79 9.77 2005-2007 25.72 2.94 8.90 1.84 7.42 2008-2010 30.95 2.75 8.73 2.30 9.16 2011-2013 32.33 2.77 10.24 2.81 9.81 Fuel expors (% of merchandise expors) 2002-2004 4.88 2.52 6.21 53.89 10.29 2005-2007 7.33 1.93 13.66 62.03 10.05 2008-2010 9.53 1.86 16.01 65.99 10.38 2011-2013 9.67 1.57 19.12 69.72 11.93 Ores and meals expors (% of merchandise expors) 2002-2004 8.42 1.74 4.32 7.33 17.53 280
Comparing Spillover Effecs among Emerging Markes wih A Higher (Lower) Share of Commodiy Expors: Evidence from he Two Major Crises 2005-2007 10.49 1.99 7.49 7.73 26.85 2008-2010 13.88 1.45 6.48 5.62 28.92 2011-2013 17.09 1.33 3.38 4.96 29.46 Commodiy expor (% of merchandise expors) 2002-2004 45.62 9.20 23.04 66.21 40.25 2005-2007 47.31 7.34 31.71 74.35 46.14 2008-2010 58.11 6.51 32.86 76.06 50.28 2011-2013 62.70 6.16 34.69 79.34 53.06 Manufacure expors (% of merchandise expors) 2002-2004 52.60 90.60 75.26 22.30 59.35 2005-2007 50.55 92.44 67.18 17.40 53.70 2008-2010 40.46 93.37 64.45 16.01 49.46 2011-2013 35.18 93.75 62.96 15.36 46.38 Commodiy expors defined as he sum of agriculural raw maerials, food expors, fuel expors, and ores and meals expors o oal merchandise expors (all in percenages). 5. Concluding Remarks The sudy empirically compares he spillover during he Global Financial Crisis and he European Sovereign Deb Crisis ino emerging economies ha have a higher (lower) share of commodiy expors. The BRICS counries are emerging economics ha no only have good economic ies bu have also regisered rapid growh as compared o many developed and developing economies over he las decade. In addiion, hey seem o arac large capial invesmen o heir growing financial markes as a resul of his rapid globalisaion process. However, one of he greaes disadvanages of he BRICS counries is heir reliance on high-income economies o sell heir producs and for invesmen inflow, which makes heir economies vulnerable and sensiive o policy changes and shocks ha may arise in high income counries (see Table 1). This makes i necessary o invesigae he degree o which hose emerging sock markes have been affeced by he crises and o deermine wheher rade profiles maer in undersanding he exen of spillover. We also consider wheher he emerging economies provide he benefi of porfolio diversificaion during urmoil in he developed markes. In order o precisely gauge he impac of he wo crises, weekly reurn daa from 6 January 2003 o 23 March 2015 is used afer being divided ino hree subsamples sable period, GFC/pos, and ESDC/pos period. We used weekly price indexes o minimise he cross-counry differences and he end-of-week 281
Murad A. Bein, Gulcay Tuna effec. A mulivariae GARCH framework is used in sudying he volailiy spillovers among each counry wih he USA and he EU indexes and o accoun for he ime variabiliy of he condiional correlaions. A dynamic srucure is included by using he DCC model of Engle (2002). For he rade profiles, we consider yearly daa from 2002-2013. We uilise hree-year averaging o conrol for business cycles and o reduce he possible disorions caused by he crises. The findings of he paper are as follows. Firs, during he GFC a significan increase in he condiional correlaion is noiced wih all he BRICS sock markes, while during he ESDC here is also an increase, albei lower han during he GFC. In addiion, we observed an increase in correlaion during he ESDC wih he Brazilian sock markes. Therefore, among he BRICS sock markes, he Brazilian marke is he leas affeced by ESDC. Furhermore, he dynamic evaluaion significanly increased from 2007 and remained high during he ESDC. Second, we found ha rade profiles can help in explaining he spillover and condiional correlaion beween emerging and developed markes. Among he BRICS economies, Brazil, Russia and Souh Africa highly depend on commodiy expors. The resuls show ha hese counries are more affeced and have a higher level of correlaion wih he developed economies. Furher, Brazil and Russia are he mos volaile markes when compared o he oher BRICS economies. Manufacuring expor-oriened counries such as China and India exhibi a lower correlaion wih he developed counries. In paricular, China has he highes manufacuring expors and he lowes correlaion. In addiion, even hough Russia has a lower correlaion han India during he sable period, during he wo crises i is observed ha he Russian sock markes have a higher correlaion. This is because commodiy prices such as food, agriculural and oil prices are more volaile in he inernaional marke as compared o manufacuring goods. The resuls have imporan implicaions for boh policy makers and invesors. Firs, from foreign invesors poin of view, he GFC and he ESDC have subsanially reduced he benefis of diversificaion in he BRICS economies, especially hose wih a higher share of commodiies (Brazil, Russia, and Souh Africa). In addiion, invesors who are willing o inves in he BRICS economies should also consider a hedging echnique agains he adverse effecs of exchange raes and sock price changes. Our sudy also emphasises ha policy makers in he BRICS counries should work oward esablishing and promoing rade and invesmen wih each oher and wih oher developing economies. This could be one way o avoid being adversely affeced by anoher shock o he developed economies. 282
Comparing Spillover Effecs among Emerging Markes wih A Higher (Lower) Share of Commodiy Expors: Evidence from he Two Major Crises REFERENCES [1] Ahmad, W., Sehgal, S. and Bhanumurhy, N.R. (2013), Eurozone Crisis and BRIICKS Sock Markes: Conagion or Marke Inerdependence. Economic Modelling, 33, 209-225; [2] Aloui, R., Aissa, M.S.B. and Nguyen, D.K. (2011), Global Financial Crisis, Exreme Inerdependences and Conagion Effecs: The Role of Economic Srucure? ; Journal of Banking and Finance, 35, 130-14; [3] Bein, A.M. and Tuna, G. (2015), Volailiy Transmission and Dynamic Correlaion Analysis beween Developed and Emerging European Sock Markes during Sovereign Deb Crisis. Romanian Journal of Economic Forecasing, 18(2), 61-80; [4] Chiang, T.C., Jeon, B.N. and Li, H. (2007), Dynamic Correlaion Analysis of Financial Conagion: Evidence from he Asian Markes. Journal of Inernaional Money and Finance 26: 1206-1228; [5] Dimiriou, D., Kenourgios, D. and Simos, T. (2013), Global Financial Crisis and Emerging Sock Marke Conagion: A Mulivariae Fiaparch Dcc Approach. Inernaional Review of Financial Analysis, 30, 46-56; [6] Dajčman, S. and Kavkler, A. (2011), A Comparaive DCC-GARCH and Rolling Wavele Correlaion Analysis of Inerdependence beween he Slovenian and European Sock Markes; Economic Compuaion and Economic Cyberneics Sudies and Research; ASE Publishing; 45(4), 99-118; [7] Dajčman, S. (2013), Forbes and Rigobon s Mehod of Conagion Analysis wih Endogenously Define Crisis Period - An Applicaion o Some of Eurozone s Sock Markes. Engineering Economics, 24(4), 291-299; [8] Engle, R.F. (2002), Dynamic Condiional Correlaion - A Simple Class Of Mulivariae GARCH Models. Journal of Business and Economic Saisics, 20, 339-350; [9] Forbes, K.J. (2002), Are Trade Linkages Imporan Deerminans of Counry Vulnerabiliy o Crisis?; NBER Working Paper No. 8194. Cambridge, Mass.: Naional Bureau of Economic Research; [10] Forbes, K. and Rigobon, R. (2002), No Conagion, Only Inerdependence: Measuring Sock Marke Co-Movemens. Journal of Finance, 57, 2223-2261; 283
Murad A. Bein, Gulcay Tuna [11] Federal Reserve Board of S. Louis. (2009), The Financial Crisis: A Timeline of Evens and Policy Acions. hp://www.slouisfed.org/financial-crisis/fullimeline (las accessed 22/08/15); [12] Harkmann, K. (2014), Sock Marke Conagion from Wesern European o Cenral and Easern Europe during he Crisis Years 2008-2012. Easern European Economics, 52(3), 55-65; [13] Hosking, J. (1980), The Mulivariae Pormaneau Saisic. Journal of American Saisical Associaion, 75, 602-608; [14] Jarque, C.M. and Bera, A.K. (1987), A Tes for Normaliy of Observaions and Regression Residuals. Inernaional Saisical Review, 55(2), 163-172; [15] Ning, C. (2010), Dependence Srucure beween he Equiy Marke and he Foreign Exchange Marke - A Copula Approach. Journal of Inernaional Money and Finance, 29, 743-759; [16] Popa, I., Tudor, C. and Paraschiv, D. (2015), Shocks Spillover wih Emerging Easern European Markes. Economic Compuaion and Economic Cyberneics Sudies and Research; ASE Publishing: 49(1), 41-58; [17] Syllignakis, M.N., Koureas, G.P.( 2011), Dynamic Correlaion Analysis of Financial Conagion: Evidence from he Cenral and Easern European Markes.. Inernaional Review of Economics and Finance 20, 717 732; [18] Tamakoshi, G. and Hamori, S. (2013), An Asymmeric Dynamic Condiional Correlaion Analysis of Linkages of European Financial Insiuions during he Greek Sovereign Deb Crisis. European Journal of Finance, 19(10), 939-950; [19] Űlkű, N. and Demirci, E. (2012), Join Dynamics of Foreign Exchange and Sock Markes in Emerging Europe. Journal of Inernaional Financial Markes, Insiuions and Money, 22, 55-86; [20] Wason, J. (1978), A Sudy of Possible Gains from Inernaional Invesmen. Journal of Business Finance and Accouning, 5, 195-206. 284