Dynamic linkage beween real exchange raes and sock prices: Evidence from developed and emerging Asian markes 1 By omoe Moore Economics and Finance Brunel Universiy Uxbridge Middlesex, UB8 3PH Unied Kingdom E-mail: omoe.moore@brunel.ac.uk el: +44 189 526 6642 Fax: +44 189 526 9770 and Ping Wang Birmingham Business School Universiy of Birmingham Edgbason, Birmingham B15 2 Unied Kingdom Email: p.wang.1@bham.ac.uk el: +44 121 414 6675 Fax: +44 121 414 6238 1 his is a revised version of he paper presened a he Asian Finance Associaion Inernaional Conference, held a Hong Kong in July, 2010. hanks are due o he paricipans of his conference for heir helpful commens. he auhors are also graeful o he edior and wo anonymous referees for heir useful and consrucive commens and suggesions. he usual disclaimer applies. 0
Absrac his paper invesigaes he sources of he dynamic relaionship beween real exchange raes and sock reurn differenials in relaion o he US marke for he developed and emerging Asian markes. We, firs, derive he dynamic condiional correlaion (DCC) of he wo series, and hen DCC is regressed on he rade balance and he ineres rae differenials. In general, he rade balance is found o be a main deerminan of he dynamic correlaion for he Asian markes, whereas he ineres rae differenial is he driving force for he developed markes. he laer seems o reflec he high capial mobiliy. Keywords: Real exchange raes; Sock reurn differenials; Dynamic condiional correlaion; rade balance; ineres rae differenials JEL Classificaion: C22, G15 1
1. Inroducion During he pas decade, a growing lieraure has emerged in an aemp o explore he relaionship beween exchange raes and sock prices, see, for example, Abdalla and Murinde (1997), Ajayi e al (1998), Granger, e al. (2000), Smyh and Nandha (2003) Phylakis and Ravazzolo (2005),Moore (2007a and 2007b) and Pan e al. (2007). Empirically, he presence of a significan relaionship beween he wo variables is found elsewhere. For example, Haemi-J and Irandous (2002) repor ha he Granger causaliy is unidirecional, running from sock prices o exchange raes, and an increase in Swedish sock prices is found o be associaed wih an appreciaion of he Swedish Krona. Moore (2007b) finds a coinegraion relaionship beween sock prices and exchange raes for Poland and Hungary by aking ino accoun he poenial srucural breaks in he wo markes. See also Bahmani-Oskooee and Sohrabian (1992) and Nieh and Lee (2001) for he coinegraion approach. Oher sudies include Malliaropulos (1998), who examined he link beween inernaional sock reurn differenials and deviaion from relaive purchasing power pariy (PPP) for four major OECD counries using he VAR approach. he volailiy spillovers beween sock reurns and changes in exchange raes are examined in he work of Kanas (2000). here are, iner alia, wo srands of classical heories suggesing a relaion beween sock prices and exchange raes. Flow-oriened models of exchange raes focus on he curren accoun or he rade balance, and asser ha he currency movemen will affec he inernaional compeiiveness and he balance of rade posiion, which in urn affecs he counry s real income/oupu and herefore sock prices (Dornbusch and Fischer 1980). his is based on he radiional macroeconomic view. On he oher hand, sock-oriened models of exchange raes, or porfolio balance approaches, predic ha he innovaions in he sock price affec exchange 2
raes via he capial accoun (Branson 1983 and Frankel 1983): he performance of he sock marke may affec he demand for money, wih he subsequen changes in ineres raes causing exchange raes o appreciae or depreciae. he relaionship beween he wo series has been heoreically and empirically esablished, however, he issue of driving forces behind he linkage remains o be unexplored. his paper aims a invesigaing he dynamic relaion beween exchange raes and sock prices and exploring he sources behind he linkage. A wo-sep esimaion procedure is employed. In he firs procedure, we derive he dynamic condiional correlaion (DCC) advanced by Engle (2002) of he wo series, and in he second sep, a linear regression model is specified, where he esimaed DCC is regressed on he poenial deerminans of he correlaion. Our sample counries include he emerging Asian counries of Indonesia, Malaysia, Souh Korea, he Philippines, Singapore and hailand 2, and he developed counries of Ausralia, Canada, Japan and he UK for he sample period from 1970s and 1980s o 2006. A relaively wider se of counries may serve o provide an unbiased insigh ino he inquiry of he relaionship beween he sock prices and exchange raes. here are a number of imporan conribuions o he lieraure. Firsly, he echnique of dynamic correlaion demonsraes a more direc indicaion of inerdependence beween sock and foreign exchange markes, where he dynamics of correlaion are modelled ogeher wih hose of he volailiy of he series. 3 In order o 2 Several sudies esablished he linkage beween sock prices and exchange raes for he Asian counries (see Abdalla and Murinde 1997, Ajayi e al 1998, Granger, e al 2000 and Smyh and Nandha 2003, Phylakis and Ravazzolo 2005 and Pan e al 2006). 3 Forbes and Rigobon (2002) argue ha he cross marke correlaion coefficiens are condiional on marke volailiy, and if such a es is no adjused for heeroskedasiciy, he esimaed correlaion coefficiens can be biased. 3
examine varying degrees of relaionship, previous sudies normally pariion he sample period ino differen phases according o srucural changes. Such a subjecive designaion of cu-off daes may no effecively describe he evoluion of he changes on financial markes over ime. By accouning for he ime-varying condiional correlaion of daa series, possible changes in condiional correlaions can be deeced when he sae of he economy changes over ime, and ha his is a rigorous approach in examining he fas-moving volaile sock and foreign exchange markes of he emerging economies. 4 Secondly, correlaion ess are conduced for he variables of real exchange raes and he sock price differenials agains he US marke. I is argued ha much empirical sudy ends o omi he US sock marke, despie he fac ha i represens he influence of world markes, and ha i has been shown ha he resuls of some of he previous sudies are invalid (Caporale and Piis 1997). Phylakis and Ravazzolo (2005) found ha i acs as a condui hrough which he real exchange rae affecs he sock marke or vice versa. hirdly, he paper clarifies he heoreical issues of he relaionship by assuming ha boh real exchange raes and sock price differenials conain permanen and emporary componens (Huizinga 1987, Fama and French 1988, Poerba and Summers 1988 and Baxer 1994). he heoreical model, which largely follows ha of Malliaropulos (1998), predics a negaive relaionship. he fourh conribuion, which is he primal objecive of his paper, is o invesigae he deerminans of he dynamic link beween he wo series. In a linear regression framework, he poenial deerminans as regressors are explicily specified. 4 he mehodology has been uilised for financial analysis, see for example, Chiang, e al (2007) and Wang and Moore (2008) for he sudy of sock marke conagion and comovemens, and also Bauisa (2006) for he invesigaion of he relaionship beween exchange raes and ineres rae differenials in six Eas Asian counries. 4
We consider he rade balance and he ineres rae differenials as possible deerminans of he linkage. he variable of rade balance is deemed o capure he exen of he open characerisics of hese economies in erms of expors and impors wih he US. Hence if he variable srongly impac on he correlaion, economic inegraion may maer for he linkage. Given ha of ineres rae differenials, he degree of capial marke inegraion beween hese counries and he US is gauged as a channel o he correlaion of sock and foreign exchange markes. If his effec is significan, financial inegraion plays a major role for he linkage. he former is based on he radiional macroeconomic view (or he flow-oriened model), while he laer is associaed wih he porfolio balance model (or he sock-oriened model). he inclusion of he maure financial markes may provide a useful comparaive sudy wih he emerging Asian markes, where foreign capial resricions end o be prevalen. In his respec, our sudy may shed a new ligh on he debae of he wo heories by addressing he sources of he correlaion in a rigorous empirical framework. he overall resuls are summarised as follows. We find a negaive dynamic relaionship beween he relaive sock prices and real exchange raes, being consisen wih he model predicion. A linear regression model reveals ha he rade balance is, in general, he main source of he dynamic correlaion for he Asian counries, whereas he ineres rae differenial is a conribuory facor for Ausralia, Canada and he UK. his appears o sugges ha in counries wih a relaively low degree of capial mobiliy, economic inegraion is likely o he main force of he linkage, supporing he flow-oriened model, whereas in counries wih he high capial mobiliy, financial inegraion is he main driving force, supporing he sock-oriened model. When, he Asian markes are modelled in he pos-asian crisis period as a subsample period, he driving force slans o he ineres rae differenials. his is 5
indicaive of he improved capial mobiliy afer he crisis for he Asian emerging marke. he paper is organised as follows. Secion 2 is for heoreical underpinning, Secion 3 is he model specificaion of he DCC from he GARCH model, and Secion 4 describes daa. Secion 5 presens he resul of he DCC esimaion, which forms a firs par of he empirical procedures, and he linear regression, he second par, is modelled in Secion 6. Conclusions are found in Secion 7. 2. heoreical underpinning Following Malliaropulos (1998), we derive a heoreical model of he relaionship beween he real exchange rae and he sock reurn differenials beween wo counries. All variables are in logarihm. We define he relaive sock price beween he domesic and foreign counries expressed in he home currency: s s * e (1) where s = he domesic sock price, * s = he foreign sock price and e = a number of domesic currency per uni of foreign currency. he real exchange rae is defined as: q e p * p (2) where p = he domesic price, * p = he foreign price. I is assumed ha he real exchange rae is composed of permanen ( q ) and emporary ( q ) facors (Huizinga 1987 and Baxer 1994): P q q P q (3a) where q P q 1 and P P q q 1. (3b) 6
he permanen componen is specified as a random walk process wih a drif, μ, and serially uncorrelaed innovaions P, whereas he emporary componen follows a firs order auoregressive process wih 0< <1 plus serially uncorrelaed innovaions. Similarly, he relaive sock price is assumed o conain boh permanen and emporary componens, and Summers 1988). P and respecively (Fama and French 1988 and Poerba P (4a) where P P P v 1 and 1. 0< <1 (4b) P and are serially uncorrelaed innovaions. We also assume ha uncorrelaed wih P and are P and. he expeced changes in he real exchange rae and sock price differenial may be defined as: * E 1 E 1( s s e ) (5) * E 1q E 1( e p p ) (6) where E 1 = expecaions formed a ime -1 given all available informaion. Equaion (5) may be analogous o he Uncovered Ineres Pariy applied o sock reurns wih a risk premium, where he risk premium E1 may conain boh a foreign exchange risk and a relaive sock reurn risk. he foreign exchange risk is one of he main risk facors in inernaional equiy invesmen. Equaion (6) implies he expeced deviaion from relaive PPP. We inroduce he real sock reurn differenial as given by * * z ( s p ) ( s p ) (7) Adding (5) and (6), and re-rearranging wih expecaions, we have E E z E q 1 1 1 (8) 7
his equaion saes ha he expeced sock reurn differenial is equal o he expeced real sock reurn differenial minus he expeced change in he real exchange rae. he unobservable expeced change can be subsiued wih a emporary componen of he series, hence from Equaions (3b) and (4b) we derive he following expressions respecively E 1) q (9) 1q ( E 1 v ( 1) (10) he expeced real depreciaion is associaed wih he emporary componen of he real exchange rae, whereas he expeced risk premium is relaed o he emporary componens of he sock price differenial. By subsiuing equaions (9) and (10) ino (8), we obain ( 1) /( 1) q 1/( 1 ( v) /( 1) ) E z (11) he equaion (11) characerises he dynamic relaionship. Since he parameers of he auoregressive erms is 0< <1 and 0< <1, he emporary componen of he relaive sock price is more likely o be negaively correlaed o emporary deviaions of he real exchange rae from PPP. A negaive relaionship means ha as sock prices increase (decreases), here would be a real appreciaion (depreciaion) of exchange raes 5. he dynamic movemen may be explained in he conex of he moneary secor, and he negaive predicion is consisen wih he view of he porfolio 5 Empirical resuls are quie mixed wih he sign and causal direcion beween exchange raes and sock prices. A significan posiive relaionship is empirically found in he sudy of Smih (1992), whereas Soenen and Hennigar (1988) find a negaive relaionship. Loudon (1993) finds ha when currency appreciaes, indusrial socks (manufacuring and service indusries) perform beer whils resource socks perform beer wih currency depreciaion for Ausralian firms. Malliaropulos (1998) finds a negaive relaionship beween he wo variables for he UK, Japan, France and Germany during he period 1973:1 o 1992:3. 8
approach o exchange rae deerminaion (Phylakis and Ravazzolo, 2005). Agens allocae heir wealh amongs a range of financial asses including money, deposis, bonds, equiies and foreign asses. An increase in he sock marke raises he value of equiies, increasing he oal wealh. Wih an assumpion of homogeneiy 6, his raises he demand for each of hese asses including money. An increase in he demand for money will cause ineres raes o go up and here may be a subsiuion from foreign securiies o domesic asses resuling in an appreciaion of he domesic currency. Also, wih a rise in he local sock marke, foreign invesors are araced o invesing ino he local marke pushing up he demand for he local currency, causing an appreciaion of exchange raes. By definiion, his effec implies a rise in, and a fall in q, forming a negaive correlaion beween he wo series. In he real secor, he associaion of he wo series can no be based on he dynamic componens due o he sluggishness of real economy. I is, however, noed ha if q has a propery of mean-revering, hen a real depreciaion above is rend level prediced by he permanen componen i.e. P q q, produces expecaions of appreciaion (Malliaropulos 1998) 7. his would cause a loss of compeiiveness in an exporing counry, hough he exen of his depends on wheher he Marshal-Lerner holds or no in he economy. If he percepion of a mean-revering naure of real exchange raes is srong amongs invesors, domesic sock prices could be dampened relaive o he foreign sock marke, since he sock prices reflec he expeced fuure cash flows in firms. his means a fall in sock prices in response o a real 6 he wealh homogeneiy implies ha any shif in an asse s share in he desired equilibrium is due o movemens eiher of yields or of oher explanaory variables, no due o he change in oal wealh. 7 For example, Huizinga (1987) and Chen and ran (1994) find ha real exchange raes follow a mean reversion owards PPP. 9
depreciaion, leading o a negaive correlaion. I is also noed ha if changes in real depreciaion are posiively correlaed wih he expecaions of inflaion, and if inflaion affecs sock reurns negaively, hen he consequence is a negaive correlaion. his effec could be srong in counries, which are prone o higher raes of inflaion. 3. DCC model specificaion he bivariae GARCH model wih dynamic condiional correlaion (DCC) specificaion (Engle, 2002) is uilized o invesigae he relaionship beween sock price differenials,, and real exchange raes, q. Le y y, y ]' be a 2 1 vecor [ 1 2 conaining he sock price differenials and real exchange rae series in a condiional mean equaion as follows, y and ~ N(0, H ) (12) 1 where is a 2 1 vecor of consan and, ] is a vecor of innovaions [ 1 2 condiional on he informaion a ime -1 ( 1 ). he error erm is assumed o be condiionally mulivariae normal wih mean zero and variance-covariance marix as H D C D (13) where D is a 2 2 diagonal marix of he ime varying sandard deviaions from univariae GARCH models wih h i, on he i h diagonal. C is a 2 2 ime-varying symmeric condiional correlaion marix. following, As indicaed, he elemens in D follow he univariae GRACH process of he h h (14) 2 i, i i i, 1 i i, 1 10
where i is a consan erm, i capures he ARCH effec, i.e. condiional volailiy and i measures he persisen of he volailiy. he evoluion of he correlaion in he DCC model is given as Q ( 1 qa qb ) Q qa 1 ' 1 qbq 1 (15) where Q { } is a 2 2 condiional variance-covariance marix of residuals wih q ij ' is ime-invarian variance-covariance marix Q = E( ), and q a and q b are nonnegaive scalar parameers saisfying q a q b 1. Because Q in (15) does no have uni diagonal elemens, i is hen scaled o ge a proper correlaion marix C, 1/ 2 1/ 2 C diag( Q ) Q diag( Q ) (16) A ypical elemen of C has he form of q q, i, j 1,2 and i j, ij q ij, / ii, jj, which is he key elemen in his mehodology, as i represens he condiional correlaion beween sock price differenials and real exchange raes. 4. Daa he daases used in his sudy are he monhly closing sock price indices and end-ofperiod nominal exchange raes, relaive o he US dollar, for he counries of Indonesia, Malaysia, Souh Korea, he Philippines, Singapore and hailand, represening he Asian emerging markes, and Ausralia, Canada, Japan and he UK for he developed markes. We examine he sock prices and exchange rae correlaion under a floaing exchange rae regime. So he saring dae for he four developed markes is from January 1973. For he six Asian emerging markes, he saring dae for each counry depends on when a floaing or managed floaing regime was adoped ogeher wih he consideraion of daa availabiliy. Specifically, he saring dae is 1983:5 for 11
Indonesia, 1980:1 for Korea, 1980:1 for Malaysia, 1986:2 for he Philippines, 1979:1 for Singapore and 1979:1 for hailand. All he daa end in December 2006. he real exchange rae is calculaed as he nominal exchange rae adjused for he domesic and foreign price levels, measured by he respecive consumer price indices. he sock price differenials are consruced by convering he US sock price ino domesic currency and hen deducing i from he domesic sock price. All he daa are rerieved from Daasream. [able 1 around here] Saionariy in he ime series is checked by applying he Augmened Dickey Fuller (ADF) es, see able 1. he resuls sugges ha i fails o rejec he null of a uni roo in he logarihm of sock price differenials and he real exchange rae, bu overwhelmingly rejec he null for he firs difference of he series. he wo variables are, herefore, differenced series for he GARCH model. 5. Esimaion resuls [able 2 around here] able 2 is he esimaed resul from he bivariae DCC-GARCH model. he model is esimaed using he quasi-maximum likelihood mehod o generae consisen sandard errors ha are robus o non-normaliy. A closer inspecion of he esimaed parameers shows ha he ARCH parameers, indicaed by, are all significan above 1 percen level for boh reurn differenials and real exchange raes for all of he markes excep for Canada. In his respec, he null hypohesis of he absence of ARCH componens is rejeced, and his indicaes he variance of he curren error erm or innovaion o be a funcion of he squares of he previous ime periods' error erms. Furhermore, he coefficiens for he lagged condiional volailiy, as shown by, are significan for all six emerging marke and Canada and UK, suggesing a 12
high persisence in shocks o he condiional volailiy. Noe also ha, given he larger magniude of coefficiens in, he persisence in volailiy is higher for he sock price differenials han ha for he real exchange rae for mos of he Asian emerging markes. In addiion o he high condiional volailiy, he mos noable poin in our resuls is he significan ime-varying correlaion. his effec is capured by he coefficiens of q a and q b, which are he parameers governing he GARCH process of he Q sequence as in equaion (15). Boh are mosly highly significan a above 1 percen level, and he condiional correlaion exhibis high persisence for all cases over he sample period. he diagnosic es is repored in he las panel, where he Ljung-Box Q- saisics for up o he 4 h orders in he levels and squares of he residuals are repored. I clearly shows ha join linear and non-linear serial correlaions in he sandardized residual have been eradicaed for boh he sock reurn differenials and changes in real exchange raes. [Figure 1 around here] Figure 1 shows he evoluion of he condiional correlaions of en counries. In all cases, he plo shows ha he DCC falls wihin he range of 0 o -1, exhibiing a negaive relaionship beween sock reurn differenials and real exchange raes hroughou he sample period and his is consisen wih he model predicion. In general, quie a volaile movemen is eviden in he Asian counries, and here are varying rends apparen wih he Asian financial crisis period around 1997-98 as a urning poin. hese emerging counies experienced sock marke liberalisaion from he mid o he end of he 1980s. Liberalisaion can ake many forms, e.g. relaxing currency resricions, reducing foreign ownership resricions, or 13
allowing capial and dividend repariaion a various poins in ime. his should have simulaed he inegraion of local markes wih he global marke, poenially narrowing he sock marke reurn differenials. An upward movemen of DCC in he 1980, excep for Malaysia, appears o reflec such experience during financial liberalisaion. hese Asian counries were heavily hi by he financial crisis of 1997. During his crisis period, a sudden dislocaion of asse demands, or he shif of porfolios from domesic o overseas asses, occurred owing o he herding behaviour of invesors or o he loss of confidence in he economy. his resuled in a decrease in he demand for domesic money and a deprecaion of he currency 8. his financial crisis characerised he coninuous devaluaion of heir domesic currencies (heir currencies fell on average by around 50% or more agains he dollar) as well as a sock marke crash, explaining he sharp fall in he correlaion. his is refleced wih DCC approaching almos perfec negaive correlaion. Noe ha hese counries were, in effec, in sof peg regimes for a long ime prior o he 1997 evens, hence, an abrup change in he correlaion srucure afer he crisis may be, in par, due o a shif o a free floaing regime, (hough subsequenly, in many cases, sof pegging was resumed or managed floaing sysem was adoped in order o sabilize he exchange rae). In he pos Asian crisis period, he level of pre-crisis correlaion was reesablished. his is indicaive of some recovery in he financial and economic 8 A crucial role in he Asian crisis was played by financial inermediaries, which were able o raise money a low raes of ineres and lend i a higher raes o finance risky invesmens. his drove up he prices of risky asses. Wih he bursing of he bubble, prices wen ino a downward spiral, in which risky asse prices plummeed, causing he insolvency of financial inermediaries, which furher conribued o fas price deflaion. Besides, heavy reliance on capial inflows as a main source of high levels of domesic credi rendered he Asian marke vulnerable. I was seen ha massive ouflows of porfolio asses and credi gave rise o he collapse of he domesic currency. 14
condiions in hese Asian counries 9, hough for Indonesia, he pre-crisis level of DCC has no fully recovered, where he crisis is said o be he greaes and mos prolonged. On he oher hand, i is noeworhy ha here is a lesser degree of a fall in he correlaion during he Asian crisis in Singapore. I is argued ha Singapore has weahered he crisis beer han mos Asian economies, where he managed exchange rae sysem allowed i o quickly depreciae he Singapore dollar in response o he loss of is expor compeiiveness arising from he collapse of he regional currencies. Overall, hese resuls are in agreemen wih oher empirical lieraure, where he inermarke relaionship inensified for a brief period around he crisis, bu hen quickly resumed heir normal pre-crisis period relaionship (e.g. Phylakis and Ravazzolo 2005). As compared wih he emerging markes, he DCC in he developed markes is less volaile and shows less flucuaion. As for Canada and he UK, i is sriking ha he paern of he DCC is remarkably similar o each oher. In he earlier period of he floaing regime he DCC demonsraes an upward movemen, i.e. an increase of correlaion, which seems o reflec a sharp rise in inflaion, hence real appreciaion wih a falling sock marke during he oil crisis period. Wih is peak in he early 1980s, i sared exhibiing a gradual decline in correlaion, i.e. he exen of negaive correlaion has increased, which is associaed wih he gradual depreciaion of heir currencies since he 1980s for boh counries. Disinguished from Canada and UK, Ausralia and Japan have shown lile rend, since boh correlaions flucuae around - 0.4 o -0.5, indicaing relaively a sable relaionship of he wo markes hroughou he sample period. 9 For example, foreign funds were reurned o hese counries. 15
6. Analyzing he sources of he dynamic linkage Linear model specificaion Having found a significan ime-varying correlaion beween he sock reurn differenials and changes in real exchange raes, we invesigae he poenial driving facors behind he linkage. We consider wo main srands of he driving forces, economic inegraion and financial inegraion, as a channel for he correlaion. he former is associaed wih inernaional compeiiveness in he real secor, where depreciaion or appreciaion of exchange raes affecs aggregae domesic demand for goods, so domesic firms marke values may be affeced. his, in urn, affecs sock prices of firms. We consider he curren accoun surplus/defici as a proxy o measure he exen of economic inegraion hrough inernaional rade as a conribuory facor o he correlaion. Financial inegraion is associaed wih inernaional porfolio behaviour via changes in ineres raes causing changes in exchange raes, e.g. inernaional invesors are araced o invesing counries, which offer higher sock reurns, and his pushes up he value of he currency wih he inermediaion of he change in ineres raes. hus, ineres rae differenials are specified as a proxy o invesigae he exen of he role of inernaional capial mobiliy and, hereby, capial marke inegraion in deermining he relaionship beween sock reurns and exchange raes. 10 In addiion o hese variables, we inroduce counry specific conrol variables by considering heir financial developmen, which is likely o impac on he condiional correlaion, independenly from hose wo driving sources. We include a 10 We, iniially, inended o use he capial accoun as a regressor, ye, due o a lack of daa availabiliy for he Asian counries in he earlier sample period, ineres raes were used. Noe ha Smih (1992) aemps o examine he relaionship beween sock markes and exchange raes wih he curren accoun posiion, by specifying exchange raes as dependen variables and he sock prices, curren accoun surplus and he ineres rae differenials as independen variables. 16
sock marke developmen variable ha is capured by he marke capialisaion relaive o GDP, which migh aler sock marke reurns. We also specify he variables of M3 and credi o he privae secor, boh a percenage of GDP, which are commonly used o measure he exen of financial developmen. he linear equaion akes he following form: DCC i, * 0i 1 idcci1 2i( cas / gdp) i 3 i( r r ) i 4i FD u (17) i he dependen variable DCC is he ime-varying condiional correlaion series for each counry i, esimaed from he previous secion 5. (cas/gdp) is he raio of he * curren accoun o GDP, and ( r r ) is he real ineres rae differenial beween he domesic counry and he US, where he lending rae and he inflaion rae, measured by he consumer price index, are used 11. FD is he sum of he raio of sock marke capializaion o GDP, he raio of M3 o GDP and he raio of credi o he privae secor o GDP. All daa are rerieved from Daasream. Since monhly daa for he curren accoun are no available, all daa are quarerly observaions. DCC was convered from monhly o quarerly ime series. Quarerly daa are, afer all, more plausible han monhly daa in erms of measuring he impac of he curren accoun on DCC, since he response o he real secor is likely o ake longer han a monh 12. 11 Morgage raes are used for lending raes. his is based on daa availabiliy; morgage raes are consisenly found for he whole sample period for hese Asian counries, whereas here is a lack of daa availabiliy for lending raes for he earlier sample period, e.g. Indonesia sars wih 2000, or he lending raes are no available for some Asian counries. Noe also ha for some counries, morgage raes are reaed as lending raes under he code 60P..ZF, in Inernaional Financial Saisics. 12 However, we acknowledge ha an analysis on monhly daa may provide differen resuls, in paricular for financial inegraion, where he dynamic movemen is much faser han ha of real secor. his can be explored by conducing a furher research on oher counries where daa are available. 17
Uni roo ess for he variables are conduced wih he ADF es, and all variables are found o be nonsaionary, a leas a he 5 percen significance level. herefore all he variables are in firs difference, as indicaed by Δ in equaion (17). Since, here is poenially an endogeneiy problem among dependen and independen variables, he insrumen variable (IV) mehod is adoped. he insrumen variables, which are orhogonal o he disurbances, are he lagged once and wice explanaory variables in levels ogeher wih he lagged dependen variable. We, preliminarily, examined he impac of he Asian crisis on he dynamic correlaion by regressing DCC on a dummy variable, i.e. dummy one for he 1997:4 onwards and zero oherwise for he Asian markes. We mosly found a significan coefficien on he dummy. Hence, we separaely esimae he model for he pos financial crisis period for he Asian counries. Linear esimaion resuls he resuls of he linear regression for he whole sample period are shown in able 3a, and hose for he pos-crisis period are in able 3b. Diagnosic es resuls indicae ha in many cases he Breusch-Godfrey serial correlaion es indicaes he absence of serial correlaion a he 5 percen level. he residuals end o rejec he null of homoskedasiciy, hence he regression is correced using robus esimaion wih Whie s heeroskedasic consisen -raios. he over-idenificaion ess sugges ha he null is no rejeced in all cases a he 1 percen level. On balance, hese diagnosic ess are quie saisfacory o warran drawing inferences from he resuls. For discussion, we, firs, concenrae on he esimaes of he whole sample periods in able 3a, hen hose of he subsample period in able 3b a he end of his secion. [able 3 around here] 18
he lagged dependen variable has been shown o be saisically highly significan in all cases, indicaing srong persisence in he dynamic correlaion. In able 3a, apar from Singapore hese emerging Asian counries have he significan coefficien of cas/gdp, suggesing ha he dynamic linkage beween sock and foreign exchange markes is closely associaed wih he rade balance. he signs on he coefficiens are negaive, excep for Malaysia, implying ha he advance of he curren accoun is associaed wih a negaive correlaion. I can be explained in ha he improvemen of he curren accoun gives a boos o share prices, and also increases he value of he domesic currency. Hence, when an economy enjoys he curren accoun surplus, a negaive correlaion can be enhanced. However, if a lower value of currency generaes price compeiiveness, he rade balance improves, hence he real oupu of a counry, which in urn affecs curren and fuure cash flows of firms, and heir sock prices can be boosed. If his effec is srong, we may observe a posiive sign on he coefficien of cas/gdp: a posiive correlaion in and q is he consequence, as in he case of Malaysia 13. he resul is indicaive of he role of he real secor as a condui of he correlaion, and his appears o reflec heir rade-led associaion wih he US markes in favouring he flow-oriened models of exchange raes (Dornbusch and Fischer 1980). On he oher hand, he coefficiens on he ineres rae differenials are no well-deermined for he Asian markes in able 3a, as we find significan coefficiens only a he 10 percen level for Indonesia, Korea and Philippine and no significan for oher Asian counries. Also, noe ha here is lile effec of financial developmen (FD) on he correlaion. his suggess ha he correlaion is less robus hrough he 13 Wheher i has a posiive or negaive sign on he rade balance depends on he direcion of causaliy beween sock and exchange markes. 19
financial inegraion. he resul seems o reflec resriced capial mobiliy ypically found in emerging markes. We also noe ha hailand and he Philippines mainained resricions on foreign exchange movemens and foreign ownership even afer hey opened heir markes o foreign invesors 14. We seem o have a differen picure for he developed counries from ha for emerging economies. For Canada, given he fac ha he coefficiens on boh cas/gdp and r-r* are saisically significan a he 5 percen level, he resuls reflec he subsanial link wih he US marke hrough boh inernaional rade and capial mobiliy in deermining he condiional correlaions. In he case of he UK and Ausralia, whereas he coefficien on he curren accoun is no significan, ha of he ineres rae differenials is significan a he 5 percen level. his reveals he sensiiviy of he ime-varying correlaion o ineres rae differenials. he resul suggess a high degree of capial mobiliy amongs he developed markes. he magniudes of he coefficien on r-r* beween Canada and he UK are no differen from each oher a 0.0039 and 0.0025 respecively, whereas Ausralia has shown a relaively large size of 0.0067. A negaive sign indicaes ha a rise in he relaive ineres rae leads o a furher negaive correlaion, and his is supporive o he porfolio approach o exchange rae deerminaion, where he equiies, being par of wealh, may affec he behaviour of exchange raes hrough he demand for money, hence ineres raes (Branson 1983, Frankel 1983 and Phylakis and Ravazzolo, 2005). his resul may imply ha financial inegraion as a channel of correlaion may be overriding ha of he economic inegraion in hese wo developed economies. In he 14 he resuls for hailand and he Philippines are inconsisen wih ha of Phylakis and Ravazzola (2005), who found ha he sock and exchange markes are conneced more hrough financial secor inegraion, raher han hrough he real secor during he period 1980-1998. However, heir es lacks rigour, as i was conduced by he Granger-causaliy echnique wihou explicily specifying imporan variables such as he curren accoun, nor he ineres rae differenials. 20
case of Japan, where significan coefficiens are observed neiher on he curren accoun balance, nor on he ineres rae differenials. Ye, likewise UK, i has a highly significan coefficien on FD, and ha he linkage of he sock and foreign exchange markes is sensiive o he developmen of financial markes, leaning o he financial secor channel of he correlaion. In he pos Asian crisis period in able 3b, here emerges some shif in he driving force of he correlaion for hese emerging markes. I is evidenced ha a number of significan coefficien on (r-r*) has increased, by conras ha of (cas/gdp) has decreased. his implies ha he correlaion has become more deermined by forces in financial secor han before. In paricular, Indonesia and Philippine have winessed a shif from economic o financial inegraion as a major driving facor of he correlaion. Souh Korea and Malaysia are more or less remains o be he same 15, whereas hailand has shown a valid role of financial inegraion as a source of he correlaion afer he crisis. he shif observed for some Asian markes may reflec he increased degree of capial mobiliy in he pos financial crisis period. Noe ha in he case of Singapore, financial markes sared developing a a much earlier sage han in oher Asian markes, wih more capial mobiliy and flexible exchange rae regimes. In able 3a for he whole sample period, we find no significan coefficiens among hese variables, ye in he pos Asian crisis period, he coefficien on (r-r*) becomes significan a he 10% level. Hence, alhough we are unable o draw a similar inference for Singapore as for oher Asian economies, he significan coefficien on he ineres rae differenials afer he crisis appears o indicae he role of financial inegraion as he main source of he correlaion. 15 In he period leading up o he Asian crisis, Malaysia had inflexible exchange raes. Wih brief spells of flexibiliy during he Asian financial crisis, he Malaysian Ringgi had gone back owards inflexibiliy afer he crisis subsided (Panaik e al 2010). 21
In he case of developed counries, we found ha here are lile shifs in he pos crisis period, since he significan coefficiens in able 3b are broadly similar o hose of he whole sample period in able 3a. I appears ha he Asian crisis may have changed he shape of he deerminans of DCC beween sock reurns and foreign exchange raes for hese Asian counries, bu no for hese developed counries. 7. Conclusion In his paper, we have examined he deerminans of he ime-varying correlaion beween sock reurn differenials and real exchange raes for he six Asian emerging markes and he four developed markes. he resuls from he empirical analysis have highlighed a number of ineresing issues. Firsly, he finding of significan ime-varying correlaion beween he wo ime series implies ha he DCC approach may be an appropriae approximaion for invesigaing he relaionship. Also, i suggess ha he US sock marke influences hese economies by bringing ogeher he foreign exchange and he local sock markes. Secondly, he evidence of he negaive relaionship beween he sock and foreign exchange markes seems o suppor he heoreical model predicion wih he assumpion of emporary and permanen componens of sock reurn differenials and real exchange raes. hirdly, he driving force of he dynamic link of he wo series is, in general, associaed wih he curren accoun balance for he Asian counries, suggesing he leading role of economic inegraion wih he US. An increase in rade balance binds he ies beween sock and foreign exchange markes for hese economies. he weak ineres rae channel, as a source of he dynamic link, seems o indicae, as ye, resricive financial markes in some emerging economies. Ye, a furher empirical analysis reveals he increased capial mobiliy for some Asian markes in he afermah of he Asian crisis. For Canada, he linkage can be fosered hrough he channels of boh inernaional rade and capial mobiliy, whils he 22
correlaion is emphasised via ineres rae differenials for he Ausralia and UK. he high degree of capial marke inegraion wih he US marke provides a possible explanaion for he laer resul, and i sresses he sensiiviy of well-developed financial markes o changes in ineres raes. his should have implicaions for policymakers in ineres-rae seing. On balance, our empirical resuls sugges wheher he sock oriened model or he flow-oriened model is suppored or no may depends on he degree of mauriy in financial markes. he invesigaion of he sources of he dynamic correlaion can be exended, for example, by augmening he linear model wih oher poenial variables, which migh affec he linkage, and also by applying i o oher counries, which may help o confirm our conrasing findings beween developed and emerging counries. References Abdalla, I. and Murinde, V. (1997). Exchange rae and sock price ineracions in emerging financial markes: evidence of India, Korea, Pakisan and he Philippines, Applied Financial Economics, 7, 25-35. Ajayi, R.A., Friedman, J. and Mehdian, S.M. (1998) On he relaionship beween sock reurns and exchange raes: ess of Granger causaliy, Global finance Journal, 9, 241-251. Bahmani-Oskooee, M. and Sohrabian, A. (1992) Sock prices and he effecive exchange rae of he dollar, Applied Economics, 24(4), 459-464. Bauisa, C.C. (2006). he exchange rae-ineres differenial relaionship in six Eas Asian counries, Economics Leers, 96, 137-142 Baxer, M. (1994). Real exchange raes and real ineres differenials. Have we missed he business-cycle relaionship? Journal of Moneary Economics, 33, 5-37. Branson, W.H., (1983). Macroeconomic deerminans of real exchange rae risk. In: Herring, R.J. (Ed.), Managing Foreign Exchange Rae Risk. Cambridge Universiy Press, Cambridge, MA. 23
Caporale, G.M. and Piis, N. (1997). Causaliy and forecasing in incomplee sysems, Journal of Forecasing, 16, 425-437. Chen B. and ran K, (1994). Are we sure ha he real exchange rae follows a random walk? A re-examinaion, Inernaional Economic Journal, 33-44. Chiang,. C., B.N. Jeon and H. Li (2007). Dynamic correlaion analysis of financial conagion: Evidence from Asian markes, Journal of Inernaional Money and Finance, 26, 1206-1228. Dornbusch, R., S. Fisher, (1980). Exchange raes and he curren accoun. American Economic Review 70, 960-971. Engle, R. (2002). Dynamic condiional correlaion: A simple class of mulivariae generalized auoregressive condiional Heeroskedasiciy models. Journal of Business and Economic Saisics 20, 339-350. Fama, F. and French, K.R. (1988). esing he predicive power of dividend yields, Journal of Finance, 48, 663-679. Frankel, J.A., (1983). Moneary and porfolio balance models of exchange rae deerminaion. In: Bhandari, J.S., Punam, B.H. (Eds.), Economic Inerdependence and Flexible Exchange Raes. MI Press, Cambridge, MA. Forbes, K.J and R. Rigobon (2002). No Conagion, Only Inerdependence: Measuring Sock Marke Comovemens, he Journal of Finance, Vol. 57, No. 5, pp. 2223 2261. Granger, C.W.J., Huang, B.N. and Yang, C.W. (2000) A bivariae causaliy beween sock prices and exchange raes: evidence from recen Asian flu, Quarerly review of Economics and Finance, 40, 337-354. Haemi-J, A and M. Irandous (2002). On he causaliy beween exchange raes and sock prices: A noe, Bullein of Economic Research 54, 197-203. Huizinga, J. (1987). An empirical invesigaion of he long-run behaviour of real exchange raes, Carnegie-Rocheser Conference Series Public Policy, 27, 149-214. Kanas, A. (2000). Volailiy spillovers beween sock reurns and exchange rae changes: Inernaional evidence. Journal of Business and Accouning, 27, 447-467. Loudon, G. (1993). he foreign exchange operaing exposure of Ausralian socks, Accouning and Finance, 33, 19-32. Ma, C.K. and Kao, G.W. (1990). On exchange rae changes and sock price reacions, Journal of Business Finance and Accouning, 17, 441-449. Malliaropulos, D. (1998). Inernaional sock reurn differenials and real exchange rae changes, 17, Journal of Inernaional Money and finance, 493-511. 24
Moore,. (2007a). he Effecs of he Euro on Sock markes: Evidence from Hungary, Poland and UK, Journal of Economic Inegraion, 22(1) March, 69-90. Moore,. (2007b). Has enry o he European Union alered he dynamic links of sock reurns for he emerging markes? Applied Financial Economics, 17, 1431-1446. Nieh, C.C. and Lee, C.F. (2001) Dynamic relaionship beween sock prices and exchange raes for G7 counries, Quarerly Review of Economics and Finance, 41, 477-490. Pan, M.S. Chi-Wing Fok,R. and Angela Liu, Y. (2007). Dynamic linkages beween exchange raes and sock prices: Evidence from Eas Asian markes. Inernaional Review of Economics & Finance, 16(4), 503-520. Panaik, I. Shah, A. Sehy, A. and Balasubramaniam, V. (2010). he exchange rae regime in Asia: From crisis o crisis, Inernaional Review of Economics and Finance, forhcoming. Phylakis, K. and Ravazzolo, F (2005). Sock prices and exchange rae dynamics, Journal of Inernaional Money and Finance, 24, 1031-1053. Poerba, J.M. and Summers, L.H. (1988). Mean reversion in sock prices: Evidence and implicaions, Journal of Financial Economics, 22, 27-59. Smih, C. (1992). Sock markes and he exchange rae: a mulicounry approach, Journal of Macroeconomics 14, 607-629. Smyh, R. and Nandha, M. (2003). Bivariae causaliy beween exchange raes and sock prices in Souh Asia, Applied Economics Leers, 10, 699-704. Soenen L. and Hennigar E. (1988). An analysis of exchange raes and sock prices he US experience beween 1980 and 1986, Akron Business and Economic Review, 19, 7-16. Wang, P. and Moore,. (2008). Financial Marke Inegraion for he ransiion economies: ime-varying condiional correlaion approach, he Mancheser School, 76(1) Sepember, 116-133. 25
able 1 Uni roo ess q q Indonesia -1.407-1.999-15.033-14.167 Korea -1.540-2.182-15.512-18.854 Malaysia -1.480-0.627-10.027-19.090 Philippine -1.136-1.565-13.395-20.694 Singapore -2.161-2.775-18.122-19.313 hailand -0.974-1.437-16.976-17.537 Ausralia -0.230-2.136-19.179-15.047 Canada -1.051-1.555-20.567-20.104 Japan -0.701-2.336-19.813-18.693 UK -2.318-2.074-18.973-18.981 * * Noe: s s e : he sock price differenial. q e p p : he real exchange rae. ADF uni roo es. Lag lengh is based on he Schwarz Informaion Crierion. Criical values are -3.44, -2.87 and -2.57 a 1%, 5% and 10% levels, respecively. 26
able 2 Esimaes from he bivariae DCC-GARCH model Indonesia Korea Malaysia Philippine Singapore hailand Ausralia Canada Japan UK Δρ (Sock reurn differenial) -0.5339 (0.5461) 0.3247 (0.3926) -0.1144 (0.3312) 0.1095 (0.5658) 0.0106 (0.2534) -0.3315 (0.4115) -0.2181 (0.2938) -0.2422 (0.1717) -0.1897 (0.2814) 0.0262 (0.2087) 7.8832** (3.4567) 7.4380** (3.3566) 2.8121 ** (1.2233) 4.9505 (3.9099) 2.3692 *** (0.9085) 3.6226 (2.4515) 23.4960*** (6.1690) 2.0134 (2.3409) 25.5727*** (7.2942) 0.7486 (0.4763) 0.2258*** (0.0492) 0.2559*** (0.0566) 0.2339*** (0.0588) 0.1397*** (0.0562) 0.2244*** (0.0361) 0.1360*** (0.0408) 0.2567*** (0.0796) 0.0728 (0.0521) 0.1942*** (0.0550) 0.0600** (0.0266) 0.77838*** 0.6805*** 0.7439*** 0.8193*** 0.7444*** 0.8244*** 0.1051 0.7635*** 0.1808 0.9031*** (0.0401) (0.0714) (0.0526) (0.0735) (0.0372) (0.0540) (0.1719) (0.2276) (0.1700) (0.0405) Δq (Real exchange rae) 0.0412 (0.2079) 0.0109 (0.0607) 0.1468*** (0.0558) -0.0815 (0.1157) -0.1704 (0.1121) 0.0437 (0.1137) -0.0977 (0.1323) 0.1019 (0.0781) 0.0203 0.1482-0.1270 (0.1337) 4.7687*** (1.0495) 0.1645** (0.0769) 0.1912 *** (0.0765) 0.7294*** (0.2031) 0.32940*** (0.1252) 1.7834*** (0.3828) 2.2794*** (0.7661) 0.1548 (0.1408) 8.2210*** (1.7519) 0.1215 (0.1195) 0.5996*** (0.1573) 0.5967*** (0.1163) 0.4560*** (0.1031) 0.3308*** (0.1123) 0.0920*** (0.0301) 0.1699*** (0.0540) 0.2633*** (0.1161) 0.05118 (0.0351) 0.1864*** (0.0672) 0.0567*** (0.0219) 0.4976*** 0.5476*** 0.5609*** 0.5905*** 0.8602*** 0.5785*** 0.3628** 0.8840*** 0.0738 0.9315*** DCC parameers (0.0851) (0.0692) (0.0744) (0.0882) (0.0350) (0.0799) (0.1700) (0.0891) (0.1510) (0.0282) q 0.0832*** 0.0809*** 0.0748 0.0728* 0.0211 0.0753** 0.0356 0.0213** 0.1026*** 0.0194*** a (0.0326) (0.0277) (0.0500) (0.0404) (0.0143) (0.0386) (0.0333) (0.0101) (0.0401) (0.0056) q 0.8528*** 0.9032*** 0.6248** 0.8586*** 0.8339*** 0.8617*** 0.7971*** 0.9684*** 0.6728*** 0.9854*** b (0.0606) (0.0336) (0.2865) (0.0766) (0.1601) (0.0626) (0.1362) (0.0201) (0.1274) (0.0070) Log likelihood -1800.29-1729.54-1599.33-1431.14-2271.58-1922.51-2166.59-1727.91-2304.16-2058.12 Diagnosic es Sock Q(4) 2.9193 5.4985 7.0461 6.0258 8.5636 0.9667 1.0445 7.2714 3.8401 3.1472 2 Sock Q (4) 1.7670 4.0687 7.1109 3.9717 3.3885 1.8921 2.8900 0.5042 3.7567 9.6383 Exchange rae Q(4) 3.3589 4.5059 15.0080 3.6942 5.8097 4.9681 30.0394 0.4368 9.553 0.4521 2 Exchange rae Q (4) 0.0459 7.2435 2.2497 0.5410 0.6285 0.0769 1.0009 0.8787 2.6691 2.7675 Noe: Sandard errors are in parenhesis. *, ** and ***: significan a he 10%, 5% and 1% level respecively. Q(4) and Q 2 (4) are he Ljung-Box Q-saisic for he 4 h orders in he levels and squares of he residuals, respecively. Under he hypohesis of no serial correlaion, he criical value is 9.49 a he 5% level. 27
able 3a Insrumen variable esimaes for he whole sample period: dependen DCC All sample Indonesia Korea Malaysia Philippine Singapore hailand Consan -0.0875*** (0.0225) -0.0867 *** (0.0207) -0.1233*** (0.0336) -0.0364* (0.0193) -0.0575*** (0.0150) -0.0478* (0.0271) DCC -1 0.90658 *** (0.0270) 0.9069 *** (0.0338) 0.5776*** (0.1110) 0.9119 *** (0.0466) 0.6533*** (0.0921) 0.8928 *** (0.0637) (cas/gdp ) -0.000503 ** (0.0002) -0.000504 ** (0.0002) 0.2484** (0.1206) -0.00049 *** (0.0002) 0.0001 (0.0225) -0.00045 ** (0.0002) (r-r*) 0.0022 * 0.0021 * 0.0121-0.0074* -0.0029 0.0043 (0.0013) (0.0012) (0.0084) (0.0040) (0.0021) (0.0090) FD 0.0068 0.00002 0.4310* -0.0161 0.0323 0.0322 (0.0170) (0.00004) (0.2228) (0.0231) (0.0306) (0.0429) Serial (1) [0.7096] [0.7228] [0.9343] [0.0391] [0.9515] [0.0172] Serial (2) [0.9042] [0.8865] [0.7834] [0.1251] [0.9906] [0.0355] Heero [0.0096] [0.0054] [0.0394] [0.0545] [0.7140] [0.0461] Over id. [0.0962] [0.2334] [0.5344] [0.2166] [0.8756] [0.0364] All sample Ausralia Japan Canada UK Consan -0.1196*** (0.0249) -0.2245*** (0.0398) -0.0148** (0.0058) 0.0226* (0.0121) DCC -1 0.7003*** (0.0620) 0.5544*** (0.0769) 0.9333 *** (0.0267) 0.9063 *** (0.0303) (cas/gd p) 0.0013 (0.0016) -1.3107 (7.2291) -0.0011 *** (0.0004) -0.1211 (0.1259) (r-r*) -0.0067** (0.0022) -0.0063 (0.0055) -0.0039 ** (0.0016) -0.0025*** (0.0009) FD -0.0082 (0.0237) -0.1459*** (0.0549) -0.00003 * (0.00002) -0.0081 *** (0.0026) Serial (1) [0.1279] [0.2846] [0.7033] [0.1259] Serial (2) [0.2151] [0.3060] [0.3558] [0.3187] Heero [0.0009] [0.0021] [0.2398] [0.0031] Over id. [0.2534] [0.7780] [0.1509] [0.0442] able 3b Insrumen variable esimaes for he pos crisis period: dependen DCC Subsample Indonesia Korea Malaysia Philippine Singapore hailand Consan 0.0821 (0.0869) -0.6152* (0.3619) -0.2819*** (0.0552) -0.0060 (0.0331) -0.0713*** (0.0264) -0.0836** (0.0372) DCC -1 0.8733*** (0.0832) 0.8753*** (0.1069) 0.4862*** (0.1247) 0.8874*** (0.0649) 0.5379*** (0.1659) 0.7976*** (0.0792) (cas/gd p) -0.00030 (0.00029) -0.00095* (0.00057) 1.2795*** (0.3148) 0.00018 (0.00026) -0.0357 (0.0513) -0.00040* (0.00023) (r-r*) 0.0027** (0.0013) 0.0172* (0.0090) 0.0106 (0.0098) -0.0094** (0.0043) -0.0044* (0.0027) -0.0121** (0.0055) FD -0.0281** (0.0123) 0.0550* (0.0345) 0.1258 (0.2929) 0.0541** (0.0276) 0.0155 (0.0391) 0.0263 (0.0449) Serial (1) [0.3292] [0.8919] [0.3162] [0.6866] [0.6987] [0.0861] Serial (2) [0.6187] [0.0845] [0.4993] [0.4126] [0.9146] [0.2293] Heero [0.1464] [0.4927] [0.0777] [0.0704] [0.0017] [0.0306] Over id. [0.6239] [0.2852] [0.3287] [0.3879] [0.8193] [0.1189] 28