Are Global Systematic Risk and Country-Specific Idiosyncratic Risk Priced in the. Integrated World Markets? Abstract

Transcription

1 Are Global Sysemaic Risk and Counry-Specific Idiosyncraic Risk Priced in he Inegraed World Markes? Absrac Empirical evidence showing significan effecs of local facors on inernaional equiy reurns while failing o find significan effecs from global sysemaic risk seems couner-inuiive in oday s inegraed world markes. This paper uses he condiional second momens esimaed from an asymmeric dynamic condiional correlaion model o measure he ime-varying world bea and counry-specific idiosyncraic risks, and ess he relaionship beween counry-level index reurns and world bea risk condiioned on posiive and negaive world marke reurns. The resuls show ha he condiional dynamic world bea risks significanly predic he crosscounry variaion in expeced index reurns, while counry-specific risk is no significanly priced. JEL Classificaion: G11, G12, G15 Keywords: World bea risk; Counry-specific idiosyncraic risk; Dynamic condiional correlaion model. i

2 1. Inroducion Mos would agree ha financial markes have become increasingly global in he pas decades. The recen financial crisis provides evidence of he srong inerlinkages among inernaional capial markes. Therefore, one would presume ha globalizaion has led o naional sock markes moving more closely ogeher and ha inernaionally raded capial asses should be globally priced. Noneheless, a large amoun of academic research devoes iself o recognizing wheher or no he world marke risk and counry-specific idiosyncraic risk are priced and coninues o poin o he significance of he effecs of local facors while failing o find significan effecs from global sysemaic risk [e.g., Cumby and Glen (1990), Ferson and Harvey (1994), Harvey and Zhou (1993), and Karolyi and Sulz (2003)]. The empirical evidence is summarized well in a recen sudy by Bali and Cakici (2010), who provide a general cross-secional es of global capial marke inegraion in an inernaional capial asse pricing model (ICAPM). Using 37 counry-level index daa and a global marke risk facor, hey show ha here is a posiive and significan relaionship beween expeced index reurns and counry-specific idiosyncraic risk, bu he relaionship beween a global-wide sysemaic risk and individual counry's expeced reurns is fla. They conclude ha he "finding ha he differences in counries' sock marke reurns can be explained by he differences in counry-specific risks is... consisen wih he view ha global sock markes are no fully inegraed." This conclusion has imporan implicaion o inernaional invesmen because i shows ha subsanial risk-reducion can be creaed from diversificaion. 1

3 The purpose of his paper is o challenge he above findings and provide more accurae evidence on global capial marke inegraion. The conribuion of he paper is wofold. Firs, on he observed weak global sysemaic effec, his paper argues ha he lieraure on condiional risk-reurn relaionship simulaed by Peengill e al. (1995) provides a possible explanaion. I saes ha he relaionship beween marke bea and realized reurns is condiional on he marke reurn. In up markes, high-bea securiies should be rewarded for bearing risk wih higher reurns han low-bea securiies, bu in down markes high-bea securiies experience lower reurns han low-bea securiies. Thus, i is necessary o pariion he daa ino up marke and down marke periods based on he sign of he realized marke excess reurn. Empirical sudies using daa from various counries mosly confirm a significan direc relaionship beween bea and reurns in up markes and a significan inverse relaionship beween bea and reurns in down markes. Among hem Flecher (2000) and Tang and Shum (2003) are of paricular ineres in ha hey use counry-level index daa and find a significan condiional relaionship beween index reurns and a world marke bea. However, Flecher uses a CAPM assuming full inegraion and does no discuss he counry-specific idiosyncraic risk, while Tang and Shum only consider exchange rae risk as he counry-specific risk. This paper uses a parial inegraion ICAPM and considers boh world bea risk and counry-specific idiosyncraic risk. The second conribuion of he paper is on he significance of he effecs of counryspecific risk on inernaional equiy index reurns, which is seldom challenged in he lieraure. One possible challenge worh exploring is he validiy of he measure of idiosyncraic risk. This issue has been raised in he lieraure of firm-level idiosyncraic risk, where monhly idiosyncraic volailiy is ofen measured by realized volailiy in he previous monh using 2

4 daily daa. 1 Spiegel and Wang (2005), Brockman e al. (2009), and Fu (2009) provide empirical evidence supporing ha, compared o he realized monhly idiosyncraic volailiy calculaed from daily daa, he condiional idiosyncraic volailiy esimaed from GARCH models using monhly daa is a more accurae proxy for expeced fuure idiosyncraic volailiy. 2 These sudies, however, all focus on he measure of idiosyncraic volailiies and do no apply ime-varying models o measure he dynamics of he sysemaic marke risk. Bali e al. (2012), on he oher hand, apply he dynamic condiional correlaion (DCC) model of Engle (2002) o consruc dynamic condiional marke beas and invesigae he significance of he condiional beas in predicing he cross-secional variaions in expeced reurns using firmlevel daa. 3 Bu hey do no discuss he idiosyncraic risk. This paper is he firs in he lieraure o consider boh dynamic idiosyncraic risk and dynamic sysemaic risk, and he firs o apply he DCC model o counry-level equiy indices and o he world marke inegraion issue. 1 This lieraure paricularly concerns he idiosyncraic volailiy puzzle raised by Ang e al. (2006). Measuring monhly idiosyncraic volailiy by realized volailiy in he previous monh using daily daa, hey find ha socks wih high idiosyncraic volailiies in he previous monh have abysmally low average monhly reurns. 2 Fu (2009) argues ha he idiosyncraic volailiy puzzle is caused by he use of he realized idiosyncraic volailiy, which is no a good predicor of he expeced idiosyncraic volailiy. Spiegel and Wang (2005), using monhly daa and EGARCH models, also find ha sock reurns are increasing wih he level of idiosyncraic volailiy. Brockman e al. (2009) apply Fu s EGARCH model o anoher se of inernaional index daa and make he same conclusion. On he oher hand, Hueng and Yau (2013) show ha realized idiosyncraic volailiies work as well as condiional idiosyncraic volailiies in predicing inernaional index reurns. 3 You and Daigler (2010) apply he DCC model o esimae he condiional correlaions among inernaional equiy markes and show ha he correlaions change over ime. 3

5 In sum, his paper revisis he relaionship among world marke risk, counry-specific risk, and expeced reurns in inernaional sock markes by proposing he following wo improvemens over he previous lieraure. Firs, his paper uses condiional second momens, insead of lagged realized second momens, o measure he sysemaic and idiosyncraic risks. The Asymmeric Dynamic Condiional Correlaion Mulivariae EGARCH (A-DCC-MV- EGARCH) model inroduced by Capiello e al. (2006) is used o model he ime-varying condiional world bea risk and o derive he condiional counry-specific idiosyncraic risk. Second, when running he Fama-MacBeh cross-secional regressions, his paper akes ino accoun he condiional relaionship beween he global sysemaic risk and he index reurns by pariioning he daa ino up marke and down marke periods based on he sign of he realized world marke reurns. The nex secion discusses he regression models ha are used o es he significance of he sysemaic and idiosyncraic risks in inernaional asse pricing. The A-DCC-MV- EGARCH model used o esimae he condiional second momens is also specified in his secion. Secion 3 describes he daa and examines he relevan sample saisics. The crosscounry risk-reurn analyses are shown in Secion 4. Secion 5 concludes he paper. 2. The Model If inernaional markes are compleely inegraed, he expeced marke reurns for counry i ( R ) depend on he covariance wih he world marke index reurns [Dumas and Solnik i, (1995)]: E R = λ Cov ( R, R ), (1) 1 i, 1 1 i, w, 4

6 where λ is he expeced world price of risk and R w, is he world reurn. For compleely segmened markes, only he variance of he index reurns of a counry affec he expeced reurns of a counry s index reurns: E R = λ Var ( R ). (2) 1 i, i, 1 1 i, where λ is he expeced price of counry i's idiosyncraic risk. i, Bali and Cakici (2010) generalize (1) and (2) and consider he following parialinegraion model: E R = λ Cov ( R, R ) + λ Var ( R ). (3) 1 i, 1 1 i, w, i, 1 1 i, To es he cross-secional predicive power of world marke risk and counry-specific risk under his parial inegraion model, hey use he following Fama and MacBeh (1973) regressions across counries in each monh : R = γ + γ Bea + γ IVOL + ε, (4) i, 0, 1, i, 1 2, i, 1 i, where γ is he price of world bea risk and 1, γ is he price of counry-specific risk. Counry 2, i's monhly world marke bea ( Bea ) is obained by regressing counry i's daily marke i, porfolio index reurns ( R ) on he daily world marke porfolio reurns ( i, d, R ): w, d, R = µ + Bea R + r, for d = 1, 2,..., D, (5) i, d, i, i, w, d, i, d, where D is he number of rading days in monh and r is he daily idiosyncraic reurn. i, d, The counry-specific idiosyncraic volailiy in monh is defined as he realized monhly 5

7 sandard deviaion of he daily idiosyncraic reurns: IVOL ( ) 2 i, = ri, d, ri, d, D d = 1. Tha is, hey run a wo-sage regression. In he firs sage, he daily daa are used in (5) o obain he esimaes of he monhly observaions for Bea and i, IVOL. Then in he second sage, hey i, use lagged realized risk measures [ Bea and i, 1 IVOLi, 1 ] o proxy for he expeced risks [ 1 i, E ( Bea ) and E 1 ( IVOL i, ) ] and run regression (4). Using counry-level aggregae marke index daa from 37 counries and a world marke porfolio index, Bali and Cakici find ha he ime-series average of he esimaed effec of IVOLi, on 1 i, ˆ R (i.e., γ 2, ) is posiive and saisically significan, while he relaionship beween average reurns on counries' sock marke indices and world marke bea (i.e., ˆ γ 1, ) is posiive bu insignifican. 4 This indicaes ha he counry-specific risk is priced and he price of his risk is he same across counries, bu he sysemaic world bea risk is no priced. When analyzing he firm-level daa in he U.S. sock marke, Peengill e al. (1995) argue ha he Fama-MacBeh mehodology canno direcly es he expeced risk-reurn relaionship implied by he CAPM. The CAPM shows ha since he expeced marke reurn mus be greaer han he risk-free rae (oherwise all invesors would hold he risk-free asses), he expeced reurn of any risky asse mus be a posiive funcion of bea. However, he Fama- MacBeh mehodology uilizes he realized reurns o proxy for he expeced reurns. Peengill e al. argue ha here mus be a non-zero probabiliy ha he realized marke reurn is smaller 4 Bali and Cakici (2010) also replace lagged idiosyncraic volailiy wih lagged counryspecific oal volailiy and reach he same conclusions on inegraion/segmenaion. 6

8 han he risk-free rae (oherwise no invesor would hold he risk-free asses). To solve his problem, hey pariion he marke ino an up marke and a down marke based on he sign of he realized marke excess reurn. A posiive risk-reurn relaionship should exis in he up marke and an inverse relaionship should exis in he down marke. In addiion, o es for a posiive risk and reurn radeoff, wo necessary condiions need o hold, namely a posiive excess marke reurn on average and a symmeric risk premium in up and down markes. Flecher (2000) and Tang and Shum (2003) adop Peengill e al.'s idea and use counry-level index daa o es he condiional relaionship beween index reurns and a world marke bea: R = γ + γ δ Eˆ Bea + γ (1 δ ) Eˆ Bea + ε, (6) + i, 0, 1, 1 i, 1, 1 i, i, where δ = 1 if R w, > 0 (an up world marke) and δ = 0 if R w, <0 (a down world marke). Flecher assumes ha he world bea is consan over ime ( Beai, = Bea ) and uses monhly i daa for he whole sample period ( ) o esimae E 1 Bea for each counry. Tang i, and Shum assume ha he world bea is consan wihin a year and is esimaed by using he daa over he pas five years in rolling regressions. Boh find evidence confirming Peengill e al.'s argumen. Bu apparenly hey do no consider counry-specific idiosyncraic risk. 5 This paper generalizes (4) and (6) and considers he following model: 5 Tang and Shum (2003) use counry indices denominaed in domesic currencies and consider exchange raes as an idiosyncraic risk facor. They include exchange rae in he firs sep of he regressions when esimaing he world bea o remove he counry-specific exchange rae effecs. Bu when esing he risk-reurn relaionship in he second sage, hey do no consider he idiosyncraic risk. 7

9 R = γ + γ δ Eˆ ( Bea ) + γ (1 δ ) Eˆ ( Bea ) + γ Eˆ ( IVOL ) + ε. (7) + i, 0, 1, 1 i, 1, 1 i, 2, 1 i, i, Unlike he aforemenioned sudies, his paper uses auoregressive condiional second momens o esimae he expeced world bea and counry-specific risk. Spiegel and Wang (2005), Brockman e al. (2009), and Fu (2009) all provide empirical evidence showing ha he condiional idiosyncraic volailiy esimaed from GARCH models using monhly daa is a more accurae proxy for expeced fuure idiosyncraic volailiy han he realized monhly idiosyncraic volailiy calculaed from daily daa. The same argumen can apply o he covariance beween individual counry index reurns and he world marke reurns [You and Daigler (2010) and Bali e al. (2012)]. Therefore, I consruc he measures of he expeced world marke bea and idiosyncraic volailiy by using he Asymmeric Dynamic Condiional Correlaion Mulivariae EGARCH (A-DCC-MV-EGARCH) model suggesed by Cappiello e al. (2006). Cappiello e al.'s asymmeric DCC model generalizes Engle's (2002) DCC model by allowing condiional asymmeries in correlaion dynamics. This modeling sraegy involves regressions in wo sages. In he firs sage a univariae GARCH model is esimaed for each asse. In he second sage, he ransformed residuals resuling from he firs sage are used o esimae a condiional correlaion esimaor. ( i, Specifically for he purpose of his paper, i is assumed ha counry i's index reurn R ) and he world marke index reurn ( R ) are condiionally bivariae normal wih zero w, hi, hiw, expeced value and covariance marix H hiw, h. In he firs sage, I use he following w, univariae EGARCH model o esimae he condiional variances h and i, h : w, 8

10 R R ln h = + ln h + +, j = i, w. j, 1 j, 1 j, κ j α j j, 1 β j γ j h j, 1 h j, 1 (8) This EGARCH model, proposed by Nelson (1991), akes accoun for he imporance of he asymmeric responses of condiional volailiies o posiive and negaive news. R j, In he second sage, le ε j, = be he sandardized reurns wih h esimaed from j, h j, he firs sage and ε = εi, ε w, '. Express he covariance marix as H = C P C, where C h, 0 i =. 0 hw, 1 1 Then P = C H C is he correlaion marix wih ones on he diagonal and he off-diagonal elemen less han or equal o one in absolue value. Engle (2002) suggess ha he diagonal elemen of he correlaion marix can be consruced by esimaing anoher univariae GARCH process. This compuaional advanage makes he DCC model more aracive han oher mulivariae GARCH specificaions such as he VEC model and he * 1 * BEKK represenaion [see Engle and Kroner (1995)]. Le P = Q Q Q 1, where q Q q q 11, 12, q 12, 22, is symmeric and Q * q11, 0 =, he marix version of he DCC model 0 q22, is given by ( ) parameers o be esimaed. ' Q = P a P b P + a ε ε + b Q where P = E( ε ε ) and a and b are ' 2, To allow for condiional asymmeric responses in correlaion dynamics o posiive and negaive news, Cappiello e al. generalize he evoluion of he correlaion o: Q = ( P a P b P g N ) + a ε ε + g n n + b Q, (9) ' 2 '

11 where [ 0] ' n = I ε < ε and N E( n n ) =. The indicaor funcion [ ] I equals one if he argumen is rue and zero oherwise, and is he Hadamard produc. To esimae he model, P and N are proxied by heir sample analogues and Q is se o equal 0 P. A sufficien condiion for Q o be posiive definie is a posiive semi-definie ( P a P b P g N ). A necessary and sufficien condiion for his o hold is a + b + ξ g < 1, where ξ is he maximum eigenvalue of P 1/2 N P 1/2. Condiioning on he parameers esimaed in he firs 1 log P + ε P ε. 2 ' 1 sage, a, b, and g can be esimaed by maximizing he likelihood: ( ) Once P is esimaed in he second sage, combining C esimaed from he firs sage yields he esimae of hiw, H. Then he ime-varying condiional correlaion is h h i, w,, he, condiional world marke bea is measured as ˆ hiw E 1 ( Beai, ) =, and he idiosyncraic reurn is h consruced as r ˆ ( ) i, = Ri, E 1 Beai, R. The condiional idiosyncraic volailiy w, E ˆ ( ) 1 IVOLi, is he squared roo of he condiional variance from anoher univariae EGARCH process (8) w, for i, r. Once E ˆ ( ) 1 Beai, and 1 i, E ˆ ( IVOL ) are obained from he A-DCC-MV-EGARCH model in he firs sage, hey are plugged ino he cross-secional regression (7) in he second sage. 3. Daa I use he same daa as hose in Bali and Cakici (2010), whose daa end in Sepember 2006, bu updae heir daa o May The daa are obained from Daasream Global indices and 10

12 include sock marke indices for 37 counries plus he world marke porfolio, all denominaed in U.S. dollars. There are 23 developed markes and 14 developing or emerging markes. 6 Panel (A) of Table 1 shows he summary saisics of he monhly marke index reurns for each counry and he world marke, including he means, sandard deviaions, and consan correlaions wih he world marke index reurns. The saring monh for each counry is shown in he second column. The sample ends in May 2012 for all counries. Compared o hose repored in Bali and Cakici, he updaed daa show slighly lower means and sandard deviaions. More ineresingly, he consan correlaions wih he world are higher in mos of he counries wih he new daa added, indicaing ha he global markes are geing more inegraed afer heir sudy. 7 Comparing across counries, he summary saisics are in general very similar o hose in Bali and Cakici in ha he emerging markes exhibi higher average reurns and higher sandard deviaions of reurns compared o he developed markes. In Panel (B) of Table 1, he firs four columns show he means, sandard deviaions, maximum values, and minimum values of he ime-varying correlaions wih he world esimaed from he A-DCC-MV-EGARCH model. As can be seen for mos counries, he 6 Ince and Porer (2006) poin ou several daa problems in Daasream. Since mos of he problems idenified in heir paper are concenraed in he smaller size firms, o make sure ha he poenial problems do no change my conclusions, I also do he analyses using an alernaive daa source - he MSCI large- and mid-cap price reurns, available from January This daase also avoids he problem from using index reurns. These resuls do no change he conclusions in he paper and are available from he auhor upon reques. 7 Even hough here I follow Bali and Cakici (2010) and preliminarily use he average correlaions wih he world marke as an alernaive measure of inegraion, noe ha i is wellknown in he lieraure ha correlaion is no a good measure of inegraion. See, for example, Dumas e al. (2003) and Carrieri e al. (2007). 11

13 variaions of he correlaions are significan. When regressing he dynamic correlaions on a consan and a linear ime rend, I find ha he slope agains he ime rend (repored in he fifh column) is posiive and saisically significan for all counries excep for Finland (an insignificanly posiive slope) and Japan (an insignificanly negaive slope). This indicaes ha he degree of inegraion wih he world is increasing over ime in mos of he markes. Panels (A) and (B) of Table 2 repor he summary saisics for he realized world marke bea, Beai,, and he condiional measure of world marke bea, 1 E ˆ ( ) 1 Beai,, respecively. The realized bea, used in Bali and Cakici, is esimaed by using daily daa wihin each monh o run ime-series regression (5). In general he realized bea in Panel (A) shows high volailiies, wih he sandard deviaion as high as he mean in mos of he markes. The firsorder auocorrelaion of he realized bea is mosly lower han 0.5, indicaing ha pas realized marke bea is no a good predicor of fuure marke bea. On he oher hand, in Panel (B), he condiional measure of bea esimaed from he A-DCC-MV-EGARCH model is more sable, wih he sandard deviaion much lower han he mean in mos of he markes. In addiion, he condiional measure of bea is very persisen, wih he firs-order auocorrelaion coefficien mosly higher han 0.9. Furhermore, he las column of Table 2 shows ha he correlaions beween he realized bea and he condiional bea are mosly lower han 0.5. Therefore, i is expeced ha using he condiional bea would yield differen resuls han hose in he previous sudies ha use he realized bea. Panels (A) and (B) of Table 3 repor he summary saisics of he realized idiosyncraic volailiy, IVOLi,, and he condiional measure of idiosyncraic volailiy, 1 E ˆ ( ) 1 IVOLi,, respecively. The realized monhly idiosyncraic volailiy is calculaed from he daily 12

14 idiosyncraic reurns esimaed in equaion (5). Compared o hose repored in Bali and Cakici, he realized counry-specific volailiies are in general lower wih he updaed daa added. Bu he cross-counry comparison is in general consisen wih hose repored in Bali and Cakici in ha he counry-specific idiosyncraic volailiy is much higher in he emerging markes han in he developed markes. The firs-order auocorrelaions are around 0.5 for mos counries, which again indicaes ha pas realized idiosyncraic volailiy is no a very good predicor of fuure idiosyncraic volailiy. The condiional idiosyncraic volailiy esimaed from he A- DCC-MV-EGARCH model, on he oher hand, generally has a higher mean and is more sable compared o he realized measure. The condiional measure is also very persisen. Finally, he las column of Table 3 shows ha he correlaions beween he realized measure and he condiional measure are mosly lower han 0.5. Therefore, i is imporan o replace he realized measures wih he condiional measures in order o es he risk-reurn relaionship. 4. Empirical Resuls The empirical work sars by revisiing Bali and Cakici's (2010) resuls in equaion (4), where he Fama and MacBeh (1973) regression is used for he exended sample period o examine he cross-counry relaionship beween expeced reurns and he realized measures of risks. Following Bali and Cakici, I include wo more conrol variables, he earnings-o-price raio and he dividends-o-price raio, in he regressions. The firs row of Panel (A) in Table 4 repors he ime-series averages of he esimaed coefficiens, heir p-values based on he Newey and Wes (1987) heeroscedasiciy- and auocorrelaion-adjused -saisics, and he ime-series averages of he R-squared. The newly added daa do no aler Bali and Cakici's conclusions: he counry-specific risk is priced and is effec on index reurns is saisically 13

15 significan. The effec of he sysemaic risk on index reurns, on he oher hand, is highly insignifican. The second row of Panel (A) shows he resuls from using he condiional measures of he world marke bea and idiosyncraic risks. The condiional measures fi he model beer by raising he R-squared by 1.7%. More imporanly, he effec of he idiosyncraic risk on he index reurns is lowered and only saisically significan a he 7.7% level. Recall ha he world markes are more inegraed afer Bali and Cakici's sudy. Therefore, one would expec he significan effec of he counry-specific risk on reurns found in heir paper o decline in he new sample. Using he condiional measure, he resuls of his paper are consisen wih his expecaion. 8 Nex I urn he aenion o he sysemaic risk. Consisen wih previous sudies, he resuls in Panel (A) of Table 4 show ha here is no significan uncondiional relaionship beween expeced reurns and he world marke bea. However, hese resuls from he Fama- MacBeh regressions are apparenly subjec o Peengill e al.'s (1995) criics and canno direcly es he expeced risk-reurn relaionship implied by he CAPM. To conduc a valid es, he Fama-MacBeh regressions should be condiioned on posiive and negaive world marke reurns. The resuls of he condiional regression (7) in he up and down world markes are shown in Panel (B) of Table 4. Consisen wih Flecher (2000) and Tang and Shum 8 I also run a regression wih condiional measures using only daa prior o he recen financial crisis (daa up o 2007). The conclusions do no change. The esimaed coefficien on he idiosyncraic risk is higher (0.105) in he pre-crisis period han ha (0.094) in he whole sample period, and is again only marginally significan a he 6.7% level. This resul suppors he argumen ha he world marke is more inegraed during he financial crisis. 14

16 (2003), when he sample is pariioned ino up and down world markes, here is a significan posiive relaionship beween index reurns and he world bea in up marke monhs and a significan and negaive relaionship in down marke monhs. Therefore, high bea indices ouperform low bea indices when he realized world marke reurn is posiive, bu incur higher losses when he realized world marke reurn is negaive. To es for a posiive sysemaic risk and reurn radeoff, he las column of Panel (B) repors he es saisics of he wo-sample -es for a symmeric risk premium in up and down marke monhs. The symmeric relaionship canno be rejeced a he convenional significance level. Combined wih he posiive mean world reurn, he evidence suppors he expecaion of a posiive reward for bearing sysemaic risk. Along wih he resuls in Panel (A), afer adding he mos recen daa when he world markes are highly inegraed, he world marke bea remains significan bu he counry-specific risk is only marginally significan. Furher suppor from a differen sample The empirical resuls so far provide more inuiive evidence ha counry-specific risk plays a lesser role in deermining inernaional equiy reurns in a more inegraed world capial marke. To furher suppor his argumen, I consider a more recen sample period when he global marke is increasingly more inegraed. Sudies have provided evidence supporing ha he degree of world financial marke inegraion has increased since he 1990s. For example, Carrieri e al. (2007) use an inegraion index o show ha emerging markes have become 15

17 more inegraed wih oher global markes afer he early 1990s. 9 Hardouvelis e al. (2006) find ha European markes became fully inegraed in he second half of he 1990s due o he formaion of he European Union. You and Daigler (2010) use ime-varying correlaions o show ha he correlaions of inernaional index markes have increased since he lae 1990s. Therefore, I consider daa saring from he 1990s o see wheher he effec of counry-specific risk on inernaional equiy reurns is insignifican in a more inegraed world capial marke. A larger pool of inernaional index daa is available saring from June The MSCI Invesable Marke Indices cover all invesable large, mid, and small cap securiies across 24 developed markes, 21 emerging markes, and 26 fronier markes, a oal of 71 markes plus he world marke porfolio. In Appendix, Tables A1-A3 repor he summary saisics of he daa for he developed markes, he emerging markes, and he fronier markes, respecively. Since he daa on he earnings-o-price raio and he dividends-o-price raio are no available for half of he counries, hese wo variables are no included in he regression. Since he resuls in Secion (II) of Table 4 show ha hese wo variables do no have significan effecs on reurns when he condiional measures are used, i should no be oo unreasonable o assume ha ignoring hese wo variables would no change he conclusions. 10 Table 5 repors he esimaion resuls. There is a significan and posiive relaionship beween index reurns and he world bea in up marke monhs and a significan and negaive 9 Bekaer and Harvey (1995, 2000) imply ha liberalizaion of he financial markes of emerging counries causes furher inegraion of inernaional sock markes. Bekaer, Harvey, and Lumsdaine (2002) show ha significan break poins exis for facors ha indicae world financial marke inegraion for several emerging markes afer he liberalizaion. 10 Flecher (2000) and Tang and Shum (2003) do no include hese wo variables in heir analyses, eiher. 16

18 relaionship in down marke monhs. The relaionships in he up and down markes are symmeric. Therefore, he sysemaic risk is priced. On he oher hand, I find no evidence of a priced counry-specific idiosyncraic risk. The relaionship beween index reurns and idiosyncraic risk is highly insignifican. This resul indicaes ha he global capial markes are highly inegraed so ha he counry-specific idiosyncraic risk is no imporan in pricing inernaional equiy indices. 5. Conclusions Empirical evidence poining o he significan effecs of local facors on global equiy reurns while failing o find significan effecs from global sysemaic risk seems couner-inuiive in oday s inegraed world markes. This paper revisis his empirical issue by providing more accurae evidence. Specifically, an asymmeric dynamic condiional correlaion mulivariae EGARCH model is used o esimae he ime-varying condiional world marke bea risk and o derive he condiional counry-specific idiosyncraic risk. In addiion, when using he Fama and MacBeh (1973) mehodology o es he significance of he relaionship beween world bea risk and counry-level index reurns, his paper akes ino accoun he condiional relaionship beween he global sysemaic risk and index reurns by pariioning he daa ino up marke and down marke periods based on he sign of he realized world marke reurns. Using 37 counry-level index daa and a global marke risk facor from 1973 o 2012, and a larger pool of daa from 71 counries for he period , his paper shows ha he condiional world bea risks significanly affec counry-level index reurns, while counryspecific risk facors are no significanly priced. The resuls, herefore, suppor inernaional financial inegraion. 17

19 References Ang, A., Hodrick, R. J., Xing, Y., and Zhang, X. (2006). The Cross-Secion of Volailiy and Expeced Reurns. Journal of Finance 61, Bali, T. G., Engle, R. F. and Tang, Y. (2012). Dynamic Condiional Bea is Alive and Well in he Cross-Secion of Daily Sock Reurns. Fordham Universiy Schools of Business Research Paper No Available a SSRN: hp://ssrn.com/absrac= Bali, T.G. and Cakici, N. (2010). World Marke Risk, Counry-Specific Risk and Expeced Reurns in Inernaional Sock Markes. Journal of Banking and Finance 34, Bekaer, G. and Harvey, C. (1995). Time-Varying World Marke Inegraion. Journal of Finance 50, Bekaer, G. and Harvey, C. (2000). Foreign Speculaors and Emerging Equiy Markes. Journal of Finance 55, Bekaer, G., Harvey, C. and Lumsdaine, R. (2002). Daing he Inegraion of World Equiy Markes. Journal of Financial Economics 65, Brockman, P., Schue, M.G. and Yu, W. (2009). Is Idiosyncraic Risk Priced? The Inernaional Evidence. Working paper. Available a SSRN: hp://ssrn.com/absrac= Cappiello, L. Engle, R. F. and Sheppard, K. (2006). Asymmeric Dynamics in he Correlaions of Global Equiy and Bond Reurns. Journal of Financial Economerics 4,

20 Carrieri, F., Errunza, V. and Hogan, K. (2007). Characerizing World Marke Inegraion hrough Time. Journal of Financial and Quaniaive Analysis 42, Cumby, R. E. and Glen, J. D. (1990). Evaluaing he Performance of Inernaional Muual Funds. Journal of Finance 45, Dumas, B., Harvey, C., Ruiz, P. (2003). Are Correlaions of Sock Reurns Jusified by Subsequen Changes in Naional Oupus? Journal of Inernaional Money and Finance 22, Dumas, B., and Solnik, B. (1995). The World Price of Foreign Exchange Risk. Journal of Finance 50, Engle R. F. (2002). Dynamic Condiional Correlaion - A Simple Class of Mulivariae GARCH Models. Journal of Business and Economic Saisics 20, Engle R. F. and Kroner, K. (1995). Mulivariae Simulaneous GARCH. Economeric Theory 11, Fama, E. and MacBeh, J. (1973). Risk, Reurn, and Equilibrium: Empirical Tess. Journal of Poliical Economy 81, Ferson, W. F. and Harvey, C. (1994). Sources of Risk and Expeced Reurns in Global Equiy Markes. Journal of Banking and Finance 18, Flecher, J. (2000). On he Condiional Relaionship beween Bea and Reurns in Inernaional Sock Reurns. Inernaional Review of Financial Analysis 9,

21 Fu, F. (2009). Idiosyncraic Risk and he Cross-Secion of Expeced Sock Reurns. Journal of Financial Economics 91, Hardouvelis, G., Malliaropulos, D. and Priesley, R. (2006). EMU and European Sock Marke Inegraion. Journal of Business 79, Harvey, C. and Zhou, G. (1993). Inernaional Asse Pricing wih Alernaive Disribuional Specificaions. Journal of Empirical Finance 1, Hueng, J. and Yau, R. (2013). Counry-Specific Idiosyncraic Risk and Global Equiy Index Reurns. Inernaional Review of Economics & Finance, 25, Ince, O. S. and Porer, R. B. (2006). Individual Equiy Reurn Daa from Thomson Daasream: Handle wih Care! Journal of Financial Research, 29, Karolyi G. A. and Sulz, R. M. (2003). Are Financial Asses Priced Locally or Globally? Handbook of he Economics of Fnance. Vol 1B: Newey, W. K. & Wes, K. D. (1987). A Simple, Posiive Semi-Definie, Heeroskedasiciy and Auocorrelaion Consisen Covariance Marix. Economerica 55, Nelson, D. (1991). Condiional Heeroskedasiciy in Asse Reurns: a New Approach. Economerica 59, Peengill, G., Sundaram, S. and Mahur, I. (1995). The Condiional Relaion Beween Bea and Reurns. Journal of Financial and Quaniaive Analysis 30,

22 Tang, G. Y. N., & Shum, W.C. (2003). The Condiional Relaionship beween Bea and Reurns: Recen Evidence from Inernaional Sock Markes. Inernaional Business Review 12, You, L. & Daigler, R. T. (2010). Is Inernaional Diversificaion Really Beneficial? Journal of Banking and Finance 34,

23 Table 1: Summary Saisics of Inernaional Marke Indices The daa are from Daasream Global indices. The sample ends in May 2012 for all counries. The dynamic correlaions wih he world are esimaed from he A-DCC-MV-GARCH model. The aserisk * indicaes saisical significance a he 5% level, and ** a he 10% level. Counry Daa sar (A) Marke Index Reurns Mean Sd. Dev. Cons. Corr. w/ WORLD (B) Dynamic Correlaion wih World Trend Mean Sd Max Min Slope x10 3 Argenina Aug ** Ausralia Jan * Ausria Jan * Belgium Jan * Brazil Jul * Canada Jan * Chile Jul * China Jul * Denmark Jan * Finland Mar France Jan * Germany Jan * Greece Jan * Hong Kong Jan * India Jan * Ireland Jan * Ialy Jan * Japan Jan Korea Sep * Malaysia Jan * Mexico May * Neherlands Jan * New Zealand Jan * Norway Jan * Philippines Nov * Poland Mar * Porugal Jan * Singapore Jan * Souh Africa Jan * Spain Mar * Sweden Jan * Swizerland Jan * Taiwan May * Thailand Jan * Turkey Jun * UK Jan * US Jan * WORLD Jan

24 Table 2: Summary Saisics of he Realized and Condiional Marke Beas The realized bea is esimaed by equaion (5) using daily daa. The condiional bea is esimaed from he A-DCC-MV-GARCH model using monhly daa. Counry (A) Realized Bea (B) Condiional Bea Corr. beween Realized Mean Sd. Auo corr. Mean Sd. Auo corr. and Condiional Beas Argenina Ausralia Ausria Belgium Brazil Canada Chile China Denmark Finland France Germany Greece Hong Kong India Ireland Ialy Japan Korea Malaysia Mexico Neherlands New Zealand Norway Philippines Poland Porugal Singapore Souh Africa Spain Sweden Swizerland Taiwan Thailand Turkey UK US

25 Table 3: Summary Saisics of he Realized and Condiional Idiosyncraic Volailiies The realized idiosyncraic volailiy is calculaed from equaion (5) using daily daa. The condiional idiosyncraic volailiy is esimaed from he DCC-MV-GARCH model using monhly daa. (A) Realized IVOL (B) Condiional IVOL Corr. beween Realized Counry Mean Sd. Auo corr. Mean Sd. Auo corr. and Condiional IVOLs Argenina Ausralia Ausria Belgium Brazil Canada Chile China Denmark Finland France Germany Greece Hong Kong India Ireland Ialy Japan Korea Malaysia Mexico Neherlands New Zealand Norway Philippines Poland Porugal Singapore Souh Africa Spain Sweden Swizerland Taiwan Thailand Turkey UK US

26 In Panel (A), (I) repors he cross-secional regression resuls from: R = γ + γ Bea + γ IVOL + γ EP + γ DY + ε, i, 0, 1, i, 1 2, i, 1 3, i, 1 4, i, 1 i, Table 4: Cross-Secional Regressions where he monhly Bea is obained by regressing counry i's daily marke reurns on he daily world marke reurns [Equaion (5)] i, and he idiosyncraic volailiy IVOL is he realized monhly sandard deviaion of he daily idiosyncraic reurns obained from he i, regression. The conrol variable EP is he naural logarihm of he earnings-o-price raio, and i, DY is he naural logarihm of he i, dividends-o-price raio in monh. Rows (II) and (III) of Panel (A) repors he cross-secional regression resuls from: R = γ + γ Eˆ ( Bea ) + γ Eˆ ( IVOL ) + γ EP + γ DY + ε, i, 0, 1, 1 i, 2, 1 i, 3, i, 1 4, i, 1 i, where he monhly observaions of E ˆ ( ) 1 Beai, and 1 i, EGARCH model [Equaions (8) and (9)]. In Panel (B), (I) repors he cross-secional regression resuls from: + R = γ + γ D Bea + γ (1 D ) Bea + γ IVOL + γ EP + γ DY + ε, i, 0, 1, i, 1 1, i, 1 2, i, 1 3, i, 1 4, i, 1 i, E ˆ ( IVOL ) are obained from he regression resuls of he A-DCC-MV- and (II) and (III) repor he cross-secional regression resuls from: + ˆ R = γ + γ D E ( Bea ) + γ (1 D ) Eˆ ( Bea ) + γ Eˆ ( IVOL ) + γ EP + γ DY + ε. i, 0, 1, 1 i, 1, 1 i, 2, 1 i, 3, i, 1 4, i, 1 i, The average inerceps, average slope coefficiens, and average R 2 are presened. The numbers in parenheses are P-values calculaed based on Newey and Wes (1987) -saisics. A P-value of indicaes ha he P-value is nonzero, bu smaller han Model (I) Realized Measures (II) Condiional Measures (A) γ 0 γ 1 γ 2 γ 3 γ 4 (B) Average R 2 γ H : γ = γ + γ P-Value (0.001) (0.549) (0.036) (0.001) (0.940) (0.000) (0.001) (0.186) (0.511) (0.077) (0.246) (0.509) (0.000) (0.000) 25

27 Table 5: Cross-Secional Regressions for 71 Markes This able repors he cross-secional regression resuls from: R = γ + γ D Eˆ ( Bea ) + γ (1 D ) Eˆ ( Bea ) + γ Eˆ ( IVOL ) + ε. + i, 0, 1, 1 i, 1, 1 i, 2, 1 i, i, The average inerceps, average slope coefficiens, and average R 2 are presened. The numbers in parenheses are p-values calculaed based on Newey and Wes (1987) -saisics. A p-value of indicaes ha he p-value is nonzero, bu smaller han γ 0 γ + 1 γ γ 1 2 Average R 2 + H :γ = γ p-value (0.237) (0.000) (0.000) (0.710) 26

28 Appendix: The daa in Tables A1-A3 are from MSCI Invesable Marke Indices. The sample ends in May 2012 for all counries. The dynamic correlaions wih he world are esimaed from he A- DCC-MV-GARCH model. A value of indicaes ha i is nonzero bu smaller han Table A1: Summary Saisics of Inernaional Marke Indices (Developed Markes) (A) Marke Index Reurns (B) Dynamic Correlaion wih World Counry Daa sar wih World Consan Mean Sd Correlaion Mean Sd Max Min AUSTRALIA May AUSTRIA May BELGIUM May CANADA May DENMARK May FINLAND May FRANCE May GERMANY May GREECE May HONG KONG May IRELAND May ISRAEL May ITALY May JAPAN May NETHERLANDS May NEW ZEALAND May NORWAY May PORTUGAL May SINGAPORE May SPAIN May SWEDEN May SWITZERLAND May UNITED KINGDOM May USA May WORLD May

29 Table A2: Summary Saisics of Inernaional Marke Indices (Emerging Markes) (A) Marke Index Reurns (B) Dynamic Correlaion wih World Counry Daa sar wih World Consan Mean Sd Correlaion Mean Sd Max Min BRAZIL May CHILE May CHINA May COLOMBIA May CZECH REPUBLIC May EGYPT May HUNGARY May INDIA May INDONESIA May KOREA May MALAYSIA May MEXICO May MOROCCO May PERU May PHILIPPINES May POLAND May RUSSIA May SOUTH AFRICA May TAIWAN May THAILAND May TURKEY May

30 Table A3: Summary Saisics of Inernaional Marke Indices (Fronier Markes) (A) Marke Index Reurns (B) Dynamic Correlaion wih World Counry Daa sar wih World Consan Mean Sd Correlaion Mean Sd Max Min ARGENTINA Nov BAHRAIN May BANGLADESH Nov HERZEGOVINA Nov BOTSWANA Nov BULGARIA Nov CROATIA Nov ESTONIA Nov GHANA Nov JAMAICA Nov JORDAN May KENYA Nov KUWAIT May LITHUANIA Nov MAURITIUS Nov NIGERIA Nov OMAN May PAKISTAN Nov QATAR May ROMANIA Nov SLOVENIA Nov SRI LANKA Nov TOBAGO Nov TUNISIA Nov UAE May ZIMBABWE Nov