On International Financial Spillovers to Frontier Markets

Transcription

1 On Inernaional Financial Spillovers o Fronier Markes Galin Todorov Prasad Bidarkoa Florida Inernaional Universiy Fall 2010 Absrac We explore he degree o which sock index reurns and condiional volailiy of 21 fronier markes were affeced by flucuaions on he American sock marke beween December 1s 2005 and January 15h We find weak, posiive reurn spillovers from US o 17 fronier markes. For four counries, Jordan, Lebanon, Nigeria, and Kenya, we find weak negaive reurn spillovers from he US, implying possible diversificaion opporuniies. For hireen markes he influence of pas local shocks is greaer han he influence of curren shocks from he US, and for sixeen markes local pas volailiy has sronger impac han volailiy from he US. Key phrases: Fronier markes; Emerging markes; spillovers; conagion; imevarying volailiy JEL codes: F36, G15, C58 Corresponding auhor: Galin Todorov, Deparmen of Economics, Florida Inernaional Universiy, FL 33199, USA, elephone:(305) , godorov@fiu.edu. Address: Deparmen of Economics, Florida Inernaional Universiy, FL 33199, USA, elephone:(305) , godorov@fiu.edu. Address: Deparmen of Economics, Florida Inernaional Universiy, FL 33199, USA, elephone:(305) , bidarko@fiu.edu.

2 1. Inroducion The main issue invesigaed in his aricle is he exen o which conemporaneous reurns and condiional volailiy of 21 fronier markes were affeced by he flucuaions in reurns and condiional volailiy on he American sock marke during he period beween December 1s, 2005 and January 15h, The World Bank defines emerging markes as markes for which GDP per capia falls bellow a cerain ime-dependen hurdle. Fronier markes are defined as emerging markes ha are invesable, bu have lower capializaion and marke liquidiy compared o he more developed emerging markes. Financial spillovers exis due o real economic and financial ies beween world economies. As a resul of he exisence of such ies, new informaion arising in one counry affecs no only local reurns and volailiy, bu also he reurns and volailiy of asses raded on oher markes. The impac of a change in reurns of one sock marke on he reurns of anoher sock marke is defined as reurns spillovers, and he impac of a change in reurns volailiy of one sock marke on he reurns volailiy of anoher sock marke is defined as volailiy spillovers. The new informaion may be absorbed immediaely by oher markes, or wih a delay, depending on he presence and number of informed invesors, informaion asymmery, level of marke liquidiy, exisence of feed back raders and herding behavior, high ransacion coss and oher poenial marke specific facors. The magniude and speed of spillovers provide valuable insigh ino he naure and swifness of disseminaion of such new informaion among counries. The size of spillovers reflecs how global invesors feel abou news, as well as heir appraisal of is impac on asse prices across markes. In order o ravel across borders, informaion needs ransmission channels. In he shor run, asse price changes are he primary channel for ransmission of financial shocks across borders. Owing o he dependence of fronier markes on common bank crediors and cross marke porfolio re-balancing by invesmen funds, financial markes and insiuions have been shown o ac as a major ool for cross-border shock ransmission (Kodres and Prisker (2002), Calvo (1999), Prisker (2001)). Deailed review of spillover ransmission channels is offered by Prisker (2001), and Kaminsky and Reinhar (2002). The grea majoriy of heoreical and empirical work has so far been concenraed on exploring spillovers among he maure and more developed emerging markes. However, he growing size and imporance of fronier markes naurally draw considerable ineres from invesors, policy makers, and academics alike. In his sudy we exend he exising lieraure by analyzing he exen o which small markes are vulnerable o shocks from he USA, as represened by he reurns and volailiy spillovers from he US marke o 21 fronier sock markes. The exposure of he smalles developing markes o financial shocks from US is a maer of subsanial concern for inernaional invesors. Fronier markes are ofen 1

3 considered very risky. Tha however, migh no sem from hose markes no being fundamenally sound; i migh jus be ha invesors do no know much abou hem. This sudy may provide financiers wih furher informaion abou poenial invesmen and porfolio diversificaion opporuniies in hose 21 small economies. Direc invesmen, projec risk evaluaion, cos of capial calculaion, asse pricing and allocaion, in addiion o he developmen of hedging echniques can poenially benefi from his aricle as well. Policy makers may ake advanage of his sudy since undersanding iner-marke connecions could provide for more informed decisions, improve macroeconomic managemen, as well as advance heir abiliy o ime, predic, and evaluae suscepibiliy of a counry o shocks from abroad. Improved assessmen of he naure and origin of a financial shock and possible subsequen economic downurn may faciliae he adapaion of more appropriae ani-crisis echniques and hus alleviae, or a leas shoren, he suffering of hose mos affeced by he deerioraion of economic condiions. Exploraion of poenial iner-marke linkages may also aid academics in shedding more ligh on he oucome of marke liberalizaion on capial flows and mobiliy. Enhanced awareness of marke co-movemen may expand heir undersanding of he significance of he srucure and poenial effecs of free cash flows as well as any subsequen resricions. Furhermore, his sudy may assis academics in forecasing and evaluaing he reacion of inernaional financial markes o global and local shocks, as well as in achieving deeper undersanding of he shock ransmission mechanisms. Las bu no leas, when analyzing hese economies, we should no be confused wih he meaning of he word small: i refers only o he per capia income, no o he number of people living in hose economies. The populaion of Argenina alone is close o 40 million, Romanian populaion is over 20 million, and in any one of hose small economies live people sruggling o survive every day. This paper will ry o answer he quesion of how quickly, and how badly were hose markes affeced by he recen economic downurn and may even provide some insighs abou fuure macroeconomic policy course. The evidence presened in his aricle suggess ha spillovers from USA o fronier markes are raher weak, albei saisically significan. One possible implicaion is ha he global economic downurn worked is way in hese counries hrough a differen ransmission channel, or ha he economic deerioraion in hose fronier markes is due mainly o an idiosyncraic counry specific shock. Furhermore, our resuls pu forward he noion ha i is no he lagged US marke reurns ha have influence, raher i is he expeced US marke reurns ha have impac on fronier marke reurns. The inference is ha despie differen opening hours, non synchronous rading, sale orders, informaion asymmery and oher possible small marke inefficiencies, he fronier markes absorb news from he USA almos immediaely a he ime of he news release. This 2

4 leads us o he conjecure ha fronier marke migh no a all be inefficien, raher, heir real and financial secors migh be resilien o influence from he US. From he perspecive of a global financier, his finding means ha here migh exis some diversificaion benefis from invesing in hese developing economies. From he perspecive of a developing counry policy maker i means ha a poenial increase in capial flows from he US does no necessarily increase local marke volailiy even in imes when he US economy is deerioraing. In secion 2 of his aricle we proceed wih relevan lieraure review on financial spillovers; in secion 3 we presen he daa and offer some descripive saisics; secion 4 develops he empirical models; in secion 5 we repor and discuss our resuls; secion 6 concludes. 2. Lieraure Review Inerdependency and correlaions among world financial markes have been invesigaed since he mid 1960s, however, his area of inernaional finance gained mos of is populariy afer he sock marke crash of 1987 wih King and Wadhwani (1990) esablishing he unidirecional impac of he US sock reurns on oher markes, Hamao, Masulis, and Ng (1990) confirming he unidirecional impac of US marke reurn volailiy, and Eun and Shim (1989) examining he inerdependency across nine sock markes. Harvey (1995) found ha adding emerging marke asses o a porfolio significanly enhances is opporuniy se. He showed ha he exposure of emerging markes o common facors is low and ha i is more likely ha local variables have sronger influence han do global variables. Aggarwal, Inclan, and Leal (1999) suppor he las claim, reporing ha imporan local evens in each counry, raher han global facors, end o be associaed wih sudden changes in volailiy. Bekaer and Harvey (1997) ascerain ha capial marke liberalizaion increases reurn spillovers across markes bu does no affec marke volailiy, while Tanizaki and Hamori (2009) deermine ha marke volailiy is increasing in he amoun of informaion available. From asse allocaion poin of view, increasing spillover effecs are generally associaed wih an increase in cross marke correlaions and hus reduced opporuniies for cross border porfolio diversificaion (Bekaer, Harvey (2000), (2003)). The lieraure on financial spillovers received a significan boos afer he 1997 Asian crisis. Sola, Spagnolo, and Spagnolo (2001) find unidirecional volailiy spillovers from Thailand, he counry where he crisis originaed, o mos of he counries in he region. Caporale, Cippollini, and Spagnolo (2003) confirm unidirecional spillovers in reurns from he Thai sock marke. Kim (2005) and Gebka and Serwa (2006) affirm he significan influence of he American marke on he Asian markes a all imes, and Gebka and Serwa (2007) find significan conemporaneous spillovers from he USA during he same period. 3

5 Eger and Kocenda (2009), as well as Fadhlaoui, Bellalh, Dherry, and Zonaouri ( 2009) find some evidence for shor erm spillovers, bu no long erm relaions, beween developed and emerging markes of Cenral and Easern Europe. Yu and Hassan (2006) and Al-Kulaib, Najand, and Mashayekh (2009) find no spillovers from he USA o he Middle Eas, Norh African, and Gulf Cooperaion Counries, while Chen, Fir, and Rui (2000) sugges high correlaion of Lain American counries wih world markes and hus limied poenial for porfolio diversificaion. Psillaki, and Margariis (2008) claim no long erm relaionship, bu some shor erm inerdependence beween he USA sock marke and he French and German sock markes. Pollard, Sapra, and Canarella (2007), on he oher hand, find significan impac from boh he USA marke reurns and volailiy on he Canadian and Mexican markes. Beirne, Caporale, Schulze-Ghas and Spagnolo (2008) sugges ha maure markes influence he condiional variances in many emerging markes. Sgherri and Galesi (2009) clarify for 27 counries ha asse prices are he main channel of ransmission of financial shocks inernaionally in he shor run. 3. Daa In his aricle we use daily MSCI Barra index closing prices o explore he reurns on he US sock marke and 21 fronier markes for he period from December 1s, 2005 o January 15h, The daily reurns for each counry are calculaed as follows: R i = ln(p i /P i(-1) ) where P i is he value for each counry s index closing price a ime. We choose daily daa since i will beer accoun for he sock marke dynamics and provides greaer insigh on cross-marke ineracions.the counries included are: Argenina, Bahrain, Bulgaria, Croaia, Esonia, Jordan, Kazakhsan, Kenya, Kuwai, Lebanon, Mauriius, Nigeria, Oman, Pakisan, Qaar, Romania, Saudi Arabia, Slovenia, Sri Lanka, Tunisia, and Unied Arab Emiraes. The counries and duraion of he period under sudy are chosen such ha he longes index series is available for he greaes number of counries. Lihuania, Serbia, Ukraine, Bangladesh, Trinidad and Tobago, Jamaica, Boswana and Ghana are also defined by he World Bank as fronier markes, bu are no included in his sudy due o lack of daa for a sufficienly long period. MSCI indices are esablished consisenly across counries and hus provide an adequae ground for exploraion of iner-marke relaions. They are value-weighed and calculaed wih he dividends reinvesed. In order o avoid double couning, sock prices of companies se up abroad are no included. All indices are in US dollars, which provides addiional comparabiliy across markes and implicily akes care of currency marke effecs. 4

6 Descripive saisics for he US and 21 fronier markes are repored in Table 1.1. The saisics include annualized arihmeic and geomeric means, median, maximum, and minimum geomeric reurns for each counry, as well as skewness, excess kurosis, Jarque-Bera, Ljung-Box (6), and Arch (6) es saisics. Annualized arihmeic mean reurns range from 0.09% for Kazakhsan o -0.08% for UAE and % for he US marke. The annualized geomeric mean reurns range from % for Kazakhsan o % for UAE and % for he USA. The annualized sandard deviaion of geomeric reurns ranges from 8.013% for Kazakhsan o % for Tunisia and % for he USA. The mean reurns are low as anicipaed, considering ha he period under sudy incorporaes he economic downurn of The sandard deviaion is quie high for he fronier markes, as expeced (Harvey (1995)). I could be resuling from various liquidiy effecs or heerogeneous informaion ses of invesors. The sandard deviaion is quie high for he US reurns as well, represening he urmoil on he US marke. The range of geomeric reurns is quie narrow, amid he high sandard deviaion, and no ha disan from he US marke reurns implying ha he fronier markes migh be well inegraed wih he res of he world, despie he possibiliy for segmenaion. The null hypohesis for no skewness is rejeced a 5% significance level for Kazakhsan, Kenya, Mauriius, and Sri Lanka; he values for skewness are posiive for all four counries. All counries display high and significan excess kurosis, possibly due o o ime variaion of condiional variance, as well as significan non normaliy as represened by he Jarque-Bera es saisic. The reurns for all counries are highly serially correlaed a 6 lags. No saisically significan auo-correlaion is exhibied by he reurns from Argenina, Bahrain, Kuwai, Romania, and Saudi Arabia. The summary saisics provided in Table 1.1 offer an insigh ha conrass wih he exising lieraure. The average reurns and sandard deviaion for fronier markes are no much differen han for he USA. One implicaion could be ha hese markes, however young, may no represen a considerable diversificaion opporuniy for foreign invesors. This could be due o he fac ha hese markes are economically and financially inegraed wih he US marke. Anoher possibiliy is ha he fronier marke indices under sudy may be comprised mainly of inernaionally operaing companies for which shocks from he US have greaer impac han local marke shocks. These indices may or may no ruly represen he financial and economic secors of he respecive counries, in which case we can infer nohing abou inegraion. 4. Mehodology In he Daa secion we have demonsraed he presence of subsanial deviaions from normaliy and considerable lepokurosis in he counry daa series. One ype of models ha can capure hese characerisics is he Arch ype models. Arch ype models were firs inroduced by Engle (1982) o accoun for he influence of changing volailiy 5

7 in ime series. Engle (1982) represens he condiional variance as a linear funcion of lagged squared residuals. The basic specificaion for an Arch model has he form R = a + u u = h 1/2 z z iidn(0, 1) h = b 0 + b 1 u 2-1 Because only one lag of squared residual is incorporaed, a model such as he above is called Arch(1). Bollerslev (1986) modified he Arch model by allowing lagged values of he condiional variance o be incorporaed in he condiional variance equaion. A basic GARCH (p,q) model describes he condiional variance as R = a + u u = h 1/2 z z iidn(0, 1) p q h = c 0 + c i u 2 -i + b i h -1 i=1 i=1 In order o avoid negaive condiional variance he parameers of he variance equaion mus be non-negaive, c 0 > 0, c i 0, b i 0. GARCH models are shown o beer accoun for fa ails and volailiy clusering resuling from ime variaion in condiional volailiy. I has also been found ha GARCH (1,1) specificaion is he mos appropriae for capuring hese effecs ( Bollerslev (1986), Enders (2001), Hamilon (1994)). In his paper we uilize wo separae models o describe he evoluion of reurns in emerging markes. Firs, we uilize univariae GARCH (1,1) model o specify marke reurns as a funcion of own pas values; hen we specify a bivariae model ha accouns for he impac of reurns and volailiy from he US marke. 4.1 UNIVARIATE MODEL 6

8 The univariae model is defined as follows: R j = b i=1 b i R j -i + u j u j = h j z z iidn(0, 1) R j log index reurn for counry j. h j = a 0 + a 1 u j2-1 + a 2 h j -1 b 0 he porion of he reurns explained by facors oher han shocks and pas reurns. b i parameer represening he impac of own counry lagged reurns on curren reurns. u j represens he unexplained porion of he counry reurns. h j condiional variance for counry j. I is inerpreed as a proxy for marke senimen owards news originaing wihin he counry. a 0 GARCH regression consan; represens he porion of he condiional volailiy explained by facors oher han he lagged excess reurns and lagged condiional volailiy. I is he porion of he marke senimen ha is due neiher o marke senimen owards news in he pas, nor o excess reurns. a 1 parameer represening he impac of lagged excess reurns. a 2 parameer represening he impac of marke senimen owards pas news originaing wihin he counry. The above specificaion is esimaed for reurns from all 21 fronier markes, as well as he Unied Saes. The number of lagged reurns included in he bes-fiing model is deermined using he Schwarz-Bayesian Crierion (SBC). Appendix 1 presens deails on he esimaed facor loadings of he univariae models for each counry and Table 1.2 repors a summary of he bes-fiing models; discussion is offered in secion BIVARIATE MODEL In his secion we upgrade he bes-fiing univariae model o a bivariae model in order o accoun for he effec of he USA reurns and condiional volailiy on he marke reurns of individual counries. The bivariae model is defined as follows: 7

9 R j = b i=1 b i R j -i + 12 i=0 c i R us -i + u j u j = h j z z iidn(0, 1) h j = a 0 + a 1 u j2-1 + a 2 h j -1 + a 3 h us + a 4 h us -1 where he variables and parameers addiional o he univariae model are defined as follows: R us -i log index reurns for he US marke represening shocks originaing from he USA. When i = 0 we consider shocks on he US marke ha occur on he same day as he reurns of he counry under sudy. When i > 0, we consider pas shocks o he US marke. c i parameer represens reurn spillovers from he US marke o he marke of he counry under sudy. When i = 0, c 0 represens he impac on reurns of conemporaneous US shocks, and when i > 0, c i represens he impac on reurns of pas US shocks. h us condiional variance for he USA marke. I is inerpreed as a proxy for marke senimen of raders on he USA marke owards news originaing in he USA and is derived from he univariae model of US marke reurns. a 3 & a 4 parameers represening he impac of marke senimen of US invesors owards conemporaneous news originaing in he USA on he senimen of he individual fronier counry invesors. We choose o consider a GARCH model wihou any asymmery, alhough he asymmeric impac on financial series volailiy of negaive news has long been sudied (Glosen, Jagannahan, Runkle (1993)). For emerging markes, however, no asymmery effecs have been shown and any poenial asymmery is assumed o ener hrough he idiosyncraic shock u j (Bekaer, Harvey, and Ng(2005)). The above specificaion is esimaed for reurns from all 21 fronier markes. The number of lagged US reurns included in he bes-fiing bivariae model is deermined using he Schwarz-Bayesian Crierion (SBC). Appendix 2 presens deails on he esimaed facor loadings of he bivariae models for each counry and Table 1.3 repors a summary of he bes-fiing models; discussion 8

10 is offered in secion HYPOTHESIS TESTING In his ex we invesigae wheher he counry s own lag marke reurn effecs b i and he US curren and lagged marke reurn effecs c i have any saisical significance. This is done by using LR es o explore he null hypohesis of wheher b i = 0 and c i = 0, where he number of lags i of local and US reurns is deermined using he SBC. Furher, we apply he LR es o examine he impac of he volailiy spillover effecs a 3, and a 4 ; he hypoheses esed are a 3 = 0 and a 4 = 0. We also sudy he possibiliy for homoscedasiciy of he error erm; we do ha by applying he LR es o he null hypohesis: a 1 = a 2 = a 3 = a 4 = 0. A he 5% significance level and he null for homoscedasiciy is rejeced in favor of he Garch specificaion. The es saisics derived from he significance ess performed on he bes-fiing bivariae models are presened in Table 1.4 and a summary is repored in Table Empirical Resuls 5.1 UNIVARIATE MODEL Appendix 1 presens deailed resuls from he esimaion of univariae models for he US and each of he 21 fronier markes. For each counry we invesigae 12 models, where each model is specified based on a differen number of included own counry lagged reurns. The number of lagged reurns under consideraion ranges from 1 o 12, and he bes-fiing model is deermined using he Schwarz-Bayesian Crierion (SBC). A summary of he coefficiens from he esimaion he bes-fiing univariae models for all counries is presened in Table 1.2 According o he Schwarz-Bayesian crierion (SBC), for he US and 20 ou of he 21 fronier markes, wih Kenya being he only excepion, he bes-fiing univariae model is one ha includes only one period lagged home reurns, and is defined as follows: R j = b 0 + b 1 R j -1 + u j u j = h j z z iidn(0, 1) h j = a 0 + a 1 u j2-1 + a 2 h j -1 One possible implicaion ha can be inferred from his finding is ha he influence of shocks from home is no more persisen for he fronier markes han for he USA 9

11 marke. For Kenya, he bes-fiing univariae model is R j = b 0 + b 1 R j -1 + b 2 R j -2 + u j wih he oher equaions as described above. Table 1.2 suggess magniude of he regression consan b 0 greaer han 0.1 for more han half of he counries. The presence of a significan posiive consan means he uncondiional expeced reurns are posiive. This suggess opimisic oulook of invesors abou he fuure. The consan capures laen facors deermining he mean of he reurns, as well as unobservable variables affecing he reurns, bu omied in our model. The exisence of such facors may be due o poenial unavailabiliy of daa oher han lagged reurns, or poor qualiy of he available daa. Poenial laen facors could be poliical insabiliy, governmen regulaions, corrupion, inernaional compaibiliy of accouning sandards, as well as differen regional facors. For several counries, Table 1.2 repors lower magniude of he consan, relaive o heir peers. The relaively low magniude of he consan implies low, or even negaive mean reurns for hese counries, low or negaive expeced reurns, and hus relaively pessimisic prospecs abou he fuure. The low magniude may also imply ha here is more daa available for hese counries, as well as beer qualiy of he daa, which is fully capured by he index reurns. One implicaion of such possibiliy could be ha markes in hese counries are more efficien, which resuls in he weaker presence of laen facors and unobservable variables. Anoher possibiliy is ha while home markes as a whole may be unable o incorporae all relevan informaion, he reurns of he companies having mos weigh in he MSCI index are resilien o local facors oher han pas reurns. This could be because he companies in quesion operae mosly inernaionally, here is more and beer qualiy daa available for hem, or hey are large enough o no be affeced by home marke processes. According o Table 1.2, for mos of he counries he influence of pas reurns b i is below 0.1. For eigh counries, he impac of pas reurns is greaer han 0.1. The greaer he impac of pas reurns, he longer i akes for hose markes o evaluae he full effec of pas shocks. One possibiliy is ha he economies of hese counries are no ha well diversified, and a shock o one secor rickles o oher secors and ha rickle akes ime o be evaluaed. The magniude of he variance consan a 0 for mos of he counries is beween zero and one, as presened in Table 1.2. For he USA he value of he variance consan is 0.086, which is he lowes for all counries. Since he US marke has been shown o be efficien in incorporaing informaion (Fama (1998)), he effec of any unobservable variables or laen facors is absorbed immediaely hrough a change in reurns. For 10

12 he fronier markes, he lowes consan value is for Kenyan reurns: For four markes he magniude of he consan is greaer han one. Those markes are Argenina (1.711), Lebanon (2.5), Mauriius (1.549), and Pakisan (1.4). Table 1.2 furher indicaes he influence of one period lagged evens on curren marke senimen a 1 is beween 0.5 and 1, for almos all counries, including he US (0.919). Considering ha he period under sudy incorporaes periods of grea marke urbulence and uncerainy, i seems reasonable o assume ha he impac of any pas news is being coninuously reassessed. A high coefficien of impac may imply no only abundance of informaion, bu also abundance of imporan informaion, such as announcemen of srucural reforms, or lack here of, announcemens of new governmen policies or regulaions, oudaed saisical informaion. The only counry wih relaively low impac of pas local news is Esonia (0.087). Considering ha Esonia is a small, expor oriened economy, i is realisic o believe ha he index daa is composed of predominanly expor oriened companies, for which local informaion is relaively unimporan relaive o global news and hus he coefficien of impac of pas local news is relaively low. 5.2 BIVARIATE MODEL Appendix 2 presens deailed resuls from he esimaion of bivariae models for each of he 21 fronier markes and a summary of he resuls is presened in Table 1.3. For each counry we invesigae 13 models, where each model is specified based on a differen number of included US lagged reurns. The number of lagged reurns under consideraion ranges from 0 o 12, and he bes-fiing model is deermined using he Schwarz-Bayesian Crierion (SBC). According o he Schwarz-Bayesian crierion (SBC), for 16 ou of he 21 fronier markes he bes-fiing bivariae model is one ha includes only conemporaneous US marke reurns and no US lagged reurns. specified as follows: R j = b 0 + b 1 R j -1 + c o R us u j = h j z + u j The model is z iidn(0, 1) h j = a 0 + a 1 u j2-1 + a 2 h j -1 + a 3 h us + a 4 h us -1 where h us is derived from he uni-variae model for US reurns. For four counries, Bahrain, Kuwai, Oman, and Tunisia, Table 1.3 suggess he bivariae model conains one period US lagged reurns, along wih he conemporaneous 11

13 reurns: R j = b 0 + b 1 R j -1 + c 0 R us + c 1 R us -1 + u j wih he oher equaions he same as above. For he Kenyan marke, he bivariae model conains wo period lagged domesic reurns and conemporaneous reurns from he US. This model is as follows: wih oher equaions as above. R j = b 0 + b 1 R j -1 + b 2 R j -2 + c 0 R us + u j Figures ploing he observed reurns, as well as he reurns prediced by he besfiing bivariae model for each counry are presened in Appendix 3. Each figure reveals he progression of he ploed variables over ime and illusraes he degree o which he variance of he observed reurns is explained by our bes-fiing models. 5.3 DISCUSSION OF SPILLOVERS The cener piece of he bivariae model, summarized in Table 1.3, is he invesigaion of reurn and volailiy spillovers from he US o he local markes as measured by he parameers c 0, a 3, and a 4. The greaes reurn spillovers from he USA are o Argenina (0.957), Kazakhsan (0.237), and Romania (0.329). For majoriy of he counries he impac coefficiens fall beween 0.0 and 0.1, wih 4 markes experiencing negaive spillovers: Jordan (-0.021), Lebanon (-0.059), Nigeria (-0.027), and Kenya (-0.028) a lag one. For Argenina, Kazakhsan, and Romania he high impac coefficiens may indicae deep economic iner-dependence beween hose counries and he US, raher han financial relaions only. The economies of hese four counries are predominanly expor oriened and rade relaions migh be one reason for economic inegraion. If he rade expor of hose economies is predominanly o he USA, hen hey migh no be well diversified and hus vulnerable o shocks from he American marke. One implicaion is ha in hese four counries, inernaional invesors may find lile or no scope for diversificaion. No such is he siuaion wih he markes experiencing negaive impac from he US: Jordan, Lebanon, Nigeria, and Kenya. The negaive relaion beween he US reurns and reurns in hose counries implies low degree of iner-dependence beween local economies and he US economy and hus presence of possible diversificaion opporuniies from he sandpoin of inernaional porfolio managers. From a policy maker s poin of view, hose counries seem resilien o shocks from he US marke. One implicaion is ha downurn in USA should no be blamed for recession in hose counries, and any economic slow down is more likely o be home grown raher han impored from US, a leas no hrough his channel. I is ineresing o noe ha alhough he local and US marke reurns occur on he same dae, he markes under sudy open before he US marke and have lile or no overlap in rading hours. Mos of hose markes open seven or more hours before he US 12

14 marke opens, wih he excepion being Argenina opening one hour ahead, Nigeria and Tunisia five hours ahead, and Slovenia six hours ahead. One imporan implicaion of his finding is ha i migh no be he acual US reurns ha maer, bu he expeced US reurns. Any announcemens made in he USA afer closing of he sock exchanges on day one, and before opening on day wo, are absorbed by he US marke on day wo. Throughou he rading day on local fronier markes, invesors observe hose announcemens, and incorporae hem in heir asse valuaion immediaely, while hey only ge incorporaed in he US reurns on he nex rading day in he US, which in mos cases sars jus afer closing of local markes. To beer describe such a siuaion, we could say ha i is he US overnigh reurns ha affec he local markes, raher han he acual daily reurns, where he overnigh reurns are defined as he change in price beween closing on day one and opening on day wo. The overnigh reurns form he expeced reurns, and are refleced in he US daily reurns on he nex day, so hose reurns seem o affec he local marke reurns despie he fac ha local markes may close for he day before he US marke opens. Similar discussion and more deails on he inernaional ransmission of overnigh reurns can be found in Hamao, Masulis, and Ng (1994), Lin, Engle, and Io (1994), and Baur and Jung (2006). For hireen of he weny one fronier markes, Table 1.3 repors he impac coefficien of lagged home reurns b 1 is greaer han he impac coefficien of he curren US reurns c 0. This implies ha for he period under sudy for hose markes, he average impac of domesic shocks is sronger han he average impac of shocks from he US marke. This suggess ha he economies of hose counries may be less vulnerable o US marke deerioraion relaive o heir peers. The opposie is rue for he remaining eigh markes. For hose, he average influence of shocks from he US is greaer han he average influence of domesic shocks, and heir economies may be more vulnerable o US marke shocks relaive o heir peers. The impac coefficiens a 3 and a 4 represening volailiy spillovers give us he scale of ransmission of marke senimen from he US sock marke o he local markes of he counries under sudy. In accordance wih lieraure on emerging markes, Table 1.3 repors ha volailiy spillovers are weaker relaive o reurn spillovers. Curren period spillover parameer a 3 for mos of he counries falls beween 0.00 and The low magniude of impac coefficiens imply weak, albei saisically significan, cross-marke ransmission of marke senimen from he US marke. The degree of ransmission of marke senimen migh also be indicaive of he presence of feed back rading and herding behavior. The low parameer values imply very limied presence of hose inefficiencies. Excepion are Argenina (0.084), Bahrain (0.134), Bulgaria (0.158), Croaia (0.103), Mauriius (0.33), Romania (1.836), and Saudi Arabia (0.122). Similarly for one period lagged spillovers, majoriy of he counries fall in he range beween 0.00 and Excepions are Bahrain (0.109), Bulgaria (0.112), Kenya (0.119), Lebanon 13

15 (0.066), Nigeria (0.161), Romania (0.436), and Saudi Arabia (0.28). For hese markes, and especially for Romania, here is sronger indicaion for cross-border ransmission of marke senimen and hus more prominen presence of herding behavior and feed back rading. I is worh exploring wheher he marke senimen of local marke invesors h j is affeced mosly by he curren and lagged marke senimen of US invesors h US and h US -1, or by he local invesors senimen owards domesic lagged evens h j -1. For mos of he counries, Table 1.3 indicaes magniude of he impac coefficien of lagged own marke news a 1 beween 0.5 and 1.0. This resul implies ha pas local news and local marke senimen on average have sronger impac on local invesors compared o boh curren and lagged news from he US marke. For several counries like Kenya, Lebanon, Nigeria, Romania, and Saudi Arabia, he coefficien capuring he effec of local news on marke senimen is no significanly differen from null. For hese counries, he senimen of US invesors is more imporan han local facors. For five counries, he value of variance consan a 0 is quie high: for Kenya, for Lebanon, for Nigeria, for Romania, and for Saudi Arabia. Implicaion is ha for hese fronier markes here is significan influence of inernaional and domesic facors oher han hose included in he Garch specificaion. One possible explanaion is ha hose economies are mosly expor oriened and heir expors are sensiive o a wide variey of inernaional facors. To summarize, for half of he counries, he influence of local shocks and marke senimen is sronger han he influence semming from he US marke. The inerdependence of mos of he fronier markes wih he US marke is weak, albei significan. The weak inerdependence can be aribued o flucuaions in he relaive significance of marke specific shocks versus shocks from he US. When shocks from US dominae domesic shocks, markes will move closer ogeher and will appear more inegraed. When domesic shocks dominae shocks from he US, he markes will move furher apar and will appear more segmened. In addiion, fronier markes could be underrepresened in global porfolios and hus be insulaed from porfolio re-balancing. Las bu no leas, he US and fronier marke indices may have significanly differen srucure which may furher reduce he cross-border marke co-movemen. Furhermore, we find ha local markes absorb informaion from US as soon as i appears which implies fronier markes are highly informaionally efficien. 6. Conclusion In his aricle we examine he degree o which he reurns and condiional volailiy of 21 fronier markes were affeced by he flucuaions in reurns and condiional volailiy of he American sock marke during he period beween December 1s, 2005 and January 15h, Using Schwarz-Bayesian crierion we find ha for seveneen counries he bes-fiing model is one ha includes only he conemporaneous US reurns, and for 14

16 four counries, he bes-fiing model includes one period lagged US reurns as well. We find weak, albei significan, mosly posiive reurn spillovers from he US marke. For four counries, Jordan, Lebanon, Nigeria, and Kenya, we find weak negaive spillovers, implying possible diversificaion opporuniies. For hireen markes, he influence of pas local shocks is greaer han he influence of curren shocks from he US, and for sixeen markes local pas volailiy has sronger impac han volailiy from US. We find ha fronier markes incorporae new informaion as soon as i arrives and for mos of he counries local informaion is weighed heavier relaive o informaion from he USA. The research presened in his aricle may be exended using ime-varying parameer echniques o beer accoun for marke dynamics and possible swiching dominance overime of domesic and US shocks. Oher exensions could be he invesigaion of he effecs of incomplee informaion and fa ails on marke inerdependence, as well as an empirical assessmen of various shock ransmission channels across fronier markes. 15

17 References Aggarwal R, Inclan C, Leal R. Volailiy in emerging sock markes. Journal of Financial and Quaniaive Analysis. 1999;34(1):33. Alkulaib Yaser A., Najand Mohammad MA. Dynamic linkages among equiy markes in he Middle Eas and Norh African counries. Journal of Mulinaional Financial Managemen. 2009;19: Baur, Dirk, Jung RC. Reurn and volailiy linkages beween he US and he German sock marke. Journal of Inernaional Money and Finance 2006;25: Beirne J, Caporale G, Schulze-Ghaas M, Spagnolo N. Volailiy spillovers and conagion from maure o emerging sock markes. IMF WP08-286; Bekaer G, Harvey C, Ng A. Marke inegraion and conagion. Journal of Business. 2005;78(1). Bekaer G, Harvey C. Emerging markes finance. Journal of Empirical Finance. 2003;10(1-2):3 56. Bekaer G, Harvey C. Foreign speculaors and emerging equiy markes. Journal of Finance. 2000;55(2): Bekaer, Geer HC. Emerging equiy marke volailiy. Journal of Financial Economics. 1997;43: Bollerslev T. Generalized auoregressive condiional heeroskedasiciy. Journal of Economerics. 1986;31(3): Calvo G. Conagion in emerging markes: When Wall Sree is a carrier. IEA Conference Volume Series.Vol 136. Cieseer; 2004: Caporale G, Piis N, Spagnolo N. Volailiy ransmission and financial crises. Journal of Economics and Finance. 2006;30(3): Chen G, Firh M, Meng Rui O. Sock marke linkages: evidence from Lain America. Journal of Banking & Finance. 2002;26(6): Eger B, Kocenda E. Inerdependence beween Easern and Wesern European sock markes: evidence from inraday daa. Economic Sysems. 2007;31(2): Enders W. Applied economeric ime series. Wiley & Sons;1995. Engle R. Auoregressive condiional heeroskedasiciy wih esimaes of he variance of U.K. inflaion. Economerica. 1982;50(4): Eun CS, Shim S. Inernaional ransmission of sock marke movemens. Journal of Financial and Quaniaive Analysis. 1989;24(2):241. Fama E. Marke efficiency, long-erm reurns, and behavioral finance. Journal of Financial Economics. 1998;49(3): Fadhlaoui Kais, Bellalah Makram, Dherry Armand ZM. An empirical examinaion of inernaional diversificaion benefis in cenral european emerging markes. Inernaional Journal of Business. 2009;14(2). 16

18 Gebka B, Serwa D. Are financial spillovers sable across regimes? Evidence from he 1997 Asian crisis. Journal of Inernaional Financial Markes, Insiuions and Money. 2006;16(4): Gebka B, Serwa D. Iner-regional spillovers beween emerging capial markes around he world. Research in Inernaional Business and Finance. 2007;21: Hamao Y, Masulis R, Ng V. Correlaions in price changes and volailiy across inernaional sock markes. Review of Financial Sudies. 1990;3(2): Hamilon J. Time series analysis. Princeon Univ Pr; Harvey C. Predicable risk and reurns in emerging markes. Review of Financial Sudies. 1995;8(3): Kaminsky G, Reinhar C. Financial markes in imes of sress. Journal of Developmen Economics. 2002;69(2): Kim S. The spillover effecs of US and Japanese public informaion news in advanced Asia-Pacific sock markes. Pacific-Basin Finance Journal. 2003;11(5): King, M., Wadhawani S. Transmission of volailiy beween sock markes. Review of Financial Sudies. 1990;3(1):5-33. Kodres, Laura E, Prisker M. A raional expecaions model of financial conagion. Journal of Finance. 2002;57(2): Lin W, Engle R, Io T. Do bulls and bears move across borders? Inernaional ransmission of sock reurns and volailiy. Review of Financial Sudies. 1994;7(3): Pollard S, Sapra S, Canarella G. Asymmery and spillover effecs in he Norh American equiy markes. econsor.eu. 2007;1: Prisker M. The channels for financial conagion. Inernaional financial conagion. 2001;(202): Psillaki M, Margariis D. Long-run inerdependence and dynamic linkages in inernaional sock markes: evidence from France Germany and he US. Journal of Money, Invesmen and Banking. 2008;(4). Sgherri S, Galesi A. Regional financial spillovers across Europe: a global VAR analysis. IMF WP09-23; Sola M, Spagnolo F, Spagnolo N. A es for volailiy spillovers. Economics Leers. 2002;76(1): Tanizaki H, Hamori S. Volailiy ransmission beween Japan, UK and USA in daily sock reurns. Empirical Economics Yu, Jung-Suk, Hassan M. Kabir. Global and regional inegraion of he Middle Eas and Norh African (MENA) sock markes. The Quarerly Review of Economics and Finance. 2008;48:

19 Table 1.1: Disribuional Characerisics of Annualized Daily Percenage Counry Logarihmic Reurns. Counry Arim. Geomeric Median Max Min S.Dev. Skewness Excess Jarque-Bera Ljung-Box Arch Mean Mean Kurosis Q (6) (6) Argenina Bahrain Bulgaria Croaia Esonia Jordan Kazakhsan * Kenya * Kuwai Lebanon Mauriius * Nigeria Oman Pakisan Qaar Romania Saudi Arabia Slovenia Sri Lanka * Tunisia UAE Unied Saes Noe: The es saisics are based on 1077 daily observaions of counry reurns beween December 1s, 2005 and January 15h, 2010 as given by MSCI Barra counry indices. All counry indices are in US currency. The disribuion of he Jarque-Bera es saisics approximaes q 2 (2), wih 5.99 criical value a he 5% level. The Ljung-Box and Arch es saisics are based on 6 lags of he reurns and are asympoically disribued as q 2 (6) wih criical value of a he 5% level. The null hypohesis of no excess kurosis is rejeced for all counries and * indicaes he null hypohesis for no skewness is rejeced a 5% level. 18

20 Table 1.2. Summary of he bes-fiing univariae models for he US and each of he 21 fronier markes. Univariae Model Counry b 0 b 1 b 2 a 0 a 1 a a Argenina (0.14) (0.03) (0.11) (0.02) (0.02) Bahrain (0.09) (0.04) (0.11) (0.02) (0.02) Bulgaria (0.11) (0.04) (0.13) (0.02) (0.02) Croaia (0.11) (0.03) (0.31) (0.05) (0.03) Esonia (0.1) (0.03) (0.09) (0.02) (0.02) Jordan (0.09) (0.03) (0.05) (0.01) (0.01) Kazakhsan (0.17) (0.03) (0.22) (0.02) (0.02) Kenya (0.09) (0.04) (0.04) (0.042) (0.01) (0.01) Kuwai (0.1) (0.04) (0.05) (0.01) (0.00) Lebanon (0.1) (0.04) (0.59) (0.07) (0.07) Mauriius (0.09) (0.04) (0.37) (0.06) (0.06) Nigeria (0.08) (0.03) (0.173) (0.04) (0.04) Oman (0.09) (0.04) (0.06) (0.01) (0.02) Pakisan (0.12) (0.03) (0.27) (0.03) (0.03) Qaar (0.122) (0.04) (0.17) (0.02) (0.02) Romania (0.18) (0.03) (0.06) (0.00) (0.00) S. Arabia (0.13) (0.04) (0.08) (0.01) (0.01) Slovenia (0.1) (0.03) (0.35) (0.07) (0.05) Sri Lanka (0.08) (0.04) (0.06) (0.02) (0.03) Tunisia (0.07) (0.04) (0.12) (0.03) (0.03) UAE (0.134) (0.04) (0.1) (0.01) (0.01) USA (0.07) (0.03) (0.02) (0.01) (0.01) Noe:The esimaion is done via ML uilizing 1077 daily observaions of counry reurns beween December 1s, 2005 and January 15h, 2010 as given by MSCI Barra counry indices. 19

21 Table 1.3: Summary of he bes-fiing bivariae models for each of he 21 fronier markes. Bivariae Model Counry b 0 b 1 b 2 c 0 c 1 a 0 a 1 a 2 a 3 a 4 Argenina (0.12) (0.03) (0.05) (0.44) (0.07) (0.04) (0.314) (0.31) Bahrain (0.09) (0.04) (0.03) (0.04) (0.47) (0.00) (0.03) (0.00) (0.07) Bulgaria (0.11) (0.04) (0.05) (0.31) (0.07) (0.03) (1.21) (1.219) Croaia (0.11) (0.04) (0.04) (0.54) (0.09) (0.05) (0.05) (0.00) Esonia (0.1) (0.03) (0.04) (0.12) (0.03) (0.02) (0.65) (0.65) Jordan (0.09) (0.03) (0.03) (0.07) (0.02) (0.01) (0.01) (0.00) Kazakhsan (0.17) (0.03) (0.06) (0.23) (0.02) (0.02) (0.00) (0.00) Kenya (0.09) (0.03) (0.03) (0.03) (0.47) (0.00) (0.06) (0.02) (0.04) Kuwai (0.11) (0.04) (0.04) (0.04) (0.64) (0.00) (0.02) (0.00) (0.08) Lebanon (0.1) (0.03) (0.03) (0.87) (0.00) (0.13) (0.00) (0.042) Mauriius (0.08) (0.04) (0.03) (0.48) (0.06) (0.08) (0.00) (0.000) Nigeria (0.09) (0.03) (0.03) (0.57) (0.00) (0.05) (0.00) (0.05) Oman (0.09) (0.04) (0.03) (0.03) (0.04) (0.03) (0.03) (0.00) (0.07) Pakisan (0.12) (0.03) (0.02) (0.27) (0.03) (0.03) (0.00) (0.000) Qaar (0.12) (0.04) (0.04) (0.22) (0.03) (0.02) (0.00) (0.00) Romania (0.16) (0.03) (0.06) (1.59) (0.00) (0.04) (0.00) (0.00) S. Arabia (0.16) (0.04) (0.05) (1.30) (0.00) (0.02) (1.19) (1.19) Slovenia (0.1) (0.03) (0.04) (0.47) (0.12) (0.03) (2.34) (2.35) Sri Lanka (0.08) (0.04) (0.03) (0.11) (0.05) (0.05) (0.19) (0.19) Tunisia (0.08) (0.04) (0.02) (0.02) (0.39) (0.00) (0.05) (0.00) (0.03) UAE (0.14) (0.04) (0.05) (0.14) (0.02) (0.01) (0.3) (0.3) Noe: The esimaion is done via ML uilizing 1077 daily observaions of counry reurns beween December 1s, 2005 and January 15h, 2010 as given by MSCI Barra counry indices. All counry indices are in US currency. 20

22 Table 1.4: LR ess of he coefficiens of he bes-fiing model of counry reurns wih spillovers LR Tes Saisics H 0 : b 1 = 0 c 0 = 0 b 1 = c 0 = 0 a 1=0 a 2 = 0 a 3 = 0 a 4 = 0 a 1 = a 2 = a 3 = a 4 = 0 Counry (j) (i-max, = a 3 = a 4 = 0 k-max) Argenina (1,0) Bahrain (1,1) Croaia (1,0) Esonia (1,0) Jordan (1,0) Kazakhsan (1,0) Kenya (2,0) Kuwai (1,1) Lebanon (1,0) Mauriius (1,0) Nigeria (1,0) Oman (1,1) Pakisan (1,0) Qaar (1,0) Romania (1,0) Saudi Arabia (1,0) Slovenia (1,0) Sri Lanka (1,0) Tunisia (1,1) UAE (1,0) Noe: The bes-fiing model is R J i-max=2 = b 0 + b i=1 i R-1 J a 0 + a 1 h J -1 + a 2u 2J -1 + a 3h USA + k-max=1 c k=0 k R-1 USA + u J, uj h = J z, z N(0, 1), hj = + a 4 h USA -1.The esing echnique used is LR, where he LR saisic is approximaely chisquared. The criical value wih 1 degree of freedom and p-value=.05 is 3.84;wih 2 degrees of freedom and p-value=.05 he criical value is 5.99; wih 3 degrees of freedom and p-value =.05 he criical value is 7.82; and wih 4 degrees of freedom and p-value=.05 he driical value is