WORKING PAPER NO. 48 FINANCIAL MARKET INTEGRATION IN EUROPE: ON THE EFFECTS OF EMU ON STOCK MARKETS BY MARCEL FRATZSCHER

Transcription

1 ROPAN CNTRAL BANK WORKING PAPR SRIS WORKING PAPR NO. 48 FINANCIAL MARKT INTGRATION IN ROP: ON TH FFCTS OF M ON STOCK MARKTS BY MARCL FRATZSCHR March

2

3 ROPAN CNTRAL BANK WORKING PAPR SRIS WORKING PAPR NO. 48 FINANCIAL MARKT INTGRATION IN ROP: ON TH FFCTS OF M ON STOCK MARKTS BY MARCL FRATZSCHR * March * uropean nversy Insue Deparmen of conomcs Bada Fesolana 56 San Domenco (FI) Ialy; marcel.frazscher@ue. I am graeful o Mke Ars Rober Brooks Rober Faff Phlpp Harmann Søren Johansen Gabrel Perez-Quros John Wllamson and he referee and semnar parcpans a he uropean Cenral Bank for commens.the paper was wren whle I was vsng he Research Deparmen of he uropean Cenral Bank n he summer of. I hank he uropean Cenral Bank for s hospaly. The vews expressed n hs paper are hose of he auhor and do no necessarly reflec hose of he uropean Cenral Bank.

4 uropean Cenral Bank Address Kasersrasse 9 D-63 Frankfur am Man Germany Posal address Posfach D-666 Frankfur am Man Germany Telephone Inerne hp:// Fax Telex 4 44 ecb d All rghs reserved. Reproducon for educaonal and non-commercal purposes s permed provded ha he source s acknowledged. The vews expressed n hs paper are hose of he auhors and do no necessarly reflec hose of he uropean Cenral Bank. ISSN 56-8

5 Conens Absrac 5 Inroducon 7 Leraure on fnancal negraon 8 3 Modelng negraon:theorecal movaon and emprcal mehodology 3. Theorecal movaon:an uncovered asse reurn pary (AP) relaon 3. mprcal mehodology:trvarae GARCH models o measure fnancal marke negraon 3.3 Asymmery and hreshold effecs n he shock ransmsson Specfcaon ess 4 4 Tesng for fnanal negraon 5 4. The daa 5 4. The resuls 6 5 Illusrang me-varyng fnancal negraon: Rollng esmaon and recursve esmaon echnques 9 6 xplanng me-varyng fnancal negraon:the role of M 4 6. Modelng me-varyng negraon 5 6. The role of exchange raes The role of M Prncpal componen analyss Indvdual varable analyss Specfcaon and robusness ess 3 7 Conclusons 37 References 38 Daa Appendx: Defnons and daa sources 4 uropean Cenral Bank Workng Paper Seres 4 CB Workng Paper No 48 March 3

6 4 CB Workng Paper No 48 March

7 Absrac Ths paper analyzes he negraon process of uropean equy markes snce he 98s. Is cenral focus s on he role ha M and specfcally changes n exchange rae volaly has played n hs process of fnancal negraon. Buldng on an uncovered neres rae pary condon o measure fnancal negraon a rvarae GARCH model wh me-varyng coeffcens yelds hree key resuls: frs uropean equy markes have become hghly negraed only snce 996. Second he uro area marke has ganed consderably n mporance n world fnancal markes and has aken over from he S as he domnan marke n urope.and hrd he negraon of uropean equy markes s n large par explaned by he drve owards M and n parcular he elmnaon of exchange rae volaly and uncerany n he process of moneary unfcaon. JL classfcaon: C3 F3 G5 Keywords: fnancal negraon sock markes M exchange rae volaly GARCH model mevaraon. CB Workng Paper No 48 March 5

8 6 CB Workng Paper No 48 March

9 Inroducon There s growng evdence ha real convergence n urope has no only been a pre-condon for successful moneary unon bu ha moneary unon self has srenghened real negraon among s members (e.g. Frankel and Rose 997). However far less aenon has so far been gven o he queson of wheher M has caused a subsanal ncrease n fnancal negraon among uropean markes and f so whch elemens of M have been cenral o hs negraon process. Ths paper analyzes he naure and he changes n he negraon of uropean sock markes snce he md-98s. I addresses wo relaed quesons: frs how srongly negraed are uropean sock markes and has hs degree of negraon nensfed over me? And second wha role has he drve owards M played n he process of uropean fnancal marke negraon? Whle here s ample evdence ha full convergence n uropean bond markes and money markes had been acheved by he md- o lae 99s (e.g. Frankel 994) s far less clear wheher and o wha exen uropean equy markes have become more negraed. The answer o hese quesons has mporan mplcaons boh for nvesors porfolo allocaon decsons and for polcy-makers n meeng he challenges of uropean negraon and shapng polcy responses o more negraed and nerdependen fnancal markes n urope. To address he quesons he paper bulds on an uncovered neres pary condon appled o asse prces o defne fnancal marke negraon and employs a rvarae GARCH model for 6 OCD counres o es he mpled negraon hypoheses emprcally. Addressng he frs of hese quesons yelds wo key fndngs. Frs uropean equy markes have become hghly negraed wh each oher only snce 996.And second he uro area marke has aken over from he S he role as he mos mporan marke n explanng equy reurns n mos ndvdual uropean markes. Ths leads us o he second queson: wha explans he ncreased role of he uro area fnancal marke? Whle here has been some research on he negraon of uropean equy markes and s me varaons (e.g. Longn and Solnk 995 Bodar and Redng 999 Hardouvels e al. 999) no sysemac aemp has ye been underaken o explan whch facors have been drvng he me varaons n negraon. Ths s he man conrbuon hs paper ams o make. I compares he relave mporance of he hree M pllars of exchange rae sably real convergence and moneary polcy convergence n explanng he me varaons of equy marke negraon n urope. The paper develops a GARCH mehodology wh me-varyng coeffcens o analyze and compare he role of hese hree facors. The resuls ndcae ha M has ndeed fundamenally alered he naure of fnancal negraon n urope. I s found ha was n parcular he reducon and elmnaon of exchange rae volaly and o some exen also moneary polcy convergence ha has played a cenral role n explanng he ncreased fnancal negraon among M members. Moreover he shock ransmsson across equy markes s found o be asymmerc.e. negave shocks are more srongly ransmed large shocks have a sronger mpac han small shocks and hese asymmery and hreshold effecs have become larger over me. Afer revewng some of he leraure on fnancal negraon n secon secon 3 develops he emprcal mehodology o es for fnancal negraon. Secon 4 provdes emprcal ess of he model for a se of 6 OCD counres. Dfferen mehodologes for measurng me-varaons n negraon are dscussed and llusraed n secon 5. Secon 6 hen employs a GARCH model CB Workng Paper No 48 March 7

10 wh me-varyng coeffcens o es whch facors (exchange rae volaly real convergence or moneary polcy convergence) explan he ncreased nerdependence among urope s fnancal markes.the fndngs are summarzed and some polcy mplcaons are dscussed n secon 7. Leraure on fnancal negraon The felds of nernaonal macroeconomcs and nernaonal fnance have developed dfferen bu relaed mehodologes o es for fnancal negraon. In nernaonal macroeconomcs much work has ulzed neres rae pary condons o es for fnancal negraon of money markes (e.g. Frankel and MacArhur 988 Frankel 99 Lemmen and jffnger 996): Covered neres rae pary (CIP): = + ( f s ) + k + k + k w ncovered neres rae pary (IP): + k = + k w + ( s + k ) + k + k + k + k = [ w + ( f s )] + [ ( s ) ( f s )] wh as nomnal neres raes f he forward rae and s he spo rae. IP s a broader defnon of fnancal negraon han CIP because no only measures a counry premum.e. drec barrers o capal flows (measured by CIP he frs erm on he r.h.s. n he las IP equaon) bu also allows for an exchange rae rsk premum as mpedmen o negraon (he second erm on he r.h.s.). There s a broad consensus ha mos of he devaon from IP n developed markes s due o exchange rae rsk prema whereas counry prema have become smaller or dsappeared over me (Frankel 99). On he conrary much of he nernaonal fnance leraure has employed a capal asse prcng model (CAPM) n esng for fnancal marke negraon (Bekaer and Harvey 995 Dumas and Solnk 995 Ferson and Harvey 99 Hardouvels e al. 999): ( r ) = λ β + λ β w w d d wh r as he excess reurn on he local porfolo λ he marke rsk premum β w he rsk of porfolo relave o world porfolo w defned as β w = cov - [r r w ] / var - [r ] and β dw analogously for he domesc marke porfolo d. The null hypohess of full negraon requres λ d =.e. he local porfolo s solely prced relave o he global porfolo w. Thus he basc nuon of he CAPM s ha expeced local reurns (r ) n a fully negraed marke depend only on nondversfable nernaonal facors. Regardng he emprcal mplemenaon of hese conceps varous economerc mehodologes have emerged over he years. arly aemps o es for nernaonal lnkages of equy markes have mosly focused on aheorecal VAR models (Kng and Wadhwan 99 Koch and Koch 993 um and Shm 993) and generally found rsng cross-marke correlaons and growng regonal nerdependence. More recen research on fnancal marke negraon has been conduced n a generalzed ARCH (GARCH) framework n order o ake no accoun he exsence of ARCH effecs n daa of hgher frequency. For nsance Ln and Io (994) and Koumos and Booh (995) An alernave measure s he Feldsen-Horoka condon whch saes ha n a fully fnancally negraed counry an exogenous shf n savngs should have no effec on nvesmen. However emprcal ess produce ofen couner-nuve resuls such as mplyng a hgher degree of fnancal negraon for emergng markes han for developed counres. See e.g. Frankel (99) for a dscusson of hs ssue. A que dsnc mehodology used o deec sable long-run relaonshps across fnancal markes s based on conegraon analyss (e.g. Dcknson Rchards 996). I s less relevan for hs paper here snce conegraon analyss does no yeld much nformaon abou he dynamcs and me-varaons n he degree of negraon. 8 CB Workng Paper No 48 March

11 fnd compellng evdence for some prce spllovers and volaly spllovers beween he London Tokyo and New York sock markes. A very mporan bu ofen gnored ssue s ha fnancal negraon may exhb srong varaons over me. Mos research esng for fnancal negraon has eher gnored hs ssue enrely or has looked a dfferen sub-perods o oban nformaon abou he dynamcs of negraon (e.g. Longn and Solnk 995 Bodar and Redng 999). Alhough comparng dfferen sub-perods may yeld a frs proxy for long erm changes masks much of he me varaon snce he degree of negraon may ofen change frequenly and exhb hgh volaly. A more promsng approach s one n whch me-varyng coeffcens are modeled explcly hrough nsrumenal varables. Three ypes of nsrumenal varables have been used n he leraure. Frs here has been ample evdence ha he removal of legal and non-legal barrers o capal flows have rased fnancal negraon subsanally no only n developed counres bu also for a number of emergng markes n recen years (e.g. Bekaer and Harvey 995 Ng ). Second numerous sudes have shown ha he degree of real negraon measured by he correlaon of busness cycles has a srong effec on fnancal negraon (Fama and French 989 Ferson and Harvey 99 Jagannahan and Wang 996). Moreover he degree of fnancal negraon ends o be hghes durng perods when boh counres or he domnan counry are n recesson (rb e al. 994 Ragunahan e al. 999). And hrd here s evdence ha exchange rae uncerany can have a large effec on fnancal negraon because exchange rae rsk s an mporan source of rsk prced on capal markes (e.g. Dumas and Solnk 995 Bodar and Redng 999 Hardouvels e al. 999). A common feaure of mos of hese nsrumenal varable approaches o negraon s ha hey only look a a very small se of varables ha may parly explan he me varaon n negraon. Such a narrow approach can be very problemac no leas because he ncluded varables may really pck up effecs of oher excluded varables. Ths danger s parcularly prevalen for he analyss of uropean fnancal markes for whch a reducon n overall exchange rae volaly concded wh real convergence and convergence n moneary polcy durng he perod leadng up o M. One herefore needs o carefully dsngush beween hese dfferen sources of negraon and n parcular one needs o ake no accoun ha wha deermnes he degree of fnancal negraon may no only be a counry s own economc performance bu also he degree of real and fnancal convergence wh oher economes. ndersandng he rue facors behnd he me varaons n negraon n uropean equy markes herefore requres a broader approach ha compares he role of hese dfferen sources.ths s he am of he remander of he paper. CB Workng Paper No 48 March 9

12 3 Modelng negraon:theorecal movaon and emprcal mehodology 3. Theorecal movaon: An uncovered asse reurn pary (AP) relaon Followng he approach commonly used n nernaonal macroeconomcs I use an uncovered neres rae pary (IP) condon as he sarng pon for measurng fnancal negraon. From he IP condon one can derve a smple uncovered asse reurn pary (AP) relaon for he degree of fnancal negraon of he local equy marke wh he world equy marke w: or n ex pos form [ r ] [ r ] + [ s ] = w (.a) r = rw + s (.b) wh r as he excess reurn of he marke porfolo n local currency n excess of he rsk-free uro neres rae and s he spo exchange rae of counry.ths ex pos AP relaon saes ha for a fully negraed marke asse reurns r are equal o asse reurns of he world marke r w afer accounng for exchange rae changes s. Jus as he IP condon he AP relaon of (.b) may ofen no be sasfed n realy. In addon o expecaonal errors here are a leas wo ypes of reasons for why (.b) may no hold. Frs (.b) gnores rsk prema ha are prced n he marke such as relaed o dfferences n volaly of marke reurns and exchange raes and second here may be oher barrers o cross-counry nvesmen ha preven markes from beng fully negraed. A more general formulaon of equaon (.b) o es he degree of negraon of marke wh world markes w can herefore be wren as r = φ rw + χ () wh φ measurng he correlaon beween local reurns r and global reurns r w and χ as a vecor of counry specfc facors such as busness cycle condons dfferen ypes of rsks and expecaons whch can drve a wedge beween local and world marke reurns. χ can be decomposed no a share ha s expeced from pas nformaon - [r ] and an unexpeced share.e. a conemporary dosyncrac shock: χ + = β [ r ] (3) Analogously world marke reurns r w can be expressed as a funcon of avalable nformaon from he pas - [r w ] and a conemporary nnovaon w φ + r w = βw [ rw ] w w (4) Subsung equaons (4) and (3) back no () yelds r = β w [ rw ] + w w + β [ r ] + (5) Ths relaon n essence says ha local reurns r are deermned by nformaon avalable from he pas boh local and worldwde (he expeced share of he reurns) and by conemporary shocks orgnang boh locally and n he world marke (he unexpeced share). CB Workng Paper No 48 March

13 quaon (5) now allows us o measure he degree of negraon of marke wh he world marke w: Defnon of marke negraon: Marke s more negraed he sronger domesc reurns r depend on conemporaneous world marke shocks w wh w as he measure of he degree of negraon of marke. Much of he research on fnancal marke negraon of he 99s has focused on he parameers β w and β as measures of negraon. The danger of usng he parameers β w and β o deduc nformaon abou fnancal marke negraon s ha hey also conan nformaon abou marke effcency. Defnon of marke effcency: Marke s more effcen he faser marke-relevan nformaon are ncorporaed no asse prces. nder fully effcen markes nformaon from me - should no affec reurns n perod.e. β w = and β =.Therefore β w and β are measures of he effcency of marke. The mporance of he ranson from equaon () o equaon (5) should now become apparen. Changes n he parameer φ of () can be eher due o changes n marke negraon or due o changes n marke effcency. quaon (5) allows o dsngush beween hese wo facors n explanng varaons n local reurns r. The ranson owards full marke effcency (β w = and β =) for nsance can be perfecly compable wh a move owards marke negraon (a rse n w ). Noe ha here s a far amoun of ambguy abou defnng he erm negraon. For money markes cross-marke negraon s generally defned as he absence of cross-marke arbrage opporunes so ha he law of one prce holds (e.g. Frankel 99). Ths defnon n essence mples no only he absence of barrers o capal flows bu also ha nvesors wll underake capal ransacons o elmnae arbrage opporunes ha arse. For equy markes measurng arbrage opporunes s more dffcul as he perssence of he equy home bas puzzle underlnes. Wha he equy home bas puzzle mples s ha he elmnaon of formal barrers o capal flows such as capal conrols and ransacon coss has no been suffcen o nduce crossborder capal flows as predced by varous asse prcng models. Accounng for facors such as rsk averson and unceranes e.g. exchange rae rsk has been shown o explan some par of he equy home bas alhough he puzzle of he equy home bas remans largely unsolved (e.g. Lews 999). The defnon of fnancal negraon of equy markes used n hs paper here herefore s a srcer one n ha no only looks a he openness of equy markes bu drecly measures he exen o whch shocks are ransmed across equy markes. 3 The ransmsson of a shock no only requres marke openness and he absence of barrers o capal flows bu also requres ha capal acually flows across markes n order o ake advanage of marke opporunes. 3. mprcal mehodology: Trvarae GARCH models o measure fnancal marke negraon quaon (5) serves as he sarng pon for he emprcal model. 4 A key dffculy wh esng (5) emprcally s ha he daa generang process (DGP) s unobservable.e. s mpossble o know 3 In conras Chen and Klez (995) and Ayuso and Blanco () ry o measure he absence of abrrage opporunes drecly by deermnng he dsance beween sochasc dscoun facors across makes.e. hey measure negraon as he exen o whch rsks are prced dencally n dfferen markes ndependen of an underlyng asse prcng model. 4 A smlar seup s used for esng fnancal negraon of emergng markes by Bekaer and Harvey (997) by Flemng and Lopez (999) for he analyss of S reasury blls by Karoly and Sulz (996) for Japanese frms negraon wh he S marke and by Ng () for he neracon of Pacfc-Basn equy markes wh hose of Japan and he S. CB Workng Paper No 48 March

14 CB Workng Paper No 48 March he full se of nformaon ha nvesors have used n each perod o form her expecaons. One herefore has o use nsrumens X - of pas marke fundamenals o proxy hs nformaon se. Moreover o compare and evaluae he relave mporance of he uro area marke wh ha of he S I furher dsngush beween regonal shocks orgnang n he uro area (denoed by subscrp ) and global shocks comng from he res of he world proxed hrough shocks from he S marke (subscrp ). Thus he condonal frs momens of he rvarae GARCH model n vecor forma are esmaed as (6) wh (7) (8) where r j s a 3x vecor of excess reurns ω j- a 3x vecor of condonal mean reurns ha are hemselves a funcon of pas nformaon on local regonal and global fundamenals X. The local nnovaon µ s explaned hrough conemporaneous global shocks and regonal shocks as well as a purely dosyncrac componen. Noe ha uro area shocks are allowed o affec S reurns on he same calendar day ( - ) whereas S shocks affec uropean markes only on he followng day ( - - ) due o he dfferences n radng mes. The measures of negraon n he model (6)-(8) are he me-varyng coeffcens - and -. They measure he conemporaneous dependence of marke reurn r on regonal shocks and global shocks. Ths degree of negraon s allowed o change over me. Afer esng he GARCH model wh consan coeffcens n secon 4 n order o oban he benchmark case secons 5 and 6 wll hen allow o change over me and es whch facors can accoun for he me varaons. Due o condonal heeroskedascy n he dosyncrac nnovaon j wh he 3x vecor of condonal varance of j s modeled as (9) whch mples ha he condonal varance s deermned by s own pas varance own pas squared shock and by conemporaneous squared exernal nnovaons. 5 Noe agan ha s consdered as conemporaneous because he S marke opens and closes afer he uropean markes and herefore affecs hese only on he followng calendar day. j j j r µ ω + = + = = j µ µ µ µ = = j X X X β β β β β β β ω ω ω ω ) ( ~ j j j N Ω = = j 5 The coeffcens for he effec of oher counres /regons pas condonal varance and pas squared nnovaon are resrced o zero n (9). mprcal ess showed ha hese coeffcens were mosly no sgnfcanly dfferen from zero and herefore were excluded from he fnal model n order o reduce he large number of parameers o be esmaed va maxmum lkelhood.

15 3 CB Workng Paper No 48 March Fnally he GARCH model s jonly mplemened for he daa va maxmum lkelhood esmaon of he log lkelhood funcon () where θ s he parameer vecor o be esmaed T he number of observaons and he me varyng condonal varance-covarance marx. The Smplex algorhm s used o ge nal values for he maxmzaon problem. To oban he parameer esmaes numercal maxmzaon s employed hrough he algorhm developed by Bernd Hall Hall and Hausman (974). 3.3 Asymmery and hreshold effecs n he shock ransmsson One parcular feaure of fnancal negraon s ha ends o be asymmerc - negave shocks are more srongly ransmed han posve ones - and ha hey are ofen skewed - large shocks have a bgger effecs han small ones. The asymmery has been shown o parly resul from he neracon of leverage and volaly feedback effecs (Campbell and Henschel 99 Bekaer and Wu 997): he effec of a posve shock on asse prces hrough mproved leverage of frms (leverage effec) s parly offse by he ncrease n volaly (volaly feedback effec) whereas he wo effecs re-enforce each oher for negave shocks. Large shocks end o have bgger spllover effecs because hey are ofen unexpeced and hus consue news ha need o be ncorporaed no asse prces. To es hs hypohess n he GARCH framework one can exend he condonal reurn and he condonal varance equaons (7) and (9) n he followng way: () () wh S A as an ndcaor funcon ha akes he value of f he nnovaon s negave and S T as an ndcaor funcon ha s equal o f he shock s larger han s sandard devaon over he full sample perod.the above specfcaon allows o dsngush beween four ypes of shocks: ( ) = + = T T L ' ln ) ln( ) ( π θ = = T T T A A A j S S S S S S τ τ τ τ δ δ δ δ ν ν ν ν µ µ µ µ = = T T T A A A A j S S S S S S η η η η ϕ ϕ ϕ ϕ ς ς ς ς

16 Prce effecs Volaly effecs Small posve shocks ν ζ Large posve shocks Small negave shocks Large negave shocks ( ν τ ) ( ς + η) + ( ν δ ) ( ς + ϕ) + ( ν τ δ ) ( ς + ϕ + η) + + The specfcaon used n equaons (9) and () s smlar o he one ha was frs suggesed by Glosen Jagannahan and Runkle (993 GJR) alhough should be noed ha here s some dsagreemen abou he approprae specfcaon o model asymmeres and hreshold effecs Specfcaon ess If he benchmark GARCH model (6)-(9) correcly specfes he DGP hen he followng orhogonaly condons hold: [ Ω ] = k l = k l k l (3) whch saes ha he dosyncrac shocks of he hree markes are ndependen. If he unexpeced share µ of he reurns n counry s solely explaned by regonal and global shocks hen he dosyncrac nnovaons are no only orhogonal o regonal and global shocks bu are also ndependen from shocks occurrng n oher local markes n [ Ω ] = n n (4) The valdy of hese condons consue mporan specfcaon ess of he model because f he nnovaons are no ndependen hen he negraon measures - may be based. 7 If he condons are fulflled hen he condonal varance of he unexpeced share µ of he reurns and s condonal covarances for he local marke wh global and regonal shocks can be derved from equaon (7) as [ µ Ω ] = = + + (5) [ µ µ Ω ] = = k k k k k = (6) These relaons are mporan no only o check wheher he model s correcly specfed bu hey also allow us o derve varance raos ha provde a goodness-of-f measure.e. a es for how much of he local reurn varance s explaned by regonal facors and by world facors: 6 Henry (998) compares he sascal properes of varous GARCH specfcaons for S daa and concludes ha he exponenal GARCH he quadrac GARCH and he GJR specfcaon are he mos approprae ones and are smlar n ha hey produce resuls ha are closes o he daa generang process. Due o s smplcy I decded o use he GJR specfcaon alhough he GARCH model was also esed and produced smlar resuls. 7 Forbes and Rgobon (999) and Boyer Gbson and Lorean (999) emphasze he exsence of anoher ype of bas f he condonal correlaon coeffcen ρ k s used as a measure of fnancal negraon. Gven he defnons of he correlaon coeffcen as ρ k = k / k and he regresson coeffcen of (7) as k = k / k one ges he relaon beween boh as ρ k = k ( k / ). Boh papers argue ha ρ k may be a based measure of fnancal negraon f a rse n ρ k over me s due o an ncrease n he rao of sandard devaons k / whereas he rue degree of negraon measured by k may acually say consan or even fall. Ths s he case because under he condon k < he rao k / rses as volaly ncreases.the analyss employed n hs paper here s no open o hs bas because drecly uses he regresson coeffcen k as a measure of fnancal marke negraon. 4 CB Workng Paper No 48 March

17 VR = VR = [ ] [ ] (7) (8) 4 Tesng for fnancal negraon 4. The daa The emprcal analyss s conduced for a se of 6 counres some of whch are par of he uro area (Ausra Belgum Fnland France Germany Ialy Neherlands Span) some of whch have no adoped he uro ye (Denmark Sweden K) and fve counres from ousde he (Ausrala Canada Japan Norway Swzerland). The am of akng such a broader sample s o analyze wheher here are dfferences n negraon beween uro members and oher counres. The daa on sock marke reurns are he marke ndces calculaed by Daasream Inernaonal and have daly frequency from January 986 o June or 3783 daly observaons for mos counres and a somewha shorer me perod for a few counres where some of he me seres sared laer. The reurn ndex for he uro area s he weghed average of hose markes ha have joned he uro. The uro area ndex used for he emprcal esmaon for ndvdual uro area counres excludes hs counry from he ndex n order o focus only on shocks ha are exernal o each marke. Hence he uro area reurn ndex r for each ndvdual marke s calculaed as r = wk rk wk k k wh k only ncludng uro area markes excep for marke and w k as he weghs reflecng he share of marke capalzaon of marke k n he oal uro area marke. There are wo man reasons for choosng Daasream ndces: frs hey are broader measures of sock marke reurns snce hey also nclude frms wh smaller capalzaon and herefore provde a more accurae presenaon of he whole marke. Second hey end o be more homogenous hus allowng comparsons across markes. Moreover s mporan o emphasze ha unlke n oher sudes menoned n secon all daa used here s denomnaed n domesc currency because usng a common currency would have prevened an explc analyss of he effec of exchange rae changes and uncerany. 8 (9) 8 The underlyng assumpon of usng reurns n domesc currences s ha nvesors are able o hedge a leas some of her foregn exchange exposure. sng reurns denomnaed n a common currency would assume ha nvesors are no able o hedge any of her exposure.ths may also nroduce a bas n ha a hgh degree of negraon may smply be due o a smlary n exchange rae changes raher han drec fnancal negraon. The rvarae GARCH model was neverheless also esed usng common currency reurns bu yelded very smlar resuls o hose obaned from usng domesc currency reurns. CB Workng Paper No 48 March 5

18 Table Descrpve Sascs of quy Reurns Daly Daa //986 /3/ Average Sandard Skewness Kuross LB() Reurn rror Belgum * 4.83 * 8.7 * Denmark * 9.45 *.88 * Fnland * 5.86 * * France * 5.73 * 5.8 * Germany * 9.65 * * Ialy * * Japan *.648 * * Neherlands * 8.4 * * Norway * * * Ausra *.35 * 46. * Sweden * * * Span * 5.86 * 67.9 * Swzerland *.65 * 8.96 * K *.98 * * SA * * 3.43 * Noe: * Indcaes sgnfcance a he 5% level. LB() s he Ljung-Box es for lags. Daa samples for Denmark Fnland Ialy Span are based on a somewha shorer me perod due o daa avalably. A frs look a he daa characerscs (Table ) shows ha here s ndeed srong evdence ha negave shocks are more frequen han posve shocks (negave skewedness) ha large shocks are more common han expeced sascally (excess kuross) and ha equy reurns are auocorrelaed. There s srong emprcal evdence ha sock reurns are predcable on he bass of pas nformaon (Harvey 99 Ferson and Harvey 993 De Sans and Gerard 997) denoed by he vecor X - n equaon (8).Ths vecor s These conrol varables reflec he busness cycle and agens belefs abou an economy s prospecs. The varables are commonly used n he leraure and are all lagged one perod: own excess reurns r he change n shor-erm neres raes sr (3-day urocurrency raes as proxes for rsk-free raes) he change n he erm srucure s he change n he dvdend yeld n excess of he shor-erm neres rae dy and a Frday dummy o cover end-of-he-week effecs Df. As ndcaed n equaon (8) for he local marke boh he own conrol varables as well as hose of he regonal and global markes are ncluded. 4. The resuls ( r sr s dy Df ) X = Tables and 3 presen he resuls for he rvarae GARCH model (6)-(9) wh consan coeffcens.e. wh coeffcens ha are assumed o be me-nvaran. Table gves a summary of he resuls aggregang coeffcens over he uro area versus he non-uro area and for dfferen sub-perods. Three srkng resuls emerge concernng (a) he overall degree of negraon (b) he relave mporance of he uro area versus ha of he SA and (c) he changes n hese relaons over me. () 6 CB Workng Paper No 48 March

19 Table GARCH Summary Resuls for Sock Marke Inegraon The GARCH model for counry whou asymmery and hreshold effecs as defned n equaons (6)-(9) s r = ω + µ µ + = + = (6.a) (7.a) (9.a) where r s he excess reurn ω - he condonal mean reurn ha s a funcon of pas nformaon on local regonal and global fundamenals X. The local nnovaon µ s explaned hrough conemporaneous global shocks - and regonal uro-area shocks as well as a purely dosyncrac componen. The measures of negraon are he coeffcens and and are esmaed as consan coeffcens over each sample perod. Table lss weghed averages of he negraon coeffcens wh he weghs beng GDP shares. The GARCH model wh asymmery and hreshold effecs of equaons () and () and s negraon measures are accordngly: Prce effecs Volaly Prce effecs Volaly effecs effecs Small posve shocks: V - ζ Small negave shocks: ( ν ) δ + ( ς + ϕ) Large posve shocks: ( ν ) τ + ( ς + η) Large negave shocks: ( ν τ δ ) ( ς + ϕ + η) + + /986-6/ OWN FFCTS FROM RO-ARA RTRN SPILLOVRS FROM SA VOLATILITY SPILLOVRS FROM RO-ARA FROM SA prce volal. oal posve shock negave shock oal posve shock negave shock oal posve shock negave shock oal posve shock negave shock sm a l large sm a l large sm a l large sm a l large sm a l large sm a l large sm a l large sm a l large β ALL counres uro-area Non-uro-area /986-7/99 ALL counres uro-area Non-uro-area /99-7/993 ALL counres uro-area Non-uro-area /993-4/998 ALL counres uro-area Non-uro-area /998-6/ ALL counres uro-area Non-uro-area CB Workng Paper No 48 March 7

20 Table 3 GARCH Resuls for Sock Marke Inegraon Indvdual Counres The able descrpon s dencal o ha for Table wh he dfference ha he coeffcens n Table 3 are for ndvdual counres and asymmery and hreshold effecs are no lsed n order o preserve space. PRIC SPILLOVRS VOLATILITY SPILLOVRS FROM RO-ARA FROM SA FROM RO-ARA FROM SA TOTAL SB-PRIODS TOTAL SB-PRIODS TOTAL SB-PRIODS TOTAL SB-PRIODS 986- /86-7/9 8/9-7/93 8/93-4/98 5/98-6/ 986- /86-7/9 8/9-7/93 8/93-4/98 5/98-6/ 986- /86-7/9 8/9-7/93 8/93-4/98 5/98-6/ 986- /86-7/9 8/9-7/93 8/93-4/98 5/98-6/ uro-area: Ausra.37 **.9 **.4 **.33 **.364 **.8 **.95 **.7 **.3 **. **.43 **.57 **.9 *.4 *.78 **. -.5 ** **.66 ** Belgum.94 **.94 ** **.547 **.78 **.75 **.5 **.35 **.4 **.94 **.46 **.5 *.39 **.8 *.5 **.76 ** Fnland.855 **.665 **.96 **.73 **.695 **.77 **.47 **.79 *.769 **.3..9 ** France.474 **.88 ** **.98 **.354 **.38 **.335 **.4 **.35 **.44 **.33 *.99 *.36 *.65 *.9 **.66 ** Germany.466 **.389 **.93 **.385 **.995 **.44 **.375 **.365 **.573 **.399 **.94 **.7 **.9 **.9 **.3.8 **.9 ** -.7 ** Ialy.496 **.8 ** **.977 **.33 **.39 **.58.3 **.94 **.48 **.57 **.48. *.55 * ** Neherlands.4 **.6 ** **.87 **.395 **.367 **.345 **.495 **.4 **.4 ** **.58 **.8 **.5 **.83 * ** Span.43 **.6 **.7 *.656 **.843 **.34 **.39 **.359 **.373 **.56 **.53 **.33 **.9.8 **.9 **.5 *.4 * -.38 ** ** Non-uro-area: Denmark.46 **. **.49 *.35 **.35 **.33 **.5 **.65 **.368 **.368 **.55 **.73 *.7.53 **.7. * *. * Norway.46 **.7 **..437 **.56 **.4 **.53 **.475 **.434 **.3 **.59 **..7 **.5 **.63 **.8.46 ** Sweden.539 **.338 **.84 *.7 **.873 **.464 **.458 **.45 **.5 **.43 **.3 **.46 **. *.56 **.8.7 *.5 * * Swzerland.45 **.74 **.3 *.554 **.75 **.339 **.346 **.394 **.387 **.54 **.5 **.68 **.7.3 **.9 **.37 **.63 **.4.8 * K.33 **.85 ** **.656 **.8 **.4 **.368 **.66 **.339 **.3 ** **.45 *. **.45 ** ** Japan.5 **.9 ** **.3 **.75 **.58 *.34 **.365 **.8 * **. **.4.36 **.33 ** * Canada. * **.679 **.43 **.98 **.48 **.535 **. -.4 *.. *.3 **.6 **.7 **.49.4 *.4 ** Ausrala.97 ** *.8 **.459 **.48 **.486 **.56 **.37 **.3 * **.39.3 **.98 **.3.5 *.5 * p-values are lsed n small numbers below he coeffcens. ** * o he rgh of a coeffcen show wheher s sgnfcan a he % level and he % level respecvely. Frs he large sze of he coeffcens are a frs ndcaon ha equy markes n urope are hghly negraed (Table ): a shock of % n he uro area leads on average o a change n reurns of.344% n oher markes whereas shocks n he SA have a smlar mpac of.359%.there s also srong evdence ha he shock ransmsson s hghly asymmerc and exhbs hreshold effecs: large negave nnovaons n he uro area have a coeffcen of.373 whereas he correspondng coeffcen for small posve shocks s only.69. As a rough esmae he resuls ndcae ha on average negave shocks have a 5% larger effec han posve nnovaons and large shocks have abou a % bgger mpac han small ones. Second whle he SA s clearly he domnan marke ousde he uro area s no longer he only domnan marke whn he uro area. By comparng he perods wh 993- (Table 3) one can see ha he S marke was he mos mporan one durng he lae 98s and early 99s bu ha he uro area marke has become he domnan force for ndvdual uro area counres snce he md-99s. Ths holds for all of he egh uro area counres bu no for markes ousde of he uro area. Thrd here are some furher srkng changes over me n he degree and naure of fnancal negraon. No only has he negraon of equy markes wh hose of he uro area markes and he SA become sronger over me bu also he mporance of own pas shocks has become 8 CB Workng Paper No 48 March

21 sgnfcanly smaller. Boh of hese fndngs underlne how much more negraed and also effcen equy markes have become durng he 99s. Fourh he degree of negraon has srongly ncreased whn he uro area snce he announcemen of he uro members n May 998 (Table 3). The sze of he coeffcens for he reurn spllover from he uro area o ndvdual uro area markes has more han doubled from.445 for he full sample o.9 for he perod May 998- (Table ). Ffh a srkng fndng s ha fnancal marke negraon whn urope was sgnfcanly lower durng he RM crss of (Table 3). Ths fndng s of parcular neres because s conrary o he fndng of ncreased fnancal nerdependence found durng he Lan Amerca and Asa currency crses (e.g. Bag and Goldfajn 998). Fnally no only has he ransmsson of shocks become sronger bu mporanly asymmerc and hreshold effecs have become larger over me. In oher words no only do shocks n one equy marke rase volaly n oher markes more now han before bu s n parcular he ransmsson of large and negave shocks ha has ncreased more proporonally han he spllover of small and posve nnovaons. For nsance a large negave reurn shock n he uro area had a hgher mpac of only.93 (=.5 -.3) han a small posve nnovaon durng Ths dfference rose o.53 (= ) n 998- (Table ). Overall hese fndngs gve srong suppor o he hypohess ha ndvdual markes whn he uro area have become ncreasngly negraed over me and ha he degrees of asymmery and hreshold effecs have nensfed. 5 Illusrang me-varyng fnancal negraon: Rollng esmaon and recursve esmaon echnques The dsadvanage of lookng a longer me horzons and even a sub-samples as presened n secon 4 s ha such an analyss provdes only a very general pcure of he overall negraon process. However he degree of negraon may ofen change on a regular possbly even a daly bass. The purpose of hs sub-secon s o llusrae he volaly n he negraon of sock markes. To capure he me-varaon I use rollng esmaon and recursve esmaon echnques. For he rollng esmaons I ake a -monh regresson wndow sarng from he perod January 986- January 987 and move hs -monh wndow forward by one monh a a me. For he recursve esmaon I sar wh he same regresson for January 986-January 987 bu hen keep he sarng dae fxed and only move he end dae forward by one monh a a me.9 Boh rollng and recursve mehods have some shorcomngs bu hey provde a good frs proxy of he volaly of he parameers of he sysem. 9 Kalman flerng s an ofen-used ype of recursve mehod n whch he model s updaed each perod alhough requres que resrcve assumpons abou he updang and smoohng of he me-varyng parameers. McKenze Brooks and Faff (999) provde a dscusson and evaluaon of alernave ways of measurng me varaons n sock markes. CB Workng Paper No 48 March 9

22 Fgure Comparng Mehodologes for smang Tme-Varyng Coeffcens for Inegraon wh he uro Area Neherlands The GARCH model esmaed for counry s r = ω + µ µ + = + = (6.a) (7.a) (9.a) where r s he excess reurn ω - he condonal mean reurn ha s a funcon of pas nformaon on local regonal and global fundamenals X. The local nnovaon µ s explaned hrough conemporaneous global shocks - and regonal uro-area shocks as well as a purely dosyncrac componen. The measures of negraon are he coeffcens and and are esmaed as consan coeffcens over each sample perod. ß s a measure of marke effcency and measures he dependence of reurns r I on own pas reurns r I-. Fg. shows only he coeffcen for he negraon of he Neherlands wh he uro area usng hree dfferen sample perods/echnques. The heavy doed lne shows for he enre sample perod The hn lne presens he subperod analyss for four dfferen sub-perods and he hck lne shows he esmaes usng -monh rollng esmaon wndows Toal: 986- Sub-perods Rollng smaon CB Workng Paper No 48 March

23 Fgure Tme-Varyng Inegraon uro Area Average GARCH -Monh Rollng smaes The fgure descrpon s dencal o ha for Fg. wh he dfference ha Fg. also shows he coeffcens for negraon wh he SA and ß for he effec of own pas reurns r -. Fg. shows he coeffcen esmaes no for an ndvdual counry bu he coeffcen average for he uro area counres wh each counry beng weghed by s GDP share n he uro area uro-area Own pas shock SA Fgure llusraes he shorcomng of usng longer me perods o analyze fnancal negraon. Fgure shows parameer for conemporaneous spllovers from he uro area o he Neherlands for dfferen esmaon mehodologes. By comparng he parameers becomes clear ha even usng he sub-perod analyss of secon 4 sll hdes a lo of nformaon abou he rue me varaon of. The esmae of he -monh rollng esmaons shows how hghly volale he degree of fnancal negraon can be. Despe he volaly n negraon Fgure as well as Fgure for he uro-area average and Fgures 3.a-3.p for each ndvdual counry (Fgures and 3 also nclude parameers for conemporaneous spllovers from he SA and ß for he effec of pas own nnovaons) yeld a number of mporan resuls. Frs he resuls show ha he uro area has become he domnan marke for mos ndvdual uro area markes snce he md-99s. The S marke was he mos mporan one for all ndvdual uropean markes ll he md-99s bu whle s mpac has remaned roughly consan n he long run he effec of uro area shocks on ndvdual markes has ofen more han doubled beween he early and lae 99s. CB Workng Paper No 48 March

24 Fgure 3 Tme-Varyng Inegraon Indvdual Counres: GARCH -Monh Rollng smaes The fgure descrpon s dencal o ha for Fg. wh he dfference ha Fg. 3 also shows he coeffcens for negraon wh he SA and ß for he effec of own pas reurns r -. Fg. 3 shows he coeffcen esmaes for each ndvdual counry. Fg. 3.a: Fg. 3.b: Fg. 3.c: Tme-Varyng Inegraon Tme-Varyng Inegraon Tme-Varyng Inegraon Ialy Neherlands Span uro area Own pas shock SA uro area Own pas shock SA uro area Own pas shock SA Fg. 3.d: Fg. 3.e: Fg. 3.f: Tme-Varyng Inegraon Tme-Varyng Inegraon Tme-Varyng Inegraon K Japan Norway uro area Own pas shock SA uro area Own pas shock SA uro area Own pas shock SA Fg. 3.g: Fg. 3.h: Fg. 3.: Tme-Varyng Inegraon Tme-Varyng Inegraon Tme-Varyng Inegraon Ausra Belgum Denmark uro area Own pas shock SA uro area Own pas shock SA uro area Own pas shock SA CB Workng Paper No 48 March

25 Fgure 3 (connued) Fg. 3.j: Fg. 3.k: Fg. 3.l: Tme-Varyng Inegraon Tme-Varyng Inegraon Tme-Varyng Inegraon Fnland France Sweden uro area Own pas shock SA uro area Own pas shock SA uro area Own pas shock SA Fg. 3.m: Fg. 3.n: Fg. 3.o: Tme-Varyng Inegraon Tme-Varyng Inegraon Tme-Varyng Inegraon Swzerland Ausrala Canada uro area Own pas shock SA uro area Own pas shock SA uro area Own pas shock SA Fg. 3.p: Tme-Varyng Inegraon Germany uro area Own pas shock SA Second he srkng feaure of he rollng esmaons s he hgh degree of volaly of negraon whn he uro area ( ).The dynamcs of negraon for mos uro area markes show he same behavor: a low degree of negraon durng and 995 a very rapd ncrease beween and hen a levelng off or even slgh decrease n Thrd here are some mporan dfferences n he mng of fnancal negraon across uro area markes. Fgure 4 llusraes ha counres ha nally were consdered unlkely canddaes o jon CB Workng Paper No 48 March 3

26 he uro such as Ialy experenced a somewha laer ncrease n fnancal negraon han more probable canddaes such as he Neherlands and Span. Whle he Neherlands and Span experenced a que seady ncrease n negraon afer a rough n a subsanal rse n negraon occurred for Ialy only snce lae 997/early 998 when became clear ha also Ialy would jon he uro. Fgure 4 Comparng Tme-Varyng Inegraon wh uro area: Ialy Neherlands Span. GARCH -Monh Rollng smaes The fgure descrpon s dencal o ha for Fg. wh he dfference ha Fg. 4 shows only he coeffcen esmae usng -monh rollng esmaons and for a shorer me perod (Sepember March ) n order o emphasze he dfferen negraon dynamcs across counres Ialy Neherlands Span So far one can only speculae abou he explanaon of hese dsnc feaures bu he mng of hese evens makes exchange rae volaly and credbly a srong canddae: he declne and low degree of negraon n and 995 may be explaned hrough urbulence n he RM as a number of counres dropped permanenly or emporarly ou of he RM n and a furher bu mlder dsurbance n 995. Smlarly he rapd ncrease n he negraon parameer snce may be due o he sably and credbly of he RM leadng up o he adopon of he uro.the followng secon provdes a more sysemac es of hs hypohess. 6 xplanng me-varyng fnancal negraon:the role of M Wha explans he rapd ncrease and he volaly n he degree of fnancal negraon n urope? Was M he drvng force behnd hs process? And more specfcally how mporan was he reducon and evenual elmnaon of exchange rae volaly and uncerany hrough he RM and he uro? 4 CB Workng Paper No 48 March

27 6. Modelng me-varyng negraon To explan he me varaon n negraon I now relax he assumpon of consan parameers and allow he negraon parameers - and - of equaon (7) o change over me. The behavor of hese parameers s modeled as a funcon of a vecor Z of underlyng economc and fnancal varables ha deermne he decsons of nvesors: ' = ψ Z () ' = ψ Z () As menoned n secon he choce of varables o be ncluded n he vecor Z s a conroversal ssue. One would lke o fnd varables ha help explan he degree and he changes n he ransmsson of shocks from one marke o anoher. The followng wo sub-secon wll es dfferen specfcaons of hs vecor of nsrumenal varables. 6. The role of exchange raes The cenral hypohess o be esed s ha exchange raes played an mporan role n he fnancal negraon process n urope. The exsence of exchange rae uncerany can funcon as an mporan devce for marke segmenaon. The more volale and unpredcable exchange raes are and he more cosly hedgng agans such uncerany s he sronger he degree of marke segmenaon and he lower he degree of correlaon across markes. Thus a more volale exchange rae of counry rases he naonal rsk premum as nvesors requre a hgher reurn o compensae for ncreased uncerany. Analogously he reducon or elmnaon of currency rsk as enaled n M and he nroducon of he uro may rase he degree of fnancal negraon across counres. To es hs hypohess he behavor of he negraon parameer s expressed only as a funcon of exchange rae volaly: = ψ vol (3) = ψ vol (4) wh vol as he volaly of he daly DM exchange rae changes of counry over he pas hree monhs and vol as s S dollar exchange rae volaly. Oher mporan lmaons o he negraon of fnancal markes can be he exsence of ransacon coss and he presence of governmen resrcons on marke ransacon such as lmng he access o foregn exchange. For mos of Wesern urope s markes such consrans had been abolshed by he lae 98s. xchange rae volaly for Germany n equaon (3) s measured vs-à-vs oher uro area counres weghed by her GDP shares. CB Workng Paper No 48 March 5

28 Table 4 GARCH Model wh Tme-Varyng Inegraon: xchange Rae Volaly The GARCH model esmaed for counry s r = ω + µ µ + = + = (6.a) (7.b) (9.a) where r s he excess reurn ω - he condonal mean reurn ha s a funcon of pas nformaon on local regonal and global fundamenals X. The local nnovaon µ s explaned hrough conemporaneous global shocks - and regonal uro-area shocks as well as a purely dosyncrac componen. The measures of negraon are he coeffcens - and - and are esmaed as me-varyng coeffcens as a funcon of exchange rae volaly vs-à-vs he DM (vol ) and vs-à-vs he S$ (vol ) wh he excepon ha Germany's exchange rae volaly (vol ) s measured vs-à-vs a GDP weghed average of oher uro-area currences: = ψ vol = ψ vol (3) (4) DM S$ DM S$ xchange Raexchange Rae xchange Raexchange Rae Volaly Volaly Volaly Volaly vol vol vol vol Ausra -.73 ** -.64 * Norway.4 *.9 * Belgum Span * Denmark * -.56 * Sweden -.35 ** Fnland -.9 ** -.8 Swzerland France -.54 **.8 K * Germany -.57 *.3 Japan.73 * -. ** Ialy -.34 ** -.36 ** Canada -.6 ** Neherlands ** -.4 Ausrala Robus sandard errors are lsed n small numbers below he coeffcens. ** * o he rgh of a coeffcen show wheher s sgnfcan a he % level and he % level respecvely. Table 4 for he rvarae GARCH model shows ha exchange rae volaly has a lo of power n explanng he me varaon of negraon. Frs almos all he coeffcens have a negave sgn ndcang ha hgher exchange rae volaly leads o a lower degree of negraon. Second s convncng ha currency volaly vs-à-vs he DM s sgnfcan and large mosly only for he counres whn he uro area whereas played less of a role for counres ha were less dependen on he uro area.and hrd s he volaly n he DM exchange rae ha s he mos 6 CB Workng Paper No 48 March

29 mporan varable whereas he S dollar exchange rae dd no seem as mporan n explanng he ransmsson of shocks from he S. Wha hese fndngs ndcae s ha he mporan role of he S equy marke n he world s only o a smaller exen explaned by exchange rae volaly and due o oher facors such as s large and domnan marke sze. Varous ypes of sensvy analyses were conduced. The resuls were robus o alernave specfcaons of he exchange rae volaly measure such as usng a shorer or longer me horzon. I was furher esed wheher he nroducon of he uro n January 999 or he announcemen of s members n May 998 consued a srucural break ha alered he relaon beween exchange rae uncerany and fnancal negraon. However varous specfcaons such as usng dummes for he perod afer he nroducon of he uro as well as lookng a subperods pror o 999 showed ha he resuls of Table 4 are que robus and confrmed he mporance of he effec of exchange rae volaly on fnancal negraon. 6.3 The role of M The fndngs of secon 6. ndcae ha exchange raes may have been a very mporan explanaory varable for he me-varyng dynamcs of fnancal negraon n urope. However he danger wh only lookng a exchange raes s ha gnores oher mporan facors ha may have been drvng he negraon process. Ths danger s parcularly srong for an analyss of fnancal negraon n urope because uropean economes have experenced sgnfcan real convergence and adhered o srcer moneary and fscal polces n he process of uropean negraon. As dscussed n secon above real convergence can have an mporan effec on fnancal negraon because asse reurns reflec o some exen he busness cycle. Havng more smlar busness cycles and beng more nerdependen hrough rade may rase he degree o whch shocks are beng ransmed across fnancal markes. Smlarly snce fnancal markes are very sensve o changes n moneary polces a hgh degree of fnancal negraon n urope may a leas n par be explaned hrough he convergence of moneary polces among uropean and n parcular uro area counres. Snce ncreased exchange rae sably wen hand n hand wh real and moneary convergence and were par of he same process a poenal problem herefore s ha he exchange rae coeffcens of equaons (3)-(4) may pck up effecs of excluded varables.a superor approach s o nclude a broader range of varables ha also reflec real and moneary convergence. I focus on varables ha are commonly used n he analyss of opmal currency areas (OCA) one se measurng he degree of convergence of counry wh he uro area and anoher se relang o he convergence wh he SA.The daa appendx lss he defnons of he real convergence and moneary convergence crera used n hs paper Prncpal componen analyss The problem wh he emprcal mplemenaon s ha because real and moneary convergence as well as he move owards more sable exchange raes concded here s a hgh degree of correlaon among many of he OCA crera. One way of mnmzng hs problem of mulcollneary and avodng spurous resuls s o form prncpal componens from he OCA crera. Through prncpal componen analyss he me-seres varables are lnearly ransformed no an equal number of prncpal componens ha are orhogonal o each oher. Prncpal componens are formed separaely for he real convergence crera and he moneary polcy convergence varables. To furher reduce he problem of mulcollneary among he See for nsance Ars and Zhang (998) and Bayoum and chengreen (998). CB Workng Paper No 48 March 7