The Spillover Effecs of U.S. and Japanese Public Informaion News in Advanced Asia-Pacific Sock Markes Suk-Joong Kim Absrac School of Banking and Finance The Universiy of New Souh Wales UNSW SYDNEY 2052 Ausralia Tel: +61 2 9385-4278 Fax: + 61 2 9385-6347 Email: s.kim@unsw.edu.au This paper invesigaes he naure of informaion leadership of he U.S. and Japan in he advanced Asia-Pacific sock markes. Insead of jus relying on reurn and reurn volailiy spillovers from major markes, specific and disaggregaed news evens are also uilized. In paricular, he aim is o examine he naure of spillover effecs of scheduled announcemens of he U.S. and Japanese macroeconomic variables in he advanced Asia-Pacific sock markes of Ausralia, Hong Kong and Singapore for he period 2 January 1991 o 31 May 1999. The invesigaion reveals ha boh U.S. and Japanese announcemen news elici significan firs and second momen influences on he reurns of he oher markes, in general, and ha here is a complex array of significan marke responses o various news announcemens. There is also srong evidence of markes responding differenly o bad news announcemens compared o overall news (including boh good and bad news) announcemens which indicae ha he informaion conen of each economic announcemen is a source of radable informaion raher han he ac of releasing economic figures. Thus, his paper conribues o he lieraure by shedding ligh on he imporan drivers of he documened informaion leadership of he U.S. and Japanese sock markes. JEL Classificaion: G14, G15, F36 Keywords: Informaion spillovers, Announcemen news Published in Pacific Basin Finance Journal, 2003, Vol. 11, No. 5, 611-630 Elecronic copy of his paper is available a: hp://ssrn.com/absrac=448120
1. Inroducion The exisence of financial marke linkages among advanced equiy markes is well documened. Empirics sugges, in general, i) significan inerdependence of marke movemens (Hamao, Masulis and Ng, 1990, Theodossiou and Lee, 1993, Koumos and Booh, 1995), ii) informaion leadership role of he U.S. marke (Eun and Shim, 1989, Arshanapalli and Doukas, 1993, Connolly and Wang, 2000, and o a less exen, Bae, Karolyi and Sulz, 2000), and iii) inensificaion of iner-marke linkages afer he 1987 global sock marke crash (Arshanapalli and Doukas, 1993). Ineresingly, researchers repor a negligible role of he Japanese marke in informaion leadership among he markes of he U.S. and Wesern Europe and an absence of significan marke linkages beween Japan and oher major markes 1. Similar paerns exis for Asia-Pacific sock markes in erms of significan firs and second momen reurn spillover effecs (Janakiramanan and Lamba, 1998; Pan, Liu and Roh, 1999; Arshanapalli, Doukas and Lang, 1995), increasing marke linkages and informaional leadership of he U.S. afer he 1987 crash (Arshanapalli, Doukas and Lang, 1995) and for he pos-1997 Asian crisis period (Chow, 1999) 2. The U.S. marke has been providing a significan leading influence in he Asia-Pacific Markes (Arshanapalli, e al, 1995; Ghosh, Saidi and Johnson, 1999; Janakiramanan and Lamba; 1998, Lin and Pan, 1997; Liu, Pan and Fung, 1996). The oher major economic influence in he Asia-Pacific region is coming from Japan and he influence of he Japanese sock marke movemens on he res of he markes in he region is hus of imporance. A number of sudies repor significan spillover effecs from 1 Bae and Karolyi (1994), however, repor ha he degree of marke linkages beween he U.S. and he Japanese sock markes were significanly undersaed when good and bad marke reurns were no invesigaed separaely. 2 Kaminsky and Schmukler (1999) repor ha large sock marke movemens in Asia were riggered by local and neighboring counry news and herding insincs of he markes during he 1997-1998 crisis period. 1 Elecronic copy of his paper is available a: hp://ssrn.com/absrac=448120
boh he U.S. and Japan o he Asia-Pacific markes. These include Cha and Cheung (1998), Liu and Pan (1997) and Ng (2000). However, despie close economic linkages (especially from he lae 1980s) beween Japan and oher regional counries, he Japanese influence had no been very srong unil he onslaugh of he financial crises in he Eas Asian counries in 1997 (see Chow, 1999; Ghosh, Saidi and Johnson, 1999). Given he leadership role of he U.S. and Japanese sock markes for he Asia-Pacific markes, a logical nex sep in he analysis of he marke linkages would be o invesigae he impac of disaggregaed informaion flows from hese wo major markes. I is of grea ineres o examine he degree o which differen ypes of informaion ha move he U.S. and he Japanese markes direcly affec he Asia-Pacific markes. Sock markes, in general, are influenced by boh public and privae informaion. The laer can be an exension of he former in ha heerogeneous inerpreaions of he implicaion of public informaion may lead o differen responses from marke paricipans. Scheduled announcemens of macroeconomic daa in he U.S. and Japan injec new informaion regarding heir respecive domesic economic condiions o he markes, and o he exen ha marke paricipans are caugh by surprise and heir opimal responses are dependen on heir inerpreaion of he news, significan marke movemens will resul. These include i) significan adjusmens o he curren equilibrium prices of financial asses, and eiher ii) heighened levels of volailiy due o divergen inerpreaions of he news in he marke and he resuling increased volumes of rade, or iii) lower levels of volailiy due o he release of informaion ha reduces he degree of marke uncerainy. To he exen ha he U.S. and Japanese sock marke movemens have an influence in he Asia-Pacific markes, hese informaion evens are wached carefully for heir impac on he marke condiions in he region. Significan informaion conen of hese announcemens will hus direcly be facored ino he prices in he Asia-Pacific markes. 2
Alhough here is some empirical evidence of he U.S. announcemen news spillover effecs in foreign financial markes, hey are confined o advanced markes 3 and here has been no published research on he news spillover effecs of U.S. and he Japanese macroeconomic announcemens in he Asian sock markes. This paper aims o address his imporan research issue by invesigaing he disaggregaed informaion spillover effecs from he U.S. and Japanese scheduled economic announcemen evens in he advanced Asia- Pacific sock markes of Ausralia, Hong Kong and Singapore. The main findings of his paper are ha; i) boh he U.S. and Japanese announcemen news elici significan firs and second momen influences on he reurns of he oher markes, and ha here is a complex array of significan marke responses o various news announcemens, ii) here is also srong evidence ha markes responded differenly o bad news announcemens compared o overall news (including boh good and bad news) announcemens which indicaes ha he informaion conen of each economic announcemen was a source of radable informaion raher han an ac of informaion release. The res of he paper is organized as follows: Secion 2 discusses he economeric modeling and daa consrucion issues, secion 3 repors and analyses he esimaion resuls, and secion 4 concludes his paper. 2. Daa and economeric mehodology The sock markes invesigaed are he U.S. marke and four advanced Asia-Pacific sock markes of Ausralia, Japan, Hong Kong and Singapore. Daily index observaions (open, 3 Becker, Finnery and Friedman (1995) on he U.K. sock fuures; Becker, Finnery and Kopecky (1995) on he German and Japanese deb markes; Connolly and Wang (2000) on he U.K and he Japanese sock markes; and Kim (1998), Kim and Sheen (2000) on he Ausralian foreign exchange and deb markes. 3
high, low and close) of hese markes were obained from Commodiies Sysems, Inc. for he period 2 January 1991 o 31 May 1999. The indexes are he All Ordinaries, Nikkei 225, Hang Seng, Sraigh Times and Dow Jones Indusrial, respecively for each counry. Figure 1 shows he ime line of he marke rading hours of he Asia-Pacific and he U.S. markes, and he iming of he U.S. and he Japanese macroeconomic informaion releases. While here are overlaps beween rading hours of he Asia-Pacific markes, he U.S. marke is closed when he non-u.s. markes are operaing. The informaion flow from he Japanese marke, which can be regarded as regional informaion, is hus conemporaneous while U.S. marke news (overnigh in he Asia-Pacific), which consiues global informaion, can influence he Asia- Pacific markes when hey open hree o four hours afer he U.S. marke closes. The spillover effecs of he disaggregaed flow of informaion from he U.S. and Japan, in he form of he news conen of he scheduled announcemens of macroeconomic variables, on he firs and second momens of Asia-Pacific marke reurns are invesigaed over holding periods ha include hese announcemens. The daes and magniudes of each of he scheduled public informaion releases in he U.S and Japan during he sample period were obained from Money Marke Services. In addiion, (median) marke survey expecaions of he magniudes of he economic daa from he same source were used o proxy he expeced componens of he economic announcemens. The announcemen variables considered are ones ha represen economic aciviies and ones ha conain informaion on inflaion, and only hose variables wih unbiased marke expecaions are included in his paper. 4 They are balance of paymen (BOT), real GDP growh rae (GDP), reail sales growh rae (RET), unemploymen rae (UE), producer price index inflaion (PPI) and consumer price index inflaion () for he U.S.; 4 Marke survey generaed expecaions of economic announcemen variables are said o be unbiased if here is no persisen deviaions in eiher direcion from he corresponding realized acual variables. Economerically, his is esed by running a simple regression of he acual on he expeced variable and esing for he zero inercep and uni slope coefficien. The las row of Table 1 repors he p-values of his unbiasedness hypohesis. 4
rade balance (TB), curren accoun balance (CAB), unemploymen rae (UE), money supply growh rae (MS), wholesale price index inflaion (WPI) and consumer price index inflaion () for Japan. Table 1 repors he deails of hese announcemen daa 5. All, excep for he US GDP daa, announcemens were made monhly and he announcemens cover he period January 1991 o May 1999 for he U.S. and early 1991 o mid-1999 for he Japanese announcemens. The good and bad economic announcemens 6 were equally likely, in general, during he sample periods. There were some excepions; good news announcemens were more common for he U.S. GDP daa announcemens (65% of he ime), and bad news announcemens were more frequen for he U.S. reail sales and Japanese unemploymen news (61% and 75% of he ime, respecively). However, he marke expecaions were shown o be unbiased overall which is due o he fac ha hese higher frequency observaions were smaller in magniudes. In all cases, he survey of marke expecaions are shown o be unbiased, as repored in he las row of Table 1. The exac iming of he U.S. and Japanese macroeconomic announcemens in relaion o he marke operaing hours in he Asia-Pacific are also shown in Figure 1. The U.S. announcemens are made one hour before he U.S. marke opens (08:30 Easern Sandard Time) and while all he Asia-Pacific markes are closed. The marke reurns ha capure he U.S. informaion releases are hus calculaed over he closing price on he calendar day before he announcemens and he opening price one calendar day afer for he Asia-Pacific markes Open Close (overnigh reurn on calendar dae +1, ln( P+ 1 / P ) 100 ), and over he closing price observed he day before he announcemen day and he opening price on he day of announcemen for he U.S. marke (overnigh reurn on calendar dae, Open ln( P / P ) 100 ). The Japanese informaion releases are made a various imes Close 1 5 The announcemen variables repored and used in he paper are raw announcemen daa as opposed o being sandardized wih marke expecaions. 6 Good (bad) announcemens are larger (smaller) han expeced announcemens of economic aciviy variables and lower (higher) han expeced inflaion variables (, WPI and PPI) and unemploymen rae. 5
hroughou he announcemen day; four announcemens jus before he Japanese marke opening and he oher wo while he marke is open. Thus, he Japanese announcemens are made eiher jus before marke opening or while markes are open in he Asia-Pacific markes and while he U.S. marke is closed. The impac of he Japanese economic announcemens are hen examined over he period beween he closing price on he day before he announcemen and he closing price on he day for he Asia-Pacific markes (daily reurn on Close Close calendar dae, ln( P / P 1 ) 100), while he period for he U.S. marke is over he closing price one calendar day before he announcemens and he opening price on he same Open Close calendar dae as he Japanese dae (overnigh reurn on calendar dae, ln( P / P 1 ) 100 ). The summary saisics of he index reurns of he U.S. and he Asia-Pacific markes over daily and overnigh holding periods are shown in Table 2. The means are fairly close o zero and he daily variances are spread over he wo sub-holding periods. Alhough here is no immediaely noiceable paern in he skewness, a considerably higher kurosis is observed during he overnigh period in all cases, expec for he US, which suggess higher frequencies of exreme observaions. The reurn series hus exhibi significan skewness, lepokurosis, highly significan linear and non-linear serial correlaions, and asymmeric responses of volailiy o innovaions. These characerisics are ypical of high frequency financial reurn series. Empirical modeling of he reurns mus accoun for hese momen characerisics and various researchers have shown ha daily sock price movemens, along wih many oher higher frequency financial ime series, can be adequaely modeled by a family of generalized auoregressive condiional heeroskedasic (GARCH) models. I has been repored ha exponenial GARCH models wih an appropriae disribuional assumpion explain he daily sock price movemens well 7. In his paper, parsimonious MA (moving average) - 7 see Bollerslev, e. al (1992) for an exensive survey of empirical papers. 6
EGARCH(1,1) models are used o model he reurn series wih asymmeric response characerisics, and hey are shown below. q H, = α i, c + α i,hol i, + ε i, + α i, k ε i, k k = 1 R i HOL (1a) Where: ε i, = zi, hi, ~ (0, hi, ), z i, ~ iid(0,1) H H ε ε i, 1 i, 1 ln hi, = βic, + βihol, HOLi, + βih, ln hi, 1 + βi, ε1 + βi, ε2 h i, 1 hi, 1 2 π H Close Close R i, = Sock index reurns over holding periods of daily (H = D, ln( P / P 1 ) 100 ) and Open Close overnigh (H = ON, ln( P / P 1 ) 100 ) for five markes, i = Ausralia, Hong Kong, Japan, Singapore and he U.S. HOL i, = (1b) Seasonal dummy ha akes he number of days beween wo successive observaions. 1 for normal weekdays, 3 for Mondays and 2 or higher for days immediaely following marke closures due o holidays. h, = Condiional volailiy of H i R,. H i q = Number of moving average erms included in he condiional mean equaion o remove serial correlaion in he esimaed sandardized residuals, z. The influences of he U.S. and he Japanese daa releases are hen invesigaed by including a news dummy (USNEWS j, and JPNEWS j, ) for each variable in he condiional mean and variance equaions of he sock index reurns [(1a) and (1b) above] measured over he period surrounding he daa announcemens, as discussed above, and examining he sign and significance of he esimaed coefficiens 8. The news dummies ake he value of one if here was a news conen in he announcemen (i.e. acual figure announced was differen from he marke expecaions) and zero oherwise. 9 8 In place of he announcemen news dummies, magniudes of surprises measured by acual daa released minus he MMS survey expecaions were also ried. The resuls of he magniude esimaions are much he same as he announcemen dummy esimaions repored in his paper, and so hey are no repored o conserve space. 9 This requires some measure of marke expecaions on he announcemens, and he mos common approach is o uilize marke based survey expecaions provided by Money Marke Services. We follow his approach and proxy marke expecaions by he survey of he U.S. and he Japanese marke expecaions of heir scheduled informaion releases generaed by Money Marke Services in he U.S. and Japan. 7
In addiion o he general marke responses o he overall news announcemens, he effecs of bad news announcemens are also invesigaed for heir poenially differenial impac on he markes. In general, here is considerable evidence of sock markes showing higher levels of volailiy and more inense volailiy spillovers beween markes during periods of crisis. Bad (worse han expeced) economic daa announcemens would be perceived o have differen informaion conen han good ones hus eliciing differen marke responses. Bad news announcemen days are when he magniudes of he announced daa were smaller han he marke expecaions for he economic aciviy variables (BOT, GDP and RET for he U.S.; and TB, CAB for Japan) and when higher han expeced inflaion daa (PPI and for he U.S.; and WPI, and MS for Japan) and unemploymen daa were announced. The asymmeric news effecs are hen modeled by including no only he overall news dummy variables bu also he bad news dummy for each variable in boh he condiional mean and variance equaions. The bad news dummies ake he value of one on he days of bad news announcemens and zero oherwise 10. The final form of he models o be esimaed are: H Bad Bad i = + αi, j j, + αi, j j, j= BOT j= BOT for he US news effecs R M () ( USNEWS USNEWS ), j= TB Bad Bad = M ( ) + ( α i j JPNEWS j, + α, JPNEWS, ) for Japanese news effecs j= TB H Bad Bad i = + βi, j j, + βi, j j j= BOT j= BOT ln h V ( ) ( USNEWS USNEWS ) i j for he US news effecs j (2a) (2b) Bad Bad βi, j j, βi, j i, j j= TB j= TB for Japanese news effecs = V () + ( JPNEWS + JPNEWS ) 10 There also exiss a poenial for asymmeric effecs of bigger shocks having larger impac on he volailiy. This was modeled by he EGARCH models wih above average news dummy for each variable ha akes he value of one on days of larger han sample average (in magniudes) surprise and zero oherwise. However, he esimaion resuls, in general, do no differ significanly from he general esimaion of announcemen dummies, and so hey are no repored in he paper o save space. Ineres readers may obain he resuls from he auhor. 8
Where M(.) and V(.) are he righ hand sides of (1a) and (1b), respecively. Holding periods relevan for he invesigaion of he U.S. announcemens are Open Close Non-U.S. markes: H = ON (Overnigh) a +1, (ln( P 1 / P ) 100) Open Close U.S. marke: H = ON a, (ln( P / P 1 ) 100) Holding periods relevan for he invesigaion of he Japanese announcemens are Close Close Non-U.S. markes: H = D (Daily) a, (ln( P / P 1 ) 100) Open Close U.S. marke: H = ON a, (ln( P / P 1 ) 100) j = BOT, GDP, RET, UE, PPI and for he U.S. news and TB, CAB, UE, MS, WPI and for he Japanese news announcemens. + α i, j and The signs of he coefficiens for he announcemen dummies in he mean equaion, α Bad i, j, are o be inerpreed as he general direcion of price movemens in he relevan markes following he informaion releases and so represening an average effec of each announcemen. The volailiy responses o he informaion evens, shown by β i, j and β Bad i, j, depend on wheher he daa release adds o he exising informaion heerogeneiy in he marke (β s > 0) or i resolves informaion asymmery by providing level informaion playing field (β s < 0). Some macroeconomic news announcemens may increase he heerogeneiy of beliefs and hus furher disurb a financial marke. This migh occur for a low macroeconomic saisic for which a widespread consensus develops relaively easily abou is imporance and relevance. In he days approaching he nex announcemen, he marke may sele owards some degree of homogeneiy of beliefs. When surprises are revealed, he homogeneiy evaporaes giving rise o unusually high ransacion volumes and hus condiional price volailiy. In ime, beliefs abou he fundamenal implicaions of he previous announcemens begin o converge. The volume effec is represened by a posiive announcemen coefficien in he condiional volailiy equaion (β s > 0). By conras, some 9
oher ypes of macro news announcemens may end o almos immediaely sele a marke. For hese macroeconomic saisics, some individual paricipans in asse markes may have a poor undersanding or convicion abou heir imporance and relevance, while ohers may have relaively beer knowledge or convicion. Leading up o hese announcemens, rading occurs based on he diversiy of knowledge or beliefs abou he possible value ha will be conained in he announcemen. The release of new informaion, hus, adds o curren informaion ses and so may have he effec of reducing he degree of informaion asymmery in he marke. Afer he announcemens, he bigger he surprise, he less likely are he ill informed o rade, and he more likely is a price adjusmen reflecing he knowledge or convicion of he oher group. Thus he surprise in such announcemens reduces volailiy by sidelining hose less able or unwilling o ake a differen posiion 11. This effec is shown by a negaive announcemen coefficien in he condiional volailiy equaion (β s < 0).. 3. Empirical resuls Tables 3 and 4 repor he esimaion resuls of he U.S. and Japanese announcemen news spillover effecs, respecively 12. The resuls repored are he Quasi-ML esimaions of (2a) and (2b) where six news variables are included o pick up he overall impac of he 11 A good example of an imporan paricipan who migh ac wih knowledge and convicion is he cenral bank. Afer a large macroeconomic surprise, he cenral bank may adjus is policy insrumen o affec he condiional mean of, say, he shor erm ineres rae, bu i may also decide o demonsrae an exra degree of firmness in is sance by acing o reduce he volailiy of ha rae (i.e. by smoohing ). If marke paricipans believe ha is a credible sance, hey will be less willing o rade. 12 The EGARCH modeling of he index reurns are shown o be successful in all cases. Significan negaive asymmeric effec (negaive β ε1 ) and volume effec (posiive β ε2 ) are presen. Tha is, higher condiional volailiy was eviden when he previous day s reurn was unexpecedly lower (a negaive innovaion) and larger in magniude. The coefficien for he lagged volailiy, β h, is close o one in all cases suggesing ha shocks o he volailiy were fairly persisen. The holiday effec is also significanly presen. In general, marke reurns were significanly lower and condiional volailiy higher on days immediaely following holidays. The diagnosics of he esimaion are repored in he boom panels of Tables 3 and 4. I is eviden ha he EGARCH models are effecive in addressing he ime series properies documened in Table 2. The skewness and kurosis are significanly reduced, linear and non-linear serial correlaions eliminaed, and asymmeric volailiy responses o innovaions are removed (excep for Singapore). 10
informaion releases and six bad news announcemen dummies are used for worse han expeced announcemens 13. U.S. informaion spillover effecs The overall U.S. news announcemens, in general, had a posiive effec on reurns in all non-u.s. markes, excep for he BOT news in Hong Kong and Singapore, as evidenced by posiive and saisically significan news coefficiens, α i,j s (13 posiive and only 3 negaive coefficiens). This suggess ha he U.S. news, on average, improved marke senimens in he Asia-Pacific leading o upward adjusmens of earnings forecass for he markes. On he oher hand, he spillover effec is mosly negaive for he bad news announcemens (11 negaive and only 6 posiive news coefficiens, α, Bad i j s). This suggess ha worse han expeced U.S. economic performances generaed negaive senimens on he likely fuure performances of he Asia-Pacific markes 14. Ineresingly, he BOT news elicied he opposie marke responses in all bu he Ausralian marke. Overall, BOT news significanly dragged all he markes down, excep for Ausralia, whereas unexpecedly bad daa had he opposie impac. This suggess ha worse han expeced U.S. rade performances were perceived o sugges beer han expeced expor performances of he Asia-Pacific counries o he U.S. goods marke, and o he exen ha expor oriened firms weigh heavily in he Asia-Pacific marke indexes (less so for Ausralia) his would have had an adverse impac. The BOT daa, however, had a negaive impac on marke reurns in he U.S. and his 13 The models were also esimaed for he smaller sample ha ends jus before he onse of he Asian financial crisis of July 1997. The resuls are essenially he same as hose repored in his secion excep for marginally larger news coefficiens, in general. 14 This is consisen wih he informaion leadership of he U.S. marke repored in he lieraure. The U.S. informaion leads he marke movemens in he Asia-Pacific and so posiive (negaive) marke senimens in he U.S. would promoe (dampen) index reurn performances in he regional markes. 11
downurn apparenly overwhelmed he Ausralian marke. The RET news had a consisen impac in Ausralia and Japan; overall news was associaed wih posiive reurns and bad news wih negaive reurns. The overall UE news had a similar impac. Thus, news on he U.S. economic condiions ended o have a posiive spillover effec overall and an adverse influence when he news was bad. The wo inflaion variables affeced he regional markes in much he same way [overall (bad) news was associaed wih posiive (negaive) reurns] perhaps due o he ineres rae implicaions of an unexpeced inflaion shock 15. However, he responses of he Singaporean marke o he PPI news were in conras o hose of he oher markes. A negaive coefficien, alhough insignifican, for he overall PPI news, is observed as is he case in he U.S. marke. Turning o he news effec on he marke volailiy, srong U.S. news spillover effecs are shown in all he markes. Alhough only a volailiy reducing effec (β i,j < 0) was significan in he U.S. for he overall news announcemens, individual Asia-Pacific markes reaced differenly o he overall news as evidenced by evenly disribued (10 each) significan posiive and negaive news coefficiens, β i,j s. The announcemens of worse han expeced U.S. macroeconomic performances, however, had he effec of significanly lowering he level of condiional volailiies, in general (12 negaive and 6 posiive coefficiens). These announcemens apparenly fed enough informaion o reduce he level of uncerainy in he minds of marke paricipans regarding he healh of he U.S. economy and is global influence. This migh indicae ha unexpeced economic downurns in he U.S. cleared clouds of doub concerning possible U.S. policy responses so as o reduce he level of uncerainy regarding heir impac in hese markes. 15 Unexpecedly lower U.S. inflaion would negae a need for he U.S. Federal Reserve o raise ineres raes hus having a marke boosing impac, and he opposie applies for higher han expeced inflaion announcemens. 12
In sum, here is srong evidence of U.S. announcemen news spillover effecs in he Asia-Pacific. The marke responses o news and he degree of impac of each announcemen ended o vary across he markes. Collecively, however, he U.S. news effecs are significanly presen in reurns and reurn volailiies of all he markes considered as evidenced by he joinly significan news coefficiens (repored in he las hree lines in he las panel of Table 3). Japanese informaion spillover effecs The Japanese announcemen news spillover effecs are repored in Table 4. The firs panel repors he news coefficiens in he mean equaions. There is no immediaely discernable paern of responses o overall news across he markes as shown by evenly disribued significan posiive and negaive news coefficiens (6 and 5 for he overall news, and 6 and 8 for he bad news). The overall rade balance news is significan only in Japan and he U.S., and he negaive coefficien suggess a significan drop in he marke reurns in hese wo markes. In conras, worse han expeced announcemens significanly increased he reurns in all markes excep in Ausralia. This migh sugges ha unexpecedly bad Japanese rade performances implied unexpecedly beer rade performances of he U.S. leading o improved marke oulook, and o he exen ha he U.S. marke had a leading influence in he Asia-Pacific markes, his posiive response of he U.S. marke was followed by similar marke movemens in he region. Good Japanese exernal balance performances would hen be inerpreed as a bad news for he U.S. since some of he surpluses would be a he expense of he U.S. and so he expeced fall in he U.S. equiy marke would cause a downurn in hese markes. Also, increasing Japanese rade surpluses migh have caused poliical discomfor leading o negaive senimens in he region. Ineresingly, here is some evidence 13
of worse han expeced curren accoun balance daa having a marke dampening effec in Hong Kong and Singapore. Apparenly, worse han expeced curren accoun balance implied lower han expeced capial ouflows from Japan and hese wo markes were adversely affeced as a resul. Ineresing resuls are also found for he unemploymen rae news announcemens. Overall unemploymen news lowered he reurns in Hong Kong and Singapore, while unexpeced unemploymen had a posiive reurns spillover effec in Ausralia and Hong Kong. The opposie paern is shown for he U.S. marke. A possible explanaion is a porfolio rebalancing of inernaional invesors beween Japan and oher Asian markes. An announcemen of an unexpeced increase in unemploymen would lead o declining confidence in he Japanese marke leading o a conracion in he foreigners invesmens in Japan. This would benefi oher regional markes if he invesors desired a sable regional porfolio. In he case of he U.S., a lower marke reurn following bad Japanese unemploymen daa migh be an indicaion of he poenially negaive implicaion for he U.S. exporers in he Japanese goods marke. The significan news coefficiens for he hree inflaion relaed variables (MS, WPI and ) are posiive for he overall news and negaive for he bad news announcemens. The negaive influence of unexpeced Japanese inflaion may be explained by he adverse impac of he resuling weakness in he Yen on he exernal balance posiions of he rading parners of Japan. Significan Japanese news spillover effecs on marke volailiies are also presen in all markes. In general, overall and bad news announcemens had opposie effecs in all markes suggesing ha wheher an informaion release had a volailiy increasing or decreasing effec depends on is informaion conen. Overall rade balance news significanly reduced volailiy in Ausralia, Hong Kong and Singapore, while bad news had he opposie impac. Ineresingly, however, he paerns of news effecs in Japan and he U.S. are opposie o hose of he oher markes. The volailiy reducing effec of worse han expeced Japanese 14
rade balance news in hese wo markes could have been due o he lower probabiliy of rade conflic beween hem. This resul is similar o he case of he U.S. rade balance news repored above, indicaing ha he Japanese and he U.S. rade balance daa were an imporan source of informaion for marke paricipans in he Asia-Pacific. The curren accoun balance news, in general, had he opposie impac compared o he rade balance news. The overall news raised he volailiy in Japan, Hong Kong and Singapore, while he bad news significanly reduced i in Hong Kong. This suggess ha he services componen in he curren accoun obviously had a differen news conen han he rade balance componen. This migh be due o he role of Japan as a provider of capial and he services componen of he Japanese curren accoun was of some news value o he marke wachers in he region. Overall news on he WPI inflaion (and MS) increased volailiy in all markes while he inflaion news significanly decreased i in all markes excep for Ausralia. The opposie resuls are shown for he bad daa announcemens. This indicaes ha hese wo inflaion daa were perceived o convey differen economic condiions of Japan. The WPI inflaion (and MS) news could have caused a higher rading aciviy, due o he injecion of radable informaion on he Japanese economy, in hese markes leading o a higher level of volailiy. On he oher hand, he volailiy reducing effec of he inflaion news migh be explained by apparen resoluion of uncerainy regarding possible official policy responses o consumer inflaion. This suggess ha he consumer inflaion figures, in general, would lead o a consensus of marke opinion on he direcion of he Japanese economy, whereas he wholesale inflaion news added o he level of uncerainy in he Japanese marke. This news subsequenly had an adverse impac in he oher markes, which were increasingly being linked o he Japanese marke. However, he unexpeced inflaion in he form of WPI and MS had he volailiy reducing effec in general, whereas he unexpecedly high inflaion raised volailiy. 15
In sum, Japanese announcemen spillover effecs are found o be significan o varying exens in all five markes examined. In general, he markes made disincive responses o each ype of news (overall or bad) announced, and here were common responses o some announcemens (BOT and inflaion news) across he markes. The news effecs are also joinly significan in he firs and second momens of reurns series in all cases 16. 4. Conclusions This paper invesigaed he spillover effecs of scheduled economic announcemen news from he U.S. and Japan in he advanced Asia-Pacific sock markes of Ausralia, Hong Kong and Singapore. Specifically, reurn and reurn volailiy of he sock markes were examined o ascerain he presence and he naure of he spillover effecs. Two ypes of news announcemens were used; he ones ha relae o economic aciviies and he ohers on inflaion. Boh he U.S. and he Japanese announcemen news spillovers were significan in general, and hey had varying individual influences on he firs and second momens of marke reurns. In general, he announcemens of worse han expeced performances of he U.S. economy had a negaive reurns impac in he U.S. and heir spillover effecs in he Asia- Pacific were in he same direcion. The overall announcemen news ended o have a posiive effec, however. The spillover effecs on marke volailiies also exhibied asymmeric responses beween he overall and bad news announcemens. Alhough, mosly volailiy reducing effecs were found for worse han expeced daa announcemens, overall news announcemens exhibied boh volailiy increasing and decreasing spillover effecs. I appears ha he degrees of marke uncerainy and heerogeneiy of informaion ownership were lessened afer he markes were surprised wih bad news from he U.S. The Japanese 16 See he las hree lines in Table 4 where he join significance hypohesis esing resuls are repored. 16
news announcemens also provided valuable informaion for he oher sock markes. Significan firs and second momen spillover effecs were repored in he marke reurns in he oher Asia-Pacific and he U.S. The marke responses were, in general, mixed and posiive and negaive effecs were relaively evenly spread across he markes. In addiion, here was no disinc paern of overall/ bad news being consisenly associaed wih eiher posiive or negaive spillover effecs. However, here was some evidence of bad news announcemens producing posiive spillover effecs on he reurn volailiies, wih overall news increasing marke volailiies on average. In sum, i has been found ha he advanced Asia-Pacific sock markes were responsive o disaggregaed flows of informaion, in he form of scheduled economic announcemen news, from he U.S. and Japan. This sheds imporan ligh on he underlying drivers of he informaion leadership of hese wo major markes hrough he radiional channel of sock reurn and volailiy spillovers documened in he lieraure. The evidence, hus, suppors he view ha marke paricipans in he Asia-Pacific sock markes closely follow he marke developmens in he U.S. and Japan and hey incorporae he direc impac of disaggregaed informaion flows as well as indirec flow of aggregaed informaion in erms of he usual reurn and volailiy spillovers. 17
Acknowledgemen I wish o hank an anonymous referee and Andrew Karolyi (edior) for heir invaluable commens and suggesions ha grealy improved he qualiy of his paper. The remaining errors are my own. This projec was funded by an ARC small gran. 18
References Arshanapalli, B. and J. Doukas, 1993, Inernaional sock marke linkages: Evidence from he pre- and pos-ocober 1987 period, Journal of Banking and Finance 17, 193-208. Arshanapalli, B., J. Doukas and L. Lang, 1995, Pre and pos-ocober 1987 sock marke linkages beween U.S. and Asian markes, Pacific Basin Finance Journal 3, 57-73. Bae, K. H. and A. Karolyi, 1994, Good news, bad news and inernaional spillovers of sock reurn volailiy beween Japan and he U.S., Pacipic Basin Finance Journal 2, 405-438. Bae, K. H., A. Karolyi and R. Sulz, 2000, A new approach o measuring financial conagion, Working paper, Dice Cener a Fisher College of Business, Ohio Sae Universiy, Forhcoming in Review of Financial Sudies. Becker, K. G., J. E. Finnery and J. Friedman, 1995, Economic news and equiy marke linkages beween he U.S. and UK, Journal of Banking and Finance 19, 1191-1210. Becker, K. G., J. E. Finnery and K. J. Kopecky, 1995, Domesic macroeconomic news and foreign ineres raes, Journal of Inernaional Money and Finance 14, 763-783. Bollerslev, T., R. Y. Chou and K. F. Croner, 1992, ARCH Modelling in Finance: A Review of he heory and Empirical Evidence, Journal of Economerics 52, 5-59. Cha, B. and Y.-L. Cheung, 1998, The impac of he U.S. and he Japanese equiy markes on he emerging Asia-Pacific equiy markes, Asia-Pacific Financial Markes 5, 191-209. Chow, H.-K., 1999, Co-Movemen in Asian sock indices and he impac of he Asian financial crisis, PACAP Conference, July 1999, Singapore. Connolly, R. A. and F. A. Wang, 2000, On sock marke reurn co-movemens: macroeconomic news, dispersion of beliefs, and conagion, Rice Universiy Working paper, Forhcoming in Pacific Basin Finance Journal. 19
Engle, R.F. and V. K. Ng, 1993, Measuring and Tesing he Impac of News on Volailiy, Journal of Finance 48, 1749-78. Eun, C. S. and Shim, S, 1989, Inernaional Transmission of Sock Marke Movemens, Journal of Financial and Quaniaive Analysis 24:2, 241-56. Garman, M. and M. Klass, 1980, On he esimaion of securiy price volailiy from hisorical daa, Journal of Business 53, 67-78. Ghosh, A., R. Saidi and K. H. Johnson, 1999, Who moves he Asia-Pacific sock markes - U.S. or Japan? Empirical evidence based on he heory of coinegraion, The Financial Review 34, 159-170. Hamao, Y. R., R. W. Masulis and V. K. Ng, 1990, Correlaions in price changes and volailiy across inernaional sock markes, The Review of Financial Sudies 3, 281-307. Janakiramanan, S. and A. Lamba, 1998, An empirical examinaion of linkages beween Pacific-Basin sock markes, Journal of Inernaional Financial Markes, Insiuions and Money 8, 155-173. Kaminsky, G. and S. L. Schmukler, 1999, Wha riggers marke jiers? A chronicle of he Asian crisis, The Journal of Inernaional Money and Finance 18, 537-560. Karolyi, G. A. and R. Sulz, 1996, Why do markes move ogeher? An invesigaion of U.S.- Japan sock reurn comovemens, Journal of Finance 51, 951-986. Kim, S.-J., 1998, Do Ausralian and he U.S. macro economic news announcemens affec he USD/AUD exchange rae? Some Evidence from EGARCH esimaions, Journal of Mulinaional Financial Managemen 8, 233-248. 20
Kim, S.-J. and J. Sheen, 2000, Inernaional linkages and macroeconomic news effecs on ineres rae volailiy - Ausralia and he U.S., Pacific Basin Finance Journal 8, 85-113. Koumos, G. and G. Booh, 1995, Asymmeric volailiy ransmission in inernaional sock markes, Journal of Inernaional Money and Finance 14, 747-762. Liu, Y. A., P. M.-S. and H.-G. Fung, 1996, Inernaional ransmission of sock price volailiy: evidence from he U.S. and six Pacific Basin markes, Journal of Mulinaional Financial Managemen 6, 81-94. Liu, Y. A. and M.-S. Pan, 1997, Mean and volailiy spillover effecs in he U.S. and Pacific- Basin sock markes, Mulinaional Finance Journal 1, 47-62. Ng, A., 2000, Volailiy spillover effecs from Japan and he U.S. o he Pacific-Basin, Journal of Inernaional Money and Finance 19, 207-233. Pan, M.-S., Y. A. Liu and H. J. Roh, 1999, Common sochasic rends and volailiy in Asian-Pacific equiy markes, Global Finance Journal 10, 161-172. Theodossiou, P. and U. Lee, 1993, Mean and volailiy spillovers across major naional sock markes: furher empirical evidence, Journal of Financial Research 16, 337-350. 21
Figure 1: Trading hours and imes for informaion releases JAPAN: CAB, TB, MS, WPI (23:50 GMT, 08:50 TT) JAPAN:, UE (01:30 GMT, 10:30 TT) GMT Time Line 0 3 6 9 12 15 18 21 24 Ausralia (GMT+10) Japan (GMT+9) 10:00 16:00 09:00 11:00 12:30 15:00 U.S.: All Releases 13:30 GMT, 08:30 EST Hong Kong (GMT+8) Singapore (GMT+8) 10:00 12:30 14:30 15:30 09:00 12:30 14:00 16:00 U.S. (EST:GMT-5) 09:30 16:00 U.S. macroeconomic variables: Balance of Trade (BT), Gross Domesic Produc growh rae (GDP), Reail Sales growh rae (RET), Unemploymen rae (UE), Producer Price Index inflaion (PPI), Consumer Price Inflaion (). Japanese macroeconomic variables: Trade Balance (TB), Curren Accoun Balance (CAB), Unemploymen rae (UE), Money supply growh rae (MS), Wholesale Price Index inflaion (WPI), Consumer Price Inflaion (). Sock marke Indexes: All Ordinaries, Nekkei 225, Hang Seng, Sraigh Times and Dow Jones Indusrial 22
Table 1: Summary of scheduled informaion releases from he U.S. and Japan Balance of Paymen (BOT) Gross Domesic Produc (GDP) Unemploymen Rae (UE) U.S. Reail Sales Growh (RET) Consumer Price Index () Producer Price Index (PPI) Trade Balance (TB) Curren Accoun Balance (CAB) Unemploymen Rae (UE) Money Supply Growh (MS) Wholesale Price Index (WPI) Frequency of Announcemens Monhly Quarerly Monhly Monhly Monhly Monhly Monhly Monhly Monhly Monhly Monhly Monhly Japan Consumer Price Index () Source: Acual and MMS expeced announcemen figures Money Marke Services Inernaional Money Marke Services Inernaional Uni of Measuremen $ US billion % change in GDP from previous quarer Unemploymen Rae, % % change of gross reail sales from previous monh % change in from previous monh % change in PPI from previous monh Yen billion Yen billion Unemploymen Rae, % % change in M3 from previous monh % change in WPI from previous monh % change in from previous monh Announcemen Time Daa Period See Figure 1 See Figure 1 January 1991 o May 1999 Toal Number of Announcemens wihin Daa Period 100 34 101 89 99 86 99 56 60 99 96 98 No. (% of oal) of good announcemens (a) 43 (43%) 22 (65%) 45 (45%) 35 (39%) 42 (42%) 44 (51%) 46 (46%) 24 (43%) 15 (25%) 47 (47%) 49 (51%) 45 (46%) No. (% of oal) of bad announcemens (b) 57 (57%) 12 (35%) 56 (55%) 54 (61%) 57 (58%) 42 (49%) 53 (53%) 45( 57%) 45 (75%) 52( 52%) 47 (49%) 53 (54%) Tes of unbiased expecaions (c) 1.02 (0.38) 1.01 (0.75) 0.99 (0.57) 0.94 (0.33) 0.97 (0.4) 0.92 (0.4) 0.94 (0.35) 0.95 (0.44) 1.02 (0.69) 0.96 (0.13) 0.99 (0.66) 0.95 (0.06) Jan-91 o Mar-99 Nov-93 o Mar-99 May-94 o Mar-99 Jan-91 o Apr-99 Apr-91 o Apr-99 Jan-91 o Mar-99 (a) (b) Good announcemens are beer han expeced announcemens of economic aciviy variables and lower han expeced inflaion variables (, WPI and PPI) and unemploymen rae. (c)the unbiasedness is esed via Acual = a + b Expeced + e. 'b' is repored wih he p-value of he hypohesis of a=0 and b=1 in he bracke. 23
Table 2: Saisical properies of overnigh and daily sock marke index reurns Ausralia Japan Hong Kong Singapore Daily (a) Overnigh (b) Daily Overnigh Daily Overnigh Daily Overnigh Daily Summary Saisics U.S. Overnigh Mean 0.0373 0.0012-0.0179 0.0109 0.0633 0.0042 0.0228 0.0281 0.0633 0.0243 Variance 0.6563 0.0421 2.0373 0.1528 3.0738 0.6819 1.7091 0.3546 0.7333 0.2558 Skewness -0.3059-1.0645 0.2321-1.4664 0.0727 0.1688 0.5039 0.1632-0.4533-0.0725 Excess Kurosis 5.5458 32.8117 2.7963 19.3741 11.2988 34.3461 12.0775 48.9803 6.8727 5.4996 Tes of Univariae iid (c) Q(20) : χ 2 (20) 35.98 ** 52.55 24.95 44.33 *** 52.83 *** 62.89 *** 100.61 *** 61.71 *** 27.45 38.51 *** {0.0155} {0.0001} {0.2035} {0.0014} {0.0001} {0.0000} {0.0000} {0.0000} {0.1230} {0.0077} Q 2 (20): χ 2 (20) 426.27 *** 58.20 426.27 *** 426.27 *** 426.27 *** 426.27 *** 426.27 *** 426.27 *** 426.27 *** 426.27 *** {0.0000} {0.0000} {0.0000} {0.0000} {0.0000} {0.0000} {0.0000} {0.0000} {0.0000} {0.0000} E-N: χ 2 (3) 214.25 *** 38.15 21.38 *** 10.38 ** 165.39 *** 149.73 *** 72.49 *** 8.12 ** 384.44 *** 62.43 *** {0.0000} {0.0000} {0.0001} {0.0156} {0.0000} {0.0000} {0.0000} {0.0436} {0.0000} {0.0000} Sock marke indexes are All ordinaries, Nikkei 225, Hang Seng, Sreigh Times, and Dow Jones Indusrial Close Close (a) ln( P / P 1 ) 100 Open Close (b) ln( P / P 1 ) 100 (c) Box-Pierce Q-es of serial correlaion for linear and non-linear (squared) reurns. E-N is Engle and Ng (1993) 's join sign bias es. *, * and ***: Significance a 10, 5 and 1%, respecively. 24
Table 3: The impac of U.S. public informaion releases H Bad Bad Ri = M () + ( αi, j USNEWS j, + αi, j USNEWS j, ), M ( ) j= BOT j= BOT H Bad Bad i = + βi, j j, + βi, j j j= BOT j= BOT ln h V ( ) ( USNEWS USNEWS ) = righ hand side of (1a),, V ( ) = righ hand side of (1a) Open Close Holding periods are H = ON (Overnigh) a +1, for he Non-U.S. marke: [ln( P+ 1 / P ) 100], and H = ON a, Open Close [ln( P / P 1 ) 100] for he U.S. marke, j = BOT, GDP, RET, UE, PPI and. Ausralia Japan Hong Kong Singapore US Coeff p-value Coeff p-value Coeff p-value Coeff p-value Coeff p-value Condiional Mean Equaions αc 0.0000 {0.9845} -0.0181 {0.4937} -0.0145 * {0.0578} 0.0168 *** {0.0044} 0.0110 {0.1080} αhol -0.0075 *** {0.0014} -0.2567 *** {0.0000} -0.0632 {0.1078} -0.0693 *** {0.0002} -0.0142 {0.3734} αbot 0.0074 {0.1975} -0.0892 {0.3720} -0.0557 *** {0.0002} -0.0875 *** {0.0017} -0.0793 ** {0.0362} αgdp 0.0094 {0.1037} -0.0069 {0.9513} 0.0453 {0.1781} 0.0239 ** {0.0127} 0.0116 {0.7632} αret 0.0080 ** {0.0184} 0.2220 ** {0.0121} 0.0831 {0.3239} -0.0127 {0.2900} 0.0019 {0.9554} α UE 0.0373 *** {0.0000} -0.0010 {0.9887} 0.0784 * {0.0626} 0.0643 *** {0.0035} 0.0202 {0.5421} αppi 0.0262 *** {0.0000} 0.5075 *** {0.0000} 0.2037 *** {0.0092} -0.0022 {0.9203} -0.0455 {0.2686} α 0.0098 *** {0.0011} 0.2727 *** {0.0098} 0.0752 ** {0.0488} 0.0660 *** {0.0001} 0.0465 {0.1807} αbot-bad -0.0277 *** {0.0100} 0.1240 {0.5584} 0.2879 ** {0.0252} 0.1531 *** {0.0014} -0.0976 * {0.0687} αgdp-bad -0.0319 *** {0.0033} 0.3274 * {0.0627} 0.0141 {0.7807} 0.0882 ** {0.0391} -0.0612 {0.3256} αret-bad -0.0088 ** {0.0361} -0.3930 *** {0.0052} 0.0328 {0.7147} -0.0526 {0.2264} 0.1026 * {0.0713} αue-bad -0.0352 *** {0.0000} 0.1022 {0.6947} 0.1315 {0.1123} -0.0137 {0.5091} -0.0014 {0.9801} αppi-bad -0.0258 *** {0.0001} -0.4120 ** {0.0138} -0.4663 *** {0.0000} 0.1081 ** {0.0254} 0.0886 {0.2593} α -Bad -0.0713 *** {0.0007} 0.0904 {0.7238} -0.0815 {0.1413} -0.1487 *** {0.0000} 0.0757 {0.1955} Condiional Variance Equaions βc -0.1145 *** {0.0000} -0.0305 *** {0.0000} -0.0591 *** {0.0000} -0.0430 *** {0.0000} -0.0820 *** {0.0000} β ε1-0.0456 {0.2070} -0.0815 *** {0.0000} -0.0598 * {0.0551} -0.0071 * {0.0730} -0.0314 ** {0.0403} β ε2 0.3237 *** {0.0000} 0.1258 *** {0.0000} 0.1361 ** {0.0265} 0.1508 *** {0.0000} 0.1377 *** {0.0000} βh 0.9437 *** {0.0000} 0.9847 *** {0.0000} 0.9810 *** {0.0000} 0.9912 *** {0.0000} 0.9561 *** {0.0000} βhol 0.0651 *** {0.0000} 0.2805 *** {0.0000} 0.0458 {0.5467} 0.3715 *** {0.0000} 0.1314 *** {0.0000} β BOT 0.7501 *** {0.0000} -0.1070 *** {0.0000} -0.2042 {0.1687} -0.0523 * {0.0795} -0.1555 *** {0.0000} βgdp 0.4751 *** {0.0000} 0.0312 {0.2955} 0.1721 {0.1469} 0.0212 {0.5838} -0.1858 *** {0.0000} βret -1.5379 *** {0.0000} 0.0780 *** {0.0002} 0.8328 *** {0.0000} -0.3833 *** {0.0001} -0.0781 *** {0.0000} βue 0.3423 *** {0.0000} -0.0274 * {0.0935} -0.0211 {0.9185} 0.4445 *** {0.0000} 0.0723 {0.1398} βppi -0.3904 *** {0.0012} 0.0163 {0.3879} 0.4975 *** {0.0003} -0.1839 *** {0.0009} 0.0242 {0.1910} β 0.7648 *** {0.0010} 0.0581 ** {0.0466} -0.2053 {0.3216} 0.7024 *** {0.0000} 0.0331 {0.4314} βbot-bad 0.8877 *** {0.0000} -0.0067 {0.8908} 0.7986 *** {0.0000} -0.2674 *** {0.0000} 0.2709 *** {0.0000} βgdp-bad -0.1936 ** {0.0323} -0.4102 *** {0.0000} 0.1150 {0.3648} -0.2995 *** {0.0000} 0.1098 {0.2188} βret-bad -0.4399 *** {0.0000} -0.1545 ** {0.0200} -0.7129 *** {0.0002} -0.0531 {0.1367} 0.0461 {0.1660} βue-bad -0.6109 *** {0.0000} -0.2789 *** {0.0055} 0.1667 {0.4365} -0.6324 *** {0.0000} -0.4214 *** {0.0083} βppi-bad -0.7088 *** {0.0000} 0.0984 {0.1567} 0.3767 * {0.0912} -0.1242 {0.2908} 0.1432 ** {0.0339} β-bad 0.7470 *** {0.0000} -0.0913 {0.1712} -0.1911 {0.4410} -0.3279 {0.2036} 0.0487 {0.4131} Diagnosics Log-L 2996.1 1259.9 71.3 989.8 666.0 q 5 2 3 0 8 Skewness -1.4899-0.2140-0.0688-1.1688-0.3242 Kurosis 28.6630 15.7940 8.2497 31.7499 2.7568 Q(20) 21.68 {0.3579} 22.62 {0.3080} 13.80 {0.8406} 12.56 {0.8955} 13.97 {0.8322} Q 2 (20) 15.64 {0.7389} 8.13 {0.9910} 23.57 {0.2615} 18.42 {0.5599} 12.22 {0.9084} E-N 1.05 {0.7892} 1.30 {0.7286} 4.56 {0.2070} 9.43 ** {0.0241} 2.42 {0.4905} H0: α news =0 390 *** {0.0000} 131 *** {0.0000} 128 *** {0.0000} 354 *** {0.0000} 17 {0.1571} H0: β news =0 95022 *** {0.0000} 1086 *** {0.0000} 8890 *** {0.0000} 1174346 *** {0.0000} 1132 *** {0.0000} H0: α news =β news =0 160231 *** {0.0000} 1971 *** {0.0000} 19803 *** {0.0000} 1448276 *** {0.0000} 1246 *** {0.0000} Tess of whie noise on he sandardized residuals of he esimaions, z, are repored in he las panel. Q(20) and Q 2 (20) are Q-ess of linear and non-linear serial correlaions of z, and E-N is Engle and Ng es of join sign bias for z. Hypohesis esings include join significance ess of news effecs on he mean, condiional variance, and on boh mean and variance. *, * and ***: Significance a 10, 5 and 1%, respecively. 25
Table 4: The impac of Japanese public informaion releases H Bad Bad R = M () + ( α JPNEWS + α JPNEWS ) M ( = righ hand side of (1a),, ) i i, j j, i, j j, j= TB j= TB H Bad Bad ln hi = V ( ) + ( βi, j JPNEWS j, + βi, j JPNEWSi, j ), V ( ) = righ hand side of (1a) j= TB j= TB Close Close Holding periods are H = D (Daily) a, [ln( P / P 1 ) 100] for he non-u.s. markes and H = ON a, Open Close [ln( P / P 1 ) 100] for he U.S. marke, j = TB, CAB, UE, MS, WPI and. Ausralia Japan Hong Kong Singapore US Coeff p-value Coeff p-value Coeff p-value Coeff p-value Coeff p-value Condiional Mean Equaions α c 0.0250 * {0.0865} 0.0276 *** {0.0012} 0.0833 *** {0.0024} 0.0585 *** {0.0062} 0.0104 {0.1585} α Hol -0.0487 {0.1782} -0.2457 *** {0.0000} -0.1241 ** {0.0160} -0.1441 *** {0.0001} 0.0051 {0.7717} α TB 0.0048 {0.9428} -0.2648 *** {0.0095} -0.1109 {0.1811} -0.0688 {0.3343} -0.1649 *** {0.0000} α CAB 0.1521 {0.1468} 0.2963 ** {0.0322} -0.0862 {0.5475} -0.1316 {0.2100} -0.0915 * {0.0571} α UE -0.1236 {0.1763} -0.0515 {0.7477} -0.3390 *** {0.0008} -0.1674 *** {0.0069} 0.1572 *** {0.0041} α MS 0.0630 {0.3572} 0.0375 {0.6692} 0.3005 *** {0.0002} 0.0439 {0.4680} 0.0089 {0.8094} α WPI 0.0032 {0.9666} -0.0224 {0.8167} 0.0625 {0.5228} 0.0800 {0.2141} 0.0307 {0.4890} α 0.1322 * {0.0570} 0.1673 {0.1104} 0.1952 ** {0.0345} 0.2107 *** {0.0001} -0.0004 {0.9917} α TB-Bad 0.0477 {0.6696} 0.5369 *** {0.0012} 0.2528 * {0.0655} 0.2404 *** {0.0027} 0.1969 *** {0.0001} α CAB-Bad -0.1546 {0.3801} -0.3520 {0.2005} -0.5546 * {0.0587} -0.3809 *** {0.0031} 0.0461 {0.5012} α UE-Bad 0.2452 * {0.0670} -0.0108 {0.9673} 0.7762 *** {0.0001} 0.2102 {0.1060} -0.3009 *** {0.0001} α MS-Bad 0.0214 {0.8165} -0.2484 * {0.0870} -0.3575 ** {0.0258} -0.1719 ** {0.0485} -0.1251 ** {0.0349} α WPI-Bad -0.0310 {0.7772} 0.1128 {0.4116} 0.2402 {0.2490} 0.0307 {0.7151} 0.0880 {0.2870} α -Bad -0.1400 {0.1639} -0.4655 ** {0.0273} -0.1046 {0.3004} -0.1019 {0.3162} 0.0280 {0.7043} Condiional Variance Equaions β c -0.0858 *** {0.0000} -0.0163 *** {0.0000} 0.0157 *** {0.0001} -0.0748 *** {0.0000} -0.0938 *** {0.0000} β ε1-0.0832 *** {0.0000} -0.0750 *** {0.0000} -0.0845 *** {0.0000} -0.0646 *** {0.0000} -0.0449 ** {0.0308} β ε2 0.1182 *** {0.0000} 0.0994 *** {0.0000} 0.2078 *** {0.0000} 0.2073 *** {0.0000} 0.1420 *** {0.0000} β h 0.9490 *** {0.0000} 0.9861 *** {0.0000} 0.9712 *** {0.0000} 0.9798 *** {0.0000} 0.9483 *** {0.0000} β Hol 0.3085 *** {0.0000} 0.1243 *** {0.0000} 0.0640 ** {0.0382} 0.5082 *** {0.0000} 0.0430 *** {0.0000} β TB -0.1479 *** {0.0000} 0.0758 *** {0.0051} -0.4362 *** {0.0000} -0.5925 *** {0.0000} 0.0514 {0.1061} β CAB -0.0229 {0.6718} 0.0433 * {0.0618} 0.1573 *** {0.0013} 0.1532 *** {0.0020} -0.3654 *** {0.0000} β UE 0.0563 {0.2908} -0.2355 *** {0.0019} 0.3504 * {0.0617} -0.0265 {0.6324} 0.4800 *** {0.0000} β MS 0.0028 {0.8999} -0.0556 {0.3786} 0.1780 ** {0.0483} 0.0877 ** {0.0202} 0.1158 *** {0.0000} β WPI 0.0529 * {0.0522} 0.0573 *** {0.0000} 0.2977 *** {0.0000} 0.3286 *** {0.0000} 0.3619 *** {0.0000} β 0.1180 *** {0.0000} -0.1476 *** {0.0000} -0.1425 ** {0.0212} -0.2567 *** {0.0000} -0.3300 *** {0.0000} β TB-Bad 0.2781 *** {0.0000} -0.1105 ** {0.0134} 0.5316 *** {0.0000} 0.4300 ** {0.0147} -0.1449 *** {0.0003} β CAB-Bad -0.0598 {0.6160} 0.0103 {0.8892} -0.2610 ** {0.0235} -0.0983 {0.3016} 0.0811 {0.4151} β UE-Bad -0.0639 {0.4854} 0.3317 *** {0.0003} -0.2367 * {0.0502} 0.0012 {0.9928} -0.2031 ** {0.0262} β MS-Bad -0.0762 {0.2170} 0.0464 {0.3241} -0.3063 *** {0.0000} -0.2829 *** {0.0000} -0.1354 * {0.0666} β WPI-Bad -0.2491 *** {0.0019} 0.2509 *** {0.0019} -0.1311 {0.1849} 0.0268 {0.7727} -0.2238 ** {0.0146} β -Bad 0.0178 {0.7612} 0.2543 *** {0.0000} -0.0842 {0.1415} -0.1371 {0.1263} 0.3652 *** {0.0002} Diagnosics Log-L -505.1-1660.8-1865.1-1113.1 680.4 q 1 0 1 2 8 Skewness -0.2234 0.1432-0.1803 0.3912-0.3267 Kurosis 1.1340 1.7480 2.3443 2.5675 2.6815 Q(20) 19.15 {0.5118} 12.49 {0.8983} 18.59 {0.5483} 29.20 * {0.0839} 11.72 {0.9253} Q 2 (20) 18.58 {0.5495} 15.20 {0.7646} 6.92 {0.9969} 20.42 {0.4318} 12.76 {0.8876} E-N 4.98 {0.1733} 1.57 {0.6670} 3.44 {0.3293} 5.91 {0.1159} 3.37 {0.3387} H0: α news =0 38 *** {0.0002} 79 *** {0.0000} 132 *** {0.0000} 65 *** {0.0000} 219 *** {0.0000} H0: β news =0 210 *** {0.0000} 1223 *** {0.0000} 6924 *** {0.0000} 1014 *** {0.0000} 1299 *** {0.0000} H0: α news =β news =0 339 *** {0.0000} 1451 *** {0.0000} 8795 *** {0.0000} 1811 *** {0.0000} 1724 *** {0.0000} Tess of whie noise on he sandardized residuals of he esimaions, z, are repored in he las panel. Q(20) and Q 2 (20) are Q-ess of linear and non-linear serial correlaions of z, and E-N is Engle and Ng es of join sign bias for z. Hypohesis esings include join significance ess of news effecs on he mean, condiional variance, and on boh mean and variance. *, * and ***: Significance a 10, 5 and 1%, respecively. 26