Causal Relationship between the United States, Hong Kong and China s stock markets. Kwok Sin Hang Sindy Finance Option

Transcription

1 Causal Relaionship beween he Unied Saes, Hong Kong and China s sock markes BY Kwok Sin Hang Sindy Finance Opion An Honours Degree Projec Submied o he School of Business in Parial Fulfilmen of he Graduaion Requiremen for he Degree of Bachelor of Business Adminisraion (Honours) Hong Kong Bapis Universiy Hong Kong April 2005

2 Conen Acknowledgemen... Absrac...2. Inroducion Lieraure Review Daa and Preliminary Resuls Mehodology... I) Procedures of he Granger Causaliy Tes... II) Granger Causaliy Relaionship Empirical Resuls...3 I) The pre-enering period...7 II) The pos-enering period Limiaions Conclusion and Discussion...22 References...24 Appendix I...26 Uni Roo Tes for Hang Seng Index...26 Uni Roo Tes for Shanghai A Index...27 Uni Roo Tes for Shanghai B Index...29 Uni Roo Tes for Shenzhen A Index...30 Uni Roo Tes for Shenzhen B Index...32 Uni Roo Tes for S&P 500 Composie Index...33 Appendix II...34 Granger Causaliy Tess Resuls for he pre-enering period...34 Granger Causaliy Tess Resuls for he pos-enering period...36

3 Acknowledgemen I would like o ake his opporuniy o express my sincere hanks owards my supervisor, Dr. Joseph K W FUNG. During our discussion, he no only has given his invaluable advice for my projec, bu also has showed his paience and encouragemen, ha s really suppor me a lo. Besides, I would like o hank o my classmaes, especially Mr. Dick Tsao and Mr. Kay Wong, o help me on he echnical problems I encounered, and sress I suffered. Wihou heir help, his projec may no be finished so smoohly. Lasly, I would like o hank all he saffs in Hong Kong Bapis Universiy who each me a lo abou he academic suffs, which are useful for my whole life.

4 Absrac This paper examines he causal relaionships beween Shanghai A, Shanghai B, Shenzhen A, Shenzhen B, Hang Seng Index and he S&P 500 Composie Index. In order o find ou wheher here exis such relaionships before and afer China became a member of he World Trade Organizaion, suiable hypohesis were made and esed by he Granger mehodology. And he invesigaion was divided ino wo periods, obviously he pre-enering period and he pos-enering period. I is ineviably ha here is a feedback relaionship beween he Hang Seng Index and S&P 500 Composie Index, as hey are boh imporan indices in he world financial marke. Moreover, he Granger causaliy resuls indicaed ha here were significan changes in he Chinese sock markes for he wo periods. The A sock markes had leading roles o he Hong Kong marke and oher Chinese markes in he pre-enering period, while he B sock markes Granger caused all he Hong Kong marke, Chinese A markes and even he US marke in he pos-enering period. I showed ha he China markes which are open o he foreign invesors had developed a lo afer China became a member of he WTO. I is believed o be caused by a higher marke efficiency brough by his open rade policy. Furhermore, hese resuls give a grea suppor o he asserion ha he Chinese sock markes are becoming more inegraed o he global economy. 2

5 . Inroducion The economic reforms sared in 978 led o he rebirh of he sock markes in China. The Shanghai Sock Exchange and he Shenzhen Sock Exchange are he wo major emerging capial markes in China. The Shanghai Sock Exchange commenced operaions in December 990 while he Shenzhen Sock Exchange opened is marke in April 99. Alhough hese sock exchanges only operae for a shor period of ime when compare wih hose in he Unied Saes and Hong Kong, he rapid expansion of he China markes reflecs China s significan economic growh, hus simulaed many inernaional invesors o include Chinese socks in heir porfolio. Hence he knowledge of he linkage beween Chinese markes and oher foreign markes enables porfolio managers o make more inelligen decisions. As he US and Hong Kong are he wo major inernaional rading parners of China, and also he op wo direc invesors in China, here are very close relaionships beween China and he US and beween China and Hong Kong. Wih 997 handover of Hong Kong o China, he securiies indusries of he wo sides ough o inegrae ino one marke. However, he Basic Law governing Hong Kong mainains unchanged Briish Law sysem as well as he life syle of Hong Kong capialis sociey, a leas for he nex 50 years. This had ulimaely srenghened he relaionship beween China and Hong Kong. 3

6 Moreover, being a new member of he World Trade Organizaion (WTO) in December 200 evenually showed ha China is gradually opening is marke o he world. Pracically, his enrance can promoe he rade beween China and oher counries, and informaion is supposed o be sen more efficienly among he world. So wih he sronger efficiency in he marke, more and more invesors are ineresed and confiden in he China marke which has grea business opporuniies. I would probably enrich he local sock markes. Therefore, his paper is going o invesigae he causaliy of he Chinese sock markes wih each oher and wih he US and Hong Kong sock markes, in erms of he sock reurns, afer he enrance of China ino he WTO o see wheher he Chinese sock markes have inegraed ino he global economy. 2. Lieraure Review There are quie a number of researches on he inerrelaionship beween he world sock markes. However, i appears ha hey have no provided consisen resuls. The sizes and signs of correlaion coefficiens varied depending on he choice of markes, he sample period chosen, he frequency of observaions, ec. A. G. Malliaris and Jorge L. Urruia (992) wroe a paper analyzing he lead-lag relaionships for six major inernaional sock indexes, including he New York S&P 4

7 500, Tokyo Nikkei, London FT-30, Hong Kong Hang Seng, Singapore Srais Times, and Ausralia All Ordinaries, for he ime periods before, during and afer he Ocober 987 marke crash. A dramaic increase in unidirecional and bidirecional causaliy is observed in he monh of he crash, compared wih he periods before and afer he crash. In he crash monh, New York showed feedback wih London and Hong Kong. London led Hong Kong, and Hong Kong led he oher Asian markes. This phenomenon probably suggesed ha he crash sared simulaneously in all he sock markes, making he marke crash of Ocober 987 seems o have been an inernaional crisis of he equiy markes. Afer analyzing he crash in Ocober 987, Fok Ka Man in HKBU (200) invesigaed he relaionship beween he Unied Saes sock marke and he Hong Kong sock marke, for he periods before, during and afer he 997 Asian financial crisis. I showed ha boh Dow Jones Indusrial Average (DJIA), NASDAQ composie Index and S&P 500 have subsanial impac on he Hong Kong sock marke, shown in he Hang Seng Index (HSI). I is clear ha he US markes led he Hong Kong marke in he pre-financial crisis period and he pos- financial crisis period,,bu he relaionship was less significan during he crisis period which was believed o be ha he regional facors had much impac on Hong Kong han ha from he US. Concerning he Chinese sock markes, Marin Laurence, Francis Cai and Sun Qian 5

8 (997) invesigaed he weak-form efficiency and causaliy ess in Chinese sock markes in erms of daily sock reurns. I concluded ha here is a causal relaionship from B socks markes o he A sock markes, meaning he foreign markes exer a significan influence on he markes only open o Chinese naionals. Also, he US sock marke exhibis a srong causal relaion o all he four Chinese sock markes and he Hong Kong marke. This suggesed ha he Chinese sock markes are gradually being inegraed ino he global economy. Cheng F. Lee and Oliver M. Rui (2000) examined wheher rading volume conains informaion o predic sock reurns in he China s sock markes, he US and he Hong Kong sock markes. Surprisingly hey have an opposie conclusion from ha of Marin Laurence, Francis Cai and Sun Qian (997). They summarized ha he US and he Hong Kong financial marke informaion conained in reurns, volailiy and volume have very weak predicive power for Chinese financial marke variables. The China s financial marke is independen of he world financial marke. 3. Daa and Preliminary Resuls The daa in his sudy will be he daily observaions for he four Chinese sock marke indices, one US and one Hong Kong sock index. The indices for he Chinese marke are he Shanghai A (SHA), he Shanghai B (SHB), he Shenzhen A (SZA), and 6

9 is he Shenzhen B (SZB), where A shares are for domesic invesors and B shares are for foreign invesors. For he US marke, S&P 500 Composie Index (S&P) is used; and for he Hong Kong marke, he major sock index Hang Seng Index (HSI) is used. The causal relaionship beween all hese six sock exchanges is examined. Daily sock indices from January, 999 o December 3, 2004 are used in he analysis. In order o observe he effec of China geing ino he WTO o all hese sock exchanges, wo periods are divided as follows: Pre-enering period: January, 999 o December 0, 200 Pos-enering period: December, 200, o December 3, 2004 The daily sock reurns are calculaed using he coninuously compounded formula. R P = ln( P ) where PBB he price of he index in day and ln is naural logarihm. Since here is ime difference among he world, and he rading hours of differen markes are differen, so he ime-zone difference is needed o ake ino concern. As shown in Figure, from he Hong Kong ime perspecive, he Hong Kong marke opens a 0:00 and closes a 6:00, whereas he Chinese sock markes open a 09:30 and close a 5:00, while he US marke opens a 22:30 and closes a 05:00 in he following day. The Hong Kong marke and he Chinese markes have common ime inerval, bu no wih ha of he US marke. Thus he causaliy beween Hong Kong 7

10 and China is of + day, while ha beween Hong Kong and US, and China and US may be observed on he same rading day. Figure. Trading Hours for Hong Kong marke, Chinese marke and US marke Marke Time Tables Sock Markes Hong Kong Time (Hours) Local Time (Hours) Hong Kong 0:00-2:30, 4:30-6:00 0:00-2:30, 4:30-6:00 Unied Saes 22:30-05:00(he following day) 09:30-6:00 China 09:30-:30, 3:00-5:00 09:30-:30, 3:00-5:00 As he reurns a ime in China ( R ) and Hong Kong ( R CN HK ) affec reurns in US ( R US ) on he same calendar day, a Granger regression invesigaing if China or Hong Kong is leading US looks as follows. R R CN HK q p US = a0 + i R i + 0 α β R + ε q p US = a0 + i R i + 0 i i CN i α β R + ε HK i On he oher hand, he US reurns ( R ) a ime may affec reurns in China ( US CN R + ) and Hong Kong ( HK R + ) a ime +. Thus afer adjusing for ime-zone differences, a Granger regression posulaing ha US is leading China or Hong Kong looks like his. R R US US q p CN = a0 + i R i + α β R + ε q p HK = a0 + i R i + i i US i α β R + ε US i 8

11 is is is Moreover, he rading hours of he Chinese sock markes and he Hong Kong sock marke are of similar ime period. Thus he lead-lag relaionship beween hem can be found in successive days by he following regression models. R R HK CN q p CN = a0 + i R i + α β R + ε q p HK = a0 + i R i + i i HK i α β R + ε CN i The vecor auoregressions we used for causaliy ess assume ha he variables in he sysem are saionary. So before esimaing he relaionship among differen markes, a uni roo es is performed for each of he six sock indices using he augmened Dickey-Fuller (ADF) es, o es for he saionariy of he sock reurns of differen markes. The daa used for he uni roo es are hose of he wo periods, ha is from January, 999 o December 3, The augmened Dickey-Fuller (ADF) es consiss of esimaing he following regression model: R = β + β + δr + α R + ε 2 m i i where εbb a pure whie noise error erm and R = R R ), ( 2 R = ( R R ) 3, ec. 2 2 We es wheher δ= 0. If i is zero, we conclude ha RBB nonsaionary and has a uni roo in i. Bu if i is negaive, we conclude ha RBB saionary and do no have a 9

12 is uni roo. The following hypohesis is se up for esing each sock marke s reurns. HB0B: δ= 0, RB Bis nonsaionary HBB: δ< 0, RBB saionary Table presens he ADF saisics for he esing. Table. Uni Roo Tess for he sock reurns Index Augmened Dickey-Fuller es saisic HB0B: nonsaionary HSI Rejeced SHA Rejeced SHB Rejeced SZA Rejeced SZB Rejeced S&P Rejeced Noe: HSI is Hang Send Index, SHA is Shanghai A, SHB is Shanghai B, SZA is Shenzhen A, SZB is Shenzhen B, and S&P is S&P 500 Composie Index. The criical values for wih observaions are: a 0%, a 5 %, and a %. The ADF es saisics es he hypohesis of a uni roo in each series. Saisically insignifican (δ= 0) imply accepance of he null hypohesis. As shown in Table, all 0

13 and represens be causes is series, in he indices are significan (δ< 0) a % level, 5% level, and 0%level, hus we rejeced HB0 Bin all he ess and concluded ha all he indices are saionary. 4. Mehodology In his paper, Granger Causaliy Tess are used o invesigae lead-lag relaionships among he four Chinese sock markes and wih ha of he US and Hong Kong. Le YBB XBB he reurn series for any wo markes ou of he six. YBB XBB he Granger sense if presen value of X can be prediced by using pas values of Y han by no doing so, considering also oher relevan informaion, including pas values of X. More specifically, X is said o cause Y, provided some αbb no zero in Equaion (2). I) Procedures of he Granger Causaliy Tes The seps involved in implemening he Granger causaliy es are as follows. Firs, using he echnique of vecor auoregression (VAR), esimae a resriced equaion of order p Y p = δ + α Y + ε 0 i i () where YBB he reurn in marke Y a ime. The lag-lengh p is deermined using he Akaike informaion crierion (AIC). Nex, esimae an unresriced equaion which includes he pas daa of he XBB

14 which is q p = + β i X i + Y σ 0 α Y + µ (2)* *Specific regression model is used for each sock marke, considering he i i corresponding ime-zone difference. The lag-lengh q of XB Bis also deermined using AIC, given he value of p from he resriced equaion. Then run he resriced regression and unresriced regression respecively o obain he resriced residual sum of squares, RSSBRB and unresriced residual sum of squares, RSSBURB. The F-saisic below F = ( RSS RSS UR R RSS UR ) / q /( n p q ) where n is he sample size, provides a formal es for causaliy. If he compued F value exceeds he criical F value a he chosen level of significance, we concluded ha here is a causaliy relaionship from marke X o marke Y. II) Granger Causaliy Relaionship Considering he following pair of regressions: q p 0 + i X i + α i Y = a β Y i + e (I) q p 0 + iy i + λi X = c δ X i + u (II) 2

15 A unidirecional causaliy from X o Y is indicaed if he esimaed coefficiens on he lagged X in Equaion (I) are saisically differen from zero as a group (i.e., β i 0) and he se of esimaed coefficiens on he lagged Y in Equaion (II) is no saisically differen from zero (i.e., δ i = 0 ). Conversely, a unidirecional causaliy from Y o X exiss if he se of lagged X coefficiens in Equaion (I) is no saisically differen from zero (i.e., β i = 0 ) and he se of lagged Y coefficiens in Equaion (II) is saisically differen from zero (i.e., δ i 0 ). Feedback causaliy exiss when he ses of X and Y coefficiens are saisically differen from zero in boh regressions. ( β i 0 & δ i 0 ) Finally, independence is suggesed when he ses of X and Y coefficiens are no saisically significan in boh regressions. ( β i = 0 & δ i = 0 ) 5. Empirical Resuls This secion presens he resuls of he ess of he causaliy relaionship beween he four Chinese sock indices, Hang Seng Index and S&P 500 Composie Index. The main resuls of he Granger Causaliy ess among he six sock exchanges are presened in Table 2a and Table 2b. 3

16 Table 2a. Granger Causaliy es resuls for he pre-enering period Pre-enering Period X o Y p q n F-value Significan Causal Level Relaionship HSI o SHA Yes HSI o SHB No HSI o SZA No HSI o SZB No HSI o S&P Yes SHA o HSI Yes SHA o SHB No SHA o SZA No SHA o SZB Yes SHA o S&P No SHB o HSI No SHB o SHA No SHB o SZA No SHB o SZB No SHB o S&P No SZA o HSI Yes SZA o SHA No SZA o SHB No SZA o SZB Yes SZA TO S&P No SZB o HSI No SZB o SHA Yes SZB o SHB No SZB o SZA No SZB o S&P No S&P o HSI Yes S&P o SHA No S&P o SHB No 4

17 S&P o SZA No S&P o SZB No Noe: HSI is Hang Send Index, SHA is Shanghai A, SHB is Shanghai B, SZA is Shenzhen A, SZB is Shenzhen B, and S&P is S&P 500 Composie Index. X o Y indicaes X leads Y. The values of significan level are chosen a 5%. Yes represens he causal relaionship being significan a 5%. From Table 2a, SZA o HSI and SZA o SZB showed unidirecional causaliies; HSI and SHA, SHA and SZB, and HSI and S&P showed feedback causaliies; while he remaining ses of X o Y markes are independen o each oher. Table 2b. Granger Causaliy es resuls for he pos-enering period Pos-enering Period X o Y p q n F-value Significan Causal Level Relaionship HSI o SHA No HSI o SHB No HSI o SZA No HSI o SZB No HIS o S&P Yes SHA o HSI Yes SHA o SHB No SHA o SZA No SHA o SZB No SHA o S&P No SHB o HSI Yes SHB o SHA Yes SHB o SZA Yes SHB o SZB No SHB o S&P No 5

18 SZA o HSI Yes SZA o SHA No SZA o SHB No SZA o SZB No SZA TO S&P No SZB o HSI Yes SZB o SHA No SZB o SHB No SZB o SZA No SZB o S&P Yes S&P o HSI Yes S&P o SHA No S&P o SHB No S&P o SZA No S&P o SZB No Noe: HSI is Hang Send Index, SHA is Shanghai A, SHB is Shanghai B, SZA is Shenzhen A, SZB is Shenzhen B, and S&P is S&P 500 Composie Index. X o Y indicaes X leads Y. The values of significan level are chosen a 5%. Yes represens he causal relaionship being significan a 5%. From Table 2b, SHA o HSI, SHB o HSI, SHB o SHA, SHB o SZA, SZA o HSI, SZB o HSI, and SZB o S&P all showed unidirecional causaliies; HSI and S&P showed feedback causaliies; while he remaining ses of X o Y markes are independen o each oher. To help undersand he resuls in a more concise manner, all he resuls in Table 2a and Table 2b are summarized in Table 3a and 3b. 6

19 I) The pre-enering period For he period before China became a member of he World Trade Organizaion, Shenzhen A showed leading roles in he cases comparing wih Hang Seng Index and Shenzhen B respecively. For Shenzhen A Granger-caused Hang Seng Index, as Hong Kong hangover o China in 997, here was a closer relaionship beween Hong Kong and China, so as he relaionship in heir sock markes, hence he performance of he Shenzhen A affeced he HSI s performance. For he A sock markes reurns leading he B sock markes reurns, i is believed ha i was due o he weak marke efficiency in he China marke. Table 3a. Summary of causal relaionships for he pre-enering period X Y HSI SHA SHB SZA SZB S&P HSI Y N N N Y SHA Y N N Y N SHB N N N N N SZA Y N N Y N SZB N Y N N N S&P Y N N N N Noe: HSI is Hang Send Index, SHA is Shanghai A, SHB is Shanghai B, SZA is Shenzhen A, SZB is Shenzhen B, and S&P is S&P 500 Composie Index. Yes represens he causal relaionship being significan a 5%. As he B shares socks are for foreign invesors, hey found ha i was more difficul, relaive o he domesic invesors, o acquire and access reliable informaion 7

20 abou he local Chinese firms. Thus he performance of he A sock markes could be a good predicor o he B sock markes. This happened beween SZA and SZB. On he oher hand, feedback causal relaionships exised beween Hang Seng Index and Shanghai A, Shanghai A and Shenzhen B, Hang Seng Index and S&P 500 Composie Index respecively. I showed ha here were close iner-relaionship beween each pair of hem. As menioned before, A sock markes in China had a leading role o HSI due o he srenghened relaionship beween hem afer he hangover of Hong Kong o China, his happened also beween Shanghai A and Hang Seng Index. A he same ime, SHA showed feedback o he performance of HSI, indicaing ha Hong Kong, as a financial leader in Asia, does posiively influenced he decision of he domesic invesors in Shanghai A. For Shanghai A and Shenzhen B, he siuaion was similar o ha of he Shenzhen A and Shenzhen B. These resuls indicaed ha Shenzhen B invesors relied on he performance of he A sock markes. The main difference beween hem was ha Shenzhen B performance also worked as a facor of decision o he Shanghai A invesors, bu no o he Shenzhen A invesors. Thus here was only feedback in he case beween SHA and SZB. And for he feedback beween Hang Seng Index and S&P 500 Composie Index, i was obvious from he pas journals and from he markes ha here was very close relaionship beween he Hong Kong sock marke 8

21 and he US sock markes, as hey are boh very imporan sock exchanges in he world financial marke. Ineresingly, Shanghai B had no apparen influence on all he oher five sock exchanges. All F-values are saisically insignifican. I showed ha i had no leading role in he China, US and Hong Kong markes. I had a very weak predicive power in all hese markes during he esed period. Also, all Chinese sock markes are independen o he US S&P 500 Composie Index, his showed ha he Chinese sock markes did no follow he performance of he US marke, which is one of he imporan financial markes in he world, and neiher had any predicive power o he US marke. II) The pos-enering period Afer China became a member of he WTO on December, 200, he causal relaionships among he six sock exchanges had some changes. Firs, he predicive power of HSI o he Chinese sock markes seen in he pre-enering period disappeared, indicaing ha he Chinese sock markes were independen o he Hong Kong marke in he pos-enering period. Second, he resuls showed ha Hang Seng Index had a lagging posiion in all he cases wih he oher five exchanges, which means ha boh he China and US markes 9

22 led he Hong Kong marke. The previous rading day performance in hese markes srongly influenced he rading day performance of Hong Kong in he following days. And Hang Seng Index only showed feedback o he S&P 500 Composie Index, bu no for he Chinese sock markes, indicaing ha he Hong Kong marke and he US marke affec each ohers, which was he same as he period before being a member of he WTO. Table 3b. Summary of causal relaionships for he pos-enering period X Y HSI SHA SHB SZA SZB S&P HSI N N N N Y SHA Y N N N N SHB Y Y Y N N SZA Y N N N N SZB Y N N N Y S&P Y N N N N Noe: HSI is Hang Send Index, SHA is Shanghai A, SHB is Shanghai B, SZA is Shenzhen A, SZB is Shenzhen B, and S&P is S&P 500 Composie Index. Yes represens he causal relaionship being significan a 5%. Besides, he sock exchanges in China ha open o he foreigners, ha is he Shanghai B and Shenzhen B, showed leading saus in differen ess. Shanghai B led boh Hang Seng Index, Shanghai A and Shenzhen A, while Shenzhen B led S&P 500 Composie Index. This phenomenon proved ha he B sock markes in he China marke experienced a leading role wihin he local sock markes, and also 20

23 o he US and Hong Kong markes in he pos-enering period. I is hus believed ha here was a posiive effec o he Chinese sock markes wih he saus of being a member of he WTO. Informaion were more easily available o he foreign invesors, hus hey could make heir invesmen decision more concisely and confidenly, which made he B sock markes in China having a increased predicive power no only o he China marke, bu also he Hong Kong and US markes in he pos-enering period. 6. Limiaions There are hree main limiaions in his paper.. The exac dae of becoming a WTO member, December, 200, is used as he separaion day for he wo periods of daa. However, his news was announced some days before, he influence of i on he sock markes may already be refleced in he indices. Thus he resuls in he pos-enering period may no be ha much significan for comparison. 2. Abou 3 years daa is used for each period, i may no be sufficien enough o observe he real siuaion. However, i is sill a shor period of ime for China being a member of he WTO. In order o have a comparaive ime lengh for comparison, such ime period was chosen for his paper. 2

24 3. When using he Akaike informaion crierion (AIC) o find ou a suiable lag lengh for he Granger causaliy ess, a maximum 0 lagged erms was used for he rial and error in each case. Though he causaliy relaionship migh be refleced in more han 0 days, i was reasonable o do so as in mos of he cases, informaion could be ransferred and refleced wihin 0 lagged erms. 7. Conclusion and Discussion This paper analyzed possible causal relaionships among he Chinese sock markes, he Hong Kong sock marke and he US sock marke before and afer China became a member of he World Trade Organizaion. Lead-lag relaionships among six sock exchanges reurns were invesigaed. Granger Causaliy ess resuls showed no changes in he lead-lag relaionship beween he Hang Seng Index in he Hong Kong marke and he S&P 500 Composie Index in he US marke. They have a feedback relaionship in regardless of he even ha China became a member of he WTO. Bu o he Chinese sock markes, here were significan influences resuled. A he ime China no ye become a member of he WTO, he A sock markes in China, ha is he Shanghai A and Shenzhen A, mosly led he oher markes jus like he Hang Seng Index and Shenzhen B. This showed ha foreigners could 22

25 hardly ge he informaion accuraely from he China markes han he domesic ciizens could, hus he performance of he A sock markes became one of he beer predicion means o he Chinese markes o he foreign invesors. However, afer China became a member of he WTO, he Granger Causaliy ess resuls alered. The B sock markes in China had a leading role owards mos of he oher sock exchanges, jus like Shanghai B led Hang Seng Index and he wo A sock markes, and Shenzhen B Granger caused S&P 500 Composie Index. These resuls probably suggesed ha he Chinese sock markes were gradually being inegraed ino he global economy as a consequence of becoming a member of he World Trade Organizaion. 23

26 References. A. G. Malliaris and Jorge L. Urruia, The Inernaional Crash of Ocober 987: Causaliy Tess, Journal of Financial and Quaniaive Analysis, Vol. 27, No. 3, Sepember 992, p Bwo-Nung Huang,,Chin-Wei Yang and John Wei-Shan Hu, Causaliy and coinegraion of he sock markes among he Unied Saes, Japan, and he Souh China Growh Triangle, Inernaional Review of Financial Analysis 9 (2000) p Chao, Tien-Yo, Hong Kong and China, Inernaional Financial Law Review: Supplemen London: Mar 993, p. 24 (9 pp.) 4. Cheng F. Lee and Oliver M. Rui, Does Trading Volume Conain Informaion o Predic Sock Reurns? Evidence from China s Sock Markes, Review of Quaniaive Finance and Accouning, Vol. 4, No. 4, June 2000, p Fok Ka Man, The Relaionship beween Unied Saes sock marke and Hong Kong sock marke, Honours Degree Projec, Marin Laurencec, Francis Cai and Sun Qian, Weak-form Efficiency and Causaliy Tess in Chinese Sock Markes, Mulinaional Finance Journal, Vol., No. 4, December 997, p

27 7. Niu Tiehang, Sock marke inegraion in Hong Kong and China, Journal of Conemporary China Princeon, Vol.6, Issue 6, November 997, p Ser-Huang Poon, Han Lin, Taking sock a he new millennium: A sudy of welve sock markes, Managerial Finance Paringon:200, Vol. 27, Issue /2, p Damodar N. Gujarai, Basic Economerics, fourh ediion, McGRAW HILL 0. Bill Hung, Deparmen of Economics, HKBU EVIEWS Applicaions of various economeric models hp:// 25

28 Appendix I Uni Roo Tes for Hang Seng Index Null Hypohesis: HSI has a uni roo Exogenous: Consan Lag Lengh: 0 (Auomaic based on AIC, MAXLAG=23) -Saisic Prob.* Augmened Dickey-Fuller es saisic Tes criical values: % level % level % level *MacKinnon (996) one-sided p-values. Augmened Dickey-Fuller Tes Equaion Dependen Variable: D(HSI) Mehod: Leas Squares Dae: 04/09/05 Time: 3:30 Sample (adjused): Included observaions: 563 afer adjusmens Variable Coefficien Sd. Error -Saisic Prob. HSI(-) C R-squared Mean dependen var.84e-05 Adjused R-squared S.D. dependen var S.E. of regression Akaike info crierion Sum squared resid Schwarz crierion Log likelihood F-saisic Durbin-Wason sa Prob(F-saisic)

29 Uni Roo Tes for Shanghai A Index Null Hypohesis: SHA has a uni roo Exogenous: Consan Lag Lengh: 20 (Auomaic based on AIC, MAXLAG=23) -Saisic Prob.* Augmened Dickey-Fuller es saisic Tes criical values: % level % level % level *MacKinnon (996) one-sided p-values. Augmened Dickey-Fuller Tes Equaion Dependen Variable: D(SHA) Mehod: Leas Squares Dae: 04/09/05 Time: 3:30 Sample (adjused): Included observaions: 543 afer adjusmens Variable Coefficien Sd. Error -Saisic Prob. SHA(-) D(SHA(-)) D(SHA(-2)) D(SHA(-3)) D(SHA(-4)) D(SHA(-5)) D(SHA(-6)) D(SHA(-7)) D(SHA(-8)) D(SHA(-9)) D(SHA(-0)) D(SHA(-)) D(SHA(-2)) D(SHA(-3))

30 D(SHA(-4)) D(SHA(-5)) D(SHA(-6)) D(SHA(-7)) D(SHA(-8)) D(SHA(-9)) D(SHA(-20)) C 6.88E R-squared Mean dependen var 4.82E-06 Adjused R-squared S.D. dependen var S.E. of regression Akaike info crierion Sum squared resid Schwarz crierion Log likelihood F-saisic Durbin-Wason sa Prob(F-saisic)

31 Uni Roo Tes for Shanghai B Index Null Hypohesis: SHB has a uni roo Exogenous: Consan Lag Lengh: 3 (Auomaic based on AIC, MAXLAG=23) -Saisic Prob.* Augmened Dickey-Fuller es saisic Tes criical values: % level % level % level *MacKinnon (996) one-sided p-values. Augmened Dickey-Fuller Tes Equaion Dependen Variable: D(SHB) Mehod: Leas Squares Dae: 04/09/05 Time: 3:29 Sample (adjused): Included observaions: 560 afer adjusmens Variable Coefficien Sd. Error -Saisic Prob. SHB(-) D(SHB(-)) D(SHB(-2)) D(SHB(-3)) C R-squared Mean dependen var -3.58E-05 Adjused R-squared S.D. dependen var S.E. of regression Akaike info crierion Sum squared resid Schwarz crierion Log likelihood F-saisic Durbin-Wason sa Prob(F-saisic)

32 Uni Roo Tes for Shenzhen A Index Null Hypohesis: SZA has a uni roo Exogenous: Consan Lag Lengh: 20 (Auomaic based on AIC, MAXLAG=23) -Saisic Prob.* Augmened Dickey-Fuller es saisic Tes criical values: % level % level % level *MacKinnon (996) one-sided p-values. Augmened Dickey-Fuller Tes Equaion Dependen Variable: D(SZA) Mehod: Leas Squares Dae: 04/09/05 Time: 3:29 Sample (adjused): Included observaions: 543 afer adjusmens Variable Coefficien Sd. Error -Saisic Prob. SZA(-) D(SZA(-)) D(SZA(-2)) D(SZA(-3)) D(SZA(-4)) D(SZA(-5)) D(SZA(-6)) D(SZA(-7)) D(SZA(-8)) D(SZA(-9)) D(SZA(-0)) D(SZA(-)) D(SZA(-2)) D(SZA(-3)) D(SZA(-4)) D(SZA(-5))

33 D(SZA(-6)) D(SZA(-7)) D(SZA(-8)) D(SZA(-9)) D(SZA(-20)) C -5.26E R-squared Mean dependen var -8.97E-07 Adjused R-squared S.D. dependen var S.E. of regression Akaike info crierion Sum squared resid Schwarz crierion Log likelihood F-saisic Durbin-Wason sa Prob(F-saisic)

34 Uni Roo Tes for Shenzhen B Index Null Hypohesis: SZB has a uni roo Exogenous: Consan Lag Lengh: 7 (Auomaic based on AIC, MAXLAG=23) -Saisic Prob.* Augmened Dickey-Fuller es saisic Tes criical values: % level % level % level *MacKinnon (996) one-sided p-values. Augmened Dickey-Fuller Tes Equaion Dependen Variable: D(SZB) Mehod: Leas Squares Dae: 04/09/05 Time: 3:29 Sample (adjused): Included observaions: 556 afer adjusmens Variable Coefficien Sd. Error -Saisic Prob. SZB(-) D(SZB(-)) D(SZB(-2)) D(SZB(-3)) D(SZB(-4)) D(SZB(-5)) D(SZB(-6)) D(SZB(-7)) C R-squared Mean dependen var -3.44E-05 Adjused R-squared S.D. dependen var S.E. of regression Akaike info crierion Sum squared resid Schwarz crierion Log likelihood F-saisic Durbin-Wason sa Prob(F-saisic)

35 Uni Roo Tes for S&P 500 Composie Index Null Hypohesis: S_P has a uni roo Exogenous: Consan Lag Lengh: 0 (Auomaic based on AIC, MAXLAG=23) -Saisic Prob.* Augmened Dickey-Fuller es saisic Tes criical values: % level % level % level *MacKinnon (996) one-sided p-values. Augmened Dickey-Fuller Tes Equaion Dependen Variable: D(S_P) Mehod: Leas Squares Dae: 04/7/05 Time: 00:05 Sample (adjused): Included observaions: 563 afer adjusmens Variable Coefficien Sd. Error -Saisic Prob. S_P(-) C -6.83E R-squared Mean dependen var -2.72E-07 Adjused R-squared S.D. dependen var S.E. of regression Akaike info crierion Sum squared resid Schwarz crierion Log likelihood F-saisic Durbin-Wason sa Prob(F-saisic)

36 Appendix II Akaike informaion crierion (AIC) Resuls for he pre-enering period Y X HSI SHA SHB SZA SZB S&P * * * * * * SHA HSI SHB SZA SZB S&P * * * * * * SHB HSI SHA SZA SZB S&P * * * * * *

37 SZA HSI SHA SHB SZB S&P * * * * * * SZB HSI SHA SHB SZA S&P * * * * * * S&P HSI SHA SHB SZA SZB * * * * * * * indicaes he chosen value for he lagged lengh 35

38 Akaike informaion crierion (AIC) Resuls for he pos-enering period Y X HSI SHA SHB SZA SZB S&P * * * * * * SHA HSI SHB SZA SZB S&P * * * * * * SHB HSI SHA SZA SZB S&P * * * * * *

39 SZA HSI SHA SHB SZB S&P * * * * * * SZB HSI SHA SHB SZA S&P * * * * * * S&P HSI SHA SHB SZA SZB * * * * * * * indicaes he chosen value for he lagged lengh 37