Sources of Investor Sentiment and Price Deviations of Cross-Listed Shares: Evidence from Chinese A- and H-shares

Transcription

1 Sources of Invesor Senimen and Price Deviaions of Cross-Lised Shares: Evidence from Chinese A- and H-shares Qinqin Wu, Robin K. Chou, Ying Hao, and Jing Lu * Absrac We es how differences in he local, global, and oal senimen indices in China and Hong Kong affec he price deviaions of A- and H-shares cross lised on he respecive markes. We also examine he effec of imporan insiuion reforms in China on he price deviaions of A- and H-shares. We find ha invesor senimen, paricularly local senimen, significanly affecs price deviaions. Higher oal and local senimen in China relaive o ha in Hong Kong increases A-share prices relaive o H-share prices. The impac of local senimen differences on price deviaions remains significan and is reversed afer he fourh year, indicaing a significanly longer senimen effec han previously found in he lieraure. Keywords: Asse pricing; Cross-lised shares; Invesor senimen; Price deviaions JEL Classificaion: G12, G14. * Qinqin Wu, Ying Hao, and Jing Lu are a he School of Economics and Business Adminisraion, Chongqing Universiy, China; and Robin K. Chou is a he Deparmen of Finance, College of Commerce, Naional Chengchi Universiy, Taiwan. Address for correspondence: Jing Lu, Deparmen of Finance, School of Economics and Business Adminisraion, Chongqing Universiy, China. Tel: ; Fax: ; lujing@cqu.edu.cn. 1

2 1. Inroducion A growing lieraure indicaes ha invesor senimen has an imporan impac on asse pricing (Baker and Wurgler, 2006, 2007; Baker e al., 2012; Brown and Cliff, 2004, 2005; Firh e al., 2015; Lee e al., 1990; Sein, 1996; Yu and Yuan, 2011). In paricular, Baker and Wurgler (2006) find ha invesor senimen serves as a conrarian predicor of marke reurns and has a greaer impac on socks ha are difficul o value and arbirage. Baker e al. (2012) exend he research by decomposing he oal senimen index ino local and global componens and examine heir influences on sock prices. Based on a validaion es using a sample of hree Siamese win companies cross-lised on he US and UK sock markes, hey find ha he relaive prices of hese sock pairs are affeced by differences in invesor senimem across differen markes. Compared wih he univariae analysis framework in Baker e al. (2012), we exend he research on he effecs of cross-marke invesor senimen on price deviaions of cross-lising shares. Our paper makes he following conribuions o he lieraure. Firsly, afer conrolling for informaion asymmery, demand elasiciy, liquidiy and risk aversion, we find ha differences in invesor senimen across markes are an imporan facor affecing price deviaions of A- and H-shares, which help explains he puzzle of discouns on he foreign cross-lised Chinese equiy shares. Secondly, we find ha insiuional reforms on he Chinese sock markes help miigae he effecs of he cross-marke differences in invesor senimen on price deviaions. Finally, our sudy show ha he cross-marke differences in invesor senimen have a long lasing effec on he A- and H-share price deviaions. We choose Chinese A- and H-share cross-lised companies as our sample. 1 A-shares of 1 B-shares of Chinese companies are lised on eiher he Shanghai Sock Exchange (SHSE) or he Shenzhen Sock Exchange (SZSE), priced in US dollars (on he SHSE) or Hong Kong dollars (on he SZSE). Before 2001, B-shares are raded by non-chinese ciizens. We do no include B-shares in our ess because domesic invesors of China have been permied o rade B-shares since February 2001, which led o a non-sric segmenaion 2

3 Chinese companies are lised on eiher he Shanghai Sock Exchange (SHSE) or he Shenzhen Sock Exchange (SZSE), priced in Renminbi (RMB), and are available o Chinese ciizens and domesic insiuions (including qualified foreign insiuional invesors). H-shares of he same Chinese companies are lised on he Hong Kong Sock Exchange (HKSE), priced in Hong Kong dollars (HKD), and raded by invesors from all over he world excep hose from China. 2 The underlying asses of cross-lised shares have he same cash flow sreams, and, herefore, heir prices should heoreically saisfy he law of one price. Neverheless, he lieraure shows ha he foreign share prices of cross-lised companies are higher han he domesic share prices in almos all counries excep China (Gagnon and Karolyi, 2010, 2011; Gu and Reed, 2013). For example, analyzing daa of 506 cross-lised shares, Gagnon and Karolyi (2010) find ha foreign share prices are higher on average han domeic share prices by approximaely 38%. However, he prices of Hong Kong cross-lised H-shares are consisenly significanly lower han hose of domesically lised A-shares in China. For example, for a sample of 86 cross-lised companies, he he average closing price of H-shares on June 19, 2015 is lower han ha of A shares by approximaely 47.75%. The price anomaly of Chinese cross-lised shares is glaring, causing he lieraure o refer o he maer as a puzzle in he Chinese sock marke (Fernald and Rogers, 2002). Facors proposed o explain his phenomenon based on he raional financial heory include informaion asymmery (Bailey, 1994; Chakravary e al., 1998; Chan e al, 2008), differenial demand elasiciy (Sulz and Wasserfallen, 1995), differenial liquidiy (Amihud and Mendelson, 1986; Bailey, 1994; Longsaff, 1995), and differenial risk aversion (Ma, 1996). Froo and Dabora (1999) es six hypoheses for he price deviaions of cross-lised beween A- and B-shares (Chan e al., 2008). 2 Saring on November 17, 2014, hrough he Shanghai-Hong Kong Sock Connec, invesors in China and Hong Kong can rade and sele eligible shares on he oher marke via he exchange and clearinghouse in heir local marke. Our enire sample period precedes he implemenaion of his sock connec, which ensures a sricer segmenaion beween hese wo markes. 3

4 shares, including he discreionary use of dividends by paren companies, differences in paren expendiures, voing righs, currency flucuaions, ex-dividend dae iming, and ax-induced invesor heerogeneiy. They find ha only he las hypohesis can parially explain he price deviaions. We aemp o explain he price deviaions of A- and H-shares by he cross-marke invesor senimen. In addiion o he daa from six counries as used in Baker e al. (2012), we include daa from he China and Hong Kong sock markes o consruc he oal, local and global senimen indices. Specifically, following Baker e al. (2012), we consruc oal senimen indices for eigh sock markes including China, Hong Kong, Canada, France, Germany, Japan, he Unied Kingdom, and he Unied Saes. We use volailiy premium (PVOL), urnover (TURN), oal number of iniial public offerings (IPOs; NIPO), and he firs-day reurns of IPOs (RIPO) as invesor senimen proxy variables o consruc eigh oal senimen indices for hese counries or regions. We hen consruc a global senimen index wih hese eigh oal senimen indices and hen decompose he oal senimen indices of China and Hong Kong ino wo local senimen indices. We use 62 cross-lised companies wih A- and H-shares o invesigae he influence of he oal, local, and global senimen differences on he price deviaions of A- and H-shares. We find ha he oal invesor senimen is a significan facor influencing he price deviaions of A- and H-shares, wih he effecs mainly driven by local invesor senimen, a finding ha has no o dae been comprehensively examined. Meanwhile, we find ha insiuional reforms of rading regulaions end o ease he effecs of invesor senimen on sock prices. Alhough he Chinese sock marke has undergone reform and developmen for more han 20 years, i remains an emerging marke wih peculiar insiuional feaures, such as individual invesor-dominaed rading, excessive governmen inervenions, and a consan revolving supervision sysem. I is difficul o sudy all insiuional reforms on he China sock marke, so ha we focus on he reforms ha are 4

5 likely o simulaneously affec inernaional, as well as domesic invesors, such as exchange rae sysem reform, spli share srucure reform, and Qualified Foreign Insiuional Invesors (QFII) and Qualified Domesic Insiuional Invesors (QDII) schemes. They play a crucial role in decreasing fracion and giving more freedom o he capial marke. Li (2012) shows ha hese reforms significanly srenghen he link beween China s financial markes and oher markes around he world. We hus furher invesigae wheher he insiuional reforms in he Chinese capial marke affec he price deviaions of A-and H-shares. We find ha he exchange rae sysem reform is helpful in decreasing A- and H-share price deviaions whereas he spli share srucure reform increases price deviaions. Before he insiuional reforms, he oal senimen differences beween China and Hong Kong posiively affec he price deviaions of A- and H-shares. We find ha local senimen differences beween China and Hong Kong are mainly responsible for he effec, while global senimen differences have lile effec on he A- and H-share price deviaions. Wih he exchange rae sysem reforms, he spli srucure reform, and he QFII and QDII schemes, he marginal effecs of he oal and local senimen differences on he price deviaions of cross-lised shares significanly decline, indicaing ha insiuional reforms have a moderaing effec on invesor senimen. We also sudy how firm characerisics affec he relaions of senimen differences and price deviaions of A- and H-shares. Following Chang e al. (2012) s mehod, we include he ineracion erms of senimen differences and a dummy variable ha equals one if he firm is more influenced by invesor senimen. We find ha oal senimen differences have a larger effec on price deviaions for small firms and firms wihou dividends. Nex, we invesigae he lasing effecs of senimen differences on price deviaions of A- and H-shares. We regress he price deviaions on up o six lagged years of senimen differences. The resuls indicae ha he local senimen differences beween China and Hong Kong have a long-lasing effec on he price deviaions of A- and H-share, which is reversed 5

6 in he fourh year. This effec is much longer han ha previously found in he lieraure. The impac of he lagged oal senimen and lagged global senimen differences is no significan. We hink ha he long-lasing effec of local senimen is caused by he individual-rader dominaed A-share invesor srucure. China s local marke senimen is dominaed by noise invesors, which ends o drive prices furher away from fundamenals. Thus, due o limis o arbirage, raional invesors may be unable or relucan o arbirage he price deviaions away due o significan risk and ransacion coss (Barberise al., 1998; De Longe al., 1990; Shleifer and Vishny, 1997). Thus, local senimen differences beween China and Hong Kong have a long-lasing effec on price deviaions of A- and H-shares. On he oher hand, lagged global senimen does no have a significan effec on he price deviaions of A and H shares possible due o marke segmenaions and capial resricions. We hus find ha he senimen of an individual invesor-dominaed marke has a long-lasing effec on sock prices likely due o more noise rading on he marke (Baker e al., 2012; Chuang and Susmel, 2011; Schmeling, 2007). Finally, as a robusness check, we invesigae he senimen differences on price deviaions of A- and H-shares by indusry. Based on he indusry classificaions of he China Securiies Regulaory Commission, we classify our sample ino seven indusries. We find ha he effecs of senimen differences on price deviaions are, in general, consisen among indusries. We also rerun our ess by excluding finance indusry, cuing our sample ino wo subperiods and using P/E raio as oal senimen proxy, respecively, and find similar resuls. To he bes of our knowledge, his paper is he firs examining he effecs of cross-marke invesor senimen on he discoun of foreign cross-lised Chinese equiy shares. Sudies ha are relaed o ours include Grossmann e al. (2007), Arquee e al. (2008) and Baker e al. (2012). Grossmann e al. (2007) sudy price deviaions of American Deposiary Receips (ADRs) and find ha no only ADRs bu also he underlying securiies of ADRs are 6

7 more affeced by ha American consumer index, a proxy for invesor senimen, han by local senimen. However, Grossmann e al. (2007) do no decompose heir senimen index ino differen componens of local and global invesor senimen. Arquee e al. (2008) alernaively measure he marke senimen by he raio of he Shanghai A-Share Index price/earnings (P/E) raio o he Hang Seng China Enerprises Index P/E raio (or he S&P500 P/E raio) and measure company senimen by he company P/E raio o he Shanghai A-Share Index P/E raio. They find ha boh relaive marke senimen and company senimen can significanly accoun for he US ADR and H-share discouns of SHSE-lised firms. However, P/E raios likely capure effecs oher han invesor senimen. P/E raio as a proxy for invesor senimen is no as widely used in he lieraure, compared o he invesor senimen index by Wurgler e al. (2006) and Wurgler e al. (2012). Baker e al. (2012) decompose he oal senimen index ino global and local componens and find ha boh he oal and local senimen differences beween lising counries significanly affec price deviaions. The remainder of his paper is organizedas follows. Secion 2 describes China s capial marke and insiuional background. Secions 3 and 4 presen he research mehodology and empirical resuls, respecively. Finally, Secion 5 concludes. 2. China s capial marke and insiuional background For almos 40 years, China s capial marke experienced significan changes hrough he process of ransforming from a cenrally planned o a marke-oriened economy. Because he esablishmen of he SHSE and he SZSE is enirely planned by he governmen, he markes have been heavily influenced by he cenral governmen since heir incepion, which is very differen from he marke-led esablishmen of capial markes in mos of Wesern counries. Early IPOs in hese wo exchanges primarily served he capial needs of sae-owned 7

8 enerprises. Over ime, he exchanges became increasingly marke-oriened as privae sake holders acquired sronger bargaining power wihin he developing economy. Privae enerprises demanded equal righs and he same opporuniies of capial raising enjoyed by sae-owned enerprises, and invesors demanded legal proecions of heir invesmen righs and ineress. In addiion, inernaional invesors asked for access o he Chinese capial markes o paricipae in is rapid economic growh and o diversify heir invesmen risk. As a resul, he SHSE and he SZSE, esablished during he early 1990s, grew rapidly. According o World Federaion of Exchanges saisics, a he end of May 2015 he SHSE and he SZSE ranked hird and fifh in he world, respecively, in marke capializaion. Several imporan marke reforms, such as he exchange rae sysem reform, he spli share srucure reform, and he QFII and QDII schemes, have been insiued o improve China s marke infrasrucure and likely affec he price deviaions of A-and H-shares. Individual invesors consisenly dominae he A-share marke, which causes excessive marke volailiy and reduces marke efficiency (Allen and Shen, 2013). Foreign invesors were iniially resriced from invesing on he A-share markes. Afer joining he World Trade Organizaion in November 2001, China acceleraed he process of opening up and liberalizing is capial markes. In Augus 2003, he Chinese governmen approved he firs QFII scheme, which allowed qualified foreign insiuions o inves in he A-share and governmen bond markes. The objecive of he QFII scheme is o improve marke efficiency and corporae governance of lised companies. As of Augus 2014, 254 foreign insiuions were approved, and he oal amoun of invesmen was US$59.67 billion. Neverheless, Chinese domesic invesors were iniially confined o he A-share markes, resuling in heir inabiliy o diversify inernaionally. However, in December 2004, in an effor o make a beer use of he large amoun of foreign exchange reserves and o increase capial marke ransparency (Pioroski and Wong, 2013), China launched is QDII scheme, which allowed qualified 8

9 domesic insiuions o inves in he foreign capial markes. From Sepember 1997 o June 2005, China insiued a fixed exchange rae sysem o mainain capial inflows and rade surplus and accumulaed a large amoun of foreign exchange reserves. The RMB was ighly pegged o he US dollar (USD) a he level of 1 USD o 8.28 RMB during ha ime period. Domesic and foreign invesors widely believed ha he RMB was undervalued, resuling in srong revaluaion expecaions. Evenually, in July 2005, he Chinese governmen implemened a managed floaing exchange rae sysem wih respec o a baske of reference currencies, parially responding o marke demand and supply. From July 2005 o December 2012, he RMB appreciaed by more han 28% agains he USD. The exchange rae sysem reforms likely made he domesic capial markes more responsive o changes in inernaional economic condiions and sock marke flucuaions. In he early 1990s, o raise funds from he sock marke and a he same ime reain majoriy ownership, when iniially lised sae-owned enerprises issued addiional radable shares o he general public and reained he original shares, which were non-radable on he marke. The reained non-radable shares resuled in he coexisence of radable and non-radable shares; such a shareholder srucure is ofen referred o as a spli share srucure. As he capial marke develops, he coexisence of non-radable and radable shares becomes problemaic. Firs, under he spli share srucure, no all ypes of legal shareholders are allowed o rade, so he pricing mechanism of capial marke is disored, which induces inefficien capial allocaions. Second, because a significan amoun of non-radable shares is owned by large shareholders and managers, he capial marke is no able o provide effecive incenives or disciplinary measures, which undermines he role of markes as a corporae governance mechanism. Third, radable shares are publicly raded, while non-radable shares are ransferred privaely by agreed-upon prices beween ransacing paries. The markes for he equiy of he same firms are effecively segmened, which causes grea difficulies in firm 9

10 valuaions and hampers he fundamenal marke pricing funcions. Finally, because muliple spo prices are quoed for radable and non-radable shares, pricing equiy derivaives is difficul and can significanly impede derivaive marke developmen (China Securiies Regulaory Commission, 2012). In 2005, he Chinese governmen formally launched he spli share srucure reform, which allows non-radable shares o be raded on he sock markes by requiring shareholders of non-radable shares o pay shareholders of radable shares a premium. Li e al. (2011) describe his reform. Chen e al. (2012) argue ha he reform plays an imporan role in removing a major marke fricion in China. Due o decreased marke fricions, we expec ha he spli share srucure reform has a significan impac on he price deviaions of A- and H-shares. Tha is, an increased supply of radable shares should depress A-share prices. However, if rading ineress and demand in A-shares increases significanly afer he well-publicized spli share srucure reform, A-share prices may insead increase. Thus, how he price deviaions of A- and H-shares change afer he reform is an empirical issue. 3. Research mehod Since he lising of Tsingao Brewery s H-shares on July 15, 1993, 86 companies wih H-shares have cross lised on he HKSE. Our iniial sample includes cross-lised A- and H-shares daa from Augus 1993 o December We delee companies wih less han 30 monhly observaions, resuling in a final sample of 62 companies. Table 1 shows he sample selecion process, and Appendix A1 provides he deailed lising daes and share codes of he sample firms. [INSERT TABLE 1 ABOUT HERE] 10

11 3.1 Dependen variable The average daily price deviaions of A- and H-shares in a monh, denoed as DPRICE, is im, calculaed as DPRICE price (1) D A m 1 i, im, ln( ), H Dm 1 pricei, * R ex A H where price i, and price i, are he closing price of A- and H-shares, respecively, for firm i on day. R ex is he exchange rae of Hong Kong dollar o RMB. D m is he number of rading days in monh m. Alhough China and Hong Kong are locaed in he same ime zone, he rading hours of A- and H-shares are no compleely synchronized. 3 We use average monhly daa o preven spurious price deviaions caused by nonsynchronous rading (Gagnon and Karolyi, 2010; Kadlec and Paerson, 1999). If DPRICE is zero, hen he prices of A- and im, H-shares saisfy he law of one price. A posiive (negaive) DPRICE indicaes ha A-shares im, are rading a a premium (discoun) relaive o H-shares. Figure 1 shows ha A-share prices are a a significan premium relaive o H-share prices for almos he enire sample period. The equally weighed average price deviaion of A- and H-shares is 90.33%. The sandard deviaion is 81.41%, and he maximum and minimum price deviaions are % and 18.01%, respecively. The price deviaions of A- and H-shares are large iniially and hen become smaller wih ime. Price deviaions rise seadily from 1994 o 1998, say a a relaive high level from 1998 o 2001, and hen begin o fall gradually saring in [INSERT FIGURE 1 ABOUT HERE] 3.2 Invesor senimen indices Invesor senimen proxy variables 3 The rading ime period of A-shares is 9:30-11:30 and 13:00-15:00, while ha of H-shares is 9:30-12:00 and 13:00-16:00. 11

12 We follow Baker e al. (2012) o consruc senimen indices. We use volailiy premium, urnover, number of IPOs and he average firs-day reurns of IPOs from 1993 o 2012 as senimen proxies and consruc senimen indices for China, Hong Kong, Canada, France, Germany, Japan, he Unied Kingdom and he Unied Saes. The volailiy premium is he raio of he value-weigh average marke-book raio of high volailiy socks o ha of low volailiy socks. 4 The number of IPOs is in log forms and he firs-day IPO reurns are he log of he equally weigh average iniial IPO reurns. The urnover is he log of he oal yearly dollar rading volume divided by he average oal capializaion a he beginning and end of he year. We derend i wih hree-year moving average. To conrol for informaion relaed o macroeconomic expecaion raher han o senimen, we orhogonalize each proxy wih respec o six macroeconomic variables: consumpion growh, indusrial producion growh, inflaion, employmen growh, he shor-erm rae, and he erm premium. We use he GDP growh in Hong Kong o subsiue for indusrial producion growh because he oupu of he service secor in Hong Kong is as high as 85% of is GDP. 5 The GDP growh adequaely reflecs he economic condiions of Hong Kong. 6 Table 2 liss he senimen proxies and macroeconomic variables. Table 3 repors he summary saisics of senimen proxies. The iniial reurns of IPOs in China are significanly higher han hose in he oher counries as seen in Table 3, which reflecs he overheaed IPO markes on he A-share marke, possibly due o high invesor senimen during he sample period. [INSERT TABLE 2 AND 3 ABOUT HERE] 4 Appendix A2 gives he sample size of calculaing volailiy. 5 The raio is calculaed wih daa obained from he Chinese Naional Saisics Bureau since In addiion, we orhogonalize he senimen proxies of Hong Kong wihou he shor-erm rae and he erm premium because some of he daa are missing from 1993 o

13 3.2.2 Toal senimen indices The oal senimen indices for each marke are measured by he firs principal componen of he macro-orhogonalized senimen proxies. The resuling indices are linear funcions of he wihin-marke sandardized values of he proxies and hus have a mean of zero. The fracions of variance explained by he firs principal componens for he eigh sample markes are beween 34% and 49%, similar o hose in Baker e al. (2012). 7 The linear funcions of he eigh oal senimen indices are Toal SENTChina, TURN PVOL NIPO RIPO, (2) Toal SENTHK, TURN PVOL NIPO RIPO, (3) Toal SENTCanada, TURN PVOL NIPO RIPO, (4) Toal SENTFrance, TURN PVOL NIPO RIPO, (5) Toal SENTGermany, TURN PVOL NIPO RIPO, (6) Toal SENTJapan, TURN PVOL NIPO RIPO, (7) Toal SENTUK, TURN PVOL NIPO RIPO, (8) Toal SENTUS, TURN PVOL NIPO RIPO, (9) where Toal SENT is he oal senimen index for counry or region c, and c, TURN, PVOL, NIPO and RIPO are macro-orhogonalized urnover raios, volailiy premium, number of IPOs and he average firs-day reurns of IPOs, respecively. Figure 2 plos he oal senimen indices. We measure he senimen differences beween China and Hong Kong as 7 In Baker e al. (2012), he fracion of variance explained by he firs principal componens of oal senimen indices for Germany, France, he Unied Kingdom, Japan, Canada and he Unied Saes are 48%, 40%, 37%, 37%, 38% and 42%, respecively. The fracion of variance explained by he firs principal componens of senimen indices for he Unied Saes is 49% in Baker and Wurgler (2006). 13

14 he oal senimen index of China minus ha of Hong Kong, Toal Toal Toal SENTA H, SENTChina, SENTHK,. Figure 2 indicaes ha he oal senimen indices of China, Canada, France, and Japan reflec high invesor senimen before he global financial crisis of 2008, whereas he oal senimen indices of France, he Unied Kingdom, and he Unied Saes show paricularly low invesor senimen afer he Inerne bubble in he lae 1990s. The oal senimen index of China aligns wih he bull marke of A-shares from 1998 o 2000, he five-year bear marke from 2001 o 2005, and he rise and fall of he sock markes from 2006 o 2008 due o subprime morgages. The oal senimen index of Hong Kong also reflecs invesor senimen surrounding he Asian financial crisis in 1997 and he Inerne bubble in he lae 1990s. [INSERT FIGURE 2 ABOUT HERE] Global and local invesor senimen indices We nex consruc he global senimen index by forming he firs principal componen of he eigh oal senimen indices as SENT SENT SENT SENT Global Toal Toal Toal China, HK, Canada, SENT SENT SENT Toal Toal Toal France, German, Japan, Toal SENT SENT Toal UK, US,, (10) where Global SENT is he global senimen index. The firs principal componen of he eigh oal senimen indices accouns for 27.75% of he oal variance. We hen regress he oal Toal Global senimen index of each region on he global senimen index, SENTc, bcsent c,. Finally, he residual, c,, is he local senimen index, SENT, Local c, which shows he exen of 14

15 he local senimen componen in he sandardized oal senimen. The coefficien, b c, is he sensiiviy of global senimen in marke c. To measure he differences in local and global invesor senimen beween he A- and H-share markes, we ake he differences beween he local (global) indices of China and Hong Kong, SENT, SENT, SENT, Local Local Local A H China HK Global Global ( SENTA H, ( bchina bhk )SENT ). If b China is bigger (smaller) han b HK, hen he global senimen causes more (less) price increases in China marke relaive o hose in he Hong Kong marke. Thus, he global senimen difference is posiively (negaively) relaed o price deviaions of A- and H-shares if bchina bhk is posiive (negaive). 8 Table 4 shows he summary saisics of he oal, local and senimen indices and he correlaion coefficiens beween he global senimen index and he eigh oal senimen indices. Figure 3 plos he global senimen index and local indices of China and Hong Kong. As expeced, he paern of he global senimen index is in line wih he Inerne bubble surrounding 2000 and he global financial crisis surrounding The Chinese local senimen index suggess ha he low senimen afer 2008 is likely aribuable o exogenous shocks coming from he global financial crisis, whereas he senimen flucuaions before ha period are more likely caused by local invesors. The Hong Kong local senimen index reflecs he Inerne bubble surrounding 2000 and he global financial crisis surrounding [INSERT FIGURE 3 ABOUT HERE] [INSERT TABLE 4 ABOUT HERE] 3.3 Conrol variables 8 We find ha b China b HK is , so we expec he global senimen difference o be negaively relaed o price deviaions of A- and H-shares. 15

16 Prior lieraure shows ha facors ha affec price deviaions of cross-lised socks include informaion asymmery, demand elasiciy, liquidiy and risk aversion (Amihud, 2002; Chakravary e al., 1998; Domowiz e al., 1997; Wang and Jiang, 2004). We use firm size, demand elasiciy, illiquidiy, and risk as proxy variables o conrol for hese facors. Firm size is used o proxy for informaion asymmery, as suggesed by Chakravary e al. (1998).Invesors in China are likely o have an informaion advanage over Chinese socks compared o invesors in Hong Kong. The informaion asymmery problem is likely more significan for small firms and causes price deviaions for small firms o become larger. Firm size is measured by is oal marke value as SIZE im, D 1 m ln( MVi,), (11) D m 1 where SIZE im, is he oal marke value for firm i in monh m, MV i, is oal marke value for firm i on day. We expec company size o be negaively relaed o price deviaions. The demand elasiciy hypohesis suggess ha foreign invesors have more invesmen channels han domesic invesors (Domowize al., 1997). Thus, heir demand elasiciy is lower, which leads o a discoun of H-shares. The relaive demand is measured by he raio of he oal H-share capial o he sum of he A- and H-share capial as SHARE 1 share, (12) Dm H i. im, H A Dm 1 sharei, sharei, A H where share i, is he ousanding share of A-share on day, and share i, is he ousanding share of H-share on day. A large raio indicaes low demand elasiciy for Hong Kong invesors; we expec demand elasiciy o be posiively relaed o A- and H-share price deviaions. The liquidiy hypohesis indicaes ha he rading coss of low liquidiy socks are high, which induces a liquidiy premium in reurns and lowers sock prices for illiquid socks. We 16

17 use he Amihud illiquidiy measure (AMI), calculaed as he dollar rading volume divided by he absolue reurns (Amihud, 2002), o gauge he liquidiy of A-and H-shares. We hen calculae he relaive A- and H-share illiquidiy as he raio of A- and H-share AMI measures as AMI A im, D A m 1 volumei,, (13) D reurn A m 1 i, AMI H im, D H m 1 volumei,, (14) D reurn H m 1 i, AMI AMI AMI (15) A H im, im, im,, where volume i, is he dollar rading volume of sock i on day, reurn i, is he reurn of A H sock i on day, and AMI i, and AMI i, are he illiquidiy measures in monh m for sock i s A-and H-shares and sandardized, respecively. AMI im, is he relaive illiquidiy for sock i s A- and H-shares in monh m. We expec ha he relaive A- o H-share illiquidiy is negaively relaed o price deviaions, because a high relaive H-share illiquidiy will cause a larger discoun o H-shares. The risk of A-shares can be higher han ha of H-shares because he Chinese sock markes are sill emerging and are no as well developed as he Hong Kong sock markes. As a resul, he invesors of A-shares may require a higher risk premium, which can lower he price of A-shares and he premium of A-share prices relaive o H-share prices. We use reurn variance during a monh as a proxy for risk (Wang and Jiang, 2004), calculaed as SD, (16) 2 A, m im, 2 H, m Where and 2, are he variance of sock i s daily reurns of A- and H-shares in 2 A, m H m monh m, respecively, and SD im, is he relaive risk measure. 17

18 3.4 Insiuional variables The effec of he exchange rae sysem reform is capured by a dummy variable, which is equal o 1 afer July 2005, and zero oherwise. Similarly, he effec of he spli share srucure reform is measured by a ime dummy, which is equal o 1 afer he monh in which he non-radable shares of he firm are lised for public rading, and zero oherwise. We use he oal accumulaive amoun of he QFII and QDII schemes approved by he Chinese Adminisraion of Foreign Exchange as he proxy variable for he effec of allowing parial capial flows beween China and foreign markes. The firs QFII scheme was approved in Augus2003. QDIIs were allowed o inves in overseas sock markes in May, 2007, so we sar o include he amoun of QDII schemes from hen on. 3.5 Model Our basic regression model is DPRICE DPRICE SENT SIZE SHARE i, 0 1 i, 1 2 A H, 3 i, 4 i, AMI SD D D ( QFII QDII ). 5 i, 6 i, 7 ex 8 reform 9 i, (17) where DPRICE is he price deviaions of A-and H-shares, and i, SENT A is he differences H, in senimen indices beween China and Hong Kong (i.e., oal, local, or global senimen differences). SIZE, i, SHARE, i, AMI, and SD i, i, are proxy variables of informaion asymmery, relaive demand elasiciy, relaive illiquidiy and relaive risk, respecively. D ex, D, and QFII QDII are he exchange rae sysem reform dummy, he spli share reform srucure reform dummy, and he cumulaive amoun of he QFII and QDII schemes, respecively. Wih various insiuional changes, he sock markes in China become more liberalized and are less cosly o rade, so we expec he effecs of senimen on he price deviaions of 18

19 A-and H-shares o decrease. To verify his expecaion, in alernaive ess we include in our model he ineracion erms of senimen differences and he insiuional reform variables. Finally, considering ha he price deviaions are usually persisen (Baker e al., 2012; Sun and Tong, 2000), we include he lag erms. We use he cross-secional fixed effecs mehod and he ordinary leas square mehod (OLS), wih robus sandard errors clusered by firm, o obain robus resuls. 4. Empirical resuls 4.1 Senimen differences and he price deviaions of A-and H-shares Table 5 shows he regression resuls of he A- and H-share price deviaions on he oal senimen differences, and Table 6 shows hose of he price deviaions on local and global senimen differences. To check he robusness of our resuls, we run regressions using annual daa; Tables 7 and 8 show he resuls. Models 1 o 7 are esimaed by he cross-secional fixed effecs mehod while Model 8 is esimaed by he OLS mehod. The coefficiens of he lagged price deviaions are significanly posiive a he 1% level, showing ha he price deviaions are quie persisen. [INSERT TABLE 5 ABOUT HERE] [INSERT TABLE 6 ABOUT HERE] [INSERT TABLE 7 ABOUT HERE] [INSERT TABLE 8 ABOUT HERE] From he resuls in Tables 5 and 6, wih or wihou conrol and insiuional variables, oal and local senimen differences have a significan posiive effec on price deviaions, whereas global senimen differences have lile effec. These resuls indicae ha he higher 19

20 he oal and local invesor senimen of A-shares is compared o ha of H-shares, he larger he price deviaions are. Tables 7 and 8, which shows annual daa, provides similar resuls. Model 7 in Table 7 shows ha he coefficien on oal senimen difference is 6.30%. Given ha he sandard deviaion of oal senimen difference is , one sandard deviaion movemen in oal senimen difference changes he price deviaions by 7.25% (6.30% ), which is abou 9.21% of one sandard deviaion (78.73%) of he price deviaions. Model 7 of Table 8 decomposes oal senimen difference ino local and global senimen differences. Local senimen differences have a significan impac on price deviaions, bu global senimen differences do no. A one sandard deviaion (0.9603) increase in local senimen difference increases he price deviaions by 15.63% based on a local senimen coefficien of 16.28%, which is abou 19.86% of one sandard deviaion of he price deviaions (78.73%). Overall, we find ha he price deviaions of A-and H-shares are mainly influenced by local senimen. We conjecure ha he significan impac of local senimen on price deviaions may be due o he insiuional invesmen resricions in China. Invesors in China are no allowed o rade H-shares, and diversifying ino global markes is difficul due o various capial flow consrains. 9 Thus, global senimen affecs he Chinese markes very lile. Furher, Hong Kong and global invesors are no allowed o rade A-shares in our sample period. These facors diminish he impac of global senimen on A-share prices. We also es he effec of insiuional reforms. The coefficien on he exchange rae 9 According o Chinese laws and regulaions, including Regulaions on he Foreign Exchange Sysem of he People s Republic of China, and Adminisraion of he Selemen, Sale and Paymen of Foreign Exchange Provisions, he direc foreign invesmen of domesic invesors mus be approved by he Sae Deparmen. In addiion o financial insiuion and business enerprises approving foreign borrowing, individual invesors are no allowed o rade foreign securiies. Only financial insiuions approved by he People Bank of China can buy foreign bonds. 20

21 sysem reform is negaive, indicaing ha exchange rae reform decreases A- and H-share price deviaions, possibly due o less cross-border capial movemen consrains and less sringen foreign exchange conrols on he RMB afer he reform. Furhermore, he dividends of H-share are denominaed in RMB bu paid in HKD, so he invesors of H-shares receive he benefis of expeced RMB appreciaion. Thus hey are likely o inves more in H-shares, which decreases he price deviaions of A-and H-shares. This resul is consisen wih Arquee e al. (2008) ha he expecaion of RMB revaluaion reduces price deviaions. The dummy variable of he spli share srucure reform is posiive, which shows ha he implemenaion of his reform has a posiive impac on price deviaions. The spli share srucure reform allows non-radable shares o become radable. On he one hand, if he reform causes he demand curve o slope downward, holding he level of demand consan, he prices of A-shares would end o decrease. On he oher hand, he reform may insead generae more ineress in rading A-shares and hus increase he demand level and prices of A-shares. Thus, how demand in A-shares reacs o spli share srucure reform is an empirical issue. Our resuls suppor he laer argumen of an increased demand in A-shares following he reform. A he end of Ocober 2006, companies ha compleed he spli share srucure reform comprise more han 94% of he oal marke capializaion. In unrepored resuls, we es he daily urnover raios on he SHSE from 1991 o 2013 by he Chow es and find a significan srucure change in Namely, he urnover raios increase significanly, which coincides wih he compleion of he spli share srucure reform. Invesors rade more acively afer he reform, which increases he relaive prices of A-shares. For he effecs of he QFII and QDII schemes, we do no obain consisen resuls wih monhly and annual daa. To sudy he senimen effec changes due o insiuional reforms, we analyze he ineracion erms beween senimen differences and he insiuional reform variables. Tables 5, 6, 7, and 8 show ha he marginal effecs of senimen differences on 21

22 price deviaions decrease significanly afer he insiuional reforms. Because sandard errors change by including he ineracion erms, o gauge he marginal effecs of senimen differences, we have o recalculae he marginal effecs of senimen differences and heir sandard errors according o Models 4, 5 and 6 in Tables 5 and Panel A of Table 9 provides he resuls using monhly daa. We also recalculae he marginal effecs of senimen differences according o Models 4, 5 and 6 in Tables 7 and 8. Panel B shows he resuls using annual daa. [INSERT TABLE 9 ABOUT HERE] Considering ha he marginal effecs and sandard errors of he coefficien of he QFII and QDII schemes vary wih heir levels, we use he average level o gauge he effec of he QFII and QDII schemes. Table 9 shows ha he effecs of oal and local senimen differences on price deviaions decrease significanly due o all hree insiuional reforms. Bu he effecs of global senimen differences become srong afer he reforms. Panel B in Table 9 shows ha he marginal effec of global senimen differences are , and , respecively, afer he exchange rae sysem reform, spli share srucure reform and afer he cumulaive amoun of he QFII and QDII schemes is greaer han is sample period mean (5.1687). A one-sandard deviaion movemen in global senimen differences are associaed wih a change in he price deviaions of -4.24% (-7.98% ), -5.02% (-9.46% ) and 3.58% ( 6.74% ), respecively, 10 According o Brambor e al. (2006), aking oal senimen deviaion and exchange rae sysem reform for insance, prior o he reform he marginal effec of oal senimen deviaion is and is sandard error is in 2 parenheses. Afer he reform, he marginal effec of oal senimen deviaion is 2 exd and is sandard ex error is 2 var( 2) var( ex )Dex 2Dex cov( 2, ex), where is he coefficien of he ineracion iem of he oal ex senimen deviaions and he exchange rae sysem reform. 22

23 which are equivalen o 5.38%, 6.38%, and 4.55% of a sandard deviaion in price deviaions (78.73%). Overall, we find ha he exchange rae sysem reform, he spli share srucure reform, and he QDII and QFII schemes are helpful in alleviaing marke fricions in China and help decrease A- and H-share price deviaions. More imporanly, price deviaions beween A- and H-shares are significanly influenced by he global senimen differences, alhough is economic significance is relaively small. Among he conrol variables, he negaive coefficien on firm size (SIZE) shows ha he informaion asymmery beween foreign and domesic invesors ges worse for smaller firms and hus induces higher price deviaions beween A-and H-shares. The coefficien of H-shares relaive marke capializaion (SHARE) is posiive. More H-shares implies a low demand elasiciy of Hong Kong invesors, and demand elasiciy is expeced o be posiively relaed o he A- and H-share price deviaions. The coefficien on he relaive A- and H-share illiquidiy raio (AMI) is negaive, which indicaes ha a high relaive H-share illiquidiy will cause a larger discoun o H-shares, consisen wih our expecaion. Finally, he relaive risk of A-and H-shares, proxied by he relaive A- and H-share volailiy, is no significanly relaed o price deviaions. Baker and Wurgler (2006) sudy how invesor senimen affecs cross-secional sock reurns and show ha invesor senimen has a larger effec on socks whose valuaions are more subjecive and hard o arbirage. Therefore, we explore how firm characerisics impac he relaions beween invesor senimen differences and price deviaions of A- and H-shares in annual daa. We define a dummy variable, D fc, which equals one if firms are harder o value and hence are more likely o be affeced by senimen, and zero oherwise. Firm characerisics include firm size (ASSET), asse angibiliy (TANGIBILITY), firm profiabiliy (ROA), insiuional ownership (INSTP) and dividend paymen (DIVIP). ASSET is he oal asses a he end of each year. TANGIBILITY is gross propery, plan, and equipmen, divided 23

24 by lagged oal asses. ROA is he average reurn on oal asses for he pas hree years. INSTP is he insiuional ownership a he end of year. DIVIP is a dummy variable, which equals zero if he firm does no pay dividends, and one oherwise. All he daa are from Wind daabase. We divide he firm sample ino four groups according o each firm characerisic each year. D fc equals one if ASSET, TANGIBILITY, ROA or INSTP is in he lowes quinile, and zero oherwise. If DIVIP equals zero (one), D fc equals one (zero). Following Chang e al. (2012), we include he ineracion erms of D fc and invesor senimen differences. The resuls are repored in Table 10. The coefficiens of he ineracion erms of D fc and invesor senimen difference are posiive and significan for ASSET and DIVIP, which indicaes ha oal senimen difference has a larger impac on he price deviaions of A-and H-shares for small firms and firms ha do no pay dividends. [INSERT TABLE 10 ABOUT HERE] 4.2 Persisence of he effec of invesor senimen on price deviaions Firh e al. (2015), sudying he relaion beween corporae ransparency and invesor senimen on he Chinese sock markes, find ha he impac of senimen on marke excess reurns is reversed afer he welfh monh. Ben-Rephael e al. (2012) find ha invesor senimen is posiively relaed o he conemporaneous aggregae sock marke excess reurns in he Unied Saes, and 85% of reurns are reversed wihin four monhs. The remainder is reversed wihin 10 monhs. In addiion, hey also find ha his effec is sronger for small socks and growh socks. We nex explore he effecs of invesor senimen differences on shor- and long-erm price deviaions. We regress he price deviaions on six lags of senimen differences from one o six years. Table 11 provides he resuls. Panel B shows ha he coefficiens of he firs 24

25 hree lags of local senimen differences are posiively significan, bu hey become significanly negaive saring from lag 4. The resuls of local senimen differences from annual daa are similar; he impac of he local senimen differences on he price deviaions are reversed afer he fourh year. This impac of senimen on he prices of A- and H-shares are much longer han previously found in oher sock markes (Baker e al., 2012; Ben-Rephael e al., 2012; Firh e al., 2015). [INSERT TABLE 11 ABOUT HERE] We conjecure ha he unusually long cycle of price reversals is relaed o he individual-dominaed rading on he A-share marke. According o he SHSE Annual Saisics Repor in 2013, he proporion of he rades by individual invesors on he A-share marke is as high as 80.78%. This finding indicaes he exisence of a large amoun of noise rading as he lieraure finds ha individual invesors are noise raders (Black, 1986; Kyle, 1985). Thus, hey are more vulnerable o he impac of senimen. By conras, he majoriy of global invesors are insiuional invesors who more likely o be beer rained and disciplined. 11 In addiion, we divide our sample by he average proporion of insiuional holdings. We find ha he effecs of he lagged local senimen differences reverse wihin wo years for he high insiuional holding samples and wihin four years for he low insiuional holding samples. These resuls show ha he long-lasing effec of local senimen is a leas parially caused by he individual-rader dominaed srucure of he A-share marke. 4.3 Robusness checks As a robusness check, we examine he effecs of invesor senimen differences on price 11 According o HKSE, he urnover for foreign invesors accouns for 39% of oal marke urnover from Ocober 2013 o Sepember 2014, of which 34% comes from insiuional invesors. 25

26 deviaions in differen indusries. Guo (2013) explores he facors ha affec he price discouns of Chinese cross-lised companies. He finds ha he effecs of expeced exchange rae changes, relaive marke senimen, and relaive company senimen vary significanly among indusries. According o he indusry classificaion sandard of he China Securiies Regulaory Commission, we classify our samples ino seven indusries. We hen es a dynamic panel model for each indusry using cross-secional fixed effecs mehod. We do no invesigae he effec of insiuional reforms by indusry because he sample periods of some indusries are oo shor o cover he insiuional reform period. Table 12 shows he resuls by indusry. The effecs of senimen differences on price deviaions are generally consisen among indusries. All coefficiens of local senimen difference are posiive, and all of hem are significan. The coefficiens of global senimen difference are again, in general, no significan. All he coefficiens of oal senimen differences are posiive, and abou one-half are significan a he 5% level. [INSERT TABLE 12 ABOUT HERE] In addiion, we follow Arquee e al. (2008) and use he P/E raios of eigh markes as he oal senimen index and decompose hem ino local and global indices by he principal componen analysis. In Panel A of Table 13, we again find a significan impac of oal and he local senimen indices on he price deviaions of A- and H-shares, and once again he impac of he global senimen index on price deviaions is no significan a he 5% level. Overall, he resuls are similar. We rerun our ess by excluding he finance indusry and use he firs order difference of price deviaions as he dependen variable and find similar resuls in Panels B and C. We furher cu our samples ino wo subperiods a June 2003, and he resuls are shown in Panels D and E. I indicaes ha he global senimen significanly 26

27 impacs he price deviaions in he laer period, which is consisen wih our findings on he influences of insiuional reform mosly occurring in he laer period. Finally, cross-lised companies use differen accouning sandards o repor heir financial saemens, and differences in accouning informaion may aribue o price deviaionsof cross-lised shares (Akins e al., 2012; Bushman e al., 2004). The accouning disclosures beween China and Hong Kong are inconsisen unil China implemened new accouning sandards on January 1, We hus es wheher he differences in accouning sandards affec he price deviaions of A- and H-shares. We cu our samples ino wo subperiods a 2007 and repea all of our ess. From Panels F and G of Table 13, our findings are qualiaively similar in ha he oal and he local senimen differences have, in general, a significanly posiive affec on he A- and H-share price deviaions before and afer 2007, and he global senimen differences show a negaive effec afer [INSERT TABLE 13 ABOUT HERE] 5. Conclusions We invesigae he effecs of senimen differences in China and Hong Kong on he price deviaions of cross-lised A-and H-shares. We sudy wheher he price deviaions are impaced by oal, local, and global senimen differences. We also examine he effec of insiuional changes such as he exchange rae reform, he spli share srucure reform, and he QFII and QDII schemes. We find ha oal and local senimen differences have a significanly posiive effec on price deviaions. Global senimen differences are, in general, no significanly relaed o price deviaions, possibly due o capial resricions in China ha deer he effec of global invesor senimen. We furher documen ha he oal senimen differences have a larger effec on price deviaions for small firms and firms wihou 27

28 dividends. Insiuional reforms have a moderaing effec on invesor senimen. Afer he exchange rae reform and he implemenaion of QFII and QDII schemes, he marginal effecs of oal and local senimen differences on price deviaions decrease. Neverheless, he share segmenaion reform increases price deviaions, possibly due o increased invesor demand and ineress o paricipae in he A-share markes afer he reform. We find ha he impac of senimen differences on he price deviaions of A-and H-shares akes up o four years o reverse, which is significanly longer han previously repored in he lieraure. The A-share marke is dominaed by individual raders and hus he marke prices in China are more influenced by invesor senimen. This influence of noise rading is likely causing he effec of invesor senimen o las longer. In addiion o supporing prior sudies ha find ha senimen has a more significan impac on socks ha are difficul o arbirage (Baker and Wurgler, 2006; Baker e al., 2012), we show ha he effecs of invesor senimen end o las much longer in a marke ha is dominaed by individual invesors, a resul has ye o be repored in he lieraure. 28