- Email: [email protected]

Contents lists available at ScienceDirect

The Quarterly Review of Economics and Finance journal homepage: www.elsevier.com/locate/qref

Deregulation and liberalization of the Chinese stock market and the improvement of market efﬁciency Jui-Cheng Hung ∗ Department of Finance, Yuanpei University, No. 306, Yuanpei St. Hsin Chu 30015, Taiwan

a r t i c l e

i n f o

Article history: Received 17 July 2008 Received in revised form 1 April 2009 Accepted 2 April 2009 Available online 24 April 2009 JEL classiﬁcation: G14 G15 G18 Keywords: Variance ratio test Efﬁcient market hypothesis Chinese stock markets Deregulation

a b s t r a c t This study employs single and multiple variance ratio tests to reexamine the weak-form efﬁcient market hypothesis (EMH) of A- and B-shares on the Shanghai and Shenzhen exchanges in Chinese stock market. The study also examines the inﬂuence of the release of investment restriction of B-share markets on market efﬁciency. For the whole sample period, the weak-form EMH is only supported for Shanghai A-shares, and is not supported for the remaining shares. For the sub-sample period, the Shenzhen A-share and B-shares of both exchanges being rejected for the weak-form EMH in the earlier sample period are supported following the regulatory change. Rolling multiple variance ratio test statistic values provide additional evidence of weak-form EMH. The improvement of market efﬁciency can be explained by the increased liquidity and maturity accompanying deregulation and liberalization. © 2009 The Board of Trustees of the University of Illinois. Published by Elsevier B.V. All rights reserved.

1. Introduction The exploration of whether the asset prices follow the random walk hypothesis (RWH) is of interest to not only academics, but also, practitioners and regulators. While a random walk process requires its successive price changes to be identical and independently distributed, a martingale process allows for uncorrelated price changes with a general form of heteroskedasticity. In other words, the martingale process is a generalized version of the random walk process. Brieﬂy speaking, if an asset price follows the martingale process, its present and past prices cannot help us to forecast its future price. This is the essence of the weak-form efﬁcient market hypothesis (EMH). Investors are always

∗ Tel.: +886 3 538 1183x8621. E-mail address: [email protected] 1062-9769/$ – see front matter © 2009 The Board of Trustees of the University of Illinois. Published by Elsevier B.V. All rights reserved.

doi:10.1016/j.qref.2009.04.005

844

J.-C. Hung / The Quarterly Review of Economics and Finance 49 (2009) 843–857

interested in identifying market inefﬁciency, which the behavior of asset prices might be characterized by predictable patterns (Fama, 1970, 1991). However, regulators mainly endeavor to create an efﬁcient capital market, where newly released information can be instantly and fully reﬂected in asset prices and encourage buyers and sellers to rapidly make transactions with reasonable prices. Regulators anticipate providing efﬁcient markets for investors since this will attract international capital to their market and stimulate economic growth. The China government, through the China Securities Regulatory Commission (CSRC), announced to release investing restriction of B-share markets on 19 February 2001, which allows domestic residents to trade in B-share markets. An interesting topic that arises from this regulatory change concerns the efﬁciency of the stock markets, and the market regulators are eager to understand whether this regulatory change has improved the efﬁciency of the A- and B-share markets. The objective of this study aims to reexamine the issue of weak-form EMH of Chinese stock markets, and whether stock market efﬁciency has been improved after deregulation in the B-share markets of the Shanghai and Shenzhen stock exchanges. Numerous empirical studies have tested the weak-form EMH by using the variance ratio test of Lo and Mackinlay (1988) (hereafter, LOMAC). Although LOMAC (1989) and Liu and He (1991) asserted that the heteroskedasticity-robusted variance ratio is more powerful and efﬁcient than the Box–Pierce or Dickey-Fuller test (1979), it fails to control the joint-test size and is associated with a large probability of Type-I error, as demonstrated by Chow and Denning (1993) (hereafter, CHODE). Recently, Wright (2000) developed the ranks and signs-based variance ratio tests, which involve no size distortions and distribution assumptions. These tests thus have superior testing power to the LOMAC variance ratio test. Consequently, this study combines the variance ratio test statistics, proposed by LOMAC (1988) and Wright (2000), with the multiple test procedure by CHODE as its empirical methodology. The main contribution of this study is to investigate the weak-form EMH of Chinese stock markets and the inﬂuence of the regulatory change on market efﬁciency by applying more up-to-date methodologies to improve the robustness of the empirical results. Additionally, rolling variance ratio tests1 are carried out to provide a dynamic inspection of the regulatory change on market efﬁciency. The results indicate that the A-share market of Shenzhen and B-share markets of both exchanges become efﬁcient after the deregulation of B-share markets. It is worth to note that, as suggested by the rolling results, the market efﬁciency of Chinese stock market cannot be entirely explained by the inﬂuence of releasing investing restrictions on the B-share market. The remainder of this paper is organized as follows. Section 2 provides a brief introduction of Chinese stock market and literature review. Section 3 describes the econometric methodology. Data description and empirical results are then reported in Section 4. Finally, conclusions are presented in the last section. 2. A brief introduction of Chinese stock market and literature review In the early 1990s, for the sake of attracting international capital inﬂows while mitigating adverse international impacts on Chinese stock market, the China government designed two segmented stock markets in both Shanghai and Shenzhen exchanges.2 Each exchange trades two types of shares, known as A-shares and B-shares. A-shares are denominated in RMB (Renminbi, local Chinese currency), and can only be traded by individuals and legal persons in the People of Republic of China (PRC), which do not include residents of Hong Kong and Macau. B-shares traded in the Shanghai exchange are settled in U.S. dollars and those traded in the Shenzhen exchange are settled in Hong Kong dollars. The investors of B-shares are from Hong Kong, Macau, Taiwan, and other foreign countries. On 19 February 2001, the China government announced to release investing restriction which allows domestic residents to trade in B-share markets. Lu, Wang, Chen, and Chong (2007) and Fiﬁeld and Jetty (2008) explored the effects of these regulatory policy changes on market efﬁciency. Lu et al. (2007)

1 The technique of rolling variance ratio test is designed to accommodate the dynamic nature of the market and possible structural changes or inﬂuential outliers, and this method is also used to investigate the weak-form efﬁciency of stock indices by Tabak (2003), Belaire-Franch and Opong (2005b), Kim and Shamsuddin (2008) and Hung, Lee, and Pai (2009). 2 Reader interested in Chinese stock exchanges, please refer to Zhang and Yu (1994) and Seddighi and Nian (2004) for a more detailed account.

J.-C. Hung / The Quarterly Review of Economics and Finance 49 (2009) 843–857

845

found that the opening of the B-share market to domestic Chinese investor and the limited opening of the A-share market to foreign investors enhance market efﬁciency. Moreover, the price differentials of A- and B-share markets have been effectively decreased by the regulatory change of B-share markets, and no price spillovers between two markets existed in recent years. In line with ﬁndings of Lu et al. (2007), Fiﬁeld and Jetty (2008) adopted both parametric (LOMAC) and nonparametric (ranks-based test of Wright) variance ratio tests with disaggregated company-level data to examine the weak-form efﬁciency of the Chinese A- and B-share markets. They found that the A-share market is generally more weak-form efﬁciency than the B-share market, and the regulatory change has enhanced the pricing efﬁciency in the B-share markets by reducing the information disadvantage of foreign investors and increasing the speed of information diffusion. Among the existing literatures, there are quite a few of studies examining weak-form EMH of Chinese stock market for the last 10 years. Some studies (e.g. Darrat & Zhong, 2000; Mookerjee & Yu, 1999; Seddighi & Nian, 2004) found that both of the A- and B-share markets are not weak-form efﬁcient. However, some scholars, for instance Ma and Barnes (2001), Wang, Burton, and Hannah (2004) and Lima and Tabak (2004) concluded that the A-share markets for both Shanghai and Shenzhen exchanges are supported by weak-form efﬁciency while the B-shares markets exhibited signiﬁcant predictable components. Moreover, Laurence, Cai, and Qian (1997), Su and Fleisher (1998), Wang et al. (2004) and Cajueiro and Tabak (2006) documented that the A-share markets display greater efﬁciency than Bshares. More recently, using the dataset of Chinese stock market from 1992 to 2001, Li (2003a) found that there has been a steady movement towards efﬁciency, possibly as a result of greater levels of market liquidity and a strengthening of regulations. Similarly, Li (2003b) found that A- and B-shares have exhibited an increasing trend towards efﬁciency, although B-shares appear to be less efﬁcient than A-shares. Li (2003b) attributed this difference in the efﬁciency of A- and B-shares to the greater access to information by domestic investors as compared to their foreign counterparts. Although the conclusions of previous empirical researches are quite mixed, great majority indicate that the B-share markets are not weak-form efﬁcient before the regulatory change. The result might be attributed to the difﬁculties of accessing information about Chinese company listed on B-share than their domestic counterparts (Chakravarty, Sarkar, & Wu, 1998; Chan, Menkveld, & Yang, 2006). Some scholars asserted that the regulatory change has resulted in greater efﬁciency in the A-share markets because the Chinese government’s control of the domestic media restricts information to Ashare investors and causes them to rely on B-share investors for information (Chui and Kwok, 1998; Sjöö & Zhang, 2000; Yang, 2003). Besides, the empirical result of Fiﬁeld and Jetty (2008) suggests that the release of investing restriction on domestic resident participation in the B-share market has had a positive impact on pricing efﬁciency in the B-share market since the regulatory change appears to have reduced the information disadvantage of foreign investors and increased the speed with which information is diffused amongst them. 3. Methodology 3.1. LOMAC variance ratio test The variance ratio tests originated from the pioneering work of LOMAC. The speciﬁcations of LOMAC’s test are brieﬂy described below. Let pt denote the log value of the stock index Pt at time t, and its stochastic process is given by the following recursive equation: pt = + pt−1 + εt

εt ∼IID N(0, 2 )

(1)

where is a drift parameter, and IID N(0, 2 ) denotes that εt is independently and identically normal distributed with mean 0 and variance 2 . If the price pattern of an asset or a ﬁnancial index can be described as Eq. (1), then the RWH or the EMH is valid for them. For the empirical test, LOMAC developed tests of random walks under alternative assumptions of homoskedasticity and heteroskedasticity on εt . In order to test the RHW for these equity indices the variance ratio test (VR) statistic is used. Under the RWH null the variance of ﬁrst differences of a time series increases linearly, such that the variance of the q-th differences is simply q times the variance of the ﬁrst difference, i.e.

846

J.-C. Hung / The Quarterly Review of Economics and Finance 49 (2009) 843–857

Var(pt − pt−q ) = qVar(pt − pt−1 ). Thus, the variance ratio of the q-th difference can be deﬁned as:

(1/q)Var(pt − pt−q ) 2 (q) a k = 2 VR(q) = =1 + 2 1− q Var(pt − pt−1 ) (1) q−1

(k) ˆ

(2)

k=1

where VR(q) is the variance ratio of stock indices’ q-th difference, 2 (q) is the variance of stock indices’ q-th difference, 2 (1) is the variance of the ﬁrst-differenced stock index series, q is the difference interval, and (k) ˆ denotes the autocorrelation coefﬁcient of lag k. Following LOMAC (1988, 1989), the estimator of the q-period difference, 2 (q), is calculated as: 1 2 (pt − pt−q − q) ˆ m nq

2 (q) =

t=q

2 (1)

(3)

1 2 = (pt − pt−1 − ) ˆ nq − 1 nq

t=1

nq

where, m = (nq − q + 1)(1 − q/nq) and ˆ = (1/nq) t=1 (pt − pt−1 ) = (1/nq)(pnq − p0 ). If RWH is valid, the estimated variance ratio of different time-period q will be equal to 1. The test statistic Z1 (q) is given by: VR(q) − 1

Z1 (q) =

1/2

[(q)]

a

∼N(0, 1)

(4)

which, under the assumption of homoskedasticity, is asymptotically distributed as N(0,1). The asymptotic variance, (q), is given by: (q) =

2(2q − 1)(q − 1) 3qT

(5)

The test statistic Z2 (q), which is robust under heteroskedasticity, is: VR(q) − 1

Z2 (q) =

1/2 [∗ (q)]

where ∗ (q) =

a

∼N(0, 1)

q−1 2(q − j) 2 q

j=1

(6)

nq

2

(pt −pt−1 −) ˆ (pt−j −pt−j−1 −) ˆ

nq

ˆ and ı(j) ˆ = ı(j)

t=j+1

k=1

(pt −pt−1 − ˆ 2)

2

2

. LOMAC (1989) used

Monte Carlo simulation to demonstrate that the variance ratio test statistics are more powerful than the traditional Box–Pierce statistics, and Liu and He (1991) found the variance ratio is more powerful than Box–Pierce and Dickey-Fuller tests (1979, 1981). 3.2. Ranks-based and signs-based variance ratio tests Wright (2000) utilized the concept of nonparametric test to propose the variance ratio test statistics based on ranks and signs which attempts to enhance the testing power of the LOMAC variance ratio test. The rank-based and sign-based variance ratio tests possess two advantages. First, the LOMAC variance ratio test statistics is derived with the asymptotic theory, therefore, it is born with the problem of size distortions. However, the statistics proposed by Wright are free from this problem. Second, the statistics of Wright are more powerful than the alternatives because the distributions of most ﬁnancial data are often non-normal. Given T observations of asset returns {y1 , . . ., yT }, the test statistics R1 and R2 are deﬁned as:

R1 =

(1/Tq)

T t=q

(r1t +, . . . , r1t−q+1 )2

(1/T )

T

r2 t=1 1t

−1

× (q)−1/2

(7)

J.-C. Hung / The Quarterly Review of Economics and Finance 49 (2009) 843–857

(1/Tq)

R2 =

T t=q

(r2t +, . . . , r2t−q+1 )2

(1/T )

where r1t = r(yt ) −

T +1 2

T

r2 t=1 2t

(T −1)(T +1) /

12 ˚−1

847

× (q)−1/2

−1

, r2t = ˚−1 (r(yt )/(T + 1)), and (q) =

(8)

2(2q−1)(q−1) . 3qT

r(yt ) is the

is the inverse of the standard normal cumulative distribution rank of yt among y1 ,. . ., yT , and function. The test statistics based on the signs of returns are given by:

S1 =

(1/Tq)

T

t=q

(st +, . . . , st−q+1 )2

(1/T )

S2 =

(1/Tq)

T t=q

T

s2 t=1 t

−1

(st ()+, ¯ . . . , st−q+1 ()) ¯ 2

(1/T )

T

s () ¯ 2 t=1 t

× (q)−1/2

(9)

−1

× (q)−1/2

(10)

0.5 if xt > k, . Thus, S1 assumes a zero drift −0.5 otherwise value. If the value of the drift parameter is unknown, the procedure described in Luger (2003), based on Campbell and Dufour (1997), is applied to compute S2 . This method consists of a two-step strategy. The ﬁrst step is to establish an exact conﬁdence interval for the drift parameter , that is valid at least under the null hypothesis. Denote by y(1) ,. . ., y(T) the order statistics of the sample y1 ,. . ., yT . An exact conﬁdence interval CI (˛1 ) for with level 1 − ˛1 is given by [y(h+1) ,y(T−h) )], where h is the largest integer such that Pr[B ≤ h] ≤ ˛1 /2, for B a binomial random variable with number of trials T and the probability of success 1/2. In large samples, the binomial probabilities can be approximated with those associated with a normal distribution. The second step is to compute the S2 statistic, for each candidate value b for the drift parameter in the conﬁdence interval. The value of the S2 statistic at the aggregation interval k is then deﬁned as where (q) =

2(2q−1)(q−1) , st 3qT

= 2u(yt ,0), and u(xt , k) =

S2 (q) = inf {|S2 (q, b)| : b ∈ CI (˛1 )} where given b ∈ CI (˛1 ), S2 (k,b) is computed by deﬁning st (b) = 2u(yt ,b). The chosen S2 value is compared to the appropriate critical values for an ˛2 level test, such that the overall level of the strategy is bounded by ˛ = ˛1 + ˛2 . In the paper, we have set ˛1 = 0.01 and ˛2 = 0.04.3 3.3. Multiple variance ratio tests CHODE (1993) modiﬁed and extended the variance ratio of LOMAC into multiple setting. As indicated by CHODE, a nominal 100˛ percent critical value is not appropriate for each q selected due to failing to control overall test size for multiple comparisons, hence leading to an inappropriately large probability of Type I error. That is, the application of VR tests for multiple q values induces over-rejection of the null hypothesis, above the nominal size. CHODE proposed the multiple variance ratio test incorporated with Studentized Maximum Modulus (SMM) critical values to control overall test size for the variance ratio test statistics under different time period q. Under the null hypothesis, for a single variance ratio test, VR(q) = 1, and Mr (q) = VR(q) − 1 = 0. Now consider a set of m variance ratio tests {Mr (qi )|i = 1,. . ., m}, where {qi |i = 1,. . ., m}, and {qi ≥ 1, qi = / qj |qi ∈ N}, ∀i = / j. Under the speciﬁcation, the random walk null hypothesis consists of m sub-hypotheses: H0i : Mr (qi ) = 0 / 0 H0i : Mr (qi ) =

for i = 1, . . . , m for any i = 1, . . . , m

(10)

Rejection of any sub-hypothesis H0i will lead to the turndown of random walk hypothesis. Consider ﬁve sets of above-mentioned test statistics, {Zj (qi )|i = 1,. . ., m}, {Rj (qi )|i = 1,. . ., m} for j = 1, 2 and 3

I follow the works of Belaire-Franch and Opong (2005a, 2005b).

848

J.-C. Hung / The Quarterly Review of Economics and Finance 49 (2009) 843–857

{S1 (qi )|i = 1,. . ., m}. Since the random walk hypothesis is rejected if any of the estimated variance ratio ratios is signiﬁcantly different from one, we follow CHODE to reconstruct the test statistics under the multiple speciﬁcations. The multiple variance ratio test is based on the following inequality: Pr[max(|z1 |, . . . , |zm |) ≤ SMM(˛; m; N)] ≥ (1 − ˛)

(11)

where {zi |i = 1,. . ., m} is a set of m standard normal variates, SMM(˛; m; N) is the upper ˛ point of the SMM distribution with parameter m and N (sample size) degree of freedom. Asymptotically, when N goes inﬁnite, SMM(˛; m; ∞) = Z˛+ /2 , where ˛+ = 1 − (1 − ˛)1/m . This paper adopts six modiﬁed variance ratio test statistics, which are combined contributions of LOMAC (1988), CHODE (1993) and Wright (2000), as follows: Zj∗ (q) = max |Zj (qi )|,

for j = 1, 2

Rj∗ (q) = max |Rj (qi )|,

for j = 1, 2

Sj∗ (q) = max |Sj (qi )|,

for j = 1, 2

1≤i≤m

1≤i≤m

1≤i≤m

(12)

where the critical values of Zj∗ (q) are based on the above-mentioned SMM distribution. Following the method of Belaire-Franch and Contreras (2004), under the Assumption 0 (i.i.d. ﬁrst differences) in Wright (2000), the test statistics of Rj∗ (q) are distributed as †

†

†

max{|Rj (q1 )|, |Rj (q2 )|, . . . , |Rj (qm )|}

(13)

†

where Rj (q1 ) is the ranks-based test computed with any random permutation of the elements {yt }Tt=1 . Under Assumptions 1 and 2 (increments follow a martingale difference sequence) in Wright (2000), the test statistics of Sj∗ (q) are distributed as †

†

†

max{|Sj (q1 )|, |Sj (q2 )|, . . . , |Sj (qm )|} †

(14) † T

where Sj (q1 ) is the signs-based test compute with an i.i.d. sequence {st }t=1 , each element of which is 1 with probability 1/2 and −1 otherwise. Therefore, the exact sampling distribution of Rj∗ (q) and Sj∗ (q) (j = 1, 2) can be simulated with any arbitrary degree of accuracy. The critical values of the ranks and signs-based multiple variance ratio test statistics in the following empirical work are computed through 100,000 replications. 4. Empirical results 4.1. Data description and preliminary analysis Numerous Chinese state-owned enterprises have been authorized to issue both A- and B-share markets on the Shanghai and Shenzhen stock exchanges recently. As of June 2006, 823 companies were listed on A-share market and 54 companies were listed on the B-share market of the Shanghai Exchange, with a total capitalization value of about US$473 billion. Meanwhile, 517 companies were listed on the A-share market and 55 companies were listed on the B-share market of the Shenzhen Exchange, with a market capitalization value of about US$167 billion. Table 1 lists these statistics. The stock price data in this study were obtained from the Datastream database and a comprehensive dataset that comprises closing price indices for Chinese main stock market is used. The four Chinese stock market price indices are the Shanghai A-share index (SHAI), Shanghai B-share index (SHBI), Shenzhen A-share index (SZAI), and Shenzhen B-share index (SZBI). The sample period covers nearly 10 years, from 5 April 1996 to 30 December 2005, containing a total of 2344 daily observations for each stock index price series. To examine the impact of releasing restrictions on investment in the B-share markets on overall market efﬁciency, the sample period is divided into two sub-samples, each including 1172 observations. The ﬁrst difference of log index-prices, namely Ri,t = 100 ln(Pi,t /Pi,t−1 ), is used for the series of returns, where Pi,t is the closing price index of i share at time t.

J.-C. Hung / The Quarterly Review of Economics and Finance 49 (2009) 843–857

849

Table 1 Market size, June 2006. Exchange

Share

No. of listed securities

Total market capitalization

Shanghai

A B

823 54

RMB 3755 billion (or US$ 469.75 billion) RMB 33.1 billion (or US$ 4.14billion)

Shenzhen

A B

517 55

RMB 1280 billion (or US$ 160.13) RMB 55.8 billion (or US$ 6.98)

Note. The total companies listed on SHSE and SZSE are 833, and 572, respectively. There are 44 companies that have issued both A- and B-shares on SHSE at the same time and 42 companies on SZSE.

Table 2 lists descriptive statistics of daily returns for all indices on both the Shanghai and Shenzhen stock exchanges during the pre- and post-event periods. Skewness and kurtosis indicate that the distribution of the returns of all indices during both periods have a fat tail and sharper peak than the normal distribution. The J–B normality test signiﬁcantly rejects the hypothesis of normality. The autocorrelation coefﬁcients of lag 1–15 indicate that the coefﬁcients are larger in the pre-event period than the post-event period. Furthermore, the signiﬁcance of each autocorrelation coefﬁcient of SHBI weakens after the release of investment restrictions. Particularly, the autocorrelation coefﬁcients of SHAI and SZAI become insigniﬁcant during the post-event period. Furthermore, the Ljung–Box Q(15) statistics demonstrate that the returns of all indices during the pre-event period are strongly rejected at the 1% level, but not during the post-event period (except for SZBI). This implies that the returns are autocorrelated before the release of the investing restrictions on the B-share markets, meaning the weak-form EMH is not supported during this period. However, besides SZBI, the autocorrelation of returns was not obvious, implying the weak-form EMH might be valid following the release of investment restriction. Therefore, it appears that market efﬁciency might be enhanced following deregulation. To ensure robustness, more powerful tests are implemented in the following section. 4.2. Variance ratio tests of the stock indices associated with the Shanghai and Shenzhen exchanges This section applies the variance ratio test statistics of LOMAC (1988) and Wright (2000) to test the weak-form EHM of the A- and B-share stock indices associated with the Shanghai and Shenzhen Exchanges for the whole sample period. Six test statistics, including Z1 , Z2 , R1 , R2 , S1 and S2 , with four time intervals (q = 2, 4, 8, 16) are selected for the variance ratio tests. Table 3 lists the empirical results. Notably, the Ljung–Box Q2 (15) statistics in Table 2 are all strongly rejected at the 1% level, indicating that all returns are heteroskedastic. Consequently, the values of Z1 are not reported because this statistic is not robust to the heteroskedasticity. The variance ratios and test statistic values of Z2 , R1 , R2 , S1 and S2 are listed in the main row and the parentheses, respectively. For the A-share markets, the weak-form EMH of SHAI is rejected by q = 8, 16 of R1 and by q = 8 of R2 for different signiﬁcance levels. The null hypothesis of SZAI is not rejected only by Z2 , but is rejected by R1 , R2 , S1 and S2 . For the B-share markets, the null hypotheses of SHBI and SZBI are consistently rejected by Z2 , R1 , R2 and the S1 is around the 1% level for all time intervals. It is clear to see that if conclusions are based only on the heteroskedasticity-robusted test statistic Z2 , it will be incorrectly inferred that the A-share markets of Shanghai and Shenzhen exchanges are weak-form efﬁcient. For clarity, the multiple variance ratio test of CHODE is implemented to minimize the possibility of Type-I error. CHODE used the critical values of SMM distribution to control overall test size for the variance ratio test statistics of Z1 and Z2 for different time intervals. More recently, Belaire-Franch and Contreras (2004) proposed the method of computing critical values while extending the ranks and signs-based variance ratio tests of Wright (2000) to multiple cases. The multiple variance ratio test statistics are adopted to retest the efﬁciency of the returns of the four indices, and the results for entire sample period and corresponding critical values are listed in Tables 4 and 5, respectively. Table 4 clearly shows that the null hypotheses of SHBI and SZBI are both strongly rejected for all test statistics, except for S2∗ for SHBI, indicating that the weak-form EMH is invalid for the B-share markets of both exchanges. For the cases of SHAI and SZAI, the null hypotheses are supported by the Z2∗ statistic. However, opposite conclusions are obtained when using alternative test statistics. Unlike SZAI, the

850

J.-C. Hung / The Quarterly Review of Economics and Finance 49 (2009) 843–857

Table 2 Descriptive statistics of daily return. Index

SHAI

Panel A: pre-event period Mean SD Skewness Kurtosis J–B

0.105a 1.963 −0.458c 8.171c 1345.922c

Autocorrelations 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 Ljung–Box Q(15) Ljung–Box Q2 (15) Panel B: post-event period Mean SD Skewness Kurtosis J–B Autocorrelations 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 Ljung–Box Q(15) Ljung–Box Q2 (15)

SHBI

SZAI

SZBI

0.043 2.649 0.307c 6.333c 560.620c

0.026b 0.388 0.002 7.138c 2485.991c

0.013 0.618 0.297c 8.319c 1397.998c

0.016 −0.001 0.069 0.023 0.033 −0.052a −0.051b −0.069b −0.035b −0.010b −0.084c 0.067c 0.011c −0.012c 0.075c

0.161c −0.026c 0.055c 0.019c −0.033c −0.009 c −0.021c 0.009c 0.012c 0.000c −0.007c 0.042c 0.019c −0.011c 0.011c

−0.031 0.072b 0.058c 0.034b 0.077c −0.077c 0.005c −0.054c −0.033c 0.045c 0.015c 0.038c −0.001c 0.041c 0.034c

0.130c 0.044c 0.057c 0.038c 0.022c −0.013c −0.011c 0.009c 0.035c 0.026c −0.023c 0.030c −0.017c −0.003c −0.010c

42.774c 355.702c

40.611c 491.664c

39.767c 426.048c

33.387c 471.907c

−0.045 1.353 0.867c 9.073c 1946.451c

−0.033 1.993 0.336c 8.566c 1533.801c

−0.010 0.239 0.629c 7.927c 1262.274c

0.005 0.387 0.410c 7.616c 1072.530c

0.009 −0.029 0.025 −0.000 −0.037 −0.003 0.055 −0.013 −0.018 −0.001 −0.001 0.026 0.000 0.019 0.000

0.074b −0.013b 0.008a 0.026 0.044a 0.056b 0.006a 0.008a 0.006 0.046a 0.026a 0.034a −0.003 0.010 −0.006

0.029 −0.022 0.035 −0.001 −0.021 0.004 0.049 −0.005 −0.014 0.005 0.012 0.025 −0.005 0.012 −0.014

9.172 58.002c

19.060 465.255c

8.331 107.046c

0.102c 0.013c 0.106c 0.090c 0.041c 0.103c 0.105c 0.010c 0.045c 0.097c 0.037c 0.083c 0.059c 0.022c −0.010c 92.005c 1285.531c

Note. Superscript letters a, b and c denote signiﬁcantly at the 10%, 5% and 1% level, respectively. J–B test statistics are based on Jarque and Bera (1987) and are asymptotically chi-square-distributed with 2 degrees of freedom.

null hypothesis of SHAI is only weakly rejected by R1∗ at the 10% signiﬁcance level, and SZAI is strongly rejected by R1∗ , R2∗ , S1∗ and S2∗ for the whole sample period. Based on the preliminary analysis, the distributions of the return series are non-normal and heteroskedastic. Therefore, the nonparametric test statistics of R1∗ , R2∗ , S1∗ and S2∗ are more appropriate for application to ﬁnancial time series data.

J.-C. Hung / The Quarterly Review of Economics and Finance 49 (2009) 843–857

851

Table 3 Variance ratio test for whole sample period. SHAI

SHBI

SZAI

SZBI

q=2 q=4 q=8 q = 16

1.016 (0.426) 1.045 (0.651) 1.094 (0.922) 1.012 (0.085)

1.138 (3.893)c 1.220 (3.383)c 1.311 (3.115)c 1.405 (2.888)c

0.988 (−0.298) 1.058 (0.803) 1.164 (1.470) 1.218 (1.395)

1.120 (3.116)c 1.261 (3.712)c 1.465 (4.292)c 1.677 (4.403)c

R1 q=2 q=4 q=8 q = 16

1.011 (0.526) 1.049 (1.266) 1.112 (1.833)a 1.211 (2.325)b

1.085 (4.108)c 1.157 (4.056)c 1.255 (4.172)c 1.438 (4.816)c

1.040 (1.950)a 1.145 (3.756)c 1.287 (4.688)c 1.499 (5.484)c

1.103 (4.973)c 1.188 (4.870)c 1.308 (5.047)c 1.475 (5.226)c

R2 q=2 q=4 q=8 q = 16

1.012 (0.584) 1.047 (1.224) 1.106 (1.730)a 1.133 (1.464)

1.111 (5.376)c 1.188 (4.856)c 1.288 (4.708)c 1.437 (4.810)c

1.021 (1.021) 1.113 (2.933)c 1.245 (4.014)c 1.396 (4.354)c

1.112 (5.436)c 1.216 (5.576)c 1.371 (6.075)c 1.558 (6.138)c

q=2 q=4 q=8 q = 16

0.995 (−0.227) 0.994 (−0.144) 0.971 (−0.482) 0.985 (−0.163)

1.042 (2.045)b 1.105 (2.717)c 1.203 (3.317)c 1.446 (4.906)c

1.037 (1.797)a 1.094 (2.440)b 1.183 (2.996)c 1.329 (3.616)c

1.066 (3.202)c 1.128 (3.302)c 1.187 (3.056)c 1.289 (3.176)c

q=2 q=4 q=8 q = 16

1.000 (0.020) 1.000 (0.011) 1.000 (0.000) 0.999 (0.008)

1.021 (1.053) 1.038 (0.993) 1.054 (0.897) 1.194 (2.137)b

1.020 (0.970) 1.069 (1.799) 1.151 (2.486)b 1.303 (3.339)b

1.062 (3.036)b 1.105 (2.738)b 1.161 (2.639)b 1.252 (2.775)b

Z2

S1

S2

Note. Superscript letters a, b and c denote signiﬁcantly at the 10%, 5% and 1% level, respectively. The numbers given in the main row are variance ratio, and the test statistic values are given in the parentheses.

The results of this section are dissimilar to those of Long, Payne, and Feng (1999), Lee, Chen, and Rui (2001) and Lima and Tabak (2004). The ﬁndings of Long et al. (1999) supported RWH for both the A- and B-share markets of Shanghai exchange; meanwhile, Lee et al. (2001) rejected RWH for both the A- and B-share markets of the Shanghai and Shenzhen exchanges. However, the empirical results of Lima and Tabak (2004) rejected the weak-form EMH for B-share markets, but supported the weakform EMH for the A-share markets of both exchanges. These empirical differences may result from the different econometric methods and different datasets (data frequencies), and probably, the data periods might experience structural changes or outliers due to the events such as ﬁnancial catastrophe and regulatory changes. As mentioned by Kim and Shamsuddin (2008), the structural change can alter the nature of market efﬁciency while inﬂuential outliers adversely affect the performance of statistical tests. For this reason, the whole sample period is divided into two sub-sample periods by the regulatory change of B-share market, and more thorough investigations are carried out in the following section. Table 4 CHODE multiple variance ratio test for whole sample period.

Z2∗ R1∗ R2∗ S1∗ S2∗

SHAI

SHBI

SZAI

SZBI

0.922 2.325a 1.730 0.482 0.020

3.893c 4.816c 5.376c 4.906c 2.137

1.470 5.484c 4.354c 3.616c 3.339b

4.403c 5.226c 6.138c 3.302c 3.036b

Note. Superscript letters a, b and c denote signiﬁcantly at the 10%, 5% and 1% level, respectively. The test statistic values are given in the table. The critical values of SMM(˛, m = 4, ∞) at 10%, 5% and 1% are 2.226, 2.491 and 3.022.

852

J.-C. Hung / The Quarterly Review of Economics and Finance 49 (2009) 843–857

Table 5 Critical values of ranks and signs-based multiple variance ratio test. T = 2344, q = 2, 4, 8, 16

T = 1172, q = 2, 4, 8, 16

T = 500, q = 2, 4, 8, 16

R1∗ ˛ = 10% ˛ = 5% ˛ = 1%

2.0527 2.3434 2.9175

2.0451 2.3291 2.9134

2.022 2.303 2.863

R2∗ ˛ = 10% ˛ = 5% ˛ = 1%

2.0463 2.3433 2.9169

2.0371 2.3304 2.9034

2.022 2.298 2.861

˛ = 10% ˛ = 5% ˛ = 1%

2.0319 2.3295 2.8952

2.0447 2.3368 2.9210

2.025 2.318 2.916

˛2 = 4%

2.4179

2.4201

2.405

S1∗

S2∗

Note. Each critical value is computed through 100,000 replications.

4.3. The impact of releasing investing restriction of B-share markets to market efﬁciency Chinese stock markets recently have undergone some regulatory changes. In response to globalization, regulators must liberalize markets by removing trading restrictions. The sample period is divided into two sub-samples to compare market efﬁciency before and after permitting domestic residents to invest in B-shares. Table 6 Variance ratio test before releasing investment restrictions. SHAI

SHBI

SZAI

SZBI

1.016 (0.308) 1.059 (0.609) 1.125 (0.880) 0.974 (−0.138)

1.160 (3.398)c 1.241 (2.802)c 1.293 (2.194)b 1.333 (1.772)a

0.967 (−0.593) 1.054 (0.550) 1.176 (1.168) 1.194 (0.922)

1.125 (2.352)b 1.269 (2.799)c 1.402 (2.732)c 1.484 (2.311)b

R1 q=2 q=4 q=8 q = 16

1.002 (0.060) 1.037 (0.670) 1.105 (1.219) 1.139 (1.079)

1.135 (4.616)c 1.232 (4.247)c 1.329 (3.802)c 1.563 (4.378)c

1.037 (1.267) 1.170 (3.109)c 1.340 (3.938)c 1.522 (4.056)c

1.171 (5.843)c 1.306 (5.588)c 1.451 (5.221)c 1.647 (5.033)c

R2 q=2 q=4 q=8 q = 16

1.006 (0.206) 1.044 (0.809) 1.113 (1.310) 1.076 (0.593)

1.151 (5.167)c 1.241 (4.416)c 1.316 (3.653)c 1.459 (3.571)c

1.007 (0.256) 1.121 (2.219)b 1.281 (3.249)c 1.404 (3.142)c

1.160 (5.466)c 1.301 (5.500)c 1.450 (5.203)c 1.613 (4.768)c

q=2 q=4 q=8 q = 16

1.003 (0.088) 1.006 (0.109) 1.006 (0.069) 1.017 (0.131)

1.085 (2.893)c 1.165 (3.015)c 1.266 (3.077)c 1.566 (4.397)c

1.404 (2.601)c 1.194 (3.546)c 1.348 (4.021)c 1.526 (4.086)c

1.136 (4.646)c 1.243 (4.436)c 1.343 (3.966)c 1.450 (3.497)c

q=2 q=4 q=8 q = 16

0.999 (0.029) 0.999 (0.015) 1.000 (0.000) 0.998 (0.014)

1.060 (2.074)b 1.099 (1.827) 1.116 (1.353) 1.324 (2.522)b

1.017 (0.613) 1.127 (2.327)b 1.264 (3.057)b 1.433 (3.372)b

1.128 (4.412)b 1.211 (3.873)b 1.295 (3.413)b 1.370 (2.879)b

Z2 q=2 q=4 q=8 q = 16

S1

S2

Note. Superscript letters a, b and c denote signiﬁcantly at the 10%, 5% and 1% level, respectively. The numbers given in the main row are variance ratio, and the test statistic values are given in the parentheses.

J.-C. Hung / The Quarterly Review of Economics and Finance 49 (2009) 843–857

853

Table 7 CHODE multiple variance ratio tests before releasing investment restrictions.

Z2∗ R1∗ R2∗ S1∗ S2∗

SHAI

SHBI

SZAI

SZBI

0.880 1.219 1.310 0.131 0.029

3.398c 4.616c 5.167c 4.397c 2.522b

1.168 4.056c 3.249c 4.086c 3.372b

2.799b 5.843c 5.500c 4.646c 4.412b

Note. Superscript letters a, b and c denote signiﬁcantly at the 10%, 5% and 1% level, respectively. The numbers given in the main row are variance ratio, and the test statistic values are given in the parentheses. The critical values of SMM(˛, m = 4, ∞) at 10%, 5% and 1% are 2.226, 2.491 and 3.022.

The results of single and multiple variance ratio tests before the release of investment restrictions are listed in Tables 6 and 7, respectively. The critical values of ranks and signs-based multiple variance ratio tests for 1172 observations are also listed in Table 5. As indicated by the single and multiple variance ratio test statistics, the results of SHBI, SZAI and SZBI are basically the same as those for the whole sample period. The weak-form EMH of these three indices thus are rejected before the release of investment restrictions. For SHAI, however, the statistics of both the single and multiple variance ratio tests indicate market efﬁciency. As suggested by many empirical studies (Cajueiro & Tabak, 2006; Laurence et al., 1997; Lima & Tabak, 2004; Su & Fleisher, 1998; Wang et al., 2004), the B-share markets are less efﬁcient than the A-shares. Based on the results of this section, if traditional method (Z2∗ ) is used alone, one can obtain the same conclusion in line with previous studies. However, in light of R1∗ , R2∗ , S1∗ and S2∗ , the conclusion of previous studies is only valid for the Shanghai exchange. There is no evidence indicating that the B-share market of Shenzhen exchange is less efﬁcient than its A-shares. Table 8 Variance ratio tests after releasing investment restrictions. SHAI

SHBI

SZAI

SZBI

1.009 (0.256) 0.997 (−0.051) 0.985 (−0.136) 0.993 (−0.046)

1.064 (1.299) 1.032 (0.353) 1.066 (0.480) 1.123 (0.642)

1.030 (0.807) 1.038 (0.531) 1.060 (0.531) 1.104 (0.633)

1.086 (1.737)a 1.136 (1.469) 1.292 (2.011)b 1.355 (1.715)a

R1 q=2 q=4 q=8 q = 16

1.001 (0.040) 1.015 (0.266) 1.026 (0.306) 1.117 (0.908)

1.012 (0.425) 1.029 (0.530) 1.089 (1.031) 1.151 (1.176)

1.020 (0.693) 1.055 (1.010) 1.108 (1.255) 1.247 (1.919)a

1.023 (0.787) 1.038 (0.698) 1.097 (1.122) 1.136 (1.057)

R2 q=2 q=4 q=8 q = 16

1.002 (0.073) 1.009 (0.173) 1.023 (0.264) 1.089 (0.690)

1.032 (1.088) 1.032 (0.587) 1.095 (1.098) 1.167 (1.297)

1.024 (0.832) 1.055 (1.015) 1.106 (1.222) 1.209 (1.623)

1.044 (1.509) 1.066 (1.200) 1.165 (1.909)a 1.216 (1.676)a

q=2 q=4 q=8 q = 16

0.989 (−0.380) 0.982 (−0.328) 0.932 (−0.790) 0.946 (−0.420)

0.997 (−0.088) 1.035 (0.640) 1.114 (1.319) 1.284 (2.209)b

0.999 (−0.029) 0.994 (−0.109) 1.015 (0.178) 1.115 (0.895)

0.994 (−0.205) 1.000 (0.000) 0.999 (−0.010) 1.067 (0.518)

q=2 q=4 q=8 q = 16

0.999 (0.029) 0.997 (0.046) 0.998 (0.019) 0.997 (0.018)

0.990 (0.321) 0.999 (0.015) 1.000 (0.004) 1.017 (0.134)

0.999 (0.029) 0.999 (0.015) 1.003 (0.039) 1.032 (0.253)

0.997 (0.087) 1.000 (0.000) 0.999 (0.009) 1.058 (0.453)

Z2 q=2 q=4 q=8 q = 16

S1

S2

Note. Superscript letters a, b and c denote signiﬁcantly at the 10%, 5% and 1% level, respectively. The numbers given in the main row are variance ratio, and the test statistic values are given in the parentheses.

854

J.-C. Hung / The Quarterly Review of Economics and Finance 49 (2009) 843–857

Table 9 CHODE multiple variance ratio tests after releasing investment restrictions.

Z2∗ R1∗ R2∗ S1∗ S2∗

SHAI

SHBI

SZAI

SZBI

0.256 0.908 0.690 0.790 0.046

1.299 1.176 1.297 2.209a 0.321

0.807 1.919 1.623 0.895 0.253

2.011 1.122 1.909 0.518 0.453

Note. Superscript letter a denotes signiﬁcantly at the 10%. The numbers given in the main row are variance ratio, and the test statistic values are given in the parentheses. The critical values of SMM(␣, m = 4, ∞) at 10%, 5% and 1% are 2.226, 2.491 and 3.022.

Tables 8 and 9 list the single and multiple variance ratio tests following the release of investment restrictions, and the corresponding critical values are listed in Table 5. The tables show that the weakform EMH still holds for SHAI. However, as indicated by the multiple variance ratio tests, the null hypotheses are not rejected for SZAI and SZBI. Though SHBI is rejected by S1∗ at the 10% signiﬁcance level, the statistical signiﬁcance for the second sub-period reduces markedly compared with the results for the ﬁrst sub-period. Clearly, the degrees of efﬁciency of these three markets are increased, and the weak-form EMH became valid after domestic residents were permitted to invest in B-share markets. Chakravarty et al. (1998) and Chan et al. (2006) mentioned that the possible explanation for greater inefﬁciency of the B-share markets is information asymmetry. In particular, foreign investors in the B-

Fig. 1. Time-varying R1∗ and R2∗ statistic values of SHBI with a ﬁxed window-size of 500 observations. The horizontal lines represent 5% critical values, and the vertical line indicates the time of releasing investing restriction of B-share.

J.-C. Hung / The Quarterly Review of Economics and Finance 49 (2009) 843–857

855

share markets may have an information disadvantage relative to domestic investors trading A-shares due to language barrier, different accounting standards and the lack of reliable information about local ﬁrms. On the other hand, for the possible reason that the regulatory change has resulted in greater efﬁciency in the A-share market, Chui and Kwok (1998) and Yang (2003) have argued that the Chinese government’s control of the domestic media restricts information to A-share investors and causes them to rely on B-share investors for information. The following presents single and multiple variance ratio tests based on the rolling-window technique to present another perspective on the market efﬁciency of SHBI and SZBI. Furthermore, this method can be used to examine the robustness of the above results. The time-varying statistical values are calculated using a ﬁxed window-size of 500 observations, and only R1∗ and R2∗ are illustrated in Figs. 1 and 2 because they are robust to distribution assumptions and the joint-test size is controlled. In the ﬁgures, the horizontal lines represent critical values at the 5% level and the vertical line indicates the date of the release of investment restrictions on the B-share market. For both SHBI and SZBI, the time-varying statistic values of R1∗ and R2∗ exceed the 5% critical value before the release of investment restrictions, consistent with the preceding results. After the release of investment restrictions, it is interesting to note that the statistics of R1∗ and R2∗ do not gradually decrease as one expected. On the contrary, the statistics of both SHBI and SZBI increase instantly. By looking at graphs of SHBI and SZBI, one can ﬁnd that both indices went through an impressive rise to reduce the perceived price differentials compared with A-shares, and therefore might deteriorate the market efﬁciency. As indicated by Figs. 1 and 2, the statistical values lie within the upper 5% critical value as time evolves since the markets need time to respond to the regulatory change. Following the deregulation of B-share markets,

Fig. 2. Time-varying R1∗ and R2∗ statistic values of SZBI with a ﬁxed window-size of 500 observations. The horizontal lines represent 5% critical values, and the vertical line indicates the time of releasing investing restriction of B-share.

856

J.-C. Hung / The Quarterly Review of Economics and Finance 49 (2009) 843–857

the A-shares are opened to Qualiﬁed Foreign Institutional Investor (QFII) on 1 December 2002.4 The openness of A-share markets to embrace QFII may further improve the market efﬁciency because it alleviates the difﬁculties for foreign investors to access information about Chinese companies. Besides, a corresponding drop of R1∗ and R2∗ around 1650 in Figs. 1 and 2 might be explained by the openness of A-share markets. Hence, the rolling results demonstrate that the explanation for the efﬁciency of Chinese stock markets cannot be entirely ascribed to the inﬂuence of releasing investing restrictions on the B-share market. However, it is certain that the Chinese stock markets became more efﬁcient, and market capitalization was boosted following the recent deregulation and liberalization. 5. Summary and conclusions This study adopted single and multiple variance ratio tests, which combined the methods of LOMAC (1988), CHODE (1993) and Wright (2000), to investigate the weak-form EMH of the A- and B-shares markets of the Shanghai and Shenzhen exchanges in Chinese stock market. The ranks and signs-based multiple variance ratio tests of Belaire-Franch and Contreras (2004) are more powerful than the parametric multiple variance ratio test of CHODE. The ranks-based multiple variance ratio tests are exact under the independence and identical distribution assumption, while the sign-based tests are exact even under conditional heteroskedasticity. This study also examines the inﬂuence of the release of investment restrictions on the B-share markets on the market efﬁciency of Chinese stock market. Consequently, the whole sample period, ranging from 5 April 1996 to 30 December 2005, is divided into two sub-samples. For the whole sample period, it is consistently found that SHBI, SZAI and SZBI are not supported by weak-form EMH judging by the single variance ratio test statistics of R1 , R2 , S1 and S2 for all time intervals. Combined with the multiple variance ratio test of CHODE, SHAI is weakly rejected at the 10% level by R1∗ , and SHBI, SZAI, SZBI are all strongly rejected at the 1% level by R1∗ , R2∗ and S1∗ . Furthermore, SZAI and SZBI are also rejected by S1∗ at the 5% level. A more robust ﬁnding is that weak-form EMH is invalid for the B-shares on Shanghai exchange and both AB-shares on Shenzhen exchange, but is supported for the Shanghai A-share market throughout the sample period. Nevertheless, these results are not completely consistent with the past literature. For the analyses of sub-samples, the conclusions during the ﬁrst sub-period are the same as the whole sample period. However, except for SHAI, opposite results are found for the second sub-sample than for the ﬁrst sub-sample. All stock indices reveal evidence of weak-form EMH, and the change in market efﬁciency may result from the regulatory transformation associated with the release of investment restrictions on the B-share market. However, the rolling results demonstrate that the explanation for the efﬁciency of Chinese stock markets cannot be entirely ascribed to the inﬂuence of releasing investing restrictions on the B-share market. It is certain that the improvement of market efﬁciency can be explained by the increase of liquidity and maturity accompanied by deregulation and liberalization. Acknowledgement The author would like to thank Jorge Belaire-Franch for kindly sharing their computer codes. References Belaire-Franch, J., & Contreras, D. (2004). Ranks and signs-based multiple variance ratio tests. Working paper of Department of Economic Analysis, University of Valencia. Belaire-Franch, J., & Opong, K. K. (2005a). Some evidence of random walk behavior of Euro exchange rates using ranks and signs. Journal of Banking and Finance, 29, 1631–1643. Belaire-Franch, J., & Opong, K. K. (2005b). A variance ratio test of the behaviour of some FTSE equity indices using ranks and signs. Review of Quantitative Finance and Accounting, 24, 93–107.

4 The A-share markets have now been opened to international institutional investors under certain conditions, most notably that each Qualiﬁed Foreign Institutional Investor (QFII) must have funds of at least US$10bn under management and that any capital invested and any related income cannot be repatriated for 3 years; several investors have already been approved.

J.-C. Hung / The Quarterly Review of Economics and Finance 49 (2009) 843–857

857

Cajueiro, D. O., & Tabak, B. M. (2006). The long-range dependence phenomena in asset returns: The Chinese case. Applied Economics Letters, 13, 131–133. Campbell, B., & Dufour, J.-M. (1997). Exact nonparametric tests of orthogonality and random walk in the presence of a drift parameter. International Economic Review, 38, 151–173. Chakravarty, S., Sarkar, A., & Wu, L. (1998). Information asymmetry, market segmentation and the pricing of cross-listed shares: Theory and evidence from Chinese A and B shares. Journal of International Financial Institutions and Money, 8, 325–356. Chan, K., Menkveld, A. J., & Yang, Z. (2006). Are domestic investors better informed than foreign investors? Evidence from perfectly segmented market in China. Series Research Memoranda 0004, Free University Amsterdam, Faculty of Economics, Business, Administration and Econometrics. Chow, K. V., & Denning, K. C. (1993). A simple multiple variance ratio test. Journal of Econometrics, 58, 385–401. Chui, C. W., & Kwok, C. Y. (1998). Cross-autocorrelation between A shares and B shares in the Chinese stock market. Journal of Financial Research, 21(3), 333–353. Darrat, A. F., & Zhong, M. (2000). On testing the random-walk hypothesis: A model-comparison approach. Financial Review, 35, 205-124. Dickey, D. A., & Fuller, W. A. (1979). Distribution of the estimators for autoregressive time series with a unit root. Journal of American Statistical Association, 74, 427–431. Dickey, D. A., & Fuller, W. A. (1981). Likelihood ratio statistics for autoregressive time series with a unit root. Econometrica, 49, 1057–1072. Fama, E. F. (1970). Efﬁcient capital markets: Review of theory and empirical work. Journal of Finance, 25(2), 383–417. Fama, E. F. (1991). Efﬁcient capital markets: II. Journal of Finance, 46(5), 1575–1617. Fiﬁeld, S. G. M., & Jetty, J. (2008). Further evidence on the efﬁciency of the Chinese stock markets: A note. Research in International Business and Finance, 22(3), 351–361. Hung, J. C., Lee, Y. H., & Pai, T. Y. (2009). Examining market efﬁciency for large- and small-capitalization of TOPIX and FTSE stock indices. Applied Financial Economics, 19(9), 735–744. Jarque, C. M., & Bera, A. K. (1987). A test for normality of observations and regression residuals. International Statistical Review, 55, 163–172. Kim, J. H., & Shamsuddin, A. (2008). Are Asian stock markets efﬁcient? Evidence from new multiple variance ratio tests. Journal of Empirical Finance, 15(3), 518–532. Laurence, M., Cai, F., & Qian, S. (1997). Weak-form efﬁcient and causality tests in Chinese stock markets. Multinational Finance Journal, 1(4), 291–307. Lee, C. F., Chen, G. M., & Rui, O. M. (2001). Stock returns and volatility in China’s stock markets. Journal of Financial Research, 24, 523–543. Li, C. M. (2003a). China: further evidence on the evolution of stock markets in transition economies. Scottish Journal of Political Economics, 50(3), 341–358. Li, C. M. (2003b). Time-varying informational efﬁciency in China’s A-share and B-share markets. Journal of Chinese Economics and Business Studies, 1(1), 33–56. Lima, E. J. A., & Tabak, B. M. (2004). Tests of the random walk hypothesis for equity markets: Evidence from China, Hong Kong and Singapore. Applied Economics Letters, 11, 255–258. Liu, C. Y., & He, J. (1991). A variance-ratio test of random walks in foreign exchange rates. Journal of Finance, 46(2), 773–785. Lo, A. E., & Mackinlay, A. C. (1988). Stock market prices do not follow random walks: Evidence from a simple speciﬁcation test. Review of Financial Studies, 1(1), 41–66. Lo, A. E., & Mackinlay, A. C. (1989). The size and power variance ratio test in ﬁnite samples: A Monte Carlo investigation. Journal of Econometrics, 40, 203–238. Long, D. M., Payne, J. D., & Feng, C. (1999). Information transmission in the Shanghai equity market. Journal of Financial Research, 22, 29–45. Lu, C., Wang, K., Chen, H., & Chong, J. (2007). Integrating A- and B-share markets in China: The effects of regulatory policy changes on market efﬁciency. Review of Paciﬁc Basin Financial Markets and Policies, 10(3), 309–328. Luger, R. (2003). Exact non-parametric tests for a random walk with unknown drift under conditional heteroscedasticity. Journal of Econometrics, 115, 259–276. Ma, S., & Barnes, M. L. (2001). Are China’s stock markets really weak-form efﬁcient? Centre for international Economic Studies, Adelaide University Australia, Discussion Paper No. 0119. Available from the World Wide Web at: http://www.economics.adelaide.edu.au/mbarnes/China ME.doc. Mookerjee, R., & Yu, Q. (1999). An empirical analysis of the equity markets in China. Review of Financial Economics, 8, 41–60. Seddighi, H. R., & Nian, W. (2004). The Chinese stock exchange market: Operations and efﬁciency. Applied Financial Economics, 14, 785–797. Sjöö, B., & Zhang, J. (2000). Market segmentation and information diffusion in China’s stock markets. Journal of Multinational Financial Management, 10, 421–438. Su, D., & Fleisher, B. (1998). Risk, return and regulation in Chinese stock market. Journal of Economics and Business, 50(3), 239–256. Tabak, B. M. (2003). The random walk hypothesis and the behaviour of foreign capital portfolio ﬂows: The Brazilian stock market case. Applied Financial Economics, 13(5), 369–378. Wang, J., Burton, B. M., & Hannah, G. (2004). Further evidence on the comparative efﬁciency of Chinese ‘A’ and ‘B’ shares. Studies in Economics and Finance, 22(2), 20–40. Wright, J. H. (2000). Alternative variance-ratio tests using ranks and signs. Journal of Business and Economic Statistics, 18(1), 1–9. Yang, J. (2003). Market segmentation and information asymmetry in Chinese stock market: A VAR analysis. Financial Review, 38(4), 591–609. Zhang, Y., & Yu, D. (1994). China’s emerging securities markets. Columbia Journal of World Business Focus Issue: Emerging Capital Market, 29(2), 112–121.