AB shares
Upcoming SlideShare
Loading in...5
×
 

AB shares

on

  • 340 views

 

Statistics

Views

Total Views
340
Views on SlideShare
340
Embed Views
0

Actions

Likes
0
Downloads
0
Comments
0

0 Embeds 0

No embeds

Accessibility

Upload Details

Uploaded via as Microsoft Word

Usage Rights

© All Rights Reserved

Report content

Flagged as inappropriate Flag as inappropriate
Flag as inappropriate

Select your reason for flagging this presentation as inappropriate.

Cancel
  • Full Name Full Name Comment goes here.
    Are you sure you want to
    Your message goes here
    Processing…
Post Comment
Edit your comment

AB shares AB shares Document Transcript

  • A Microstructure Approach to the Domestic and Foreign Shares in the Chinese Stock Markets by Yea-Mow Chen and Yan He* San Francisco State University January 12, 2001 ________________________ * Contact author: Yan He, Department of Finance, College of Business, San Francisco State University, 1600 Holloway Avenue, San Francisco, CA 94132, Phone: 415-338-2600, Fax: 415-338-0997, Email: yhe@sfsu.edu.
  • A Microstructure Approach to the Domestic and Foreign Shares in the Chinese Stock Markets Abstract This paper finds that the average bid-ask spread is significantly higher for the foreign shares (B shares) than the domestic shares (A shares) traded in the Chinese stock markets. To explain the spread disparity between the A and B shares, we estimate the informed trading cost for each stock by using daily data. A measure of the informed trading cost is developed based on the model of George, Kaul, and Nimalendran (1991). Our test results show that the B-share market in China contains higher informed trading cost than the A-share market. When the informed trading cost is held constant, the bid- ask spread disparity between the A and B shares disappears. Therefore, the higher bid-ask spread in the B-share market can be attributed to the higher informed trading cost faced by B-share investors. JEL Classification: G15 Keywords: Ownership Restriction, Chinese Markets, Market Microstructure, Bid-Ask Spread, Informed Trading Cost 2
  • A Microstructure Approach to the Domestic and Foreign Shares in the Chinese Stock Markets 1. Introduction In many emerging stock markets, foreign investors face restrictions on owning domestic shares. It is widely documented that ownership restrictions result in price differentials among classes of shares. Bailey and Jagtiani (1994) find that foreign investors generate significant price premiums over domestic investors, using data from the Stock Exchange of Thailand. Stulz and Wasserfallen (1995) construct a demand function to explain why shares available to foreign investors sell at a premium, using data from Switzerland. Foerster and Karolyi (1999) test foreign stocks listed in the U.S., and their results support for market segmentation hypothesis and investor recognition hypothesis. Recently, Henry (2000) and Bekaert and Harvey (2000) report positive reactions in a country's equity market when the country liberalizes its stock ownership and capital market, such as reduce foreign ownership restriction and relax currency control. The newly developed Chinese stock markets implement foreign stock ownership restrictions. Class A shares are domestic shares and class B shares are foreign shares. The percentage of shares owed by foreign investors is constrained for individual firms. Throughout the first half of the 1990s, B shares were traded at a discount compared with A shares, and B-share returns were higher than A-share returns. Su (1999) explains the return premiums on the foreign-owned B shares in the Chinese stock markets by testing a one-period capital asset price model (CAPM). He argues that foreign investors are more risk-averse than domestic investors and so CAPM holds. In addition, Chui and Kwok 3 View slide
  • (1998) find that the returns on B shares lead the returns on A shares, which induces an asymmetric positive cross-autocorrelation between the returns on B and A shares. They argue that A- and B-share investors have differential access to information, and information will more often reach the B-share market before it reaches the A-share market in China. The ownership restriction in the Chinese stock markets creates two distinct groups of investors for a single firm: the domestic and the foreign investors. These two groups of investors have different access to information and bear different risk. Therefore, they may face different information asymmetry risk in submitting orders, take different strategies in posting bids and offers, and bear different costs in executing trades. In all, the ownership restriction may influence the behaviors of A- and B-share investors and cause bid-ask spread disparity between the A- and B-share markets, though the two markets have similar quoting and trading mechanisms. This paper adopts a microstructure approach to investigate the execution costs of trading A and B shares in the Chinese stock markets. It finds that the average bid-ask spread is significantly higher for the foreign shares (B shares) than the domestic shares (A shares). To explain the spread disparity between the A and B shares, we estimate the informed trading cost for each stock by using daily data. Our results indicate that the B- share market contains higher informed trading cost than the A-share market and the higher bid-ask spread of B shares can be attributed to the higher informed trading cost of B shares in the Chinese stock markets. The remainder of the paper is organized as follows. Section 2 introduces the quoting and trading structure of the Chinese stock markets and presents preliminary test 4 View slide
  • results on bid-ask spreads. Section 3 constructs a model for measuring informed trading cost. Section 4 discusses empirical methodology. Section 5 describes data samples. Section 6 presents and analyzes empirical results. Finally, Section 7 summarizes the findings of this paper. 2. Trading structure and bid-ask spreads The Shanghai Stock Exchange and the Shenzhen Stock Exchange are the major stock exchanges in China. The markets of the two exchanges are continuous and order- driven. There are no designated dealers or market makers on the exchanges. The mechanism of these exchanges is similar to that of the Stock Exchange of Hong Kong (SEHK).1 All of the order flows on the Chinese stock exchanges must be displayed on computer terminals viewable by investors on and off the exchanges. Limit orders are the only order type permitted on the exchanges. A buy limit order must give the bid price and number of shares to be purchased. A sell limit order must give the ask price and number of shares to be sold. When a trader submits a limit order, the order is entered into a computer system where orders can be matched and executed automatically. The system prioritizes orders first by price and then by time. Bid prices are arranged in priority from highest to lowest, while ask prices are arranged in priority from lowest to highest. The spread between the highest (or best) bid and the lowest (or best) ask represents the bid- ask spread. If a buyer requires an immediate fill, he will submit a limit bid that is high enough to touch the lowest posted ask. This buy order will then be executed at the best ask. We 1 Brockman and Chung (2000) describe the mechanism of the SEHK. 5
  • call this trade a buyer-initiated trade. If a seller requires an immediate fill, he will submit a limit ask that is low enough to touch the highest posted bid. This sell order will then be executed at the best bid. We call this trade a seller-initiated trade. For each trade, the party who initiates the trade bears the execution cost, while the counter party gains from the bid-ask spread as compensation for his expected loss to informed traders and as compensation for providing liquidity. This paper focuses on the investigation of the A- and B-share bid-ask spreads. Table 1 presents the summary statistics on bid-ask spreads for 82 pairs of A and B shares traded in the Chinese stock markets during January 1, 1998 to May 1, 2000. The average half-spread is 0.13% for the A shares and 1.38% for the B shares. Hence, the quoted percentage bid-ask spread is significantly higher on the B-share market than the spread on the A-share market, with a t-value of 7.97. Now the question is why the B-share market has higher execution costs than the A-share market in China. This paper will explore the characteristics of the domestic (A- share) and the foreign (B-share) investors, examine the informed trading cost component of bid-ask spreads, and explain the disparity in bid-ask spreads between the A and B shares. 3. The model of informed trading cost Bid-ask spread can be decomposed into an order-processing cost component and an informed trading cost component. The order-processing cost component is regarded as the compensation to the trader for providing liquidity services. The informed trading cost component exists because an uninformed trader may trade with an informed trader who 6
  • possesses private information. This cost component is regarded as the compensation to the uninformed trader for his expected loss to the informed trader. In practice, this cost component is approximated by the revision in the trader's expectation of the value of a stock resulting from the submission of an order. Stoll (1989) and George, Kaul and Nimalendran (1991) develop covariance models to estimate the proportion of bid-ask spread due to informed trading cost. Their models assume that all the stocks in a market have the same proportion of bid-ask spread due to informed trading cost. Based on their models, we can only obtain one single estimate for the informed trading cost in the market. In this paper, we extend the model of George, Kaul and Nimalendran (1991) to estimate the informed trading cost for each individual stock. We now present the model. Let Pt be the last transaction price in day t. The last transaction can be a buyer- or seller-initiated trade. Let Qt be the indicator for the buyer- seller classification of Pt. Qt is equal to +1 for a buyer-initiated trade and -1 for a seller- initiated trade. For each trade, the party who initiates the trade bears the execution cost, while the counter party gains from the bid-ask spread. For a buyer-initiated trade, the transaction price would be at the best offer. For a seller-initiated trade, the transaction price would be at the best bid. Here, the best bid and offer refer to the best quotes displayed right before the transaction. After each trade, rational investors will incorporate new information into their bids and offers. Conditional on a buyer-initiated trade, investors will revise their expectation of the stock's value upward. Conditional on a seller-initiated trade, investors will revise their expectation of the stock's value downward. The revision in the investor's 7
  • expectation of the stock's value serves as a proxy for the informed trading cost. The difference between the transaction price and the revised value serves as a proxy for the order processing cost. Let Vt be a stock's revised value after the last trade in day t. Based on Vt, the bid and offer prices will be updated immediately. Let HSPRt be the updated half-spread. We use π to denote the order processing cost component of HSPRt. So, 1-π denotes the informed trading cost component of HSPRt. Thus, the order processing cost, π*HSPRt, can be approximated by the difference between Pt and Vt. We have Pt = Vt + π*HSPRt*Qt . (1) Equation (1) is consistent with the model of George, Kaul, and Nimalendran (1991). Since Vt is unobservable, we will find an observable variable as a proxy for Vt. According to Huang and Stoll (1997), quote midpoint is adjusted relative to the fundamental value of a stock on the basis of accumulated inventory in order to induce inventory-equilibrating trades. Since there are no dealers or market makers on the Shanghai and Shenzhen Stock Exchanges, traders have no obligation to make a market or to keep certain levels of stock and cash inventories. Thus, the inventory control cost can be treated as zero and so the quote midpoint equals the fundamental value of the stock. Let Mt be the midpoint of the best bid and offer prices immediately following Pt. We have Vt = Mt. Thus, equation (1) can be rewritten as Pt = Mt + π*HSPRt*Qt. (2a) Similarly, in day t-1, we have Pt-1 = Mt-1 + π*HSPRt-1*Qt-1. (2b) Subtract (2b) from (2a), we have 8
  • Pt - Mt - (Pt-1 - Mt-1) = π*(HSPRt*Qt - HSPRt-1*Qt-1), (3) where Pt, Mt, Pt-1, Mt-1, HSPRt, and HSPRt-1 are observable, but Qt is unobservable. We take absolute values on both sides of (3) to eliminate the trade indicator variable Qt.2 Thus, |Pt-Mt-(Pt-1-Mt-1)| = π*|HSPRt±HSPRt-1|, (4) where either |HSPRt-HSPRt-1| or |HSPRt+HSPRt-1| is computed for day t. If the transactions in day t and day t-1 have the same side of trade initiation, |HSPRt-HSPRt-1| is calculated. If the transaction in day t has the opposite side of trade initiation as the transaction in day t-1, |HSPRt+HSPRt-1| is calculated.3 The parameter π denotes the order processing cost component of bid-ask spreads, which is restricted in the range of 0 and 1. The parameter 1-π denotes the informed trading cost component of bid-ask spreads. 4. Empirical methodology 4.1. Estimate informed trading cost Based on (4), we develop the following model to estimate parameter π for each stock: |Yt| = a + π*|HSPRt ± HSPRt-1| + et, (5) where Yt and HSPRt can be in dollar or percentage terms. Yt in dollar terms is defined as the dollar change from day t-1 to day t in the difference between the last transaction price 2 If intra-day quote and trade records are available, we can estimate Qt by using the method of Lee and Ready (1991). However, it is difficult to obtain intra-day quote and trade records in the Chinese stock markets. Thus, we simplify equation (3) by taking absolute values on both sides to eliminate variable Qt. 3 The judgment for the same or opposite side of initiation in two continuous days is discussed in Section 4. 9
  • and the quote midpoint following the last transaction price, and Yt in percentage terms is defined as the dollar change from day t-1 to day t divided by the quote midpoint of the last bid and ask in day t.4 HSPRt in dollar terms is defined as the last half-spread in day t, and HSPRt in percentage terms is defined as the HSPRt in dollar terms divided by the quote midpoint of the last bid and ask in day t.5 For each t, either |HSPRt-HSPRt-1| or | HSPRt+HSPRt-1| is calculated, depending on the same or opposite side of trade initiation in days t and t-1. The judgment for the same or opposite side of initiation in two continuous days is based on the restriction 0<π<1. If |Pt-Mt-(Pt-1 - Mt-1)| is less than | HSPRt-HSPRt-1|, we assume the same side of trade initiation in days t-1 and t so that | HSPRt-HSPRt-1| is calculated for day t. If |Pt-Mt-(Pt-1 - Mt-1)| is equal to or larger than | HSPRt-HSPRt-1|, we assume the opposite side of trade initiation in days t-1 and t so that | HSPRt+HSPRt-1| is calculated for day t. In this way, we roughly satisfy the restriction 0<π<1 and minimize the standard deviation of the π estimate. In order to estimate parameter π, we regress |Yt| against |HSPRt±HSPRt-1|. Two potential problems may appear in the regression analyses. First, the dependent variable and/or the independent variable may be serially autocorrelated, which induces serial correlation in the regression residuals. Second, the regression residuals are likely to be conditionally heteroskedastic. Therefore, the t-statistics are computed with the Newey- West (1987) correction for heteroscedasticity and serial correlation in the error terms. Based on (5), we obtain the estimate for π, which represents the proportion of the quoted spread due to the order processing cost. Then, we calculate 1-π, which represents the proportion of the quoted spread due to the informed trading cost. Define 4 Yt in dollar terms = Pt-Mt-(Pt-1-Mt-1), and Yt in percentage terms = [Pt-Mt-(Pt-1-Mt-1)]/Mt. 5 HSPRt in percentage terms = HSPRt in dollar terms / Mt. 10
  • INFO = 1- π, (6) where INFO is an estimate for the informed trading cost component of bid-ask spread, and INFO can be INFO1 or INFO2. INFO1 is obtained from the regression test based on Yt and HSPRt in dollar terms, while INFO2 is obtained from the regression test based on Yt and HSPRt in percentage terms. 4.2. Explain spread disparity We attempt to explain the cross-sectional variation in bid-ask spreads, and to ascribe the difference in bid-ask spreads between the A and B shares to the difference in informed trading costs. As equation (7) shows, the bid-ask spreads are regressed against the information variables. HSPRi = λ0 + λ1INFOi - D(λ0` + λ1`INFOi) + ei, (7) where i represents an individual stock that can be either domestically owned or foreign owned, HSPRi refers to the average percentage half-spread for stock i, INFO refers to the average informed trading cost for stock i, and D is a dummy variable that is equal to 1 for domestically owned shares (A shares) and 0 for foreign owned shares (B shares). Here, INFO can be INFO1 or INFO2. Parameters λ0` and λ1` measure the differences in the intercept and coefficient between the A and B shares. The dummy intercept λ0` will be used to test whether the bid-ask spread for A shares is lower relative to the spread for B shares when the informed trading cost is held constant. In addition, we employ the Generalized Method of Moments (GMM) to estimate (7) to account for heteroskedasticity in the error term (ei). 11
  • 5. The data Data are obtained from the Great China Economic Data Base. The sample period is from January 1, 1998 to May 1, 2000. The data sample contains a total of 82 pairs of A- and B-share stocks traded on the Shanghai and Shenzhen Stock Exchanges in China. Daily data are collected from the database. The daily price is the last transaction price of each trading day. The daily bid and ask are the last posted bid and ask of each trading day. The last posted quotes are typically subsequent to the last transaction price on any given day. George, Kaul, and Nimalendran (1991) argue that the only occasions on which the transaction and quotes are measured simultaneously is when market-at- close orders are placed, and there is evidence to suggest that such orders are quite infrequent. Table 2 provides summary statistics on the characteristics of selected stocks. First, the A- and B-share samples are significantly different in total shares (SHT) as well as in publicly owned shares (SHP). In the Chinese stock markets, A shares include public, institute, and state shares, whereas B shares are all publicly owned shares. A shares are domestic shares, whereas B shares are foreign shares. For an individual firm, the number of outstanding A shares on average is much more than the number of outstanding B shares with a t-value of –6.33. In contrast, the number of publicly owned A shares is much less than the number of publicly owned B shares with a t-value of 3.71. Second, the A- and B-share samples are significantly different in trading volume (VOLUME) and turnover rate (TR). Turnover rate is defined as VOLUME divided by SHP. The daily trading volume and the turnover rate of a firm's A-share stock are much 12
  • higher than those of the firm's B-share stock with t-values of -7.12 and -16.58 respectively. Thus, A shares are much more frequently traded than B shares. According to Easley, Kiefer, O'Hara, and Paperman (1996), frequently traded stocks have low informed trading costs and so their bid-ask spreads are low. Thus, it is expected that A shares have lower informed trading costs than B shares and A-share bid-ask spreads are narrower than B-share spreads. Third, the A- and B-share samples are similar in daily price returns (RET). The daily price returns of the A and B shares are insignificantly different at the five-percent level with a t-value of 0.65.6 Therefore, the return premiums on B shares have disappeared in recent years. The domestic (A-share) and foreign (B-share) shares of a firm have begun to generate similar returns.7 This phenomenon is consistent with fundamental valuation of stocks since the A- and B-share stockholders of a firm are entitled to the same future cash flows of the firm. Finally, the A- and B-share samples are different in volatility of daily returns (VRET). The return volatility of a firm's A-share stock is lower than the return volatility of the firm's B-share stock with a t-value of 3.24. Thus, A-share traders may face lower level of price uncertainty than B-share traders, and the lower price uncertainty may be related to the lower risk of information asymmetry in the A-share market. 6 We also examine monthly returns and excess returns. It is found that the monthly price returns and the monthly excess price returns of the A and B shares are insignificantly different at the five-percent level with t-values of -1.50 and -1.08 respectively. In specific, the average monthly price return is 0.021 for the A shares and 0.026 for the B shares. The average monthly excess price return is 0.019 for the A shares and 0.022 for the B shares. The monthly excess price return is calculated as the monthly price return minus the monthly riskfree rate. For A shares, the monthly riskfree rate is computed by using 3-month RMB interest rates. For B shares, the monthly riskfree rate is computed by using 3-month T-bill rates. 7 Su (1999) finds that B-share returns are significantly higher than A-share returns, using data from the Chinese stock markets during the mid 1990s. In contrast, this study uses data during the late 1990s. 13
  • Overall, the A- and B-share stocks generate similar daily and monthly returns. The differences between the A and B shares lie in their quoting and trading characteristics. For example, the A shares are more frequently traded than the B shares, and the A-share returns are less volatile than the B-share returns. These imply that the A- and B-share investors may face different risk of information asymmetry and incur different execution costs. 6. Empirical results 6.1. Estimation of informed trading costs We estimate the order processing cost component (π) for each stock by conducting regression tests with the Newey-West correction for serial correlation and heteroskedasticity in the error terms. Table 3 presents the summary results of the regression tests. When Yt and HSPRt in dollar terms are employed, the average estimate for π is 0.689 for the A shares and 0.574 for the B shares. The estimations are statistically significant at the five-percent level with an average t-value of 13.34 for the A shares and 11.00 for the B shares. The average adjusted R-square is 67% for the A-share regression and 54% for the B-share regression. When Yt and HSPRt in percentage terms are employed, the average estimate for π is 0.681 for the A shares and 0.585 for the B shares. The estimations are statistically significant at the five-percent level with an average t- value of 13.05 for the A shares and 11.75 for the B shares. The average adjusted R-square is 65% for the A-share regression and 54% for the B-share regression. Based on the π estimates, we compute the informed trading costs for each stock. Table 4 provides summary statistics on the informed trading costs. INFO1 is calculated 14
  • based on the π estimate from the dollar term regression, while INFO2 is calculated based on the π estimate from the percentage term regression. The average INFO1 is 0.311 for the A-share sample and 0.426 for the B-share sample, and the difference between the two samples is significant at the five-percent level with a t-value of 2.23. Similarly, the average INFO2 is 0.319 for the A-share sample and 0.415 for the B-share sample, and the difference between the two samples is significant at the ten-percent level with a t-value of 1.91. Therefore, the B-share sample has significantly higher informed trading cost component than the A-share sample. The result indicates that the B-share investors face higher risk of information asymmetry than the A-share investors. The higher information content of B shares may result from the ownership restriction in the Chinese stock markets. The restrictive ownership regulation segregates investors into two distinct groups. In the A-share market, public shares are traded mostly by individual investors who are not equipped with sophisticated investment skills. These investors tend to select stocks based on rumors or market sentiments rather than private information. Thus, the buy and sell orders in the A-share market are more likely to come from uninformed traders. In contrast, B shares are traded by foreign investors who are equipped with sophisticated portfolio management skills and who have quick access to information. Thus, the buy and sell orders in the B-share market are more likely to come from informed traders. So, the B-share market contains higher risk of information asymmetry than the A-share market, and the B-share investors face higher informed trading costs than the A-share investors. 15
  • 6.2. Explanation of spread disparity Since the A- and B-share samples differ in characteristics related to the microstructure of quoting and trading, we first examine whether these differences can explain the bid-ask spread disparity between the two samples. Table 5 provides the GMM test results. The bid-ask spreads are regressed against the stock feature variables SHT, SHP, VOLUME, TR, RET, and VRET respectively. The dummy intercept λ0` measures the average amount that the spread of B shares exceeds that of the A shares after accounting for a sample characteristic. The results show that the λ0` estimates are all positive and significant. Hence, the differences in sample characteristics cannot explain the bid-ask spread disparity between the A- and B-share samples. Next we examine the effect of informed trading cost on bid-ask spreads. Table 6 reports the GMM test results. As shown, the information variable (INFO1 or INFO2) explains about 76% to 79% of the cross-sectional variation in the quoted spreads (HSPR). Consistent with microstructure theory, the informed trading cost of the B shares has a positive relation with the quoted spreads of the B shares. The higher the informed trading costs, the larger the bid-ask spreads. In addition, the effects of informed trading costs on bid-ask spreads are different between the A- and B-share stocks. The constant term λ0` measures the mean difference in bid-ask spreads between the A- and B-share samples after controlling for the informed trading costs. As shown, λ0` is not significantly different from zero at the five-percent level in all tests. When the informed trading costs are held constant, the t-value for the spread difference is 0.55 for INFO1, and 1.37 for INFO2. Hence, the spread disparity between the A and B shares can 16
  • be explained by the difference in the informed trading costs. The larger bid-ask spread of B shares can be attributed to the higher informed trading costs faced by foreign investors. 7. Conclusions There have been arguments over the price returns of the domestic and foreign shares in the Chinese stock market. Previous studies find that the foreign shares (B shares) have significantly higher returns than the domestic shares (A shares), using data of the mid 1990s. However, this study finds that the return difference has disappeared, using data of the late 1990s. Given similar returns of A and B shares, this study adopts a microstructure approach to explore the difference in quoting and trading behaviors between the two classes of shares. We find that the average quoted spread is significantly higher for the foreign shares (B shares) than the domestic shares (A shares). In order to explain the spread disparity between the A and B shares, we estimate the informed trading cost for each stock by using daily data. A measure of the informed trading cost is developed based on the model of George, Kaul, and Nimalendran (1991). After controlling for the informed trading costs, we find that the spread disparity between the A and B shares disappears. Our results indicate that the higher bid-ask spreads of B shares can be attributed to the higher informed trading costs faced by B-share investors. 17
  • References Bailey, Warren and Julapa Japtiani, 1994, Foreign ownership restrictions and stock prices in the Thai market, Journal of Financial Economics 36, 57-87. Bekaert, Geert and Campbell R. Harvey, 2000, Foreign speculators and emerging equity markets, Journal of Finance 55, 565-613. Brockman, Paul and Dennis Y. Chung, 2000, Informed and uninformed trading in an electronic, order-driven environment, Financial Review 35, 125-146. Chui, Andy C.W. and Chuck C.Y. Kwok, 1998, Cross-autocorrelation between A Shares and B Shares in the Chinese Stock Market, Journal of Financial Research, 21, 333-354. Easley, David, Nicholas M. Kiefer, Maureen O'Hara, and Joseph B. Paperman, 1996, Liquidity, information, and infrequently traded stocks, Journal of Finance 51, 1405-1436. Foerster, Stephen R. and G. Andrew Karolyi, 1999, The effects of market segmentation and investor recognition on asset prices: Evidence from foreign stocks listing in the United States, Journal of Finance 54, 981-1013. George, Thomas J., Gautam Kaul, and M. Nimalendran, 1991, Estimation of the bid-ask spread and its components: A new approach, Review of Financial Studies 4, 623-656. Henry, Peter B., 2000, Stock market liberalization, economic reform, and emerging market equity prices, Journal of Finance 55, 529-564. Huang, Roger D., and Hans R. Stoll, 1997, The components of the bid-ask spread: A general approach, Review of Financial Studies 10, 995-1034. Lee, C. M. C., and M. J. Ready, 1991, Inferring trade direction from intraday data, Journal of Finance 46, 733-746. Lee, C. M. C., and M. J. Ready, 1991, Inferring trade direction from intraday data, Journal of Finance 46, 733-746. Newey, Whitney K. and Kenneth D. West, 1987, A simple, positive semi-definite, heteroskedasticity and autocorrelation consistent covariance matrix, Econometrica 55, 703-708. Stoll, Hans R., 1989, Inferring the components of the bid-ask spread: Theory and empirical tests, Journal of Finance 44, 115-134. 18
  • Stulz, Rene M. and Walter Wasserfallen, 1995, Foreign equity investment restrictions, capital flight, and shareholder wealth maximization: Theory and evidence, Review of Financial Studies 8, 1019-1057. Su, Dongwei, 1999, Ownership restrictions and stock prices: Evidence from Chinese markets, Financial Review 34, 37-56. 19
  • Table 1 Summary Statistics on Bid-Ask Spreads This table provides summary statistics on bid-ask spreads for 82 pairs of A and B shares in the Chinese stock markets. The sample period is from 1/1/1998 to 5/1/2000. HSPR refers to the average half-spread in percentage terms. Parameter A-share B-share Mean difference (B-A) t-value HSPR Mean 0.13 1.38 7.97* (%) Std. Dev. 0.03 0.51 Min. 0.07 0.45 Median 0.12 1.27 Max. 0.26 2.42 *: a significance level of five percent or better for a two-tailed test. 20
  • Table 2 Summary Statistics on Stock Characteristics This table provides summary statistics on stock characteristics for 82 pairs of A and B shares in the Chinese stock markets. The sample period is from 1/1/1998 to 5/1/2000. SHT refers to the total shares of a security. SHP refers to the publicly owned shares of a security. For A shares, SHP does not include institute or state shares. For B shares, SHP equals SHT. VOLUME is the daily number of shares transacted. TR is the turnover rate, defined as VOLUME divided by SHP. RET is the daily price returns. VRET is the volatility of daily total price returns. Parameter A-share B-share Mean difference (B-A) t-value SHT Mean 311 132 -6.33* (1,000,000s) Std. Dev. 211 90 Min. 72 22 Median 260 108 Max. 1,107 415 SHP Mean 70 132 3.71* (1,000,000s) Std. Dev. 62 90 Min. 8 22 Median 46 108 Max. 298 415 VOLUME Mean 983 335 -7.12* (1000s) Std. Dev. 724 288 Min. 234 34 Median 779 246 Max. 4,554 1,410 TR Mean 1.93% 0.25% -16.58* Std. Dev. 0.85% 0.12% Min. 0.39% 0.02% Median 1.78% 0.23% Max. 4.05% 0.54% RET Mean 0.11% 0.11% 0.65 Std. Dev. 0.09% 0.14% Min. -0.08% -0.14% Median 0.12% 0.09% Max. 0.33% 0.54% VRET Mean 3.24% 5.16% 3.24* Std. Dev. 0.57% 1.51% Min. 2.17% 3.54% Median 3.17% 4.81% Max. 5.12% 11.60% *: a significance level of five percent or better for a two-tailed test. Table 3 21
  • Summary of Regression Tests with the N-W Correction This table presents the summary of regression tests with Newey-West correction for 82 pairs of A and B shares in the Chinese stock markets. The sample period is from 1/1/1998 to 5/1/2000. For each stock, parameter π is estimated by testing on the following model. |Yt| = a+ π*|HSPRt ± HSPRt-1| + et, where Yt in dollar terms = Pt - Mt - (Pt-1 - Mt-1), Yt in percentage terms = [Pt - Mt - (Pt-1 - Mt-1)]/Mt. Pt is the last transaction price in day t. Mt is the midpoint of bid and ask quotes after Pt in day t. HSPRt is the half-spread in day t. HSPRt can be in dollar or percentage terms. When the transaction in day t has the opposite side of initiation as the transaction in day t-1, |HSPRt + HSPRt-1| is calculated. When the transaction in day t has the same side of initiation as the transaction in day t-1, |HSPRt - HSPRt-1| is calculated. Panel A presents the cross-sectional mean of the test results based on Yt and HSPRt in dollar terms. Panel B presents the cross-sectional mean of the test results based on Yt and HSPRt in percentage terms. Parameter π represents the order processing cost, and 1-π represents the informed trading cost. The parameter and the adj. R2 are estimated from the OLS regression. The t-statistics are computed with the Newey-West correction for serial correlation and heteroskedasticity in the error terms. Panel A. Yt and HSPRt in dollar terms A-share B-share π mean 0.689 0.574 t-value on π mean 13.34* 11.00* Adj. R2 mean 67% 54% Panel B. Yt and HSPRt in percentage terms A-share B-share π mean 0.681 0.585 t-value on π mean 13.05* 11.75* Adj. R2 mean 65% 54% *: A significance level of five percent or better for a two-tailed test. 22
  • Table 4 Summary Statistics on Informed Trading Costs This table provides summary statistics on informed trading costs for 82 pairs of A and B shares in the Chinese stock markets. The sample period is from 1/1/1998 to 5/1/2000. INFO1 is obtained from the regression test based on variables in dollar terms. INFO2 is obtained from the regression test based on variables in percentage terms. Parameter A-share B-share Difference (B-A) t-value INFO1 Mean 0.311 0.426 2.23* Std. Dev. 0.149 0.074 Min. 0.140 0.339 Median 0.302 0.425 Max. 0.481 0.513 INFO2 Mean 0.319 0.415 1.91** Std. Dev. 0.143 0.064 Min. 0.151 0.345 Median 0.312 0.405 Max. 0.491 0.507 *: A significance level of five percent or better for a two-tailed test. **: A significance level of ten percent for a two-tailed test. 23
  • Table 5 GMM Tests of Bid-Ask Spreads on Stock Characteristics This table provides the GMM test results for 82 pairs of A and B shares in the Chinese stock markets. The sample period is from 1/1/1998 to 5/1/2000. The testing model is HSPRi = λ0 + λ1Xi - D(λ0` + λ1`Xi) + ei, where i =1, 2, …, 154, HSPRi refers to the average percentage half-spread for stock i, Xi represents the stock characteristic for stock i, X can be SHT, SHP, VOLUME, TR, RET, or VRET, and D is a dummy variable that is equal to 1 for A shares and 0 for B shares. SHT refers to the total shares of a security. SHP refers to the publicly owned shares of a security. For A shares, SHP does not include institute or state shares. For B shares, SHP equals SHT. VOLUME is the daily number of shares transacted. TR is the turnover rate, defined as VOLUME divided by SHP. RET is the daily price returns. VRET is the volatility of daily total price returns. λ0` measures the mean difference in bid-ask spreads between the A and B shares after controlling for a stock characteristic. B-share Difference (B-A) Adj. R2 X CONST X λ1 λ 0` λ1` X=SHT Coeff. -0.003 1.682 -0.003 84% t (-7.38*) (22.15*) (-7.31*) X=SHP Coeff. -0.003 1.672 -0.003 84% t (-7.38*) (22.05*) (-6.75*) X=VOLUME Coeff. -0.001 1.550 -0.001 82% t (-6.78*) (23.10*) (-6.70*) X=TR Coeff. -40.913 1.390 -42.522 75% t (-0.88) (11.66*) (-0.92) X=RET Coeff. 244.032 1.011 230.839 86% t (8.37*) (19.55*) (7.90*) X=VRET Coeff. 18.038 0.360 16.970 82% t (6.97*) (2.44*) (6.44*) *: A significance level of five percent or better for a two-tailed test. 24
  • Table 6 GMM Tests of Bid-Ask Spreads on Informed Trading Cost Variables This table provides the GMM test results for 82 pairs of A and B shares in the Chinese stock markets. The sample period is from 1/1/1998 to 5/1/2000. The testing model is HSPRi = λ0 + λ1INFOi - D(λ0` + λ1`INFOi) + ei, where i = 1, 2, …, 154, HSPRi refers to the average percentage half-spread for stock i, INFOi refers to the informed trading cost for stock i, INFO can be INFO1 or INFO2, and D is a dummy variable that is equal to 1 for A shares and 0 for B shares. INFO1 is obtained from the regression test based on variables in dollar terms. INFO2 is obtained from the regression test based on variables in percentage terms. λ0` measures the mean difference in bid-ask spreads between the A and B shares after controlling for an informed trading cost variable. B-share Difference (B-A) Adj. R2 INFO CONST INFO λ1 λ 0` λ1` INFO=INFO1 Coeff. 2.678 0.117 2.699 79% t (5.26*) (0.55) (5.29*) INFO=INFO2 Coeff. 1.805 0.500 1.825 76% t (2.11*) (1.37) (2.14*) *: A significance level of five percent or better for a two-tailed test. 25