Market Segmentation and Asset Pricing: Evidence from the Chinese Stock Market in 1992-2020

(1)

Market Segmentation and Asset Pricing: Evidence from the

Chinese Stock Market in 1992-2020

Thesis supervisor: Prof.Dr. R.M. Salomons

(2)

Market Segmentation and Asset Pricing: Evidence from the

Chinese Stock Market in 1992-2020

Abstract

This article studies the discount of B shares in the Chinese stock market in the past two decades. While, the segmentation of Chinese stock market and the drivers of discounts are re-investigated. The results show that the price of B shares is still lower than that of A shares. However, it can be seen that the discount of B shares has been greatly reduced. In addition, the results of panel data analysis show that the price difference is still mainly due to the lack of liquidity in the B-share market, but there are also some differences that can be explained by information asymmetry. This is in line with the research results of 20 years ago, indicating that the liquidity problem still exists in the Chinese B-share market.

(3)

1. Introduction

The stock markets are segmented to deal with the global integration of capital markets in many emerging capital markets. As pointed out in plenty of international asset pricing literature, integration causes asset prices to rise by allowing more investors access to assets while reducing their anticipated returns through lowered equilibrium risk premium. Merton (1987) took the view that increasing the size of the company's investor base will lead to the reduction in investors' anticipated returns but increase the market value of the stock in companies. In these market segments, there are usually two types of stocks, one is restricted stock that can only be held as well as traded by local investors, and the other is unrestricted stock held by local and foreign investors. Although shareholders of these two stocks basically have the same rights, in most markets, the transaction value of restricted stocks is lower than that of unrestricted stocks.

Chinese stock market began to be segmented in the 1990s. As economic reforms started from the late 1970s, China has made outstanding achievement economically as evidenced by an annual growth rate of as high as 10%. However, the development of the stock market in China fails to keep pace with the economic reforms. Since its establishment in the early 1990's, the Chinese stock market has exhibited an exponential growth. The Shanghai Stock Exchange (SHSE) and the Shenzhen Stock Exchange (SZSE) opened in the early 1990s, which marks part of the effort made by the Chinese government to cultivate capital markets purposed to serve both domestic and foreign investors. China's A-share market is limited to local investors, by contrast, the B-share market is originally accessible to overseas Chinese or foreign investors. A-shares are in the unit of Renminbi (RMB) and B-shares are in the unit of US dollar for SHSE and Hong Kong dollar for SZSE.

The advantages of international diversification have attraction to free capital across national borders. It makes investors pay more for foreign stocks than for domestic ones. As mentioned earlier, premium trading of non-restricted stocks is associated with restricted stocks in most market. However, the situation in China is very different. The fact that China’s unrestricted stocks are traded at discount prices is often referred to as the “China B-share discount puzzle”(Chan, Menkveld, and Yang, 2008). Over the years, academics have proposed various explanations for the price differences that appear in the Chinese market segmentation, the most influential of which are the information asymmetry hypothesis, the differential demand hypothesis, the differential risk hypothesis as well as the liquidity hypothesis.

(4)

that the relative illiquidity of B-shares may be primarily responsible for the price disparity. Subsequently, the B-share market was opened to investors of China in March 2001. In view of the liberalization of the Chinese stock market in 2001 and 2002, Rui, Wu, and Lee (2007) re-examined the determinants of B-share discounts, finding that information asymmetry degree and B-share discounts had eased after the market opened.

As China's economy has undergone a fast-paced development over the past 20 years and a wide range of policy reforms have been carried out, it is of practical significance to update this study. Thus, this paper replicates and updates Chen et al. (2001) 's research. Specifically, this article re-examines the pricing of dual-listed Chinese stocks theoretically and empirically. Firstly, based on the expansion of the sample period, a re-investigation of the discount of B shares was conducted. Secondly, over the most recent two decades characterized by significant changes in the Chinese economy, 21 new companies have been dual-listed on both A shares and B shares. Thus, the discount of B shares of newly listed companies was also investigated. Thirdly, the driving factors (four hypotheses mentioned above) of the discount of B shares for whole sample from 1992 to 2020 were re-analyzed.

The results are consistent with what is illustrated in previous literature, indicating that there is still a discount for B shares compared with A shares. However, the B-share discount has presented an obvious decline since ownership restrictions were lift in 2001. Due to the removal of ownership restrictions, many domestic investors flood the unrestricted stock market, thus enhancing information flows from unrestricted market to the restricted market. Although the discount has been reduced, it still exists and varies widely across firms after the event. In addition, as indicated by the panel data analysis results, price difference is still mainly due to B-share market lacking liquidity in the B-share market. Generally speaking, based on these results, although the information asymmetry hypothesis can partly explain the discount of B shares, the liquidity problem still serves as the major factor of the discount, indicating that liquidity problems remain in China's B-share market.

The rest of the paper falls into four parts. The second part summarizes the historical literature on this topic. The third part focus on describing data and research methods. The fourth part describes the empirical results, and the last part concludes the paper.

2. Literature review

(5)

2.1 China Securities Market

Chinese companies are permitted to issue five stock types on the domestic market, including tradable A shares, B shares, state-owned shares, employee shares, and legal person shares. There are two main stock exchanges in China, i.e. SHSE and SZSE, which were founded in 1990 and 1991, respectively. Both stock exchanges offer Class A common shares and Class B common shares issued by the domestic companies in China, which are specific to citizens and institutions of China. At the time of listing, the company is required to issue 25% of its total outstanding shares at minimum as A shares (Rui et al., 2007). In contrast, B shares can be purchased and traded only by overseas investors prior to 2001. Besides, its issuance is purposed to attract investment from overseas. As referred to before, the B shares are in the unit of US dollars for SHSE and Hong Kong dollars for SZSE.

2.2 Price Discounts

There are plenty of emerging capital markets imposing ownership restrictions on domestic stocks owned by foreigners. As mentioned previously, there are two major types of stocks in emerging capital markets. One is the restricted stocks (A shares) which can be held exclusively by nationals and the other is the unrestricted stocks (B shares) which can be held by investors at home and abroad. The market segmentation mainly aims at attracting investment from overseas, while it is needless to worry about losing ownership control of overseas investors. The information and expectations of domestic and overseas investors can be seen from the price of A shares and B shares, respectively (Sun and Tong, 2000). As the right and obligations of two share types are the same, theoretically they should present the same value. Nevertheless, the current research has revealed that generally the trading price of unrestricted stocks is higher than that of restricted stocks.

This study focuses on market segmentation in China, but is well rooted in the emerging market literature on this topic. By demonstrating that the difference in beta risk can assist in explaining the price premium in the Finnish market, Hietala (1989) found evidence supporting this argument. Domowitz et al. (1998), they found that market segmentation in Mexico caused the information difference across markets. There is a clear premium for unrestricted stocks in Mexico. Bailey and Jagtiani (1994) demonstrated that the factors influencing the premium of Thailand's unrestricted stocks include liquidity issues and information availability. Bailey, as found by Chung and Kang (1999), foreign stocks showed a price premium across all other market segments compared with Chinese stocks.

(6)

that other domestic investors cannot when investing in domestic stocks. In addition, the rate of return required by overseas investors will be lower compared to domestic investors. The different required rates of return result in the variation in stock prices, for which the required yield of B shares is supposed to be lower, which will lead to a premium for B shares instead of a discount. Bailey (1994) conducted analysis of eight Chinese B-share stocks from 1992 to 1993 and discovered that B-shares presented a lower trading price compared with A-shares. Chen et al. (2001) have also validated this phenomenon.

Indeed, plenty of scholars have substantiated the discount in the Chinese B-share market. Companies of China strictly issue A shares and B shares to local and overseas investors, respectively, which will be explained in following part. Nevertheless, A shares are traded at a higher price relative to B shares. Therefore, the arguments and factors believed to help in accounting for the phenomena that happened in other markets seemingly cannot explain China’s situation. In the following section, the factors driving the discount of the B-share market will be analyzed separately.

2.3 Sources of Price Differences

Asymmetric Information Hypothesis

In the stock market of China, market information asymmetry is frequently referred to for explaining the considerable discounting of B-shares. Some earlier literature attributed this difference to the different levels of information about the value of assets held by foreign and domestic investors. Overseas investors tend to have less information as a result of language barriers and the difference in accounting standards. Consequently, for compensating for the information weakness compared with domestic investors, a risk premium is demanded by overseas investors (Stulz and Wasserfallen, 1995; Chakravarty, Sarkar, and Wu, 1998).

Chakravarty et al. (1998) took the view that there is a discount in B-share prices as foreign investors have realize that it is a greater challenge to access and assess the relevant information to local Chinese enterprises than domestic investors. According to the theoretical model involving asymmetric information together with market segmentation, B shares could be traded at a lower price compared with A shares. Moreover, it is a commonplace to share manipulation and insider trading, with no codes in place as protection for investors. Due to information costs, overseas investors avert holding unknown stocks, thus causing B-shares to be discounted in China (Merton, 1987).

(7)

between the A-share and B-share returns in the Chinese stock market is used to prove this hypothesis.

In light of the dissenting viewpoints, Chen et al. (2001) examined it by conducting empirical analysis. It involved (a) causation tests regarding returns as well as return volatility, (b) time-series evidence, and (c) proof with regard to the impact of a cross-sectional corporate scale. A-share returns can benefit the prediction of B-A-share returns if the information asymmetry hypothesis is validated, or vice versa. Their results show that the asymmetric information hypothesis is incapable of explaining the difference between the two share types in terms of the price.

Subsequently, Chan, Menkveld, and Yang (2008) conducted further research by developing a number of measures of information asymmetry. As confirmed, measures of information asymmetry can help explain the cross-sectional difference of discounting. Besides, Rui et al. (2007) demonstrated that the information disadvantage and the discount of B-shares were reduced after the opening of B-share market to the domestic investors in March 2001.

Differential demand hypothesis

Premised on Stulz and Wasserfallen (1995) models, according to the differential demand assumptions, domestic investors and overseas investors hold different demand functions specific to domestic stocks. It was revealed that domestic companies differentiate domestic investors from foreign investors to obtain the maximum corporate value. In this case, it is best for the domestic companies to impose limit on the issuance of shares to overseas investors. If overseas investors hold a rigid demand for domestic shares, companies can charge them a higher price of share, thus successful in price-discrimination. Undoubtedly, companies in China are not allowed to issue shares in this way. Actually, national quotas are set by the National Development and Reform Commission (NDRC) and the China Securities Regulatory Commission (CSRC) regarding the number of new shares that are issued on an annual basis in China. Moreover, companies are subject to special approval from the CSRC for the issuance of foreign shares (Sun and Tong, 2000).

(8)

Nevertheless, bank deposits are disadvantageous in interest rates, which leaves Chinese investors with less opportunity cost and causes them to request less return. An alternative, feasible explanation was made by Sun and Tong (2000) from an economic angle. The explanation is dependent on an extraordinary situation in the China regarding foreign shares. Unlike other marketplaces, there are various markets in China for the investment from overseas (liquid H-share market and Hong Kong's “red chip” market). The demand for B shares will show great flexibility if other markets show sufficient attraction to foreign capital, thus causing the decrease in the equilibrium price of B shares under a certain stock supply. To be specific, as speculated, China’s B-share discount phenomenon is ascribed to overseas investors faced with a more flexible demand curve. In comparison, the domestic investors in China are confronted by a less flexible one. A crucial factor in the flexibility of demand lies in the substitutes. The flexibility of demand for that goods will improve if there are more substitutes for a good. Thus, it is necessary to lower the price of B shares for attracting overseas investors. As discussed by Chen et al. (2001), overseas investors may be more likely to make diversified investments and have more flexible demand for Chinese stock trading.

As suggested by the differential demand hypothesis, discounts decrease as overseas demand increases, which means that discounts are considered as a negative function specific to foreign investor demand compared with the issued shares. Measures of supply and demand may not present a clear difference in an equilibrium. However, in the case that the outstanding shares are mainly depended on firms’ supply of shares instead of investors’ demand, the discounts between firms will be a positive function of the proportion of outstanding B shares in the total outstanding shares.

As an empirical proxy for cross-company measurement of relative demand, Chen et al. (2001) applied the proportion of the issued B shares in the total issued shares that was proposed by Domowitz, Glen, and Madhavan (1997). The research results contradict the differential demand hypothesis. As referred to previously, however, more B shares will lead to a decrease in the B-share price and an increase in the A-B-share premium in the case that outstanding B-shares are mainly depended on firms’ supply of shares instead of investors’ demand.

Liquidity hypothesis

(9)

share market. The relatively illiquid B shares generate a higher expected return and illiquid B shares is priced lower for compensating investors for the higher transaction costs. Chen and Xiong (2001) found that Chinese market more supported the trading activity hypothesis, demonstrating that trading price of limited institutional shares is only 78%-86% of that of unrestricted shares.

On the other hand, Domowitz et al. (1997) gathered proof regarding the factor of information availability in the Mexican market rather than the factor of liquidity. As found, share supply contributes to the explanation power. Liquidity is limited to causing transitory impact on price premium.

Differential risk hypothesis

Finally, lots of academics held a consensus viewpoint regarding Chinese investors’ speculation fever. It can cause a potential a speculative fever as investment alternatives are limited that leads to a relatively rigid demand for A shares. Considering the capital restrictions, Chinese investors are incapable of diversifying overseas investment, which contributes to a deviation between domestic investors and overseas investors with regard to the risk exposure. This is because the risk assessment on A shares and B shares is performed against various investment criteria (Chinese market for A shares and world market returns for B shares). Besides, the findings of Ma (1996) show that Chinese investors are risk enthusiasts, and the price difference depends on the investor's attitude towards risk.

Nevertheless, Eun, Janakiramanan, and Lee (2001) argue that the B shares discount is positively associated with the covariance risk exhibited by B-shares using the Morgan Stanley world market index. In spite of this, the negative correlation with the covariance risk exhibited by A-shares was not evidenced by the Chinese market index. Fernald and Rogers (2002) also did not find evidences showing the correlation between the B-share discounts and B-share or A-share covariance risks.

(10)

3. Data and Research method

3.1 Data collection and Summary Statistics

Data collection

First, with a replication of the study by Chen et al. (2001), the details of the 68 listed companies that issue both A and B shares can be acquired on the Shanghai Stock Exchange and the Shenzhen Stock Exchange, respectively. The daily price of A and B shares can be collected at Yahoo Finance, as well as the number of shares outstanding and trading volume. Some of the mentioned companies have been delisted in the past two decades, and relevant data can be collected from CRSP1_database.

Second, compared with the research by Chen et al. (2001), the sample size has increased, 21 new listed companies have been added, and the observation period has been extended to 1992-2020. The information of new listed companies can be found on the Shanghai Stock Exchange and the Shenzhen Stock Exchange. The final sample consists of 44 stocks on the SHSE and 45 stocks on the SZSE. The final sample period is from 1992 to 2020. Moreover, the monthly price of A shares and B shares, the number of issued shares, the monthly trading volume and market value of A shares and B shares were collected in the CRSP database for panel data analysis. The panel data analysis uses monthly data from 1992 to 2020.

Summary Statistics

In the past two decades, the development of Chinese stock market is rapidly. The treasury bonds and the state-owned enterprise bonds issued in the 1980s represented the rise of Chinese capital market. In addition, the Shenzhen Stock Exchange and the Shanghai Stock Exchange were successively established in the following ten years. This study covers and extends the previous research by Chen et al. (2001), the observation interval was extended to 28 years. Since the B shares of SHSE are denominated in US dollars, the B shares of SZSE are denominated in Hong Kong dollars, and A shares of these two exchanges are quoted in Renminbi, the B shares of both exchanges are converted to RMB here.

(11)

Table 1: Descriptive Statistics of Chinese Stock Exchanges

Note: SHSEA represents the A shares of the Shanghai Stock Exchange, SHSEB represents the B shares of the Shanghai Stock Exchange, SZSEA represents the A shares of the Shenzhen Stock Exchange, and SZSEB represents the B shares of the Shenzhen Stock Exchange. The data collected annual data from 1992 to 2020. Yearly Market Capitalization is in billions, Yearly Trading Amount and Yearly Trading Volume are in millions.

Variable Obs Mean Std.Dev. Min Max

SHSEA (1992-2020)

Yearly Return (including cash

dividend reinvestment) 28 0.196 0.528 -0.58 1.995

Yearly Return (without cash

Yearly Market Capitalization

(Bil,RMB) 28 11563.18 1.828 63.168 34683.08

Yearly Trading Amount

(Mil,RMB) 28 2.01E+07 2.076 23264.83 1.32E+08

Yearly Trading Volume

(Mil,Shares) 28 1853146 2.246 386.102 1.01E+07

SHSEB (1992-2020)

(Bil,RMB) 28 79.189 2.655 3.629 198.156

(Mil,RMB) 28 88470.04 2.669 10739.64 334515

(Mil,Shares) 28 11090.05 2.778 46.754 39340.24

SZSEA (1992-2020)

(Bil,RMB) 28 5295.854 2.428 50.340 18321.71

(Mil,RMB) 28 1.56E+07 2.582 41741.97 9.35E+07

(Mil,Shares) 28 1297847 2.508 1766.318 5899656

SZSEB (1992-2020)

(Bil,RMB) 28 63.109 2.558 1.612 143.312

(Mil,RMB) 28 55295 2.848 1470.201 2.18E+05

(12)

As shown in Table 1, the total market capitalization increased significantly since 1992. Over the past 28 years, the average annual trade volume of A shares on the Shanghai Stock Exchange was 1853146 million shares, and the average annual trade volume on the Shenzhen Stock Exchange was 1297847 million shares. However, the average annual transaction volume of B shares on the Shanghai Stock Exchange is 11090 million shares, and the average annual transaction volume of B shares on the Shenzhen Stock Exchange is only 10095.7 million shares. On the whole, the average annual transaction volume of the A-share market has been nearly 150 times greater than that of the B-share market from 1992 to 2020. Likewise, the average annual trading amount of the A-share market has been approximately 248 times higher than that of the B-share market. Yet, the average annual return of B shares is slightly higher than that of A shares during 1992-2020. Overall, as suggested in Table 1, Chinese A-share market and B-share market have developed well over the past two decades. In addition, the expansion speed and scale of the B-share market are inferior to the share market and less active than A-share markets.

Figure1: Comparison of A Shares and B Shares.

Figure 1 further shows the difference between A-share market and B-share market observed every five years. Panel A shows the annual returns of the A and B shares of the of the two stock exchanges. Obviously, the return of A shares is higher than that of B shares in most periods. As mentioned earlier, foreign investors can obtain diversified returns. Therefore, the rate of return required by foreign investors will be lower. In addition, it is clear that the Chinese stock market suffered heavy losses in 2008 and 2018, which can be attributed to the global financial crisis and the China–United States trade war. Panel B shows the gap between the yearly market capitalization of the A-share market and the B-share market. What can be found is that the market value of the A-share market is much larger than the market value of the B-share market, indicating the imbalance in the segmentation of the Chinese stock market. In general, the difference between the A-share market and the B-share market is huge, reflected in terms of

-80% -60% -40% -20% 0% 20% 40% 60% 2019 2018 2013 2008 2003 1998 1993

A: Difference in Yearly Return (with divedend)

SHSE(A) SHSE(B) SZSE(A) SZSE(B)

-1 -0.98 -0.96 -0.94 -0.92 2019 2018 2013 2008 2003 1998 ₁₉₉₃ B: Difference in Yearly Market Capitalization

SHSE SZSE

(13)

rate of return, market value, liquidity and activity. Although ownership restrictions have been lifted since 2001, this phenomenon has not eased over time.

In order to ensure the consistency of the study, the choice of proxy variables in this paper follows Chen et al. (2001). In order to conduct panel data analysis, six variables were calculated and regulated by the price of A shares and B shares, the number of shares outstanding, the transaction volume and the market cap. Table 2 shows the descriptive statistics of these proxy variables. Where ((PA – PB) / PB) represents the discount level of B shares relative to A shares of the same company, (SOb / SOab) represents the ratio of issued B shares to total issued shares, (Vb / Vab) represents the ratio of B share trading volume to total share trading volume, (FSab) refers firm size, (SDa/ SDb) is the ratio of the variance of A-share returns to the variance of B-share returns, and relative turnover (Vb/SOb) / (Va/SOa) as empirical proxies that measures the relative liquidity between companies. Meanwhile, the different proxy variables will test different hypotheses, and related specific information will be shown in the next section.

Note: The data is monthly data. Where (Discount) is the discount of B share price relative to A share price, (SOb/SOab) is the ratio of issued B shares to total shares outstanding，(Vb/Vab) is the ratio of B share trading volume to total trading volume, FS (Mil) is firm size, (SDa/SDb) is the ratio of the variance of A shares and B shares, (Vb/SOb)/(VaSOa) is relative turnover rate.

Table 2: Descriptive Statistics of Panel Data

Variable Obs Mean Std.Dev. Min Max

Panel A. Total Sample

Discount 22746 0.518 0.210 -0.614 0.963 SOb/SOab 22746 0.278 0.110 0.002 0.543 Vb/Vab 22746 0.553 0.335 0.001 0.989 FS (Mil) 22746 7.736 12.933 0.263 253.351 SDa/SDb 22746 0.973 0.153 0.641 1.373 (Vb/SOb)/(VaSOa) 22746 0.400 1.156 0.031 106.134

Panel B. Shenzhen Stock Exchange

Discount 10745 0.502 0.200 -0.609 0.941 SOb/SOab 10745 0.251 0.113 0.032 0.537 Vb/Vab 10745 0.847 0.092 0.454 0.989 FS (Mil) 10745 8.504 16.598 0.263 253.351 SDa/SDb 10745 0.886 0.094 0.671 1.075 (Vb/SOb)/(VaSOa) 10745 0.403 1.368 0.024 106.134

Panel C. Shanghai Stock Exchange

(14)

As shown in Table 2, the average discount of B shares relative to A share in the Chinese stock exchange market is 51.8%, and the discount of B shares relative to A shares on the SHSE is slightly higher than that of SZSE. The Shanghai Stock Exchange achieves the highest ratio of the average B-share outstanding shares to the total outstanding shares, which is 30.2%. Shenzhen Stock Exchange achieves the maximum average ratio of B-share trading volume to stock trading volume at 84.7%. It is therefore suggested that the B shares market in the Shenzhen Stock Exchange exhibit better liquidity. Furthermore, the two exchanges have similar average relative turnover rates, which are 40.3% and 39.7% respectively.

3.2 Methodology

Price Discount

In order to comparing the discounts in A and B. The discount is defined as ((PB – PA) / PA), which represents the percentage of price premium. PB and PA represent the daily prices of these two shares of the same company, respectively. When ((PB – PA) / PA) is bigger than zero, the B shares are traded at a premium and vice versa.

By replicating and expanding the study by Chen et al. (2001), this study also provides causation tests in returns, over time evidence and proof on the impact of a cross-sectional corporate scale as they did. First, a causal test is performed on the returns to check the

asymmetric information hypothesis. This process can be interpreted as if using the past events 𝑥 to predict 𝑦 is more accurate than using the mean square error without 𝑥, then it can be said that there is a one-way causal relationship between 𝑥 and 𝑦 (Granger, 1969). Therefore, if there is a two-way causal relationship between the two stocks, then the assumption of asymmetric information holds. In addition, the AR model (binary autoregression) is used to test the causal relationship between the return of A shares and B shares.

𝑥_& = 𝛼₎+ + 𝛼_,𝑥_&-, . ,/0 + + 𝛽_,𝑦_&-, . ,/0 + 𝜀_&, 𝑦_&= 𝛾₎+ + 𝛾_,𝑥_&-, . ,/0 + + 𝛿_,𝛾_&-, . ,/0 + 𝜂_&.

Where xt is A share returns, yt is B share returns.

(15)

For the first equation, If the coefficients 𝛽_, are statistically significant, the B share return resulted in A share return. Conversely, if the coefficients 𝛽, are statistically insignificant, which

means we should reject the hypothesis null, the conclusion will be reversed. Similarly, for the second equation, if the coefficients of 𝛾_, coefficient will jointly differ from zero, then A shares return Granger-causes B shares returns. There is a two-way relationship between A share returns and B share returns if both 𝛽_, and 𝛾_, are different from zero.

The Panel Data Analysis

To verify the rest of the mentioned hypotheses, a panel data analysis conducted by Domovitz, Glen, and Madhavan (1997) is used in this study. They consider that the unobserved firm effect may be very important, it is most likely related to the observable relationship in the model. Obviously, it may be related to variables, such as the ratio of circulating B shares to A shares, market value, dividends, also the premium itself. Therefore, in this case, the results of the OLS and GLS estimates will be biased and inconsistent. In view of the particularity of the industry involved, a general method of moments (GMM) estimator can be used here. This method uses explanatory variables in the form of deviations from the mean as a tool rather than a standard regression estimator to eliminate endogenousness. This technique can obtain an effective estimate by correcting the standard error of the unknown form of heteroscedasticity in the time-varying error component. The panel data method is used to test hypotheses. The regression model is as follows: Model 1: 8𝑃:_𝑃− 𝑃< : =,,& = 𝑏)+ 𝑏08 𝑃_:− 𝑃_< 𝑃_: =_,,&-0+ 𝑏?( 𝑆𝑂_< 𝑆𝑂_:<),,&+ 𝑏D( 𝑉_< 𝑉_:<),,&+ 𝑏F(𝐹𝑆:<),,&+ bI8 𝑆𝐷_: 𝑆𝐷_<= + 𝜀_,,& K𝑃𝑎_𝑃−𝑃𝑏

𝑎 M_𝑖,𝑡is the discount of B shares relative to A shares. (PQR

PQSR),,& is the ratio of B shares issued to total share issued. (TR

TSR),,& represents the ratio of B-share traded volume to total traded volume. (𝐹𝑆_𝑎𝑏)_𝑖,𝑡 represents company size, and KPUS

PURMis the ratio of A-share return volatility to B-share return volatility.

𝑏) is the constant.

(16)

Model 2: 8𝑃:− 𝑃< 𝑃_: =_,,& = 𝑏)+ 𝑏08 𝑃:− 𝑃< 𝑃_: =_,,&-0+ 𝑏?( 𝑆𝑂< 𝑆𝑂_:<),,& + 𝑏D( 𝑉_< 𝑆𝑂_< 𝑉_: 𝑆𝑂_: )_,,& + 𝑏_F(𝐹𝑆:<),,&+ bI8 𝑆𝐷: 𝑆𝐷_<= + 𝜀_,,& K𝑃𝑎_𝑃−𝑃𝑏

𝑎 M_𝑖,𝑡is the discount of B shares relative to A shares. (PQR

PQSR),,& is the ratio of B shares issued to total shares issued. (

VR WXR

VS WXS

)_,,& is the relative turnover rate.

(𝐹𝑆_𝑎𝑏)_𝑖,𝑡 represents company size, and KPUS

PURMis the ratio of A-share return volatility to B-share return volatility.

𝑏₎ is the constant.

Differential demand hypothesis

According to Chen et al. (2001), to verify differential demand hypothesis, the ratio of issued B shares to total issued shares ((_PQPQR

SR),,& ) should be used as an empirical proxy for cross-company relative demand measurement. If the variable enters into regression significant negatively, which means that the demand curve slopes downward. This phenomenon can be explained as the increase in the supply of B shares will cause the price of B shares to fall. Thus, the best proxy variable to test this hypothesis is the ratio of outstanding B shares to the total issued share capital.

Liquidity hypothesis

In model one, followed by Chen et al. (2001) and Chuhan (1994), the ratio of B-share volume to total share volume ((_TTR

SR),,&) is used for the liquidity factor. In the liquidity hypothesis, the discount has a negative correlation with the liquidity of B-shares, indicating that the relative trading volume is inversely proportional to the price discount and directly proportional to the price premium. If the hypothesis holds, then the reason of B-share discount will be the lack of liquidity in the B-share market.

(17)

In model two, this paper also employed the relative turnover (( VR WXR

VS WXS

)_,,&) an empirical proxies for

cross-company relative liquidity indicators. According to Chen et al. (2001), turnover is defined as the ratio of traded volume to the number of shares issued, also known as relative trading volume. The turnover rate is used to measure liquidity.

Differential risk hypothesis

Previous research has shown that by observing and analyzing the volatility of stock returns, the speculative activities of Chinese investors can be discovered. Thus, the relative volatility KPU_PUS

RM is used as a proxy for the risk level. It can be explained that if a pair of A and B stocks share the same information of the company, then the volatility of any A share that exceeds the volatility of B shares can be regarded as speculative exchange. If a positive correlation between the level of risk and the discount of B shares is observed, it means that excessive speculation has led to a premium in the price of A shares relative to the price of B shares.

4. Empirical Results

4.1 Empirical Evidence on Price Discount

To roughly understand the difference between A shares and B shares in the past two decades, a general comparison of price discounts will be shown in the present section.

Fig. 2 shows the monthly discounts from the Shenzhen Stock Exchange over the past two decades. By replicating the study by Chen et al. (2001), the average discount of the 32 sample firms on the Shenzhen stock exchange is currently expressed as “Avg Discount 1992-2020”. “Avg New companies Discount 1992-2020” refers to the average discount of the 12 new listed companies from the Shenzhen stock exchange.

(18)

markets. Besides, it is worth mentioning that the discount has been greatly reduced in 2001, which is related to the Chinese stock market liberalization reforms. Rui, Wu, and Lee (2007) and Doukas and Wang (2013) provide empirical evidence for this argument. Moreover, the discount rose by nearly 10% between 2008 and 2013, probably due to the global financial crisis. Meanwhile, the trend of newly listed companies is identical to the original companies, whereas the discounts are smaller. This finding meets expectations since newly listed companies generally show better transparency.

Figure 2: The Discount of B-Shares Relative to A-Shares in SZSE.

Fig. 3 shows the monthly discounts of the Shanghai Stock Exchange from 1992 to 2020. Moreover, the average discounts of the 36 sample firms on the SHSE expressed as “Avg Discount 1992-2020” and the cross-sectional average of B-share discounts of the 9 new listed companies from the SHSE expressed as “Avg New Companies Discount 1995-2020”.

Compared with the Shenzhen Stock Exchange, the Shanghai Stock Exchange seems to be less stable. After the 20th century, the discounts plummeted by nearly 50%, which related to the unique regulatory changes in 2001. After removing restrictions on foreign ownership, B-share discounts fell sharply. Likewise, the assumption about information asymmetry has been confirmed, the discounts are getting smaller over time. Besides, the discounts for newly listed companies are also smaller than originals. Overall, the discount on the SHSE is smaller than SZSE, probably due to geographical location, language barriers, different accounting standards and differences in information sources. Chakravarty et al. (1998) interpreted this difference as less information about listed companies in Shanghai. The Shanghai Stock Exchange was dominated by state-owned enterprises earlier, and state-owned enterprises are more vulnerable to accounting and information disclosure standards.

-100% -80% -60% -40% -20% 0% 20% 40% 60% 1992/12/3 1995/12/3 1998/12/3 2001/12/3 2004/12/3 2007/12/3 2010/12/3 2013/12/3 2016/12/3 2019/12/3 SZSE Discount 1992-2020

(19)

Figure 3: The Discount of B-Shares Relative to A-Shares in SHSE.

4.2Empirical Evidence on Each Hypothesis

As mentioned above, a test of the causality tests in returns will be presented. First, this study replicates Chen et al. (2001)’s study by using the same sample data and same sample period to ensure the accuracy and consistency of the study. Second, it was re-analyzed after increasing the sample size and extending the sample period.

First of all, the study of Chen et al. (2001) was be replicated in this paper. Table 3 (Appendix) lists the result of causal relation tests based on the daily return of bivariate model from 1992-1997. The results show that, 15 stocks exhibited a one-way causal relationship between the return of A shares and B shares, and 2 stocks showed a one-way causal relationship between the returns of B shares and A shares on the Shanghai Stock Exchange. On the Shenzhen Stock Exchange, 7 stocks exhibited a one-way causal relationship between the return of A shares and B-shares, 7 stocks showed that B-share returns resulted in A-share earnings. However, only one stock from the Shanghai Stock Exchange showed a two-way causal relationship between A-share returns and B-A-share returns. In comparison with the study conducted by Chen et al. (2001), the results are slightly different. As revealed from their results, there are two stocks that display a feedback relationship between A share returns and B share returns. It is true that we all set the sampling period to 1992-1995, whereas the exact date of the initial observation may vary. In this study, the date that the identical company's A shares and B shares both have quotable

-100% -90% -80% -70% -60% -50% -40% -30% -20% -10% 0% 1992/7/1 1995/7/1 1998/7/1 2001/7/1 2004/7/1 2007/7/1 2010/7/1 2013/7/1 2016/7/1 2019/7/1 SHSE Discount 1992-2020

(20)

historical data is taken as the starting point. However, empirically, we both have concluded that there is no causal relationship between these two kinds of share returns from 1992 to 1997, which also shows that there is no information flow between A-share returns and B-share returns. The mentioned findings are inconsistent with the asymmetric information hypothesis.

Table 4 lists the results of the causality test based on a binary autoregression of daily returns after the sample size is expanded. The sample period is altered to 2000-2020. On the Shanghai Stock Exchange, 33 stocks exhibited a one-way causal relationship between A-share returns and B-share returns, and 22 stocks demonstrated a one-way causal relationship between the return of B shares and A shares. On the Shenzhen Stock Exchange, there are 26 stocks showing that share returns lead to B-share returns, while 28 stocks indicate that B shares lead to A-share returns. Of the 89 stocks, 31 stocks show the two-way relationship between these two kinds of stocks, of which seventeen come from SHSE and fourteen come from SZSE. Nine of twenty-one new companies show the feedback relationship between A share returns and B share returns. Compared with the study conducted by Chen et al. (2001), the results are changed dramatically. Nearly 50% of stocks exhibit one-way causality in both Shenzhen Stock Exchange and Shanghai Stock Exchange. 35% of stocks display two-way causality about returns. 42% of newly listed companies exhibited two-way causality between returns of A shares and B shares. The mentioned results show that in the past two decades, causality is identified between A share earnings and B share earnings, revealing that changes in A share returns can explain changes in B share returns. This finding is consistent with the two-way causal relationship between A-share returns and B-share returns predicted by the asymmetric information hypothesis. Moreover, combined with Fig. 2 and 3, evidence of the trend of B-share discounts over time is found as well. Though the main turning point occurred after the 20th century, it can be clearly found that the discount has been greatly reduced after the 20th century. The reduction in the discount on B shares can be explained as the changes in government regulations on ownership restrictions and information disclosure, more accounting information regarding listed companies can be obtained over time.

Table 4. The Result of Causal Relation Test (2000-2020).

ID Rb-Ra Signif. _Level _RelationCausal Ra-Rb Signif. _Level _RelationCausal

Panel A: Shenzhen Stock Exchange

(21)

200026 F(5,4997) 0.096 No F(5,4997) 0.000 Yes 200029 F(5,4997) 0.188 No F(5,4997) 0.002 Yes 200039 F(5,3246) 0.137 No F(5,3246) 0.005 Yes 200513 F(5,3502) 0.014 Yes F(5,3502) 0.000 Yes 200011 F(5,4997) 0.918 No F(5,4997) 0.000 Yes 200012 F(5,4997) 0.462 No F(5,4997) 0.003 Yes 200016 F(5,4997) 0.023 Yes F(5,4997) 0.364 No 200017 F(5,4997) 0.000 Yes F(5,4997) 0.004 Yes 200019 F(5,4997) 0.000 Yes F(5,4997) 0.004 Yes 200020 F(5,4997) 0.000 Yes F(5,4997) 0.012 Yes 200025 F(5,4997) 0.000 Yes F(5,4997) 0.026 Yes 200030 F(5,4997) 0.000 Yes F(5,4997) 0.746 No 200037 F(5,4997) 0.000 Yes F(5,4997) 0.006 Yes 200045 F(5,4997) 0.030 Yes F(5,4997) 0.007 Yes 200055 F(5,4997) 0.000 Yes F(5,4997) 0.122 No 200056 F(5,4998) 0.002 Yes F(5,4998) 0.444 No 200413 F(5,4998) 0.016 Yes F(5,4998) 0.015 Yes 200521 F(5,4998) 0.008 Yes F(5,4998) 0.001 Yes 200539 F(5,4998) 0.001 Yes F(5,4998) 0.056 No 200541 F(5,4998) 0.115 No F(5,4998) 0.639 No 200550 F(5,4998) 0.181 No F(5,4998) 0.028 Yes 200570 F(5,4998) 0.195 No F(5,4998) 0.000 Yes 200596 F(5,4998) 0.004 Yes F(5,4998) 0.623 No 200028 F(5,4998) 0.656 No F(5,4998) 0.000 Yes 200058 F(5,4998) 0.329 No F(5,4998) 0.000 Yes 200488 F(5,4762) 0.004 Yes F(5,4762) 0.596 No 200553 F(5,4998) 0.000 Yes F(5,4998) 0.000 Yes 200429 F(5,4998) 0.000 Yes F(5,4998) 0.012 Yes 200505 F(5,4998) 0.000 Yes F(5,4998) 0.248 No 200530 F(5,4998) 0.002 Yes F(5,4998) 0.000 Yes 200581 F(5,4998) 0.006 Yes F(5,4998) 0.017 Yes 200613 F(5,4998) 0.000 Yes F(5,4998) 0.019 Yes 200625 F(5,4998) 0.008 Yes F(5,4998) 0.083 No 200725 F(5,4723) 0.000 Yes F(5,4723) 0.000 Yes 200726 F(5,4737) 0.201 No F(5,4737) 0.001 Yes 200761 F(5,4998) 0.125 No F(5,4998) 0.001 Yes 200869 F(5,4779) 0.136 No F(5,4779) 0.455 No

Panel B: Shanghai Stock Exchange

(22)

900917 F(5,4998) 0.005 Yes F(5,4998) 0.043 Yes 900918 F(5,4998) 0.000 Yes F(5,4998) 0.581 No 900919 F(5,4998) 0.000 Yes F(5,4998) 0.000 Yes 900920 F(5,4998) 0.013 Yes F(5,4998) 0.000 Yes 900922 F(5,4998) 0.000 Yes F(5,4998) 0.008 Yes 900923 F(5,4998) 0.036 Yes F(5,4998) 0.706 No 900924 F(5,4998) 0.000 Yes F(5,4998) 0.030 Yes 900926 F(5,4998) 0.011 Yes F(5,4998) 0.253 No 900927 F(5,4998) 0.003 Yes F(5,4998) 0.000 Yes 900928 F(5,4998) 0.000 Yes F(5,4998) 0.161 No 900932 F(5,4998) 0.045 Yes F(5,4998) 0.029 Yes 900934 F(5,4998) 0.139 No F(5,4998) 0.346 No 900937 F(5,4998) 0.000 Yes F(5,4998) 0.000 Yes 900938 F(5,4998) 0.000 Yes F(5,4998) 0.000 Yes 900941 F(5,4998) 0.000 Yes F(5,4998) 0.005 Yes 900906 F(5,4849) 0.122 No F(5,4849) 0.067 No 900921 F(5,4998) 0.001 Yes F(5,4998) 0.000 Yes 900925 F(5,4998) 0.609 No F(5,4998) 0.000 Yes 900930 F(5,4570) 0.255 No F(5,4570) 0.000 Yes 900931 F(5,324) 0.246 No F(5,324) 0.841 No 900933 F(5,4998) 0.090 No F(5,4998) 0.000 Yes 900936 F(5,4649) 0.008 Yes F(5,4649) 0.131 No 900940 F(5,4998) 0.015 Yes F(5,4998) 0.007 Yes 900942 F(5,4998) 0.000 Yes F(5,4998) 0.887 No 900943 F(5,4690) 0.000 Yes F(5,4690) 0.000 Yes 900945 F(5,4998) 0.761 No F(5,4998) 0.826 No 900946 F(5,4998) 0.000 Yes F(5,4998) 0.130 No 900947 F(5,4739) 0.020 Yes F(5,4739) 0.710 No 900952 F(5,4998) 0.007 Yes F(5,4998) 0.001 Yes 900955 F(5,4670) 0.000 Yes F(5,4670) 0.889 No

Note: The model is:

Y𝑅&: 𝑅_&<[ = Y 𝑑0 𝑑?[ + Y 𝐴00(𝐿) 𝐴0?(𝐿) 𝐴?0(𝐿) 𝐴??(𝐿)[ _ 𝑅&-0: 𝑅_&-0< ` + Y 𝑢0& 𝑢?&[,

Where 𝐴,b(𝐿) = ∑Ie/0𝑎,b(𝑠)𝐿e-0, 𝑓𝑜𝑟 𝑖, 𝑗 = 1,2. The numbers in brackets in the F-statistic are the number of lags 𝑚 in the bivariate autoregression and the degree of freedom, respectively. Ra = the return of A shares, Rb = the return of B share. "Yes" stands for statistical significance and causality.

PanelDataAnalysis

(23)

Table 5. Panel Data Analysis of Model 1 (1992-1997).

Table 5 reports the results of regression on the entire sample from 1992 to 1997, in which relative trading volume was used to test the liquidity hypothesis. Panel A presents the results of the total sample, and panel B and C show the results of the Shanghai Stock Exchange and Shenzhen Stock Exchange respectively. The results of the panel analysis comply with those of the Chen et al. (2001)’s study. The coefficient of lag discount is statistically significant in all panel, indicating that there is a strong autocorrelation here. In all three panels, the coefficient of SOb/SOab is also positive and statistically significant, indicating a negative correlation between discount and demand. This result does not meet the prediction of the differential demand assumption. In addition, table 2 reports the panel regression of Model 2. In all panels, the coefficients of Vb/Vab (Table 5) and (Vb/SOb)/(VaSOa) (Table 6) are both negative and statistically significant. This finding confirmed the prediction of the liquidity hypothesis. While, although the difference risk hypothesis holds that there is a positive correlation between B share discount and risk level, it is not significant in panel C of Table 5 and the entire Table 6, indicating that the risk difference hypothesis has not been proven. In addition, the relationship between company size and discounts is positive and statistically significant, indicating that

Regression results (1992-1997)

Panel A: Total Sample

discount Coef. St.Err. p-value Sig

L.discount 0.651 0.012 0.000 *** SOb/SOab 0.688 0.065 0.000 *** Vb/Vab -0.146 0.009 0.000 *** FS (Mil) 0.014 0.002 0.000 *** SDa/SDb 0.167 0.048 0.001 *** Constant -0.119 0.042 0.004 *** Panel B: SZSE L.discount 0.793 0.016 0.000 *** SOb/SOab 0.111 0.065 0.087 * Vb/Vab -0.253 0.02 0.000 *** FS (Mil) 0.015 0.002 0.000 *** SDa/SDb -0.332 0.112 0.003 *** Constant 0.36 0.103 0.000 *** Panel C: SHSE L.discount 0.532 0.015 0.0000 *** SOb/SOab 0.293 0.119 0.0140 ** Vb/Vab -0.121 0.010 0.0000 *** FS (Mil) 0.011 0.001 0.0000 *** SDa/SDb -0.067 0.063 0.2900 Constant 0.331 0.087 0.0000 *** *** p<0.01, ** p<0.05, * p<0.1

(24)

companies with large market capitalization also have large discounts during 1992-1997. This finding does not provide empirical support for the information asymmetry hypothesis. Overall, the results achieved through the replication study comply with those of the original study, demonstrating that the discount of the B-share price is inversely associated with trading volume and relative turnover. This provides the greatest evidence support for the liquidity assumption.

L.discount 0.866 0.004 0.000 *** SOb/SOab 0.210 0.019 0.000 *** (Vb/SOb)/(VaSOa) -0.014 0.000 0.000 *** FS (Mil) 0.000 0.000 0.118 SDa/SDb 0.004 0.014 0.770 Constant 0.013 0.014 0.356 Panel B: SZSE L.discount 0.875 0.005 0.000 *** SOb/SOab 0.088 0.009 0.000 *** (Vb/SOb)/(VaSOa) -0.011 0.000 0.000 *** FS (Mil) 0.000 0.000 0.205 SDa/SDb -0.058 0.035 0.092 * Constant 0.079 0.031 0.012 ** Panel C: SHSE L.discount 0.519 0.015 0.000 *** SOb/SOab 0.385 0.120 0.001 *** (Vb/SOb)/(VaSOa) -0.009 0.003 0.002 *** FS (Mil) 0.013 0.001 0.000 *** SDa/SDb 0.014 0.064 0.826 Constant 0.170 0.087 0.050 * *** p<0.01, ** p<0.05, * p<0.1

Note: The data is monthly data. Where (L.discount) is the discount of B share price relative to A share price, (SOb/SOab) is the ratio of issued B shares to total shares outstanding， FS (Mil) is firm size, (SDa/SDb) is the ratio of the variance of A shares and B shares, (Vb/SOb)/(VaSOa) is relative turnover rate.

After expanding the sample size, the result of re-analyzing model 1 is listed in Table 7. After re-analyzing, the results are changed slightly. Similarly, the lag term (L.discount) is still statistically significant in all panels, which is consistent with the findings of Domowitz et al. (1997).

(25)

demand, the discount is down-regulated. The variable to look at is SOb/SOab, the ratio of B shares issued to total A and B shares issued. The regression estimate is significantly positive in all three panels. The regression estimate is significantly positive in all three panels, which complies with the hypothesis that the greater the supply of B shares relative to total outstanding shares, the smaller B shares discount. However, this special situation is interpreted by several scholars that the company determines the number of shares issued in China, instead of the needs of investors. In general, this result is consistent with the findings of Chen et al. (2001).

According to the liquidity hypothesis, the discount between companies is an inverse function of B share liquidity relative to A share liquidity. The coefficient of relative trading volume (Vb/SOb)/(VaSOa) is significantly negative in Table 7 for all three panels. This finding provides empirical support for the liquidity hypothesis. The negative correlation between the relative trading volume and the B-share discount proves that the active trading activity in the B-share market will reduce the B-share discount. It can be found that the liquidity problem is still the main problem in the Chinese B-share market, and the result is consistent with the findings of Bailey and Jagtiani (1994).

If the asymmetric information hypothesis holds, it should be observed that companies with high market capitalization have smaller discounts. Bailey and Jagtiani (1994) argue that companies with larger market capitalizations tend to attract foreign investors due to foreign investors can more easily obtain information about the company. The variable FS (Mil) represents the company's market value. The results show a significantly negative sign in panel A (Total sample), which is inconsistent with Chen et al. (2001)'s research. However, it is consistent with the assumption of asymmetric information, indicating that the discount of B shares is inversely related to the availability of Chinese company information over the past 20 years. Furthermore, this result also complies with those the existing causality test and the evidence of time-series (over time).

(26)

L.discount 0.869 0.004 0.000 *** SOb/SOab 0.359 0.019 0.000 *** (Vb/SOb)/(VaSOa) -0.099 0.003 0.000 *** FS (Mil) -0.003 0.001 0.008 *** SDa/SDb -0.019 0.014 0.175 Constant 0.014 0.014 0.322 Panel B: SZSE L.discount 0.838 0.005 0.000 *** SOb/SOab 0.246 0.01 0.000 *** (Vb/SOb)/(VaSOa) -0.15 0.004 0.000 *** FS (Mil) 0.013 0.001 0.000 *** SDa/SDb -0.107 0.033 0.001 *** Constant 0.077 0.031 0.012 ** Panel C: SHSE L.discount 0.864 0.005 0.000 *** SOb/SOab 0.471 0.026 0.000 *** (Vb/SOb)/(VaSOa) -0.088 0.003 0.000 *** FS (Mil) 0.000 0.002 0.886 SDa/SDb -0.045 0.019 0.019 ** Constant 0.002 0.023 0.947 *** p<0.01, ** p<0.05, * p<0.1

Note: The data is monthly data. Where (L.discount) is the discount of B share price relative to A share price, (SOb/SOab) is the ratio of issued B shares to total shares outstanding，(Vb/Vab) is the ratio of B share trading volume to total trading volume, FS (Mil) is firm size, (SDa/SDb) is the ratio of the variance of A shares and B shares.

The re-analysis results of model 2 are listed in Table 8. In Model 2, the relative trading volume is replaced by the relative turnover, and similar results are achieved. The coefficient of relative turnover is negative and of statistical significance in all three panels. This provides a strong proof of the liquidity hypothesis.

(27)

L.discount 0.869 0.004 0.000 *** SOb/SOab 0.219 0.019 0.000 *** (Vb/SOb)/(VaSOa) -0.014 0.000 0.000 *** FS (Mil) 0.001 0.001 0.258 SDa/SDb 0.005 0.014 0.748 Constant 0.006 0.014 0.696 Panel B: SZSE L.discount 0.875 0.005 0.000 *** SOb/SOab 0.106 0.010 0.000 *** (Vb/SOb)/(VaSOa) -0.011 0.000 0.000 *** FS (Mil) 0.007 0.001 0.000 *** SDa/SDb -0.031 0.035 0.377 Constant 0.037 0.032 0.250 Panel C: SHSE L.discount 0.853 0.005 0.000 *** SOb/SOab 0.359 0.025 0.000 *** (Vb/SOb)/(VaSOa) -0.024 0.001 0.000 *** FS (Mil) 0.006 0.002 0.000 *** SDa/SDb -0.014 0.019 0.460 Constant -0.016 0.023 0.487 *** p<0.01, ** p<0.05, * p<0.1

(28)

5. Conclusion

(29)

Reference

Amihud, Y., Mendelson, H., 1986. Asset pricing and the bid-ask spread. Journal of financial Economics, 17, 223-249.

Bailey, W., Chung, Y., Kang, J., 1999. Foreign ownership restrictions and equity price premiums: what drives the demand for cross-border investments? Journal of Financial and Quantitative Analysis 34, 489–511.

Bailey, W., Jagtiani, J., 1994. Foreign ownership restrictions and stock prices in the Thai capital market. Journal of Financial Economics 36, 57–87.

Bailey, W., 1994. Risk and return on China’s new stock markets: Some preliminary evidence. Pacific-Basin Finance Journal 2, 243–260.

Chakravarty, S., Sarkar, A., and Wu, L.F., 1998. Information asymmetry, market segmentation and pricing of cross-listed shares: theory and evidence from Chinese A and B shares. Working paper. Purdue University.

Chan, K., Menkveld, A.J., Yang, Z., 2008. Information asymmetry and asset prices: evidence from China Foreign share discount. Journal of Finance 63, 159–196.

Chen, G.M., Lee, B.S. and Rui, O., 2001. Foreign ownership restrictions and market segmentation in China's stock markets. Journal of Financial Research, 24, pp.133-155.

Chen, Z., Xiong, P., 2001. Discounts on illiquid stocks: Evidence from China. Working paper. Yale

University.

Chuhan, P., 1994. Are institutional investors an important source of portfolio investment in emerging markets. WPS 1243, World Bank, International Economics Department, Washington, DC.

Chui, A. C. W., Kwok, C. C. Y., 1998. Cross-autocorrelation between A shares and B shares in the Chinese stock market. Journal of Financial Research 21, 333-53.

Das, M.S., 2019. China’s evolving exchange rate regime. International Monetary Fund.

Domowitz, I., Glen, J., Madhavan, A., 1997. Market segmentation and stock prices: Evidence from an emerging market. Journal of Finance 52, 1059-85.

(30)

Doukas, J. A., Wang, L., 2013. Information asymmetry, price discovery, and the Chinese B-share discount puzzle. Pacific-Basin Finance Journal, 21, 1116-1135.

Eun, C., Janakiramanan, S., Lee, B. S., 2001. The Chinese discount puzzle. Center for International Business Education and Research. Georgia Tech University Working Paper 00/01, 22.

Fernald, J., Rogers, J. H., 2002. Puzzles in the Chinese stock market. Review of Economics and Statistics, 84, 416-432.

Froot, K.A., O’Connell, P., Seasholes, M.S., 2001. The portfolio flows of international investors. Journal of Financial Economics 59, 151–193.

Hietala, P.T., 1989. Asset pricing in partially segmented markets: evidence from the Finnish market. Journal of Finance 44, pp. 697-715.

Lin, A.Y., Swanson, P.E., 2004. International equity flows and developing markets: the Asian financial market crisis revisited. Journal of International Financial Markets, Institutions and Money 14, 55–73.

Ma, X., 1996. Capital controls, market segmentation and stock prices: Evidence from the Chinese stock market. Pacific Basin Finance Journal 4, 219-239.

Merton, R.C., 1987. A simple model of capital market equilibrium with incomplete information. J. Finance 42, 483–510.

Rui, O. M., Wu, W., Lee, B. S., 2007. Market segmentation and stock prices discount in the Chinese stock market: Revisiting B-share discounts in the Chinese stock market. Available at SSRN 1029494. Stulz, R. M., Wasserfallen W., 1995. Foreign equity investment restrictions, capital flight, and shareholder wealth maximization: Theory and evidence. Review of Financial Studies 8, 1019-57.

(31)

Appendix

Table 3. The Result of Causal Relation Test (1992-1997).

ID Rb→Ra Signif. _Level _RelationCausal Ra→Rb Signif. _Level Causal

Relation Panel A: Shenzhen Stock Exchange

200002 F(5,1181) 0.983 No F(5,1181) 0.992 No 200003 F(5,926) 0.472 No F(5,926) 0.201 No 200013 F(5,1309) 0.928 No F(5,1309) 0.960 No 200015 F(5,1150) 0.992 No F(5,1150) 0.979 No 200018 F(5,1309) 0.105 No F(5,1309) 0.788 No 200022 F(5,1198) 0.870 No F(5,1198) 0.793 No 200024 F(5,1309) 0.295 No F(5,1309) 0.016 Yes 200026 F(5,1177) 0.919 No F(5,1177) 0.603 No 200029 F(5,1020) 0.712 No F(5,1020) 0.001 Yes 200039 F(5,956) 0.188 No F(5,956) 0.397 No 200513 F(5,1072) 0.401 No F(5,1072) 0.000 Yes 200011 F(5,1282) 0.691 No F(5,1282) 0.111 No 200012 F(5,1280) 0.955 No F(5,1280) 0.056 No 200016 F(5,1282) 0.000 Yes F(5,1282) 0.117 No 200017 F(5,1309) 0.800 No F(5,1309) 0.458 No 200019 F(5,1309) 0.636 No F(5,1309) 0.000 Yes 200020 F(5,1282) 0.001 Yes F(5,1282) 0.000 Yes 200025 F(5,1165) 0.005 Yes F(5,1165) 0.311 No 200030 F(5,1309) 0.540 No F(5,1309) 0.000 Yes 200037 F(5,790) 0.628 No F(5,790) 0.038 Yes 200045 F(5,976) 0.044 Yes F(5,976) 0.222 No 200055 F(5,430) 0.000 Yes F(5,430) 0.418 No 200056 F(5,370) 0.026 Yes F(5,370) 0.280 No 200413 F(5,313) 0.388 No F(5,313) 0.809 No 200521 F(5,333) 0.054 No F(5,333) 0.504 No 200539 F(5,638) 0.701 No F(5,638) 0.537 No 200541 F(5,609) 0.002 Yes F(5,609) 0.897 No 200550 F(5,571) 0.795 No F(5,571) 0.084 No 200570 F(5,320) 0.573 No F(5,320) 0.456 No 200596 F(5,311) 0.253 No F(5,311) 0.586 No 200028 F(5,1129) 0.847 No F(5,1129) 0.000 Yes 200058 F(5,247) 0.140 No F(5,247) 0.000 Yes

Panel B: Shanghai Stock Exchange

(32)

900914 F(5,1080) 0.752 No F(5,1080) 0.174 No 900915 F(5,1006) 0.641 No F(5,1006) 0.009 Yes 900916 F(5,1056) 0.506 No F(5,1056) 0.004 Yes 900917 F(5,960) 0.231 No F(5,960) 0.549 No 900918 F(5,1006) 0.001 Yes F(5,1006) 0.259 No 900919 F(5,1038) 0.613 No F(5,1038) 0.338 No 900920 F(5,976) 0.387 No F(5,976) 0.536 No 900922 F(5,1025) 0.023 Yes F(5,1025) 0.736 No 900923 F(5,1001) 0.566 No F(5,1001) 0.182 No 900924 F(5,976) 0.636 No F(5,976) 0.018 Yes 900926 F(5,974) 0.848 No F(5,974) 0.001 Yes 900927 F(5,963) 0.183 No F(5,963) 0.015 Yes 900928 F(5,941) 0.082 No F(5,941) 0.025 Yes 900932 F(5,794) 0.473 No F(5,794) 0.403 No 900934 F(5,301) 0.969 No F(5,301) 0.902 No 900937 F(5,375) 0.133 No F(5,375) 0.394 No 900938 F(5,325) 0.069 No F(5,325) 0.511 No 900941 F(5,269) 0.764 No F(5,269) 0.380 No 900906 F(5,1393) 0.520 No F(5,1393) 0.012 Yes 900921 F(5,976) 0.469 No F(5,976) 0.029 Yes 900925 F(5,987) 0.722 No F(5,987) 0.001 Yes 900930 F(5,817) 0.911 No F(5,817) 0.668 No 900931 F(5,800) 0.062 No F(5,800) 0.000 Yes 900933 F(5,781) 0.341 No F(5,781) 0.000 Yes

Note: The model is:

Y𝑅&: 𝑅_&<[ = Y 𝑑0 𝑑?[ + Y 𝐴00(𝐿) 𝐴0?(𝐿) 𝐴?0(𝐿) 𝐴??(𝐿)[ _ 𝑅&-0: 𝑅_&-0< ` + Y 𝑢0& 𝑢?&[,