Margin-Trading and Short-Selling with Asymmetric Information

Transcription

1 Margin-Trading and Short-Selling with Asymmetric Information Yonglei Wang * November 2014 Abstract By studying China s margin-trading and short-selling program, this paper provides evidence that allowing margin-trading and short-selling has a negative effect on liquidity because information asymmetry makes uninformed investors reluctant to trade. The mechanism is further verified by the fact that this effect is more pronounced for stocks (1) without analyst coverage, (2) with larger standard deviation in analysts earnings forecasts, or (3) with fewer institutional shareholders. Furthermore, as a robustness check, I find that this program has a positive liquidity impact on the ETF funds for which asymmetric information problem is attenuated. I also find evidence that short-selling improves price discovery. JEL classification: G12, G14, G15, G18 Keywords: Margin-Trading and Short-Selling, Information Asymmetry, Liquidity, Price Discovery * Toulouse School of Economics. [email protected]. I am indebted to Augustin Landier, Sophie Moinas and Yinghua He for their invaluable guidance and support. I thank Milo Bianchi, Alexander Guembel and Sébastien Pouget for their very helpful comments. This work has also benefited greatly from conversations with Nordine Abidi, Gang Wang and from the suggestions of seminar participants at Toulouse Brown Bag seminar and finance student workshop. All remaining errors are my own. I acknowledge financial supports from the Jean-Jacques Laffont Foundation.

2 1. Introduction On March 31, 2010, in a pilot program covering ninety underlying securities, the China Securities Regulatory Commission (CSRC) started to allow margin-trading 1 and short-selling 2 to integrate more information into securities prices and provide reverse securities trading on the market... forming more proper stock prices and the internal price stability mechanism on the market... to enlarge the supply and demand of funds and securities and to increase the trading volume to some extent, thus leading to active liquidity on the securities market. Over time, the list of underlying securities was revised, and more securities were added. By the end of 2013, the underlying securities amounted to more than a quarter of the total number of stocks on the Shanghai and Shenzhen Stock Exchanges. By studying four events during which securities were added to the list (on March 31, 2010, December 5, 2011, January 31, 2013 and September 16, 2013), this paper investigates the effects of margin-trading and short-selling on liquidity and price discovery of the Chinese stock market. A large strand of empirical literature has established that margin-trading improves liquidity (Hardouvelis and Peristiani, 1992; Kahraman and Tookes, 2014). Studying short-selling, Beber and Pagano (2013) and Boehmer and Jones and Zhang (2013) find the bans on short-selling are detrimental to liquidity. It thus implies that allowing short-selling improves liquidity. However, whether and how market friction changes this result is less well understood. Since the short sellers are generally more informed (Chang, Luo and Ren, 2013), the uninformed investors may avoid trading stocks if they believe their more informed counterparties can take advantage of them (Ausubel, 1990). Once the uninformed investors stop trading, the informed ones could also decide not to trade in order to avoid revealing their information (Milgrom and Stokey, 1982). This would result in a market breakdown in the spirit of the theory in Bhattacharya and Spiegel (1991). Consequently, when the asymmetric information problem is severe, allowing short-selling could also make liquidity worse. Because of severe information asymmetry in the Chinese stock market (Chan, Menkveld and Yang, 2008), its margin-trading and short-selling program provides a quasi-experiment to test the above hypothesis. First, the market is dominated by the individual investors who have few financial skills or information. According to a recent report from the Shenzhen Stock Exchange, 3 from 2007 to 2012, on 1 It is called Rong Zi in Chinese, meaning that investors can buy stocks on margin. 2 It is called Rong Quan in Chinese, meaning that investors can sell stocks short. 3 Source: 1

3 average about 42.8% of the total market capitalization was held by individual investors; and individual investors account for 85.6% of the total transactions. Second, because of murky accounting practices, black-box exchanges and inadequate corporate governance, the market lacks transparency. Individual Chinese investors could even be deprived of their legal rights to access the information on listed stocks. For example, it is reported in 2013 that a Chinese financial journalist was arrested on allegations of defaming a state-backed company. 4 To sum up, these two characteristics make the asymmetric information problem severe in the Chinese stock market. Thus, once margin-trading and short-selling become available to more informed investors, their less informed counterparties may refuse to trade for fear of being taken advantage of. Therefore, liquidity would become worse. The paper uses stock level data covering more than 2500 Chinese listed firms to test this predcition. The stocks that are eligible for the margin-trading and short-selling program are taken as the treated stocks. One concern is that, comparing to the ineligible stocks, the treated stocks differ significantly in characteristics that may be correlated with liquidity or price discovery. This will cause endogeneity in the regression. In order to overcome this problem, I employ a matched sample estimation approach (following Beber and Pagano, 2013 and Boehmer, Jones and Zhang, 2013) to mitigate the difference between the treatment and the control groups. Consistent with the above prediction, I find that liquidity is adversely affected by the margin-trading and short-selling program. Specifically, for the stocks that are eligible for margin-trading and short-selling, the Bid-Ask spread and the Amihud illiquidity ratio significantly increase, and the turnover ratio significantly decreases when compared to the stocks in the control group. I further identify the mechanism underlying this result. As argued above, if allowing margin-trading and short-selling makes liquidity worse because information asymmetry causes uninformed investors to refuse to trade, liquidity should therefore be more adversely affected for stocks with greater information asymmetry. To test this hypothesis, I employ several proxies of information asymmetry that are also widely used in the literature. The first proxy is a dummy for the absence of analyst coverage. As Chinese individual investors have limited channels to obtain accurate information on listed stocks, they largely rely on analysts reports. Therefore, it is reasonable to assume that information asymmetry is more pronounced for stocks without analyst coverage. The second proxy is defined according to the standard 4 Source: Another example can be found here: 2

4 deviation of analysts earnings per share (EPS) forecasts (Diether, Malloy, and Scherbina, 2002). Because a higher EPS forecast standard deviation implies more dispersed opinions about a stock, we would expect information asymmetry to be more severe for that stock. The last proxy is defined according to the institutional investors holding ratio. The institutional investors, by holding the stocks, can also reduce information asymmetry in those stocks (Edmans 2009, Wang and Zhang 2009). Thus information asymmetry would be more severe if the stock has fewer institutional shareholders. Overall, the regression results support the above hypothesis. The negative liquidity impact of the margin-trading and short-selling program is more pronounced for the treated stocks (1) without analyst coverage, (2) with higher analyst EPS forecast standard deviation, or (3) with fewer institutional shareholders. These results suggest that the negative liquidity impact is due to the information asymmetry that causes uninformed investors to avoid trading. I also exam the effects of the margin-trading and short-selling program on stock return volatility similar to Foucault, Sraer and Thesmar (2011). Their paper shows that when individual investors behave as noise trader, their trading has a positive impact on stock return volatility. In line with this result, I find that, after margin-trading and short-selling are implemented, volatility and autocovariance of stock returns decline for the treated stocks relative to the controls. This implies that the uninformed investors, who stop trading, play the role of noise traders. It further confirms my argument that when margin-trading and short-selling are implemented, uninformed investors stop trading, thus making liquidity worse in the spirit of Milgrom and Stokey s no-trade theorem. As a robustness check, I further study the liquidity impact on ETF funds. Because ETF funds represent a whole industry or market, it is less likely that asymmetric information arises at ETF funds level when compared to stocks. Thus, we would expect no change or even a positive liquidity effect on the ETF funds. I find result consistent with this intuition, liquidity actually improves for the ETF Funds when margin-trading and short-selling become eligible for them. This provides further evidence that in the presence of information asymmetry, margin-trading and short-selling decrease liquidity. This paper also investigates the impact of margin-trading and short-selling on price discovery. According to Diamond and Verrecchia (1987), allowing short-selling helps negative information to be incorporated into stock price more quickly. Supporting evidence is found in several empirical studies (Bris, Goetzmann and Zhu, 2007; Beber abd Pagano, 2013). In a recent paper that studies the price efficiency impact of the same margin-trading and short-selling program, Chang, Luo and Ren (2013) 3

5 separately compare (1) the cross-autocorrelation between stock returns and lagged market returns, (2) the estimated β and R 2 from the market model, and (3) the variance ratio (Saffi and Sigurdsson, 2011) 5 in the bull market and the bear market. Focusing on the first two addition events covered by this paper, they find the evidence that short-selling activities contribute to greater price efficiency. In these previous papers, the impact of short-selling on price discovery is generally identified by the asymmetric changes of the cross-autocorrelation (or the estimated β and R 2 from the market model) between stock returns and market returns in bear and bull markets. However, the events in this paper that allow margin-trading and short-selling at the same time may make the identification of this method invalid because the impact of margin-trading on the cross-autocorrelation (or the β and R 2 ) may vary freely in bear and bull markets. So, instead of focusing on returns cross-autocorrelation (or the β and R 2 ), this paper studies the impact of margin-trading and short-selling on price discovery by investigating price reversion. If margin-trading and short-selling improve price discovery, when they are allowed, overpriced stocks attract short-selling (and thus prices drop more quickly), while underpriced ones would attract margin-trading (and prices would rise more rapidly). Consequently, there would be a quicker price reversion. Similar to Jegadeesh and Titman (1993, 2001), I construct a contrarian strategy to capture such price reversion. Specifically, each day, all stocks in the treatment group are sorted according to their past returns, and those with highest past returns are sold, those with lowest past returns are bought. Then, this portfolio, which is composed of the above short and long position, is held for the next day. I also construct the contrarian strategy based on the stocks in the control group to employ a difference-in-differences methodology. If there is a quicker price reversion as described above, the contrarian strategy built on the treatment group will become more profitable than the one built on the control group when margin-trading and short-selling are implemented. Consistently, I find a significantly increased profitability, and it is mainly from the short position, indicating that the improvement in price discovery is mainly due to short-selling. This paper makes several contributions to the literature. First, it is among the first papers to document the negative impact of margin-trading and short-selling on liquidity. Unlike the previous works that study the non-exogenous short-selling bans in which short-selling is constrained because of drops in stock prices, this paper investigates a non-endogenous change in short-selling rules from no shorting to rights to short. Meanwhile, the specificity of China s margin-trading and short-selling 5 Defined as the absolute value of the variance of monthly returns divided by four times the weekly returns minus one. 4

6 program in which mostly informed investors with strong information conduct margin-trading and short-selling also allows me to investigate the effects in a different setting. Moreover, in the previous papers that also study the margin-trading and short-selling program in China (Chang, Luo and Ren, 2013 and Sharif, Anderson and Marshall, 2014), short sellers can only borrow stocks from a limited set of qualified security companies; these qualified security companies are the only source of stocks for short-selling. On February 28, 2013, a security refinancing business was implemented to allow funds and other institutional shareholders to lend out the stocks they hold. 6 By studying this security refinancing business, this paper exams the impact of the margin-trading and short-selling program in a more deregulated environment and provides a more complete understanding of the program s effects. Second, this paper is the first one to identify the mechanism underlying the negative liquidity impact. By employing cross-sectional differences in stocks information asymmetry, this paper finds that the negative liquidity impact is due to information asymmetry that causes uninformed investors reluctant to trade. This also speaks to an ongoing discussion in China about how the rights of individual investors should be properly protected. Unlike in developed markets, serious information asymmetry in the Chinese stock market requires more attention from policy makers when they reform the stock market. Third, previous works studying short-selling in China focus on a sample of cross-listed Chinese stocks. For example, Mei, Scheinkman and Xiong (2009) study stocks with both A and B shares; 7 Chan, Kot and Yang (2010) and Fang and Jiang (2013) study stocks with both A and H shares. 8 They generally look at a small sample because the number of cross-listed Chinese stocks is limited, and these stocks usually have large market capitalization, low information asymmetry and better liquidity. In contrast, I extend the analysis along two dimensions: events and stocks. This paper studies four addition events during which a large set of securities sequentially become eligible for margin-trading and short-selling and covers all stocks falling into the program, which represent around a quarter of total stocks in China. Furthermore, because the program begins with a small list of underlying securities and gradually covers a larger part of the market, this also allows me to obtain more accurate results by checking the same treatment at the different times, thus excluding factors that would bias the results. 6 More details can be found in section 2. 7 A-shares are the stocks traded by domestic citizens, while B-shares are designed for the foreign investment and were not to domestic citizens until February Both A and B shares are traded on the Shanghai and Shenzhen Stock Exchanges. 8 H-shares refer to the shares of companies incorporated in mainland China that are traded on the Hong Kong Stock Exchange. 5

7 The rest of this paper is structured as follows: Section 2 provides a detailed overview of the margin-trading and short-selling program in the Chinese stock market; Section 3 briefly reviews the related literature and develops the testable hypotheses; Section 4 summarizes the data and methodology; Section 5 presents the descriptive evidence and regression results; Section 6 presents the robustness check; and Section 7 concludes. 2. The Chinese Margin-Trading and Short-Selling Program Although margin-trading and short-selling have been discussed in China since 2006, the first official news was announced by the CSRC in October 2008, 9 broadly stating the intention of lifting the constraints of margin-trading and short-selling in the Chinese stock market. Then, the detailed underlying securities list for the margin-trading and short-selling program came out on February 12, According to the list, starting on March 31, 2010, the first 90 blue-chip securities (50 from the Shanghai Stock Exchange, 40 from the Shenzhen Stock Exchange) became available for margin buyers and short sellers. The Chinese margin-trading and short-selling program was officially launched. At the beginning of the program, there were only six security companies which are qualified to undertake the margin-trading and short-selling business. 10 According to the administrative rules promulgated by the CSRC, only qualified investors can conduct margin-trading and short-selling, and the qualification requirements differ across the security companies which undertake these business. For example, in Haitong Securities Company, 11 qualified investors must: (1) have a trading history longer than one and a half years with that company (reduced to half a year after December 2011), (2) have capital of at least half a million RMB, and (3) pass an exam and a risk-attitude test to show they have the basic knowledge of margin-trading and short-selling. Then, these investors who wanted to conduct margin-trading (short-selling) can to go to any of these six qualified security companies, open an account, deposit a required amount of money as a guarantee, and borrow the money (securities) 12 from the company to buy (sell) the securities in the underlying securities list. Moreover, the interest rate of 9 Source: 10 The number of qualified security companies also increased with the development of the margin-trading and short-selling program. 11 Source: 12 The money or securities lent by the qualified security companies are owned by these companies themselves. The self-owned money and securities are the only source for margin-trading and short-selling. 6

8 margin-trading and rebate rate of short-selling are very high, 13 which making margin-trading and short-selling costly for investors. This high cost as well as the above requirements on investors qualification implicitly imply that only the professional traders who also have the strong information would conduct margin-trading or short-selling in the market. With the development of this program, the CSRC increased the coverage of the program by adding more underlying securities. In order to be included as underlying securities, several requirements should be satisfied: (1) the securities have been traded in the stock market longer than 3 months; (2) the market value is larger than 800 million RMB; (3) the number of shareholders is larger than 4000; (4) the average daily turnover is larger than 15% of the benchmark average daily turnover, and daily trading volume in value is not less than 50 million RMB; (5) the deviation of average daily quote change is less than 4% from the benchmark; and (6) the volatility is less than five times of the benchmark one. 14 Up to the end of 2013, there were four main addition events 15 on March 31, 2010, December 5, 2011, January 31, 2013 and September 16, 2013 after which more securities became eligible for margin-trading and short-selling, the underlying securities list largely increased. Figure 1 plots the abnormal and cumulative abnormal return estimated by using the Fama-French three factors model over the window around the addition date when the treated stocks are added into the list. Consistent with Sharif, Anderson and Marshall (2014), we can see that when margin-trading and short-selling are implemented, price largely drops. Because at the beginning of this pilot program, the investors could only borrow money or securities from the qualified security companies, the amount of funds and securities which were available for 16 margin-trading and short-selling was very limited. Then, a refinancing business was introduced on August 3, 2012 to allow more market participants to provide their funds and securities. A government-backed securities-related financial institution -- China Securities Finance Co., Ltd. (CSF) -- was established as the only institution providing margin refinancing 17 and security refinancing 18 services. Specifically, according to the duties delegated by the CSRC, CSF acts as the intermediary agency in 13 For example, at the beginning of the program, the interest rate of margin-trading and rebate rate of short-selling were both 7.86% in Haitong Securities; and the rebate rate of short-selling was even larger (9.86%) in several other security companies. 14 Source: 15 There are two more addition events on July 1, 2010 and July 29, In these two events, five and one additional stocks were separately added into the underlying securities list. I don t include these two events because the sample size is too small, and there is also an overlapping in their event windows. 16 It is called Zhuan Rong Tong in Chinese. 17 It is called Zhuan Rong Zi in Chinese. 18 It is called Zhuan Rong Quan in Chinese. 7

9 lending funds (margin refinancing) and securities (security refinancing) from the fund managers, the institutional shareholders such as insurance companies, and other market participants to the qualified security companies when the security companies own funds or securities are insufficient to facilitate their customers needs (for margin-trading or short-selling). On August 30, 2012, the margin refinancing business was first implemented for all the underlying securities (which were already eligible for margin-trading and short-selling). However, the security refinancing business was not implemented until February 28, 2013, and initially only covered a small portion of securities in the underlying securities list. I obtained detailed information on the margin-trading and short-selling program as well as the refinancing business from the Shanghai and Shenzhen Stock Exchanges from March 2010 to the end of This paper focuses on the treated stocks from four addition events on March 31, 2010, December 5, 2011, January 31, 2013 and September 16, ETF funds are excluded from my data because it is much less likely that asymmetric information arises for ETF funds when compared to stocks. In fact, in the part of robustness check, I find a positive liquidity effect on the ETF funds in the underlying securities list when margin-trading and short-selling are implemented. I report the announcement date and the effective date of each addition event as well as the number of additions in Panel A of Table 1. In Panel B of Table 1, I also report the event date and the number of stocks that are covered by the refinancing business. Although the number of stocks that are covered by the security refinancing business increased once in September 2013, the total number is still less than half the stocks that are already eligible for margin-trading and short-selling. It is worth noting that, as shown in the Panel B of Table 1, 10 out of 266 treated stocks from the January 31, 2013 addition event and 42 out of 184 treated stocks from the September 16, 2013 addition event became also eligible for the security refinancing business shortly after they were added into the underlying securities list for margin-trading and short-selling program. This will cause an overlapping event window for the addition event with the event of the security refinancing business in my study. Thus, I exclude these 52 stocks to avoid any contamination. In the next part, I will briefly review the related literature and develop the testable hypotheses. 8

10 3. Literature Review and Testable Hypotheses 3.1. Liquidity With regard to the impact of margin-trading on liquidity, the results are mostly from empirical works. For example, Hardouvelis and Peristiani (1992) found that an increase in margin requirements leads to a decrease in liquidity. So, to be in line with this finding, we would expect the introduction of margin-trading improves liquidity. In a recent paper, Kahraman and Tookes (2014) also find consistent results. By employing a regression discontinuity design, they show that liquidity is higher when stocks become eligible for margin-trading and it decreases with ineligibility. However, when it comes to the impact of short-selling on liquidity, the prediction is not clear. Diamond and Verrecchia (1987) analyzed a variant of the Glosten and Milgrom (1985) model and showed that short-selling allows informed investors to trade on bad news, thus increasing the speed of price discovery and helping to resolve the uncertainty about fundamentals. If this effect dominates the adverse selection due to the informed trading, allowing short-selling will improve liquidity. Some empirical works find evidence supporting this prediction. For example, Beber and Pagano (2013) found that the short-selling bans implemented around the world in were detrimental to liquidity. Boehmer, Jones and Zhang (2013) also found that the liquidity declined for stocks after the 2008 U.S. shorting ban. With samples from the Hong Kong stock market, Chan, Kot and Yang (2010) found that the H-share volume (relative to the A-share volume) was higher for the shortable stocks than for the non-shortable stocks. They conclude that, by including the bearish investors, short-selling eligibility allows more investors to trade, thus improving liquidity. However, other authors (Amihud and Mendelson, 1986; Easley and O'Hara, 2004) show that a higher proportion of informed traders will increase the adverse selection component of the Bid-Ask spread, and thus reduce liquidity. Several papers (Ausubel, 1990; Bhattacharya and Spiegel, 1991) also argue that the less-informed investors will be more reluctant to trade if they expect insiders will take advantage of them in the trading. As shown by Boehmer, Jones and Zhang (2008, 2012) and Chang, Luo and Ren (2013), short sellers are better informed and tend to have an advantage in the stock information. This reinforces the concern of an asymmetric information problem investigated by the above authors. Therefore, rather than encouraging more investors to trade, the introduction of margin-trading and 9

11 short-selling can prevent some from joining the market, thus decreasing liquidity. To test this point, the first hypothesis we can make is as follows: Hypothesis 1: In the presence of information asymmetry, allowing margin-trading and short-selling decreases liquidity. If the above prediction is correct, allowing margin-trading and short-selling has a negative effect on liquidity. We can further test if this result is driven by the information asymmetry that causes the uninformed investors to avoid trading. Consequently, another natural hypothesis to test is: Hypothesis 2: The more severe information asymmetry is, the worse liquidity becomes after margin-trading and short-selling are implemented Price discovery While investigating the effect of short-selling on price discovery, Diamond and Verrecchia (1987) showed that allowing short-selling would increase the speed with which prices adjust to private information, especially bad news. Consistent with this prediction, both Bris, Goetzmann and Zhu (2007) and Beber and Pagano (2013) found that when short-selling is allowed, the cross-autocorrelation between stock returns and lagged market returns decreases, and this decrease is larger in a bear market than a bull one. The impact of short-selling on price discovery is thus identified by this asymmetric change in bear and bull markets. Few studies have explored the impact of margin-trading on price discovery. Although investors can always buy stocks, margin-trading can still affect how stock prices adjust to positive information because with margin-trading less funds are needed to acquire the same amount of stocks. Thus, the identification methodology used in literature may be invalid for the margin-trading and short-selling program in China. Instead of checking returns cross-autocorrelation, this paper studies the impact of margin-trading and short-selling on price discovery by checking price reversion. Specifically, on each day within a period of 150 days before and after margin-trading and short-selling are allowed, the stocks in the treatment group are sorted according to their past returns; those with the highest past returns are sold, those with the lowest past returns are bought. If 10

12 margin-trading (short-selling) improves price discovery, the price of underpriced (overpriced) stocks will go up (down) more quickly, thus the long (short) position will generate higher returns when margin-trading and short-selling are implemented. Overall, by capturing the quicker price reversion that is driven by margin-trading and short-selling, this contrarian strategy 19 provides a way to directly measure the impact on price discovery. In order to test this intuition, I make the following hypothesis: Hypothesis 3: If margin-trading and short-selling improve price discovery, the contrarian strategy should become more profitable for the treated stocks when compared to the ones in the control group after margin-trading and short-selling are implemented. 4. Data and Empirical Strategy 4.1. Data This paper focuses on the stocks in the underlying securities list, ETF funds are excluded because the concern for information asymmetry would not be so severe as for stocks. 20 Four addition events on March 31, 2010, December 5, 2011, January 31, 2013 and September 16, 2013 are used to analyze the effects of margin-trading and short-selling on liquidity and price discovery. Panel A of Table 1 provides the detailed information on the number of treated stocks, the announcement date and the effective date for each addition event. Because, as mentioned before, 10 out of 266 treated stocks from the January 31, 2013 addition event and 42 out of 184 treated stocks from the September 16, 2013 addition event are also affected by the security refinancing business during the addition event s estimation window, I exclude these 52 stocks to avoid any contamination. In the regression, I use three variables to measure stock liquidity. The first one is the weekly averaged daily Bid-Ask spread; the second one is the Amihud illiquidity ratio (Amihud and Mendelson, 1986; Amihud, 2002) which is computed as the absolute value of daily stock return divided by trading volume in value and then averaged over one week; the last one is the turnover ratio, defined as the weekly average of the ratio of daily trading volume divided by common shares outstanding. Stock level data such as closing price, trading volume, trading volume in value, bid and ask prices are from 19 Similar to the strategy in Jegadeesh and Titman (1993, 2001). 20 The liquidity impact of margin-trading and short-selling on ETF funds is investigated in the part of robustness check. 11

13 Datastream; the data of market capitalization and common shares outstanding are from Worldscrop. Due to missing data, the number of stocks in the regression is occasionally smaller than that in Table Matching Methodology My empirical strategy relies on a matched sample estimation approach to identify the effect of the margin-trading and short-selling program on liquidity and price discovery. Each treated stock that is eligible for margin-trading and short-selling is matched with an exempt stock 21 that is traded on the same stock exchange and is closest in terms of market capitalization and trading volume in value. The criteria is computed as the squared proportional difference in market capitalization plus the squared proportional difference in the average trading volume in value over a 30-day period before the addition is announced. I chose the 30-day average trading volume in value before the announcement day in order to prevent the liquidity shock on any stock from the matching procedure. More formally, for each treated stock i its matched stock is chosen by solving the following problem: Stock j Ineligible Stocks = argmin j i (MV j MV i )/MV i 2 + (VA j VA i )/VA i 2 (1) s. t. Stock j is from the same stock exchange as Stock i. where MV is the market capitalization, VA is the average trading volume in value during the 30-days before the addition is announced. These matched stocks compose the control group in my analysis. Table 2 reports the summary statistics of the stock level controls and liquidity measurements over a 60-week event window. In each case, I report the means and standard deviations of the stock level controls separately for the treatment group, the control group and all the ineligible stocks that are not in the underlying securities list. Then, for the liquidity measurements, I also report the summary statistics separately for the periods before and after the treatment. It can be seen that, before the treatment, the matched stocks in the control group 21 The exempt stock could also become eligible for margin-trading and short-selling in the future, but it remained to be ineligible during the event window checked by the regressions. 12

14 have more similar characteristics as those of the treatment group when compared to the sample of all ineligible stocks Regression Analysis The date when margin-trading and short-selling are implemented is taken as the event date. In the regression, the week of event date is dropped. Following Boehmer, Jones and Zhang (2013) and Beber and Pagano (2013), within the event window of 60 weeks around the event date, I test my first hypothesis by the following regression: y i,t y match i,t = α pair i + β 1 Post t + γ X + ε i,t (2) where y i,t is the liquidity measurement of stock i at week t; y match i,t is the liquidity measurement of the stock that is matched for stock i; α i pair is the pair fixed effect, so any difference that does not change between the treatment and the control group can be controlled in the regression; Post t is a dummy variable equal to one after margin-trading and short-selling are allowed; X stands for several stock level variables such as log of market capitalization, lag of stock returns and stock returns volatility. Thus, X are the differences of these characteristics between a treated stock and its matched one. The strategy here is to identify the effect of margin-trading and short-selling on a particular measure of liquidity by comparing the treated stocks with the matched ones during the periods before and after margin-trading and short-selling are allowed. More simply put, a difference-in-differences methodology is employed, and the effect of margin-trading and short-selling on liquidity is thus measured by the coefficient β 1 in this specification. Moreover, because the OLS standard errors of difference-in-differences estimates can be biased if there is a serial correlation in the error terms for a given stock over time, I employ a two-way clustered standard deviation to allow for the correlation in residuals over time given a stock as well as across stocks at each t. To test the second hypothesis, I employ several proxies for stock s information asymmetry. I use the data from Wind 22 on analyst coverage and the institutional investors holding data from Factset in 22 Wind Information of Shanghai, which is one of China's largest financial information vendors. 13

15 WRDS. The first proxy used in this paper is a dummy of the absence of analyst coverage. As mentioned earlier, the stock analyst is one of the most important sources of information for individual Chinese investors. It is then natural to use no analyst coverage as the proxy for a stock s information asymmetry. We would expect the information asymmetry to be more pronounced for the stocks without analyst coverage. The second proxy is the standard deviation of analysts EPS forecasts (e.g., Diether, Malloy, and Scherbina, 2002). Because a higher standard deviation means larger dispersed information, the uninformed investors would get less accurate information and the information asymmetry should be more severe. The last proxy is one minus the institutional investors holding ratio (e.g., Wang and Zhang, 2009; Michaely and Vincent, 2012). First, the institutional investors, by holding the stocks, can incorporate more information into the prices, thus making the information asymmetry less severe. Moreover, because the institutional investors (mainly funds) in China are restricted from conducting short-selling, and their holdings cannot be lent to short sellers; 23 the available amount of one treated stock for short sellers would be less if there are more institutional shareholders in that stock. Consequently, the uninformed investors concern of trading with more informed counterparties would be lessened. In summary, either of these above two channels would make the negative effect that margin-trading and short-selling have on liquidity less pronounced if there are more institutional shareholders in the treated stocks. Generally, to identify the mechanism that results in the negative impact on liquidity, the following regression is estimated: y i,t y match i,t = α pair i + β 1 Post t + β 2 Post t Asym i + γ X + ε i,t (3) where Asym i could be (1) the dummy variable for no analyst coverage on stock i; (2) the standard deviation of analysts EPS forecasts; or (3) one minus the institutional investors holding ratio. Then, Asym i is the difference of the information asymmetry proxy between stock i and its matched one. If the negative impact of margin-trading and short-selling is due to the information asymmetry which makes the uninformed investors reluctant to trade, we would find the negative effects to be more pronounced when Asym i is larger, i.e., the information asymmetry is more severe. That means that, to 23 Notice that I exclude the stocks that are also affected by the security refinancing business during the event window. 14

16 be in line with my hypothesis, the interested coefficient β 2 should be positive for the liquidity measurement of the Bid-Ask spread and the Amihud illiquidity ratio, and be negative for the liquidity measurement of the turnover ratio. I also investigate the impact on price discovery by testing whether there is a quicker price reversion. Similar to the strategy used in Jegadeesh and Titman (1993, 2001), on each day the stocks in the treatment group are sorted according to their past J day returns into deciles. Those with the lowest past returns (losers) are bought, those with the highest past returns (winners) are sold. This portfolio, which is composed by the above long and short positions, is held for the next K days. So, on each day there are a total of K portfolios which are separately formed from the last K days. The return of this contrarian strategy is the average return of these K portfolios. The same contrarian strategy is also constructed on the stocks in the control group. Then, within a period of 300 days (or 60 weeks) around each event date, the following regression is used to capture the impact of margin-trading and short-selling on price discovery: J,K R reversal,t = α + β 1 Post t + β 2 Treat reversal + β 3 Treat reversal Post t + ε p,t (4) J,K where reversal stands for the contrarian strategy, so R reversal,t is the return of the contrarian strategy, Post t is equal to one for the period when margin-trading and short-selling are allowed for the treated stocks, and is equal to zero otherwise. Treat reversal is equal to one for the contrarian strategy that is constructed on the treatment group; is equal to zero for the contrarian strategy that is constructed on the control group. So if margin-trading helps the prices of undervalued stocks go up more quickly and short-selling helps the prices of overvalued stocks go down more quickly, there will be a quicker price reversion and we should expect greater profitability in the strategy built on the treatment group after margin-trading and short-selling are implemented. Simply, the coefficient β 3 should be positive. In the paper, different values of J and K are used to provide stronger tests. Because the strategy is composed by multi-layered portfolios, the standard errors are clustered by day to allow for the correlation in residuals over time. 15

17 5. Results 5.1. Liquidity I first present the results of the impact of margin-trading and short-selling on liquidity. By using weekly data within the event window covering 60 weeks around the addition event, I compare the change in the Bid-Ask spread between the treated stocks and the control ones. Another two measurements, the Amihud illiquidity ratio and the turnover ratio, are also used to give a more complete understanding. The results for Hypothesis 1 are reported in Table 3. As argued in Section 4, the liquidity impact is captured by the coefficient of post dummy. On each of these three liquidity measurements, I first run the simplest regression and then add more controls such as the log of market capitalization, the lag of stock returns, and stock returns volatility. In the last regression for each liquidity measurement, I further control for the year-week fixed effects to exclude the possibility that the certain common shocks drive the results. Specifically, from column (1) of Table 3, we can see that the coefficient of post period dummy is significantly positive, and this result is robust to including more controls or the time fixed effects. As shown in column (3), the Bid-Ask spread increases 3 basis points for the treated stocks compared with the control ones. When checking the impact on the Amihud illiquidity ratio and the turnover ratio, the results support the same conclusion. From column (4), (5) and (6) of Table 3, there is a significant increase in the Amihud illiquidity ratio, indicating that allowing margin-trading and short-selling causes a lager order flow s price impact for the treated stocks. Thus the liquidity is worse off. Moreover, as shown in column (9) of Table 3, the turnover ratio of the treatment group decreases 59 basis points (p-value<1%) relative to the control group. This provides the direct evidence that allowing margin-trading and short-selling reduces the trading on the treated stocks. I also repeat the above regressions on each of the four addition events to exclude the possibility that the above results are driven by any one event alone. The results are reported in Table 4. Generally, although the significance of the results may be a little weak for the addition events in 2010 and 2011, the sign of interested coefficients are consistent with the hypothesis. Figure 2 provides a visual representation of the above findings. It plots the evolution of the difference 24 of the liquidity measurement between the treated stocks and the control ones during the period covering 20 weeks before and after margin-trading and short-selling are allowed. Similar to the 24 This is the left-hand side that is used in the regression equation (2). 16

18 regression analysis, I drop the week of event date. The confidence intervals are established by regressing the difference of the liquidity measurement on weekly dummies instead of the post dummy used in equation (2). It can be seen clearly that, when margin-trading and short-selling are implemented, the Bid-Ask spread increases and the turnover ratio decreases for the treated stocks when compared to the control ones; the Amihud illiquidity ratio also slightly increases. So, allowing margin-trading and short-selling makes the liquidity of the treated stocks worse in comparison to the control ones. More importantly, from Figure 2, we can see the difference between each liquidity measurement basically does not change much before the event. This indicates that the treated stocks and control ones are not on different time trend before the event, which is crucial for the identification of difference-in-differences methods. From the above analysis, we can come to the conclusion that allowing margin-trading and short-selling in China actually has a negative effect on liquidity. This empirical result does not depend on how I measure liquidity, is not driven by any of the four addition events alone, and is robust to various controls. The next step is to identify the mechanism that drives this result. As argued by Hypothesis 2, if this negative liquidity effect is due to the information asymmetry which makes uninformed investors reluctant to trade, we should expect it to be more pronounced for the stocks with more severe information asymmetry. Tables 5 through 8 report the results on my second hypothesis. To be specific, in Table 5 the dummy of no analyst coverage is used as the proxy for information asymmetry; in Table 6 and Table 7 the standard deviation of current and next year s EPS forecasts are used separately as the proxy for information asymmetry; in Table 8 one minus institutional investors holding ratio is used as the proxy for information asymmetry. Generally, all the regressions provide consistent results supporting Hypothesis 2. For example, when using the dummy of no analyst coverage as the proxy, column (3), (6) and (9) in Table 5 show that the Bid-Ask spread and the Amihud illiquidity ratio of the stocks without analyst coverage increases 27% (=0.007/0.026) and 91%, respectively, more than that of the stocks with analyst coverage, the turnover ratio decreases 10% more for the stocks without analyst coverage. Using the standard deviation of the analysts current year EPS forecasts gives a stronger result. I also use the standard deviation of the analysts next year EPS forecasts to provide a robustness check. The result is shown in Table 7. We can see, although the significance may change between the regressions, all of them provide consistent 17

19 evidence that the larger the standard deviation of analysts EPS forecasts, the worse the liquidity. The strongest results are from the proxy according to the institutional investors holding ratio. We can see in Table 8 that when margin-trading and short-selling are allowed, the less shares are held by the institutional investors, the worse the liquidity would become. In summary, all the above results support Hypothesis 2. Margin-trading and short-selling in China make the stock liquidity worse due to the fact that in the presence of severe information asymmetry, the uninformed investors become more reluctant to trade Price discovery I further study the impact of margin-trading and short-selling on price discovery. Contrarian strategies are constructed separately on the stocks in the treatment group as well as the control group. This allows me to employ the difference-in-differences methods to identify the quicker price reversion which is caused by margin-trading and short-selling. The regression results are reported in Table 9. To save space, only the coefficient of interest, that is of the interaction between the dummy of period allowing margin-trading and short-selling and the dummy of the contrarian strategy that is built on the treated stocks, is reported. Besides the return of the contrarian strategy (long position plus short position), I also repeat the analysis separately for the long and the short position. Generally, we can see, from Panel A of Table 9, that for the addition event on September 16, 2013, the contrarian strategy that is constructed on the treated stocks becomes more profitable than that on the control ones when margin-trading and short-selling are allowed. In the case with J = 1 and K = 1, the daily return increases percentage point for the contrarian strategy built on the treated stocks than the one on the control stocks. This result suggests the quicker price reversion mentioned in Section 4. So, as predicted by Hypothesis 3, price discovery is improved. Moreover, checking the regressions separately for the long and the short position, we can see the increased profitability is mainly from the short position, indicating that short-selling indeed helps the prices of past winners go down more quickly. However, there is no evidence for margin-trading helps the prices of past losers go up more quickly. The returns of the long position actually become relatively smaller. 18

20 The results from the addition event on January 31, 2013, which are reported in Panel B of Table 9, also give the same conclusion. The price discovery is improved and is mainly driven by short-selling. However, from Panel C and Panel D, I cannot find any conclusive result. There is no evidence of an improvement in price discovery in the addition events on March 31, 2010 or December 5, Robustness Check 6.1. Security Refinancing Business On February 28, 2013, in order to facilitate short-selling, China introduced a security refinancing business. Fund managers and other institutional shareholders are allowed to lend their shares to the short sellers when the amount owned by the qualified security companies is insufficient. Although this business covers only a small part of the underlying securities which are eligible for margin-trading and short-selling, it can provide a robustness check for the results of section 5.1. If the uninformed investors became reluctant to trade because the margin-trading and short-selling program introduced more informed counterparties, we should expect the same liquidity impact from this security refinancing business because it further facilitates the short-selling activities. Specifically, I employ the same matched sample estimation approach as before and conduct the same event study within a 60 week event window to investigate the liquidity impact. The treated stocks, in this section, are the ones which become eligible for the security refinancing business. As shown by Panel B of Table 1, 10 out of 86 stocks and 42 out of 207 stocks are also affected by the addition events of section 5.1 during the event window of the security refinancing business, thus I drop these stocks to avoid any contamination. The results are reported in Table 10. In Panel A, the stocks which are eligible for the security refinancing business are compared with the stocks which are not eligible for margin-trading and short-selling; in Panel B, they are compared with the ones which are eligible for margin-trading and short-selling but not eligible for the security refinancing business. From column (1) to (3) of Table 10, we can see that the liquidity becomes significantly worse for the treated stocks; the Bid-Ask spread and the Amihud illiquidity ratio significantly increase and the turnover ratio significantly decreases after the security refinancing business is implemented. Although regressions in Panel B give weaker results, the 19

21 sign of coefficients for post dummy are all in line with a negative impact. Overall, by further facilitating short-selling, the security refinancing business makes liquidity worse. This provides further evidence that the Chinese margin-trading and short-selling program has a negative effect on liquidity because it introduces more informed investors which in turn makes the uninformed investors reluctant to trade Uniformed Investors and Stock Volatility In this section, I check the impact of the Chinese margin-trading and short-selling program on stock return volatility. According to Foucault, Sraer and Thesmar (2011), if individual investors behave as noise traders, individual investors trading should have a positive effect on the volatility and the autocovariance of stock returns. This implies that if the individual investors stop trading, the volatility and the autocovariance of stock returns should decrease. As argued in the beginning of this paper, the negative liquidity impact of the margin-trading and short-selling program is due to the fact that uninformed investors stop trading, the test in this section will provide a sense of whether these investors are individual or professional, and thus generate a more complete understanding of the behavior of them. Specifically, similar to equation (2), I regress the difference of the volatility or the autocovariance of stock returns between a treated stock and its matched one on the post dummy. The results are reported in Table 11. We can see, when margin-trading and short-selling are implemented, the volatility and the autocovariance significantly decrease for the treated stocks relative to the control ones. These results provide a hint that the uninformed investors, who stop trading, are individual investors. Indeed, taking together the dominant transaction weight by individual Chinese investors and the severe information asymmetry in the Chinese stock market, we should not be surprised to find these results. Moreover, in line with Foucault, Sraer and Thesmar (2011), these results also suggest that these investors behave as noise traders. It confirms my argument that when margin-trading and short-selling are implemented in China, the uninformed investors become reluctant to trade, the uninformed ones could also choose not to trade to avoid revealing their information, and thus liquidity becomes worse in the spirit of no-trade theorem Liquidity Impact on ETF funds 20

22 In section 4.1, I exclude ETF funds in the underlying securities list because it is much less likely that asymmetric information arises at the ETF funds level as they represent a whole industry or market. So, with less concern for information asymmetry, we should expect no change or even a positive liquidity impact on the treated ETF funds when margin-trading and short-selling are implemented. In this section, I investigate the liquidity impact on ETF funds to test above intuition. In the underlying securities list, there are ten ETF funds by the end of For three of them, 25 the data is not enough to conduct an event study because these funds became eligible for margin-trading and short-selling shortly after being established. The rest seven ETF funds are all from the addition event on December 5, They include five passive funds which simply follow different market indices, and two active funds -- one tracks the stocks with high and stable dividend payout ratio, 26 the other one tracks the stocks with better corporate governance. 27 Because the passive funds just represent the whole market, they are likely with even less information asymmetry when compared to the active funds. Thus, I first check the liquidity impact on all the treated ETF funds and then separately for the passive and active funds respectively. In the regressions, I employ two proxies for the liquidity of ETF funds -- the Amihud illiquidity ratio and the log of trading volume. Similar to the matched sample estimation approach used before, I create a match for the treated ETF funds by using the average liquidity measures which are computed on all the ETF funds that are not eligible for margin-trading and short-selling in the Chinese stock market. The dependent variables in the regressions are the differences of the corresponding liquidity measures between the ETF funds and this match. Then, the same regression as equation (2) can be used. The regression results are reported in Table 12. From column (1) and (2) in Table 12, in contrast to the liquidity impact on the treated stocks, we can see that the Amihud illiquidity ratio significantly decreases and the trading volume significantly increases for the ETF funds after margin-trading and short-selling are implemented. When checking the impact only on the passive funds, for which the concern of information asymmetry could be further attenuated, the results in column (3) and (4) are much stronger. Meanwhile, as shown in column (5) and (6), although the results are little weak, they still suggest a positive impact on the liquidity of the active funds. To sum up, allowing margin-trading and short-selling significantly improves liquidity for the treated ETF funds. This result provides another evidence suggesting that the negative liquidity impact of the 25 Including Harvest CSI 300 Index ETF, Huatai-Pinebridge CSI 300 ETF and China AMC Hang Seng Index ETF. 26 SSE 180 Corporation Administration ETF. 27 Huatai-PineBridge SSE Dividend Index ETF. 21

23 Chinese margin-trading and short-selling program on the treated stocks is due to the information asymmetry that causes uninformed investors to avoid trading. 7. Conclusion This paper studies the impact of margin-trading and short-selling on liquidity and price discovery in the Chinese stock market. By employing a matched sample estimation approach, I find that the liquidity becomes relatively worse for the treated stocks when margin-trading and short-selling are implemented. The Bid-Ask spread and the Amihud illiquidity ratio significantly increase and the turnover ratio significantly decreases for stocks in the treatment group when compared to the control group. This result does not depend on the measures of liquidity, is not driven by any of my four addition events alone, and is robust to various controls. To further explore the mechanism driving this result, I use several information asymmetry proxies. I show that this negative liquidity impact is more pronounced for the treated stocks (1) without analyst coverage; (2) with larger analyst EPS forecast standard deviation; or (3) with fewer institutional shareholders. All three pieces of evidences show that the Chinese margin-trading and short-selling program makes liquidity worse because of information asymmetry that causes uninformed investors reluctant to trade. This result strongly supports the asymmetric information argument by Ausubel (1990), and is in keeping with the spirit of the theory by Milgrom and Stokey (1982). Moreover, the investigation on the security refinancing business in China also finds a negative liquidity impact, which provides further support for the above argument. As a robustness check, I study the liquidity impact on ETF funds. Since the information asymmetry is less concerned for the ETF funds, which generally represent a whole market, we should find no change or even a positive liquidity effect. In line with this prediction, I find the liquidity significantly improves for the treated ETF funds when margin-trading and short-selling are implemented. I also study the impact of margin-trading and short-selling on price discovery by checking price reversion for the treated stocks. The contrarian strategy, which would predict larger profits if margin-trading and short-selling contribute to a quicker price reversion, is used to gauge the impact on price discovery. The results show that the contrarian strategy that is constructed on the treatment group becomes more profitable than the control one when margin-trading and short-selling are implemented. 22

24 And this profitability is mainly from the short position which is consistent with the results in the literature that short-selling improves price discovery. 23

25 Bibliography Acharya, V. V. and Pedersen, L. H Asset Pricing with Liquidity Risk. Journal of Financial Economics. 77: Amihud, Y Illiquidity and Stock Returns: Cross-Section and Time Series Effects. Journal of Financial Markets, 5: Amihud, Y. and Mendelson, H Asset Pricing and the Bid-Ask Spread. Journal of Financial Economics, 17 (2): Ausubel, L. M Insider Trading in a Rational Expectations Economy. The American Economic Review, 80 (5): Asness, C. S., Moskowitz, T. J. and Pedersen, L. H Value and Momentum Everywhere. The Journal of Finance, 68 (3): Asquith, P., Pathak, P. A. and Ritter, J. R Short Interest, Institutional Ownership, and Stock Returns. Journal of Financial Economics, 78(2): Bris, A., Goetzmann, W. N. and Zhu, N Efficiency and the Bear: Short Sales and Markets Around the World. The Journal of Finance, 62 (3): Beber, A. and Pagano, M Short-Selling Bans Around the World: Evidence from the Crisis. The Journal of Finance, 68 (1): Biais, B. and Hillion, P Insider and Liquidity Trading in Stock and Options Markets. The Review of financial studies, 7 (4): Boehmer, E., Jones, C. M. and Zhang, X Which Shorts Are Informed? The Journal of Finance, 63(2): Boehmer, E., Jones, C. M. and Zhang, X What Do Short Sellers Know? Working Paper Series. Boehmer, E., Jones, C. M. and Zhang, X Shackling Short Sellers: The 2008 Shorting Ban. The Review of Financial Studies, 26 (6): Brogaard, J., Hendershott, T. and Riordan, R High-Frequency Trading and Price Discovery. The Review of Financial Studies, doi: /rfs/hhu032. Brunnermeier, M. K. and Pedersen, L. H Market Liquidity and Funding Liquidity. The Review of financial studies, 22 (6): Battalio, R. H. and Schultz, P. H Regulatory Uncertainty and Market Liquidity: The 2008 Short Sale Ban s Impact on Equity Option Markets. Journal of Financial Economics, 66 (6): Bhattacharya, U. and Spiegel, M Insiders, Outsiders, and Market Breakdowns. The Review of financial studies. 4 (2):

26 Bai, Y., Chang, E. C. and Wang, J Asset Prices Under Short-Sale Constraints. Working paper, MIT Sloan School of Management. Chakrabarty, B. and Shkilko, A Information Transfers and Learning in Financial Markets: Evidence from Short Selling Around Insider Sales. Journal of Banking and Finance, 37 (5): Chan, K., Kot, H. and Yang, Z Effects of Short-Sale Constraints on Stock Prices and Trading Activity: Evidence from Hong Kong and Mainland China. Working paper, HKUST Business School. Chan, K., Menkveld, A. J. and Yang, Z Information Asymmetry and Asset Prices: Evidence from the China Foreign Share Discount. The Journal of Finance, 63 (1): Chang, E. C., Luo, Y. and Ren, J Short-Selling, Margin-Trading, and Price Efficiency: Evidence from the Chinese Market. Journal of Banking and Finance, 48: Chang, E. C., Cheng J. W. and Yu, Y Short-Sales Constraints and Price discovery: Evidence from the Hong Kong Market. The Journal of Finance, 62 (5): Charoenrook, A. and Daouk, H A Study of Market-Wide Short-Selling Restrictions.? working paper. Chen, H. and Singal, V Role of Speculative Short Sales in Price Formation: Case of the Weekend Effect. The Journal of Finance. 58: Chen, J., Hong, H. and Stein, J. C Breadth of Ownership and Stock Returns. Journal of Financial Economics, 66: Cohen, L., Diether, K. B. and Malloy. C. J Supply and Demand Shifts in the Shorting Market. The Journal of Finance, 62 (5): D Avolio, G. M The Market for Borrowing Stock. Journal of Financial Economics, 66 (2-3): Diamond, D. W. and Verrecchia, R. E Constraints on Short-Selling and Asset Price Adjustment to Private Information. Journal of Financial Economics, 18 (2): Duffie, D., Garleanu, N. and Pedersen, L. H Over-the-Counter Markets. Econometrica. 73 (6), Dimson, E., Karakaş, O. and Li, X Active Ownership. working paper. Diether, K. B., Lee, K. H. and Werner, I. M Short-Sale Strategies and Return Predictability. The Review of financial studies, 22 (2): Edmans, A Blockholder Trading, Market Efficiency, and Managerial Myopia. The Journal of Finance, 64(6): Edmans, A Blockholders and Corporate Governance. Working Paper Series. Easley, D. and O Hara, M Information and the Cost of Capital. The Journal of Finance, 59 (4):

27 Engelberg, J. E., Reed, A. V. and Ringgenberg, M. C How are Shorts Informed?: Short Sellers, News and Information Processing. Journal of Financial Economics. 105 (2): Engelberg, J. E., Reed, A. V. and Ringgenberg, M. C Short Selling Risk. working paper. Fang, Z. and Jiang, X Effects of Differences of Opinions and Short-Sale Constraints on the Dual Listed Chinese Shares. China Finance Review International, 3 (1): Foucault, T., Sraer, D. and Thesmar, D. J Individual Investors and Volatility. The Journal of Finance, 66 (4): Glosten, L. R. and Harris, L. E Estimating the Components of the Bid-Ask Spread. Journal of Financial Economics, 21: Goyenko, R. Y., Holden, C. W. and Trzcinka, C. A Do liquidity measures measure liquidity? Journal of Financial Economics, 92 (2): Glosten, L. R. and Milgrom, P. R Bid, Ask and Transaction Prices in a Specialist Market with Heterogeneously Informed Traders. Journal of Financial Economics, 14: Gârleanu, N. and Pedersen, L. H Margin-based Asset Pricing and Deviations from the Law of One Price. The Review of financial studies. 24 (6): Grossman, S. J. and Miller, M. H Liquidity and Market Structure. The Journal of Finance, 43 (3): Harris, L., Namvar, E. and Phillips, B Price Inflation and Wealth Transfer During the 2008 SEC Short-Sale Ban. Journal of Investment Management. Second Quarter Hardouvelis, G. A. and Perisiani, S Margin Requirements, Speculative Trading, and Stock Price Fluctuations: The Case of Japan. The Quarterly Journal of Economics, 107 (4): He, Y., Nielsson, U., Guo, H. and Yang, J Subscribing to Transparency. Journal of Banking and Finance. 44: Hong, H. and Stein, J. C A Unified Theory of Underreaction, Momentum Trading and Overreaction in Asset Markets. The Journal of Finance, 54 (6): Hong, H. and Stein, J. C Differences of Opinion, Short-Sales Constraints, and Market Crashes. The Review of financial studies, 16 (2): Hong, H. and Stein, J. C Disagreement and the Stock Market. Journal of Economic Perspectives, 21(2): Hong, H. and Kacperczyk, M Competition and Bias. The Quarterly Journal of Economics. 125 (4): Jones, C. M. and Lamont, O. A Short-Sale Constraints and Stock Returns. Journal of Financial Economics, 66 (2-3):

28 Jegadeesh, N. and Titman, S Returns to Buying Winners and Selling Losers: Implications for Stock Market Efficiency. The Journal of Finance, 48 (1): Jegadeesh, N. and Titman, S Profitability of Momentum Strategies: An Evaluation of Alternative Explanations. The Journal of Finance, 56 (2): Karl B. Diether, Christopher J. Malloy and Anna Scherbina Differences of Opinion and the Cross Section of Stock Returns. The Journal of Finance, 57 (5): Kahraman, B. and Tookes, H Leverage Constraints and Liquidity: What Can We Learn from Margin Trading? working paper. Karpoff, J. M. and Luo, X Short Sellers and Financial Misconduct. The Journal of Finance, 65 (5): Largay, J. A % Margins: Combating Speculation in Individual Security Issues. The Journal of Finance, 28 (4): Miller, E. M Risk, Uncertainty, and Divergence of Opinion. The Journal of Finance, 32 (4): Mendelson, H., and Tunca, T. I Strategic Trading, Liquidity, and Information Acquisition. The Review of financial studies, 17 (2): Mei, J., Scheinkman, J. and Xiong, W Speculative Trading and Stock Prices: Evidence from Chinese A-B Share Premia. Annals of Economics and Finance, 10: Milgrom, P. and Stokey, N Information, Trade and Common Knowledge. Journal of Economic Theory, 26 (1): Michaely, R. and Vincent, C Do Institutional Investors Influence Capital Structure Decisions? working paper Nagel, S Short Sales, Institutional Investors and the Cross-Section of Stock Returns. Journal of Financial Economics. 78: Saffi, P. A. C. and Sigurdsson, K Price Efficiency and Short Selling. The Review of financial studies, 24 (3): Spiegel, M. I. and Subrahmanyam, A , Informed Speculation and Hedging in a Noncompetitive Securities Market The Review of financial studies, 5 (2): Stambaugh, R. F., Yu, J. and Yuan, Y Arbitrage Asymmetry and the Idiosyncratic Volatility Puzzle. working paper Sharif, S., Anderson, H. D. and Marshall, B. R The Announcement and Implementation Reaction to China's Margin Trading and Short Selling Pilot Programme. International Journal of Managerial Finance, 10 (3):

29 Sharif, S., Anderson, H. D. and Marshall, B. R Against the Tide: The Commencement of Short Selling and Margin Trading in Mainland China. Accounting and Finance, DOI: /acfi Thompson, S. B Simple Formulas for Standard Errors that Cluster by Both Firm and Time. Journal of Financial Economics. 99 (1): Wang, A. W. and Zhang, G Institutional Ownership and Credit Spreads: An Information Asymmetry Perspective. Journal of Empirical Finance. 16 (4):

30 Figure 1 plots the price impact of the addition event for the treated stocks of the Chinese margin-trading and short-selling program. The day when the addition is implemented is taken as day zero in the figure. Abnormal returns (in %) are estimated by using the Fama-French three factors model. The period from -300 to -50 days to the event date is taken as the estimation window. The solid line is for the cumulative abnormal return, the dash lines are the confidence intervals in 95% level. Abnormal returns are shown by the histograms. 29

31 Figure 2 presents the evolution of the difference of the liquidity measurement between the treated stocks and the control ones during the period covering 20 weeks before and after margin-trading and short-selling are allowed. Similar to the regression analysis, I drop the week of event date. So the first week after when the treatment is implemented is week zero in this figure. The solid line stands for the difference of the liquidity measurement between the treated and the control stocks, i.e., the left side of regression equation (2). The dash lines are for the 95% confidence intervals which are established by regressing the difference of the liquidity measurement on weekly dummies instead of the post dummy used in equation (2), and standard errors are clustered by stock and week. 30

32 Table 1: Timeline of Chinese Margin-Trading and Short-Selling Program This table reports the main revisions of Chinese margin-trading and short-selling program through the end of The information is from statements by the Shanghai Stock Exchange and Shenzhen Stock Exchange. Panel A reports the main four addition events in which new stocks are added into the eligible securities list for the program. Column 1 reports the date when each addition becomes eligible. Column 2 reports the date when each additions was announced by the Shanghai and Shenzhen Stock Exchanges. Column 3 reports the number of treated stocks that are studied in this paper. Panel B reports the announcement and effective date of the refinance business. Effective Date Announcement Date Number of Treated Stocks Panel A Addition Event March 31, 2010 February 12, December 5, 2011 November 25, January 31, 2013 January 25, September 16, 2013 September 6, Cumulated 718 Panel B Margin Refinancing Business August 30, 2012 August 27, 2012 all stocks that are already eligible for the margin-trading and short-selling program Security Refinancing Business February 28, 2013 February 26, from March 31, 2010 addition event treated stocks 23 from December 5, 2011 addition event treated stocks 10 from January 31, 2013 addition event treated stocks Cumulated 86 September 18, 2013 September 16, from March 31, 2010 addition event treated stocks 64 from December 5, 2011 addition event treated stocks 80 from January 31, 2013 addition event treated stocks 42 from September 16, 2013 addition event treated stocks Cumulated

33 Table 2: Summary Statistics This table reports the mean and the standard error for each variable used in the regressions. Column 2 is for the treated stocks from the addition events, Column 3 is for the control group which includes the exempt stocks that are matched to the treated stocks according to the market capitalization and the trading volume, and Column 4 is for all the exempt stocks that are not eligible for margin-trading and short-selling. MV is the market capitalization in thousands RMB yuan. Volatility is the standard deviation of daily stock returns within a week. Ret is the stocks weekly return. Three liquidity measurements are used. Spread is the averaged daily Bid-Ask spread within one week, which is recorded as a percentage. Illiquidity is the Amihud illiquidity ratio, which is defined as the absolute value of daily stock returns divided by trading volume in value and then averaged over one week, is in basis point per million RMB yuan. Turnover Ratio is defined as the weekly averaged ratio of daily trading volume divided by common shares outstanding. For the liquidity measurements, I report the summary statistics separately for the periods before and after the treatment. Variables Treated Stocks Matched Stocks All Ineligible Stocks Ln(MV) (1.425) (1.356) (1.051) Volatility (%) (1.249) (1.348) (1.367) Ret (%) (6.127) (6.458) (6.288) Sd(forecasts of current year EPS) (0.271) (0.290) (0.380) Sd(forecasts of next year EPS) (0.256) (0.282) (0.357) Funds Holding Ratio (%) (9.931) (10.796) (9.834) Before After Before After Before After Bid-Ask Spread (%) (0.084) (0.087) (0.079) (0.076) (0.089) (0.091) Amihud Illiquidity Ratio (4.059) (2.907) (5.917) (3.613) (9.700) (8.884) Turnover Ratio (%) (1.416) (1.281) (1.432) (1.363) (1.415) (1.413) 32

34 Table 3: Liquidity Impact This table reports the impact of margin-trading and short-selling on liquidity. Three liquidity measurements are used. Spread is the averaged daily Bid-Ask spread within one week, which is recorded as a percentage. Illiquidity is the Amihud illiquidity ratio, which is defined as the absolute value of daily stock returns divided by trading volume in value and then averaged over one week, is in basis point per million RMB yuan. Turnover Ratio is defined as the weekly averaged ratio of daily trading volume divided by common shares outstanding. Post is the dummy variable for the period when margin-trading and short-selling are implemented. MV is market value of stocks. Volatility is the standard deviation of daily stock returns within one week. Lag_ret is the stock returns of the previous week. Standard errors are clustered by both stock and week. t statistics are in parentheses. *, ** and *** denote significance at the 10%, 5% and 1% level respectively. Spread Illiquidity Turnover (1) (2) (3) (4) (5) (6) (7) (8) (9) Post *** *** *** *** *** *** *** *** *** (6.290) (6.126) (5.718) (5.027) (4.814) (2.623) (-2.941) (-2.849) (-8.365) Ln(MV) ** ** *** *** (-0.719) (-0.935) (-2.322) (-2.185) (-4.405) (-5.047) Volatility *** *** *** *** (-7.521) (-7.358) (-0.659) (-0.122) (21.492) (21.719) Lag_ret * * *** *** *** *** (-1.795) (-1.647) (-6.699) (-6.507) (15.976) (16.081) Observations Stock-pair FEs Yes Yes Yes Yes Yes Yes Yes Yes Yes Year-week FEs No No Yes No No Yes No No Yes R

35 Table 4: Liquidity Impact: Results by each Event This table reports the impact of margin-trading and short-selling on liquidity for each addition event separately. Three liquidity measurements are used. Spread is the averaged daily Bid-Ask spread within one week, which is recorded as a percentage. Illiquidity is the Amihud illiquidity ratio, which is defined as the absolute value of daily stock returns divided by trading volume in value and then averaged over one week, is in basis point per million RMB yuan. Turnover Ratio is defined as the weekly averaged ratio of daily trading volume divided by common shares outstanding. Post is the dummy variable for the period when margin-trading and short-selling are implemented. MV is market value of stocks. Volatility is the standard deviation of daily stock returns within one week. Lag_ret is the stock return of last week. Standard errors are clustered by both stock and week. t statistics are in parentheses. *, ** and *** denote significance at the 10%, 5% and 1% level respectively March Addition Event 2011 December Addition Event 2013 January Addition Event 2013 September Addition Event Spread Illiquidity Turnover Spread Illiquidity Turnover Spread Illiquidity Turnover Spread Illiquidity Turnover Post * *** *** *** *** *** *** (0.812) (1.929) (-1.207) (4.603) (0.499) (-0.196) (2.617) (2.671) (-1.520) (3.776) (4.602) (-2.672) Ln(MV) * * ** *** (1.843) (-1.679) (0.665) (-0.357) (0.508) (-1.645) (-1.553) (-2.283) (-4.313) Volatility ** *** *** *** *** *** *** *** *** *** *** (-2.119) (2.926) (5.881) (-2.622) (3.390) (11.951) (-5.812) (0.081) (14.308) (-4.596) (-2.694) (14.737) Lag_ret *** *** *** *** *** *** *** *** (-0.009) (-3.889) (4.458) (-1.187) (-8.240) (7.078) (-0.671) (-3.838) (12.458) (-1.563) (-4.871) (9.422) Observations Stock-pair FEs Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes R

36 Table 5: Liquidity Impact: Analyst Coverage This table reports the regression results investigating the mechanism underlying the negative liquidity impact of margin-trading and short-selling. Three measurements are used. Spread is the averaged daily Bid-Ask spread within one week, which is recorded as a percentage. Illiquidity is the Amihud illiquidity ratio, which is defined as the absolute value of daily stock returns divided by trading volume in value and then averaged over one week, is in basis point per million RMB yuan. Turnover Ratio is defined as the weekly averaged ratio of daily trading volume divided by common shares outstanding. Asym is a dummy variable, equal to 1 if there is no analyst coverage; equal to 0, otherwise. MV is market value of stocks. Volatility is the standard deviation of daily stock returns within one week. Lag_ret is the stock return of last week. Standard errors are clustered by both stock and week. t statistics are in parentheses. *, ** and *** denote significance at the 10%, 5% and 1% level respectively. Spread Illiquidity Turnover (1) (2) (3) (4) (5) (6) (7) (8) (9) Post *** *** *** *** *** ** *** *** *** (6.089) (6.003) (5.622) (4.831) (4.630) (2.519) (-2.910) (-2.753) (-8.339) Post*Asym * * ** ** ** (1.873) (1.536) (1.828) (2.503) (2.119) (2.210) (0.061) (-0.591) (-0.661) Ln(MV) ** ** *** *** (-0.416) (-0.584) (-2.206) (-2.060) (-4.260) (-4.937) Volatility *** *** *** *** (-7.536) (-7.383) (-0.643) (-0.116) (21.507) (21.734) Lag_ret * * *** *** *** *** (-1.790) (-1.651) (-6.699) (-6.517) (15.967) (16.080) Observations Stock-pair FEs Yes Yes Yes Yes Yes Yes Yes Yes Yes Year-week FEs No No Yes No No Yes No No Yes R

37 Table 6: Liquidity Impact: St. Dev. of Analysts Current Year EPS Forecasts This table reports the regression results investigating the mechanism underlying the negative liquidity impact of margin-trading and short-selling. Three measurements are used. Spread is the averaged daily Bid-Ask spread within one week, which is recorded as a percentage. Illiquidity is the Amihud illiquidity ratio, which is defined as the absolute value of daily stock returns divided by trading volume in value and then averaged over one week, is in basis point per million RMB yuan. Turnover Ratio is defined as the weekly averaged ratio of daily trading volume divided by common shares outstanding. Asym equals to the standard deviation of analysts forecasts about current year EPS. MV is market value of stocks. Volatility is the standard deviation of daily stock returns within one week. Lag_ret is the stock return of last week. Standard errors are clustered by both stock and week. t statistics are in parentheses. *, ** and *** denote significance at the 10%, 5% and 1% level respectively. Spread Illiquidity Turnover (1) (2) (3) (4) (5) (6) (7) (8) (9) Post *** *** ** *** *** *** ** *** (5.160) (5.199) (2.251) (3.947) (3.827) (1.525) (-2.772) (-2.193) (-6.962) Post*Asym * * ** *** *** *** (0.720) (0.746) (0.411) (1.855) (1.903) (2.092) (-3.550) (-4.093) (-4.503) Ln(MV) (-1.022) (-0.981) (-1.486) (-1.640) (-1.097) (-1.057) Volatility *** *** *** *** (-6.395) (-6.440) (0.539) (0.811) (20.080) (20.598) Lag_ret ** * *** *** *** *** (-2.024) (-1.861) (-5.468) (-5.288) (15.093) (15.255) Observations Stock-pair FEs Yes Yes Yes Yes Yes Yes Yes Yes Yes Year-week FEs No No Yes No No Yes No No Yes R

38 Table 7: Liquidity Impact: St. Dev. of Analysts Next Year EPS Forecasts This table reports the regression results investigating the mechanism underlying the negative liquidity impact of margin-trading and short-selling. Three measurements are used. Spread is the averaged daily Bid-Ask spread within one week, which is recorded as a percentage. Illiquidity is the Amihud illiquidity ratio, which is defined as the absolute value of daily stock returns divided by trading volume in value and then averaged over one week, is in basis point per million RMB yuan. Turnover Ratio is defined as the weekly averaged ratio of daily trading volume divided by common shares outstanding. Asym equals to the standard deviation of analysts forecasts about next year EPS. MV is market value of stocks. Volatility is the standard deviation of daily stock returns within one week. Lag_ret is the stock return of last week. Standard errors are clustered by both stock and week. t statistics are in parentheses. *, ** and *** denote significance at the 10%, 5% and 1% level respectively. Spread Illiquidity Turnover (1) (2) (3) (4) (5) (6) (7) (8) (9) Post *** *** ** *** *** ** * *** (4.971) (4.998) (2.177) (3.879) (3.765) (1.323) (-2.556) (-1.952) (-6.234) Post*Asym ** ** ** ** *** *** (0.012) (0.018) (-0.257) (1.962) (2.059) (2.257) (-2.448) (-2.999) (-3.456) Ln(MV) * (-1.356) (-1.279) (-1.502) (-1.661) (-0.984) (-0.884) Volatility *** *** * *** *** (-6.594) (-6.508) (1.370) (1.746) (19.262) (19.834) Lag_ret * * *** *** *** *** (-1.825) (-1.703) (-7.422) (-7.096) (14.361) (14.577) Observations Stock-pair FEs Yes Yes Yes Yes Yes Yes Yes Yes Yes Year-week FEs No No Yes No No Yes No No Yes R

39 Table 8: Liquidity Impact: Institutional Shareholders This table reports the regression results investigating the mechanism underlying the negative liquidity impact of margin-trading and short-selling. Three measurements are used. Spread is the averaged daily Bid-Ask spread within one week, which is recorded as a percentage. Illiquidity is the Amihud illiquidity ratio, which is defined as the absolute value of daily stock returns divided by trading volume in value and then averaged over one week, is in basis point per million RMB yuan. Turnover Ratio is defined as the weekly averaged ratio of daily trading volume divided by common shares outstanding. Asym equals to (1-FundsHoldingRatio). MV is market value of stocks. Volatility is the standard deviation of daily stock returns within one week. Lag_ret is the stock return of last week. Standard errors are clustered by both stock and week. t statistics are in parentheses. *, ** and *** denote significance at the 10%, 5% and 1% level respectively. Spread Illiquidity Turnover (1) (2) (3) (4) (5) (6) (7) (8) (9) Post *** *** *** *** * *** *** *** (3.522) (3.350) (1.566) (3.557) (3.122) (-1.706) (-3.016) (-2.691) (-4.659) Post*Asym *** *** ** *** *** ** *** ** ** (3.153) (3.094) (2.407) (2.596) (2.714) (2.284) (-2.803) (-2.271) (-2.407) Ln(MV) *** *** *** *** (-0.889) (-1.296) (-2.733) (-2.630) (-3.499) (-4.147) Volatility *** *** * *** *** (-5.480) (-5.586) (1.489) (1.709) (18.683) (19.129) Lag_ret * ** *** *** *** *** (-1.938) (-2.033) (-4.563) (-4.398) (13.904) (13.806) Observations Stock-pair FEs Yes Yes Yes Yes Yes Yes Yes Yes Yes Year-week FEs No No Yes No No Yes No No Yes R

40 Table 9: Price discovery This table compares the changes in profitability of the contrarian strategies that are built on the treated stocks with those are built on control ones. Panel A (B, C and D) is for the sample from September 16, 2013 (January 31, 2013, December 5, 2011 and March 31, 2010) addition event. The contrarian strategy is built in a similar way as Jegadeesh and Titman (1993, 2001). On each day, the stocks in the treatment groups are sorted according to past J day return into deciles, and 10th decile (with the highest past return) are sold (the short position), 1st decile (with the smallest past return) are bought (the long position). This portfolio (long+short) is held for the next K days. So there are K portfolios held each day, and the returns for the contrarian strategy is the average return of these K portfolios. The same strategy is built for the control group. Then, with a event window covering 300 days around the event date, the return of the contrarian strategy is regressed on the dummy of post period when margin-trading and short-selling are implemented, the dummy of strategy built on treated stocks, as well as their interaction. This table reports the coefficient of the interaction, first using the return of the contrarian strategy as a dependent variable, and then separately using the return of the long or short position. Standard errors clustered by day are reported in parenthesis. t statistics are in parentheses. *, ** and *** denote significance at the 10%, 5% and 1% level respectively. Panel A: Sample from the September 16, 2013 addition event Sorting according to past J =1 day return Sorting according to past J=3 day return Sorting according to past J=5 day return Sorting according to past J=8 day return Sorting according to past J=10 day return Sorting according to past J=15 day return K=1 K=3 K=5 K=8 K=10 K=15 long+short ** ** * * (2.280) (2.058) (1.919) (1.535) (1.781) (1.109) long position ** ** *** *** *** (-1.288) (-2.096) (-2.492) (-3.047) (-2.754) (-2.987) short position *** *** *** *** *** *** (3.940) (3.756) (3.835) (3.678) (3.550) (3.148) long+short *** ** ** ** ** (2.957) (2.392) (2.127) (2.200) (1.990) (0.727) long position ** ** ** *** (-0.644) (-1.578) (-2.016) (-2.127) (-2.502) (-2.971) short position *** *** *** *** *** *** (4.077) (4.070) (3.853) (3.847) (3.823) (2.861) long+short * (1.673) (0.934) (0.696) (0.894) (0.953) (-0.336) long position * ** *** *** *** *** (-1.665) (-2.388) (-2.611) (-2.626) (-2.594) (-3.233) short position *** *** *** *** *** * (3.187) (2.738) (2.616) (2.803) (2.863) (1.927) long+short (0.871) (0.848) (0.947) (0.891) (0.635) (-0.458) long position * ** *** (-1.096) (-1.212) (-1.317) (-1.711) (-1.987) (-3.197) short position * * ** ** ** (1.887) (1.897) (2.102) (2.244) (2.065) (1.541) long+short * ** (1.946) (1.977) (1.482) (1.095) (0.785) (-0.146) long position * *** (-0.660) (-0.901) (-1.428) (-1.633) (-1.780) (-2.876) short position *** *** *** ** ** (2.862) (3.042) (2.765) (2.395) (2.064) (1.632) long+short (1.226) (0.292) (-0.098) (-0.403) (-0.777) (-0.827) long position ** ** *** *** (-0.982) (-1.625) (-2.107) (-2.385) (-2.732) (-2.746) short position ** (2.236) (1.456) (1.199) (0.984) (0.760) (0.811) 39

41 Panel B: Sample from the January 31, 2013 addition event Sorting according to past J =1 day return Sorting according to past J=3 day return Sorting according to past J=5 day return Sorting according to past J=8 day return Sorting according to past J=10 day return Sorting according to past J=15 day return K=1 K=3 K=5 K=8 K=10 K=15 long+short *** ** (3.292) (2.443) (0.740) (1.047) (1.448) (0.724) long position (-0.099) (-0.832) (-1.541) (-1.192) (-0.994) (-1.295) short position *** *** * * * (3.974) (3.192) (1.928) (1.820) (1.916) (1.631) long+short ** (2.565) (0.635) (0.452) (1.081) (0.947) (0.151) long position (-0.578) (-1.428) (-1.384) (-0.714) (-0.752) (-1.378) short position *** * (3.278) (1.680) (1.403) (1.524) (1.435) (1.200) long+short ** (2.272) (0.092) (0.758) (1.372) (1.300) (0.393) long position (-0.015) (-1.479) (-0.591) (0.045) (-0.130) (-0.696) short position ** (2.578) (1.098) (1.206) (1.338) (1.381) (0.913) long+short * (1.807) (1.288) (1.353) (1.415) (1.011) (0.224) long position (0.740) (0.616) (0.711) (0.448) (-0.033) (-0.704) short position (1.607) (1.070) (1.052) (1.275) (1.158) (0.794) long+short ** (2.510) (1.129) (0.972) (0.714) (0.434) (-0.159) long position (0.969) (-0.053) (-0.209) (-0.517) (-0.825) (-1.083) short position ** (2.207) (1.379) (1.293) (1.231) (1.164) (0.707) long+short (1.176) (0.651) (0.232) (0.241) (0.090) (-0.681) long position (-0.826) (-1.095) (-1.331) (-1.033) (-1.114) (-1.409) short position * (1.930) (1.502) (1.207) (1.061) (0.968) (0.349) 40

42 Panel C: Sample from the December 5, 2011 addition event Sorting according to past J =1 day return Sorting according to past J=3 day return Sorting according to past J=5 day return Sorting according to past J=8 day return Sorting according to past J=10 day return Sorting according to past J=15 day return K=1 K=3 K=5 K=8 K=10 K=15 long+short (1.013) (0.545) (0.201) (-0.425) (0.307) (0.215) long position * * * * * (1.777) (1.708) (1.672) (1.322) (1.731) (1.905) short position * (-0.395) (-0.981) (-1.360) (-1.585) (-1.356) (-1.657) long+short (1.111) (0.018) (0.176) (-0.282) (0.381) (0.397) long position ** ** * * * (2.168) (1.994) (1.832) (1.595) (1.691) (1.898) short position (-0.394) (-1.493) (-1.292) (-1.586) (-1.047) (-1.313) long+short (0.146) (-0.301) (-0.471) (-0.419) (0.188) (0.147) long position (0.582) (0.893) (0.937) (0.777) (0.969) (1.352) short position (-0.313) (-1.111) (-1.358) (-1.127) (-0.637) (-1.077) long+short (-0.156) (-0.432) (-0.164) (0.198) (0.283) (0.318) long position * (0.087) (0.663) (0.826) (1.005) (1.361) (1.664) short position (-0.252) (-1.036) (-0.860) (-0.587) (-0.810) (-1.124) long+short (-0.202) (-0.220) (-0.171) (0.207) (0.315) (0.217) long position (0.337) (0.418) (0.517) (1.062) (1.326) (1.445) short position (-0.526) (-0.618) (-0.631) (-0.639) (-0.752) (-1.024) long+short (-0.010) (0.141) (0.468) (0.472) (0.378) (0.205) long position (0.814) (1.025) (1.219) (1.407) (1.490) (1.552) short position (-0.655) (-0.675) (-0.470) (-0.638) (-0.841) (-1.144) 41

43 Panel D: Sample from the March 31, 2010 addition event Sorting according to past J =1 day return Sorting according to past J=3 day return Sorting according to past J=5 day return Sorting according to past J=8 day return Sorting according to past J=10 day return Sorting according to past J=15 day return K=1 K=3 K=5 K=8 K=10 K=15 long+short ** (1.067) (2.039) (1.165) (0.001) (0.112) (-0.136) long position (0.987) (1.409) (1.084) (0.710) (0.885) (0.598) short position (0.348) (0.766) (0.130) (-0.518) (-0.606) (-0.536) long+short (0.774) (0.770) (0.468) (-0.250) (-0.412) (0.175) long position (0.577) (0.866) (0.863) (0.715) (0.642) (0.785) short position (0.398) (0.132) (-0.197) (-0.742) (-0.812) (-0.496) long+short * (1.770) (0.886) (-0.139) (-0.445) (-0.772) (-0.020) long position (1.372) (0.649) (0.532) (0.688) (0.519) (0.839) short position (0.967) (0.495) (-0.517) (-0.904) (-1.042) (-0.607) long+short (0.342) (0.296) (-0.613) (-0.876) (-0.501) (0.371) long position (0.407) (0.984) (0.684) (0.979) (1.158) (1.098) short position (0.080) (-0.370) (-1.077) (-1.488) (-1.250) (-0.406) long+short (0.367) (-0.460) (-1.156) (-1.068) (-0.563) (0.190) long position (1.623) (1.112) (0.458) (0.554) (0.768) (1.070) short position (-0.750) (-1.207) (-1.459) (-1.433) (-1.067) (-0.543) long+short (-0.802) (-0.390) (-0.371) (-0.118) (0.056) (0.704) long position (0.440) (0.290) (0.217) (0.506) (0.717) (0.925) short position (-1.173) (-0.609) (-0.551) (-0.461) (-0.410) (0.095) 42

44 Table 10: Liquidity Impact: Security Refinancing Business This table reports the liquidity impact of the security refinancing business. Three liquidity measurements are used. Spread is the averaged daily Bid-Ask spread within one week, which is recorded as a percentage. Illiquidity is the Amihud illiquidity ratio, which is defined as the absolute value of daily stock returns divided by trading volume in value and then averaged over one week, is in basis point per million RMB yuan. Turnover Ratio is defined as the weekly averaged ratio of daily trading volume divided by common shares outstanding. Post is the dummy variable for the period when margin-trading and short-selling are implemented. MV is market value of stocks. Volatility is the standard deviation of daily stock returns within one week. Lag_ret is the stock return of last week. Standard errors are clustered by both stock and week. t statistics are in parentheses. *, ** and *** denote significance at the 10%, 5% and 1% level respectively. In column (1) to (3) of Panel A, the stocks which are eligible for the security refinancing business are compared with the matched ones which are not eligible for margin-trading and short-selling. In column (4) to (6) of Panel B, the stocks which are eligible for the security refinancing business are compared with the matched ones which are eligible for margin-trading and short-selling but not eligible for the security refinancing business. Panel A Panel B Spread Illiquidity Turnover Spread Illiquidity Turnover (1) (2) (3) (4) (5) (6) Post *** *** * ** (7.410) (6.251) (-1.958) (0.951) (0.472) (-2.195) Ln(MV) ** *** (0.374) (2.118) (-0.124) (-1.136) (-2.963) (-0.049) Volatility *** *** *** * *** (-3.388) (0.313) (11.562) (-7.945) (1.737) (14.156) Lag_ret *** *** *** *** *** (-3.448) (-6.296) (6.455) (-1.543) (-7.767) (9.606) Observations Stock-pair FEs Yes Yes Yes Yes Yes Yes Year-week FEs Yes Yes Yes Yes Yes Yes R

45 Table 11: Impact on Stock Return Volatility and Autocovariance This table reports the impact of margin-trading and short-selling on stock return volatility and autocovariance as in Foucault, Sraer and Thesmar (2011). Volatility is the standard deviation of the residual of the market model, that is, the time-series regression of the daily excess return of stocks on the daily excess market return. Autocov is the weekly autocovariance of the daily returns of stocks. MV is market value of stocks. Lag_ret is the stock return of last week. Standard errors are clustered by both stock and week. t statistics are in parentheses. *, ** and *** denote significance at the 10%, 5% and 1% level respectively. Volatility Autocov (1) (2) (3) (4) (5) (6) Post *** *** *** *** *** *** (-2.757) (-3.240) (-4.872) (-3.165) (-4.080) (-8.465) Ln(MV) (-0.775) (-1.404) (0.234) (-0.122) Lag_ret *** *** *** *** (6.441) (6.591) (5.485) (5.424) Observations Stock-pair FEs Yes Yes Yes Yes Yes Yes Year-week FEs No No Yes No No Yes R

46 Table 12: Liquidity Impact on ETF Funds This table reports the liquidity impact of margin-trading and short-selling on ETF Funds. Two liquidity measurements are used.. Illiquidity is the Amihud illiquidity ratio, which is defined as the absolute value of daily stock returns divided by trading volume in value and then averaged over one week, is in basis point per million RMB yuan. Volume is the weekly averaged log of daily trading volume. Post is the dummy variable for the period when margin-trading and short-selling are implemented. Volatility is the standard deviation of daily returns within one week. Lag_ret is the fund returns of the previous week. Similar to the matched sample estimation approach used in this paper, I create a match for the treated ETF funds by using the average liquidity measures which are computed on all the ETF funds that are not eligible for margin-trading and short-selling in the Chinese stock market. The dependent variables in the regressions of this table are the differences between the ETF funds liquidity measure and this average liquidity measure. Then, the same regression as equation (2) can be used. Standard errors are clustered by both fund and week. t statistics are in parentheses. *, ** and *** denote significance at the 10%, 5% and 1% level respectively. Column (1) and (2) study the liquidity impact for all the treated ETF funds; column (3) (4) and column (5) (6) study the liquidity impact for the passive and active ETF funds respectively. All Funds Passive Funds Active Funds Illiquidity Ln(Volume) Illiquidity Ln(Volume) Illiquidity Ln(Volume) (1) (2) (3) (4) (5) (6) Post *** 0.492*** *** 0.603*** *** (-3.122) (4.325) (-3.552) (7.077) (-2.693) (1.122) Volatility *** *** 1.445* 0.144*** (0.233) (3.377) (-0.390) (3.426) (1.668) (3.256) Lag_ret ** *** (-0.608) (2.437) (-0.864) (1.143) (-0.078) (2.675) Observations Fund FEs Yes Yes Yes Yes Yes Yes R

47 Appendix: Liquidity Impact on Cross-Listed Stocks According to the analysis of my paper, the negative liquidity impact of the Chinese margin-trading and short-selling program is due to the information asymmetry that causes uninformed investors to avoid trading. Then, the negative liquidity impact should be less pronounced for the stocks with less information asymmetry. In line with this prediction, this section studies the liquidity impact on the cross-listed stocks for which there is likely to be less concern for the information asymmetry. In my data, there are 61 stocks also listed on the Hong Kong Stock Exchange. I use a dummy for these cross-listed stocks and interact it with the post dummy in equation (2) to capture the difference of liquidity impact between the cross-listed stocks and the rests. The results are reported in the following table. We can see that, although the results are weak, the sign of the coefficients of the interaction part between the cross-listed stocks dummy and post dummy are all in line with my prediction. 46

48 Liquidity Impact: Cross-Listed Stocks This table reports the impact of margin-trading and short-selling on liquidity for the cross-listed stocks. Three liquidity measurements are used. Spread is the averaged daily Bid-Ask spread within one week, which is recorded as a percentage. Illiquidity is the Amihud illiquidity ratio, which is defined as the absolute value of daily stock returns divided by trading volume in value and then averaged over one week, is in basis point per million RMB yuan. Turnover Ratio is defined as the weekly averaged ratio of daily trading volume divided by common shares outstanding. Post is the dummy variable for the period when margin-trading and short-selling are implemented. Cross is the dummy variable, equals to one for the stocks which are also listed on the Hong Kong Stock Exchange; equals to zero, otherwise. MV is market value of stocks. Volatility is the standard deviation of daily stock returns within one week. Lag_ret is the stock returns of the previous week. Standard errors are clustered by both stock and week. t statistics are in parentheses. *, ** and *** denote significance at the 10%, 5% and 1% level respectively. Spread Illiquidity Turnover (1) (2) (3) (4) (5) (6) (7) (8 (9) Post 0.013*** 0.013*** 0.027*** 1.042*** 0.947*** 1.028*** *** *** *** (6.353) (6.262) (5.743) (5.059) (4.854) (2.629) (-2.976) (-2.925) (-8.384) Post*Cross ** ** (-1.299) (-1.499) (-0.697) (-2.243) (-2.500) (-0.945) (1.420) (1.555) (1.228) Ln(MV) ** ** *** *** (-0.729) (-0.951) (-2.326) (-2.190) (-4.381) (-5.006) Volatility *** *** *** 0.350*** (-7.539) (-7.368) (-0.663) (-0.128) (21.510) (21.734) Lag_ret * * *** *** 0.041*** 0.041*** (-1.799) (-1.651) (-6.703) (-6.508) (15.974) (16.086) Observations Stock-pair FEs Yes Yes Yes Yes Yes Yes Yes Yes Yes Year-week FEs No No Yes No No Yes No No Yes R