Does Short Selling Improve Stock Price Efficiency and Liquidity? Evidence from a Natural Experiment in China

Transcription

1 Does Short Selling Improve Stock Price Efficiency and Liquidity? Evidence from a Natural Experiment in China Abstract China introduced the short-selling for designated stocks in March Using this important policy change as a natural experiment, we examine the effect of short-selling constraints on stock price efficiency and liquidity. We show that the introduction of short selling significantly improves price efficiency, as measured by the differences in individual stock responses to market returns and the delay in price adjustments. Short selling also enhances stock liquidity, as measured by bid-ask spread; and reduces stock volatility. Overall, our results suggest that short selling helps stabilize asset prices, provides additional liquidity and improves market quality, even in an emerging economy with a less developed stock market than that in the U.S. Given the recent debate on the pros and cons of short-selling, and various policy considerations during financial crisis, our findings provide important evidence with respect to the effect of short-selling on the stock market. JEL classification: G12, G14, G15, G18 Keywords: Short selling; Price efficiency; Liquidity; Margin trading; Chinese market

2 1. Introduction The impact of short-selling constraints on financial market has attracted the attention of investors, policy makers, and academic researchers for a long time. Millier s (1977) seminal work shows that more pessimistic investors are excluded from the market if short selling is not allowed, leading to systematic overvaluation of the securities. As an interesting contrast, Diamond and Verrecchia (1987) predict no overvaluation, as rational investors already take this constraint into their expected return framework. Empirical studies have provided mixed evidence on the impact of short-selling constraints on stock market stability and price efficiency. Some studies report findings consistent with the overvaluation view (e.g., Autore et al., 2011; Chang et al., 2007; Chen et al., 2002); while other studies find that short selling is often associated with market crashes and price manipulation (e.g., Hong and Stein, 2003; Allen and Gale, 1992; Goldstein and Guembel, 2008). Although short selling bans are generally observed whenever a market appears to be vulnerable, they may have severe unintended consequences (Battalio and Schultz, 2013; Beber and Pagano, 2013). 1 Bris et al. (2007) point out that while short selling restrictions are nearly as old as organized exchanges, there is little empirical evidence on whether they prevent or facilitate market crashes, or whether they hinder or promote price discovery. One reason for the mixed evidence could be that the effect of short selling also depends on the development of the stock market, availability of other short instruments, investor trading behavior, and legal environment and regulatory policies. These 1 In particular, Battalio and Schultz (2013) find that short sale ban is associated with dramatically increased hedging costs for options on banned stocks. Beber and Pagano (2013) show that short-selling bans are detrimental for liquidity, slow down price discovery and fail to support stock prices. 1

3 confounding factors have made it difficult to isolate the effect of short-selling constraints on market stability and price efficiency. In this study, we take advantage of an important policy change in the Chinese stock market and use it as a natural experiment to examine the effect of short sale on price efficiency and liquidity. China launched a pilot program in March 2010 when 90 designated stocks were made eligible for margin trading and short selling. This is a remarkable reform in the Chinese stock market, as short selling has been banned since the introduction of the Chinese stock market in the early 1990s. 2 There are several advantages to examine the effect of this policy change. First, unlike many developed markets in which multiple short selling vehicles coexist, there are no other alternatives to short-sell a stock in China. Consequently, the short selling mechanism provides one major way for Chinese investors to express their bearish views. This unique institutional feature allows us to construct a clean analysis compared with previous studies that focus on other markets. Second, stocks on the designated short sale list continue to change in our sample period. We have one sample of stocks that were added to the short selling list, and another sample of stocks that were removed from the list. This interesting feature allows us to construct a more robust analysis by looking at how inclusion in and exclusion from the short sale list affect the price efficiency differently. The results from the delisted sample will serve as a natural control sample and provide corroborating evidence on whether short-selling improves price efficiency and liquidity. Third, the privilege of short selling is not available to every investor. The qualified investors are typically more sophisticated and more likely to be informed. On 2 Section 2 provides the institutional background of the short selling policy in China. 2

4 one hand, short selling by these investors is more likely to alleviate price bias. On the other hand, allowing only a small number of informed investors to short sell may help prevent the market from crashing. Macey et al. (1988), Morris and Shin (1998) and Altken (1998) warn that short selling may cause self-fulfilling prophecies and aggravate security price plummeting. Based on the practice of short selling in China, our findings suggest that such a gradual approach to introduce short selling appears to be beneficial to market stability. Fourth, many recent studies on short selling focus on how short selling bans decrease market efficiency (Kolasinksi et al., 2013;Frino et al., 2011). The decision to ban short selling is usually accompanied by a series of government interventions and market stimulus plans, as observed in the recent 2008 financial crisis. This could make it difficult to interpret the effects of short selling bans, as various market forces are in play the same time. Furthermore, the effect of short selling bans and initiations may not be symmetric, and there is very little empirical evidence available in terms of the improved price efficiency brought forth by the introduction of short selling in a market. Our study fills this gap by examining the effect of introducing short-selling on stock price efficiency and liquidity. Finally, the Chinese stock market is generally considered a speculative market (Mei et al., 2009). Price efficiency is lower than that observed in developed markets. Investor trading behaviors are sometimes incongruent with theories based on markets with higher price efficiency. Given the rapid growth of Chinese economy, it is therefore important to understand the effect of short selling on price efficiency in this important stock market. 3

5 Using Chinese market data from April 2010 to December 2013, we show that the introduction of short selling in the Chinese market significantly improves stock price efficiency, as measured by the differences in individual stock responses to market returns (Bris et al., 2007) and the delay in price adjustments (Hou and Moskowitz, 2005). Short selling also enhances stock liquidity, as measured by bid-ask spread; and reduces stock volatility. We obtain consistent results when using regression analysis and difference-indifferences analysis. Our findings have important policy implications as China continues to expand its short selling program and ensure its market operates in an efficient and orderly fashion. The insights gleaned from the experience of the Chinese market with regards to short selling could also benefit other emerging markets that are considering similar market reforms. Given the recent debate on the pros and cons of short-selling, and various policy considerations during financial crisis, our findings provide important evidence with respect to the effect of short-selling on the stock market. The remainder of this paper is organized as follows. Section 2 discusses the institutional background of short selling reform in China. Section 3 presents the related literature. Section 4 describes the variables and empirical methodology. Section 5 reports the data and our empirical analyses. We conclude in Section Institutional Background China Securities Regulatory Commission (CSRC) introduced the margin trading and short selling program in March 2010, 3 when 90 constituent stocks from the Shanghai Stock Exchange (SSE) 50 Index and the Shenzhen Stock Exchange (SZSE) Component 3 CSRC also conducted a few small pilot programs in 2007 and 2008 by picking 11 top brokerages to run a trial program, which lasted for about two years. 4

6 Index were made eligible for margin trading and short selling. The CSRC then expanded the program on December 5, 2011 by adding 278 constituent stocks from the SSE 180 Index and the SZSE 100 Index to the eligibility list. Seven ETFs were also added to the list. The CSRC substantially increased the number of eligible stocks to 500 on January 31, According to Reuters, 4 totally 74 brokerages are allowed to conduct business and over 500,000 investors are able to trade under the new program. For a firm to be eligible for short selling, it must have a minimum of 200 million tradable shares, with a public float no less than RMB800 million. In addition, the firm must have more than 4,000 shareholders. Furthermore, on a three-month rolling basis, the daily turnover must be more than 15% of the index turnover, and the daily trading volume must be more than RMB50 million. Meanwhile, only qualified investors who have a good trading record, low bankruptcy risk and minimum capital of RMB500,000 and who demonstrate basic investment knowledge are allowed to short sell. As shown in Table 1, the number of stocks eligible for short selling increases from 90 in 2010 to 278 in 2011 and to 700 in A small number of mutual funds and exchange traded funds (ETFs) are also eligible for short selling. However, due to the small sample size and different investor clientele, we do not include them in our sample when conducting empirical analysis. [Insert Table 1 Here] 3. Literature Review 4 5

7 Short selling serves multiple roles in the financial market. It allows investors to profit from expected price declines in a stock and is an important component of hedging strategies. In experimental settings, many researchers (Lei et al., 2001; Noussair et al., 2001; Ackert et al., 2002; Haruvy et al., 2007; Fellner and Theissen, 2011) find that the introduction of a short selling mechanism significantly lowers price volatility and improves price efficiency. Hong and Stein (2003) find evidence that supports the notion of asset undervaluation due to short sale constraints. They show that short sale constraints may lead to a negatively skewed return distribution and even to overall market crashes. However, Porter and Smith (1995) and Haruvy and Noussair (2006) find no significant effect on price efficiency after relaxing the short sale constraints in their experiments. Short selling activities may also be negatively related to price efficiency if they exacerbate price manipulation. Goldstein and Guembel (2008) show that trading without information is profitable only with sell orders, an observation that justifies the restriction of short selling in the market to mitigate price manipulation. Brunnermeier and Pederson (2005) also suggest that short selling is associated with possible predatory trading that could result in price overshooting and decreased asset liquidation value. In this way, the market becomes illiquid and prone to widespread crisis. Empirical studies of the relation between short selling and price efficiency produce mixed results. Some studies find significant relations between the two. For example, Asquith and Meulbroek (1995), Aitken et al. (1998) and Danielsen and Sorescu (2011) find that the introduction of short selling helps prices to incorporate negative information more quickly. Christophe et al. (2004), Boehmer et al. (2008) and Diether et al. (2009b) use actual short selling trading data and find that short sellers are indeed 6

8 informed and that their trades contain more value-relevant information. Trades initiated by short sellers improve price efficiency and help correct mispricing. Some studies also find that short selling has no significant effect on price efficiency (e.g., Brent et al., 1990; Mayhew and Mihov, 2004; Alexander and Peterson, 2008). Macey et al. (1988), Altken et al. (1998) and Morris and Shin (1998) even point out that the introduction of short selling in a market simply results in the Pygmalion effect (i.e., the self-fulfilling prophecy), which causes the price to drop and the market to crash. Many regulators seem to share this view and advocate more restrictions on short selling. This is especially true in the period following the 2008 financial crisis. In addition to price efficiency, many studies consider changes in volatility and liquidity due to short sale constraints. Frinto et al. (2011) find that the short selling ban introduced after the 2008 financial crisis has lowered market liquidity. However, Alexander and Peterson (2008), Diether et al. (2009a) and Boulton and Braga-Alves (2010) show that the effect of short selling ban on market liquidity is rather minimal. Scheinkman and Xiong (2003) show that the removal of short sale constraints significantly decreases stock volatility and turnover. Diether et al. (2009b) suggest that short sellers are more likely to engage in reverse trades, and that their participation in the market tends to stabilize prices. Chang et al. (2007) find that volatility goes up after short sale constraints are removed. Henry and McKenzie (2006) find that the introduction of short selling causes stock volatility to go up in the Hong Kong market. 4. Research Methodology and Variables Definitions 4.1. Regression Design 7

9 To examine the effect of short selling on price efficiency and liquidity, we conduct three major tests using a panel data. First, we compare the price efficiency and liquidity between firms that are eligible and firms that are ineligible for short selling by running the following regression models: Efficiency ( Liquidity ) Short _ List Controls e (1) i, t i, t 0 i, t i, t i i i, t where the dependent variable Efficiency it, and Liquidity it, refer to the price efficiency and liquidity for stock i in period t. The independent variable Short _ List, is a dummy variable that equals 1 if a stock i in period t is eligible for short selling and 0 otherwise. Controls it, are the control variables for stock i in period t, and are discussed in Section 4.3. v i and e t are the fixed effects for firm and time. it, is the error term. We then examine the difference in price efficiency and liquidity for the same stock before and after they are added to the designated short-selling list. We run the following regression models: it Efficiency ( Liquidity ) Event _ In Controls e (2) i, t i, t 0 i, t i, t i i i, t where Event it, is a dummy variable that equals 1 if a stock i is added to the eligible list at period t and 0 otherwise. In the third test, we examine the relation between shorting flow of a stock and its price efficiency by running the following regression model: Efficiency Short _ Flow Controls e (3) i, t 0 i, t i, t i i i, t 8

10 where Short _ Flow, represents the shorting flow for stock i in period t. It is defined as it the ratio of the monthly average shorting flow of a particular stock to the monthly average trading volume of the stock Measures of Price Efficiency We use two measures for stock price efficiency. The first follows the approach used by Bris et al. (2007) and is based on the cross-autocorrelation between individual stock returns and lagged market returns. The second is based on Hou and Moskowitz (2005) and measures the speed of price adjustment to market information Cross-autocorrelation Mørck et al. (2000) explore the co-movements between stock markets in different nations and find that developed markets have higher idiosyncratic risks than emerging markets. Their findings suggest that the magnitude of idiosyncratic risk can be used to measure the price efficiency of an asset. Based on their findings, Bris et al. (2007) propose that one may use cross-autocorrelations between market returns lagged one period and individual stock returns to measure the price efficiency of an individual stock as follows: Cross i, t ri, t rm, t1 Corr (, ) (4) where r it, represents the return for a stock i in period t, and rmt, 1 is the return for the market in previous period t Mørck et al. (2000) also suggest the use of R from a market model regression as a measure of price efficiency. However, there are some constraints to this measure that limit its application in the literature (Griffin et al., 2010; Hou et al., 2013). 9

11 The cross-autocorrelations are equivalent to the regression coefficients when one regresses individual stock returns against lagged market returns. A smaller coefficient suggests a lower correlation between the returns of a particular stock and lagged market returns, indicating a higher idiosyncratic risk (or higher efficiency). 6 Following Bris et al. (2007), we compute the correlation coefficient between each stock s daily returns and one-week (five trading days)lagged daily market returns. We then average the daily coefficients in each month and obtain the monthly average correlation for each stock. Because the value of the correlation coefficient is bounded between -1 and 1, we take the absolute value as a measure of price efficiency. A lower absolute value therefore represents higher price efficiency Price Delay Hou and Moskowitz (2005) use the market return as the relevant news to which an individual stock responds. They propose a measure of price efficiency using the speed of price adjustment to market information. This measure is widely used in the literature (Lin et al., 2012; Boehmer and Wu, 2013). If the price fails to promptly incorporate the market information when it becomes available, this information is absorbed at a later time, causing some delay in the price adjustment. The magnitude of a price delay can be found via regression models containing lagged market returns. Hou and Moskowitz (2005) suggest that one can regress individual stock returns against contemporaneous and four different periods of lagged market returns as follows: i, t i i m, t i, n m, tn i, t n1 4 r r r (5) 6 This measure is widely used in the literature (e.g., Lin et al., 2010; Beber and Pagano,2013). 10

12 where r it, represents the return for a stock i in period t ; r, is the return on the value- th weighted market index in period t ; rm, t n stands for the n lagged market return and, is the random error term. We estimate equation (5) and obtain the coefficient of determination mt it 2 R. We then restrict all of the coefficients on the lagged market returns to zero and obtain the coefficient of determination delay measure: 2 R for the restricted model, and arrive at the first price 2 R D1i 1 (6) i2 R i Similar to the F test, this measure captures the proportion of variation of the contemporaneous individual stock returns explained by lagged market returns. A smaller value of D 1 represents a faster incorporation of market information. Hence, the prices are more efficient. BecauseD1 cannot distinguish between shorter and longer lags or the estimate precision, Hou and Moskowitz (2005) propose another price delay measure that uses the magnitude of the regression coefficients to measure the level of the relation between individual stock returns and market returns: D2 i i 4 n1 4 in, n1 i, n (7) D 2 uses the regression coefficients from equation (5) and measures the weight of the coefficients for lagged market returns in all of the regression coefficients. A smaller D 2 represents higher price efficiency. 11

13 As we only have four years of data, it is not appropriate to compute the annual delay measure in the same way as Hou and Moskowitz (2005). Following Boehmer and Wu (2013), we create an analogous monthly measure. We regress each stock s daily returns on contemporaneous daily market returns and four lagged market returns. 7 After calculating the price delays for different trading days, we then find the monthly average for each stock Measures of Liquidity We measure liquidity with bid-ask spread. We follow the method used by Corwin and Schultz (2012) and use the daily highest and lowest trading prices to estimate the bidask spread. They suggest that the ratio between the highest and lowest prices reflects both the price volatility and bid-ask spread. The volatility portion of the ratio increases proportionally as the time interval increases, and the bid-ask spread portion remains unchanged with the return interval. Following this logic, Corwin and Schultz (2012) derive a spread estimator as a function of high-low ratios over one-and two-day intervals as follows: 2( e 1) S 1 e (8) where S stands for the bid-ask spread and calculated as follows: and can be H t j E ln 0 (9) j0 L t j 7 In the robustness test, we also use the weekly return and find results similar to those reported in the paper. 12

14 H ln L 0 tt, 1 0 tt, 1 2 (10) where 0 t j H and 0 t j L represent the observed highest and lowest stock prices on the t j trading day, respectively. 0 tt, 1 H and 0 tt, 1 L are the highest and the lowest stock prices observed in two consecutive trading days Control Variables We include the following control variables in our regression analysis. 8 (1) Exchange. Although both the Shanghai and Shenzhen Stock Exchanges are regulated by the CSRC and operate in the same market environment, the two exchanges have different disclosure requirements and different investor clienteles that may affect stock price discovery processes in different ways (Liao et al., 2012). Hence, we use a variable Exchange to control for the differences between the two exchanges, which is a dummy variable equal to 1 if a particular stock is listed on the Shanghai Stock Exchange and 0 otherwise. (2) Market Capitalization and Book-to-market Value. Fama and French (1993) show that size and market-to-book ratio are two important factors determining stock returns. We include ln( Cap ) and B/ M in the regressions, whereas ln( Cap ) is measured as the natural logarithm of market capitalization and B/ M is the book value divided by the market value of equities. Both market capitalization and book value of equities are measured at the fiscal year end. 8 When using price efficiency measures as dependent variables in the regression equations, we also include bid-ask spread as a control variable. 13

15 (3) Stock Return. It was shown that the effect of short sale constraints on price efficiency is different between the bear and bull markets. To control for market returns in our model, we define a dummy variable Stock Return, which equals to 1 when the stock price increases and 0 otherwise. (4) Turnover. The speculative nature of the Chinese market usually results in high turnover in trading. We control for stock turnover in the models and it is measured as shares traded divided by total shares outstanding. 5. Data and Empirical Results We obtain data from Genius Finance Database, which provides detailed short selling information for stocks in the Chinese market. As the initial introduction of short selling began on March 31, 2010, our main sample begins from April We also get daily stock trading information, stock returns, and financial accounting information from the China Stock Market & Accounting Research (CSMAR). Our main sample period is from April 2010 to December Descriptive Statistics We report the descriptive statistics in Panel A of Table 2. Three price efficiency measures, Cross, D 1 and D 2 have mean of , , and , respectively. They also display a significant level of variability; with the standard deviations of and The mean of bid-ask spread is , with a standard deviation of The short selling flows on both the Shanghai and Shenzhen Stock Exchanges 9 To compare the changes in different types of market price efficiency measures for the first list of 90 eligible stocks (March 2010) before and after they are put on the designated list, we also need data before April In particular, for these tests, we extend the sample period one year before April

16 account for merely % of the entire market flow, ranging from 0% in some months to a maximum of % of the entire market flow. The short interest is also low, with a mean of Compared with the short selling trading in other developed markets, 10 the short selling market in China is still in its infancy. [Insert Table 2 Here] Although the size of the short selling market is rather small in China, shorting activities has increased significantly since it became available, especially after Figure 1 shows a steady increase in the short selling flow. In December 2013, short selling flow accounts for % of the overall market trading volume, representing a significant increase of 0.055% in December 2010, and % in December It would therefore be interesting to see whether such an attempt to incorporate negative views into stock prices through short selling has any significant effect on stock price efficiency. [Insert Figure 1 Here] Panel B of Table 2 reports the correlation matrix for the three measures of price efficiency. Although Cross and D 1 (D2) are positively correlated, the correlation coefficients are relatively low ( and , respectively). It appears that Cross, D 1 and D 2 mostly capture different aspects of price efficiency while also sharing a common component. D 1 measures the 2 R of the regression and represents the explanatory power of lagged market returns. D 2 measures the weight of the coefficients 10 Saffi and Sigurdsson (2011) show that in 26 exchanges in the U.S., U.K. and Australia, the average shorting flow is about 5.75% of the overall market flow. However, there are significant variations across different markets. For example, the ratio is about 8.91% in the U.S. and 0.46% in Thailand. 15

17 for lagged market returns in all of the regression coefficients. As expected, these two price delay measures D 1 and D 2 are highly correlated. In Figure 2, we plot the difference in Cross for stocks that are eligible (Group 1) and ineligible (Group 2) for short selling. The period covers April 2009 to December There is a steady increase in price efficiency for both groups of stocks. In April 2009, the price efficiency as measured by Cross is quite similar between the two groups. However, the slope of change appears to be higher for the stocks eligible for short selling, suggesting that improvement in price efficiency is more striking for eligible firms. Price efficiency for the overall market appears to improve in general after the introduction of short selling and margin trading. [Insert Figure 2 Here] These results provide some preliminary support for our hypothesis that short selling improves price efficiency. We investigate the relation in greater detail in the next section Main Empirical Results Short Selling Eligibility, Price Efficiency and Liquidity We use Eq.(1) to test the difference in price efficiency between stocks that are eligible and those that are ineligible for short selling. As shown in Panel A of Table 3, the variable of interest, Short _ List has negative and significant coefficients across the board, indicating that the stocks price efficiency has improved after they are eligible for shortselling. For example, when Cross is used as a dependent variable, the coefficient for Short _ List is and significant at the 1% level (p<0.0001). The coefficients for Short _ List remain negative when we use D 1 and D 2 to measure price efficiency (The 16

18 coefficients are and and significant at the 1% level). These results suggest that the introduction of short selling has significantly improved stock price efficiency (higher idiosyncratic risk). The results on other control variables are generally consistent with previous findings. Price efficiency is higher for stocks listed on the Shenzhen Stock Exchange and for firms with small size, high book-to-market ratio, low bid-ask spared and small turnovers. Some of our results are consistent with those in a study by Liao et al. (2012), who find higher price efficiency in the Shenzhen Stock Exchange due to its higher regulatory standards. [Insert Table 3 Here] We then use Spread as a dependent variable to examine the effect of introduction of short-selling on stock liquidity. As shown in Panel B of Table 3, the coefficients for Short _ List is significant and negative (β = , p=0.0002). This suggests that stocks that are eligible for short selling have lower bid-ask spreads than those stocks that are ineligible for short selling. In addition, large cap stocks, low book-to-market stocks, and stocks with high turnover tend to have high liquidity Price Efficiency and Liquidity Before and After Short Selling Eligibility We next conduct a test with a focus on the differences in price efficiency and liquidity for the same group of stocks before and after they are added to the short-selling list. The regression equation is specified similar to Eq. (2). We include only 90 stocks in 2010, 189 new stocks in 2011 and 512 new stocks in 2013 in our analysis. The variable of interest is a dummy variable Event _ In, which equals 1 if a stock is added to the list of stocks eligible for short selling and 0 otherwise. 17

19 We report the results in Table 4. Panel A shows that the coefficients for Event _ In are negative and significant when three measures are used as proxies for price efficiency. The coefficients are , and when we use Cross, D 1 and D 2 as the dependent variables, respectively (these coefficients are all significant at the 1% level). These results suggest that the price efficiency of stocks improves after they become eligible for short selling. The results for the controlled variables are similar to those reported in Panel A of Table 3. [Insert Table 4 Here] In Panel B of Table 4, our dependent variable is stock liquidity measure Spread. The coefficient for Event _ In has expected negative sign and is significant at the 1% level (β = , p < 0.002). Such result indicates that a stock s bid-ask spread generally goes down after the stock becomes eligible for short selling, indicating an overall improvement in market liquidity after the introduction of short selling Short Selling Flow and Price Efficiency In this section, we examine how short selling flow is related to price efficiency. Boehmer and Wu (2013) show that one can explore the relation between short selling flow and price efficiency to better understand how information enters into asset prices and the speed of price adjustment. We run Eq. (3) where Short _ Flow is defined as the ratio of the monthly average shorting flow of a particular stock to the monthly average trading volume of the stock. The results are reported in Table 5. We note that the coefficients for Short _ Flow are negative and significant when Cross and D 2 are used as the dependent variable, and the coefficient is not significantly different from zero when 18

20 D 1 is used as a measure of price efficiency (β = , p= ). Overall, our results are consistent with Boehmer and Wu (2013), and indicate that high level of shorting flow is positively related with stock price efficiency Difference-in-differences Tests [Insert Table 5 Here] There is an endogeneity issue if the firms on the short selling eligibility list already have better price efficiency before they become eligible for short selling. These stocks are part of the main exchange index, and usually are firms with have high liquidity and large market capitalization. To mitigate this endogeneity concern, we conduct a difference-in-differences test. Using four short selling admission times (March 31, 2010; December 5, 2011; January 31, 2013 and September 16, 2013), we conduct a differencein-differences regression analysis. We use constituent stocks in the Shanghai and Shenzhen 300 index to construct both the test and matched samples to mitigate any additional noise resulting from specific firm characteristics. [Insert Table 6 Here] As shown in Table 6, the coefficients for Short _ List are negative and significant across the board when the dependent variables are ρ Cross, D1 and D2. Overall, this finding is consistent with previous results, suggesting that price efficiency improved significantly after they become eligible for short-selling Other Tests Short Selling and Stock Volatility A common concern of short-selling is its impact on market stabilization and volatility, as shown in the US stock market (Bris et al., 2007) and Hong Kong market 19

21 (Chang et al., 2007). However, international evidence (Saffi and Sigurdsson, 2011) suggests that short-selling improves market stabilization by reducing volatility. In this section, we revisit the effect of short-selling on stock volatility by conducting the following regression models: Volitility Short _ List Controls e (11a) i, t 0 i, t i, t i i i, t Volatility Event _ In Controls e (11b) i, t 0 i, t i, t i i i, t Table 7 reports the regression results. In column (1) we examine stock volatility between firms that are eligible and firms that are ineligible for short-selling. The variable of interest, Short _ List has a negative and significant coefficient (β = , p<0.0001), suggesting that volatility is lower for those eligible stocks. In column (2) we investigate stock volatility for the same group of the stocks after they are added to the short selling list. The coefficient for Event _ In is negative and significant (β = , p<0.0001). This result indicates that stock volatility is lower after they become eligible for shortselling. [Insert Table 7Here] Loss of Short Selling Eligibility, Price Efficiency, Liquidity, and Volatility Our previous evidence suggests that after stocks are added onto the shorting list, their price efficiency and liquidity improves while volatility decreases. A natural question is: what if some stocks are removed from the short-selling list? The findings from these delisted stocks will provide corroborating evidence on whether short-selling improves price efficiency and liquidity. We use 62 stocks removed from the short selling list during our sample period to see if price efficiency deteriorates after a stock is taken out from the short selling list. The regression model is specified as follows: 20

22 Efficiency ( Liquidity or Volatility ) Event _ Out Controls e i i i, t i, t i, t i, t 0 i, t i, t (12) Event _ Out is a dummy variable that equals 1 if a stock i is removed from the eligible list at period t and 0 otherwise. As shown in Panel A of Table 8, the coefficients for Event _ Out are positive and highly significant when D 1 and D 2 are the dependent variables. This suggests that after these firms are removed from the short selling eligibility list, their price efficiency decreases significantly. In Panel B and C, the coefficients for Event _ Out have expected signs but become insignificant, probably due to the relatively small sample size. [Insert Table 8 Here] 6. Conclusions The short selling mechanism is an important part of an efficient market trading system. Although the pros and cons of short selling have been debated extensively, especially since the financial crisis in 2008, there remains no conclusive evidence of the effect of short selling on price efficiency. The empirical evidence is even less for emerging markets. China introduced short selling in 2010, providing a great opportunity to examine the effect of short selling on price efficiency. Because there is no alternative way to short sell an individual stock in China, our tests are free of potential contamination from options or futures markets. We find that the introduction of short selling significantly improves price efficiency, as measured by the differences in individual stock responses to market returns and the delay in price adjustments. Short selling also enhances stock liquidity, as measured by bid-ask spread; and reduces stock volatility. 21

23 Our study adds to recent debates on the unexpected costs of limiting short selling in the market. Many exchanges have taken aggressive approaches to constraining short selling in recent years due to the fear of market crashes and price manipulation. However, such restrictions may also lower the price efficiency by preventing informed short sellers to trade. 22

24 References Ackert, L. F., N. Charupat, B. K. Church, and R. Deaves, Bubbles in Experimental Asset Markets: Irrational Exuberance No More. Working paper, Federal Reserve Bank of Atlanta. Alexander, G. J., and M. A. Peterson, The Effect of Price Tests on Trader Behavior and Market Quality: An Analysis of Reg SHO. Journal of Financial Markets 11: Allen, F., and D. Gale, Stock-price Manipulation, Review of Financial Studies, 5: Altken, M. J., A. Frino, M. S. McCorry, and P. L. Swan, Short Sellings are Almost Instantaneously Bad News: Evidence from the Australian Stock Exchange. Journal of Finance 53: Asquith, P., and L. Meulbroek, An Empirical Investigation of Short Interest. Working paper, MIT. Autore, D. M., R.S. Billingsley, and T. Kovacs, The 2008 Short Sale Ban: Liquidity, Dispersion of Opinion, and the Cross-section of Returns of U.S. Financial Stocks. Journal of Banking and Finance, 35: Battalio, R., and P. Schultz, Regulatory Uncertainty and Market Liquidity: The 2008 Short Sale Ban s Impact on Equity Option Markets. Journal of Finance 66: Beber, A., and M. Pagano, Short-selling Bans around the World: Evidence from the Crisis. Journal of Finance 68: Bhojraj, S, R. J. Bloomfield, and W. B. Tayler, 2009.Margin Trading, Overpricing, and Synchronization Risk. Review of Financial Studies 22: Boehmer, E., C. M. Jones, and X. Zhang, Which Shorts are Informed? Journal of Finance 63: Boehmer, E., and J. Wu, 2013.Short Selling and the Price Discovery Process. Review of Financial Studies 26: Boulton, J., and M. V. Braga-Alves, 2010.The Skinny on the 2008 Naked Short-sale Restrictions. Journal of Financial Markets 13: Brent, A., D. Morse, and E. K. Stice, Short Interest: Explanations and Test. Journal of Financial and Quantitative Analysis 25:

25 Bris, A., W. N. Goetzmann, and N. Zhu, Efficiency and the Bear: Short Sellings and Markets around the World. Journal of Finance 62: Brunnermeier, M. K., and L. H. Pedersen, Predatory Trading. Journal of Finance60: Chang, E. C., J. W. Cheng, and Y. Yu, Short-sales Constraints and Price Discovery: Evidence from the Hong Kong Market. Journal of Finance 62: Christophe, S. E.,M. G. Ferri, and J. J. Angel, 2004.Short-selling Prior to Earnings Announcements. Journal of Finance 59: Corwin, S. A., and P. Schultz, A Simple Way to Estimate Bid-ask Spreads from Daily High and Low Prices. Journal of Finance 67: Danielsen, B. R., and S. M. Sorescu, Why do Option Introductions Depress Stock Prices? A Study of Diminishing Short Sale Constraints. Journal of Financial and Quantitative Analysis 36: Diamond, D. W., and R. E. Verrecchia, 1987.Constraints on Short-selling and Asset Price Adjustment to Private Information. Journal of Financial Economics 18: Diether, K., K. H. Lee, and I. M. Werner, 2009a. It s SHO Time! Short-sale Price Tests and Market Quality. Journal of Finance 64: Diether, K., K. H. Lee, and I. M. Werner, 2009b.Short-sale Strategies and Return Predictability. Review of Financial Studies 22: Fama, E. F., and K. R. French, Common Risk Factors in the Returns on Stocks and Bonds. Journal of Financial Economics 33: Fellner, G., and E. Theissen, Short Selling Constraints and Divergence of Opinion: Evidence from the Laboratory. Working paper, Center for Financial Studies. Frino, A., S. Lecce, and A. Lepone, Short-sales Constraints and Market Quality: Evidence from the 2008 Short-sales Bans. International Review of Financial Analysis 20: Goldstein, I., and A. Guembel, 2008.Manipulation and the Allocational Role of Prices. Review of Economic Studies 75: Griffin, J. M., P. J. Kelly, and F. Nardari, Do Market Efficiency Measures Yield Correct Inferences? A Comparison of Developed and Emerging Markets. Review of Financial Studies 23:

26 Haruvy, E., and C. N. Noussair, The Effect of Short Selling on Bubbles and Crashes in Experimental Spot Asset Markets. Journal of Finance 61: Haruvy, E., L. Yaron, and C. N. Noussair, Traders Expectations in Asset Markets: Experimental Evidence. American Economic Review 97: Henry, O., and M. D. McKenzie, The Impact of Short-selling on the Price-Volume Relationship: Evidence from Hong Kong. Journal of Business 79: Hong, H., and J. C. Stein, 2003.Difference of Opinion, Short-sales Constraints and Market Crashes. Review of Financial Studies 16: Hou, K., P. Lin, and W. Xiong Is R2 a measure of market inefficiency? Working paper. Hou, K., and T. J. Moskowitz, 2005.Market Frictions, Price Delay and the Cross-section of Expected Returns. Review of Financial Studies 18: Kolasinksi, A. C., A. Reed, and J. R. Thornock, Can Short Restrictions Actually Increase Informed Short Selling? Financial Management 42: Lei, V., C. N. Noussair, and C. R. Plott, 2001.Nonspeculative Bubbles in Experimental Asset Markets: Lack of Common Knowledge of Rationality vs. Actual Irrationality. Econometrica 69: Liao, L., Z. Li, W. Zhang, and N. Zhu, Does the Location of Stock Exchange Matter? A Within-country Analysis. Pacific-Basin Finance Journal 20: Lin, F. Y., C. Y. Chien, D. Y. Liu, and Y. S. Huang, Short Sellings Constraints and Stock Price Behavior: Evidence from the Taiwan Stock Exchange. Investment Management and Financial Innovations 7: Lin, T. C., Dynamic Short-Sale Constraints, Price Limits, and Price Dynamics. International Journal of Managerial Finance 8: Macey, J. R., M. Mitchell, and J. Netter, Restrictions on Short Sellings: An Analysis of the Uptick Rule and its Role in View of the October 1987 Stock Market Crash. Cornell Law Review 74: Mayhew, S., and V. T. Mihov, 2004.Short Sale Constraints, Overvaluation and the Introduction of Options. Working paper, Texas Christian University. Mei, J., Scheinkman, J., and W., Xiong, W, Speculative Trading and Stock Prices: An Analysis of Chinese A-B Share Premia. Annals of Economics and Finance 10,

27 Miller, E. M., Risk, Uncertainty, and Divergence of Opinion. Journal of Finance 32: Mørck, R., B. Yeung, and W. Yu, The Information Content of Stock Markets: Why do Emerging Markets have Synchronous Stock Price Movement? Journal of Financial Economics 58: Morris, S., and H. S. Shin, Unique Equilibrium in a Model of Self-fulfilling Currency Attacks. American Economic Review 88: Noussair, C., R. Stephane, and R. Bernard, 2001.Price Bubbles in Laboratory Asset Markets with Constant Fundamental Values. Experimental Economics 4: Porter, D. P., and V. L. Smith, 1995.Futures Contracting and Dividend Uncertainty in Experimental Asset Markets. Journal of Business 68: Saffi, P. A., and K. Sigurdsson, 2011.Price Efficiency and Short-selling. Review of Financial Studies 24: Scheinkman, J., and W. Xiong, 2003.Overconfidence and Speculative Bubbles. Journal of Political Economy 111:

28 Figure 1: Dynamics of Short Selling Flow and Short Interest short-selling flow (%) short interest (%) /10 08/10 12/10 04/11 08/11 12/11 04/12 08/12 12/12 04/13 08/13 12/13 27

29 Figure 2: The Dynamic Change of Price Efficiency Group 1: Stocks eligible for short selling Group 2: Stocks not eligible for short selling Trend for group 1 Trend for group /09 08/09 12/09 04/10 08/10 12/10 04/11 08/11 12/11 04/12 08/12 12/12 04/13 08/13 12/13 Note: Group 1 represents the firms eligible for short selling. Group 2 represents the firms ineligible for short selling. 28

30 Table 1: Major Changes to the List of Stocks Eligible for Short Selling 11 Time # of firms # of firms removed # of firms on # of firms on % of firms eligible added to the list from the list the list the exchange for short selling 1. Mar 31, , Dec 5, , Jan 31, , Sep 16, In addition to the three major changes to the short selling stock list, there are two minor changes. Because there are so few, we do not include them in Table 1. 29

31 Table 2: Descriptive statistics Panel A reports descriptive statistics for main variables, and Panel B shows the correlation between three Cross measures of price efficiency. The main sample period is from April 2010 to December is based on the cross-autocorrelation between individual stock returns and lagged market returns, as in Bris et al. (2007). D 1 and D 2 are other two measures of price efficiency that estimate the speed of price adjustment to market information, as in Hou and Moskowitz (2005). Spread is a measure of stock liquidity. We follow the method used by Corwin and Schultz (2012) and use the daily highest and lowest trading prices to estimate the bid-ask spread. Short _ Flow is defined as the ratio of the monthly average shorting flow of a particular stock to the monthly average trading volume of the same stock. Short Interest is calculated as the number of shares sold short (but not yet covered) as a percentage of the total number of outstanding shares. Panel A: Descriptive statistics for main variables Variable Mean Median Std Min Max Cross D D Spread Short _ Flow (%) Short Interest (%) Panel B: Correlation matrix for measures of price efficiency Cross D 1 D 2 Cross D 1 D <.0001 < <.0001 < <.0001 <

32 Table 3: Regression analysis of the price efficiency and liquidity between firms that are eligible and ineligible for short selling This table reports the results of the following regression equations where the dependent variables are price efficiency (Panel A) and liquidity (Panel B) measures. Efficiency ( Liquidity ) Short _ List Controls e. i, t i, t 0 i, t i, t i i i, t Short _ List is a dummy variable that equals 1 if a stock i in period t is eligible for short selling and 0 otherwise. Exchange is a dummy variable equals 1 if a stock is listed on the Shanghai Stock Exchange and 0 otherwise. ln( Cap ) is the natural logarithm of market capitalization at the fiscal year end. Exchange is a dummy variable which equals to 1 if a particular stock is listed on the Shanghai Stock Exchange and 0 otherwise. ln( Cap ) is measured as the natural logarithm of market capitalization. B/ M is the book value divided by the market value of equities (both are measured at the fiscal year end). Turnover is the number of shares traded for a period as a percentage of the total shares outstanding. Stock Return is a dummy variable which equals to 1 when the stock price increases and 0 otherwise. Other control variables are defined as those in Table 2. We also consider firm and time fixed effect. The p- values are listed beneath the coefficient estimates. Intercept Short _ List Exchange ln( Cap ) B/ M Spread Turnover Stock Return Panel A: Price efficiency measures Cross D 1 D 2 Panel B: Liquidity measures Spread <.0001 < < <.0001 <.0001 < <.0001 <.0001 < <.0001 <.0001 <.0001 < <.0001 <.0001 < <.0001 <.0001 < <.0001 <.0001 < <.0001 < Stock-level Effect Yes Yes Yes Yes Month-Level Effect Yes Yes Yes Yes Observations 85,556 85,688 85,689 83,696 R

33 Table 4: Regression analysis of the price efficiency and liquidity before and after short selling eligibility This table reports the results of the following regression equations where the dependent variables are price efficiency (Panel A) and liquidity (Panel B) measures. Efficiency ( Liquidity ) Event _ In Controls e. i, t i, t 0 i, t i, t i i i, t Event _ In is a dummy variable that equals 1 if a stock i is added to the eligible list at period t and 0 otherwise. Other control variables are defined as those in Table 3. We also consider firm and time fixed effect. The p-values are listed beneath the coefficient estimates. Intercept Event _ In Exchange ln( Cap ) B/ M Spread Turnover Stock Return Panel A: Price efficiency measures Cross D 1 D 2 Panel B: Liquidity measures Spread < < <.0001 < <.0001 <.0001 < <.0001 < <.0001 < <.0001 <.0001 < <.0001 < Stock-level Effect Yes Yes Yes Yes Month-Level Effect Yes Yes Yes Yes Observations 15,387 15,404 15,404 14,876 R