Back to the Futures: When Short Selling is Banned

Transcription

1 Back to the Futures: When Short Selling is Banned George Jiang a, Yoshiki Shimizu b, and Cuyler Strong c, January 2018 a Jiang is a professor in the Department of Finance and Management Sciences, in the Carson College of Business at Washington State University. george.jiang@wsu.edu. Phone: b Shimizu is a Ph.D. Candidate in the Department of Finance and Management Sciences, in the Carson College of Business at Washington State University. yoshiki.shimizu@wsu.edu Phone: c Strong is a Ph.D. Student in the Department of Finance and Management Science, in the Carson College of Business at Washington State University, cuyler.strong@wsu.edu. Phone: Abstract

2 Back to the Futures: When Short Selling is Banned Abstract Multiple articles have examined the impact that the short-sale ban in 2008 had on equity and derivatives markets, however little research has been done on the roles that single-stock futures played during the ban period. In this paper, we fill the gap in the literature by answering three research questions. First, did the trading volume of single-stock futures increase during the ban period? Second, what are the effects of single-stock futures on price discovery of the underlying stocks during the ban period? Third, what are the effects of single-stock futures on market quality of the underlying stocks during the ban period? We find that the number of single-stock futures listed increased by 55.5 percent, and single-stock futures trading volume increased significantly during the short sale ban period. Further examination shows that the single-stock futures contribution to price discovery of the underlying stock also increased significantly during the ban period. Moreover, we find that single-stock futures help mitigate the negative effects of the ban on market quality found by Boehmer et al. (2013). JEL Classification: G01, G14 Keywords: Single-Stock Futures; Short Selling; Financial Crises; Price Discovery; Market Quality

3 1. Introduction On September 19, 2008, the SEC announced an emergency plan to temporary ban short sales of 799 financial stocks. According to the SEC, the ban was aimed at helping restore falling stock prices that [had] shattered confidence in the financial markets. On September 22, 2008, 198 non-financial stocks were added to the ban. These stocks included companies that had financial subsidiaries, including General Electrics, Ford Motor Company, and General Motors. The ban was in effect for 13 trading days and lifted on October 8, Much research has been done on the effects that the ban had on financial markets. Boehmer et al. (2013) analyze shortselling activities around the 2008 short sale ban and find that short-sell activities decreased by 77 percent during the ban. Battalio and Schultz (2011) and Grundy et al. (2012) examine the options markets during the short sale ban, and find that liquidity in options markets dried up during the ban period. Though a large body of literature examines the effect of the ban on stock and options markets, less emphasis has been put on the single-stock futures market. In 2002, the Nasdaq Liffe Markets and OneChicago Exchange started trading singlestock futures (SSFs) The new futures markets were expected to be the next big thing in financial markets. Once the SSFs market opened, growth came slowly, adding a few listings every couple of months. In 2004, the Nasdaq Liffe Markets closed and assigned the remaining contracts to the OneChicago Exchange. In August of 2008, there were only 517 stocks with SSFs listed on the OneChicago Exchange. During the short sale ban period in 2008, there were 291 new introductions, increasing the number of listings to 815, a 55.5 percent increase. Although the overall size of the SSFs market is small in the United States, we argue that the importance of the SSFs market was augmented during the short sale ban period, and that informed traders could use SSFs as a viable substitute for short sales and to circumvent the short sale ban. 1

4 In this paper, we analyze the stocks that had SSFs listed on the OneChicago Exchange during the 2008 short sale ban period to answer three research questions. Our first research question is: Did trading volume of SSFs on banned stocks increase during the short sale ban period? Easley, O Hara, and Srinivas (1998) develop a theoretical model, which predicts that when short sales are constrained, informed traders will trade more in the options market. When short-sales are restricted, a short seller could buy a put option as a substitute leading to increased demand in options markets. Contrary to this belief, Battalio and Schultz (2011) and Grundy et al. (2012) show trading volume in options markets decreased significantly during the 2008 short sale ban. The reason for this is that in order for put option writers to hedge their positions, they need to have an offsetting short. When short selling is banned, put option writers are not willing to write the options, leading to a reduction in options trading volume. Danielsen et al (2009) show that single-stock futures can be used as substitutes for short-selling. We extend the model by Easley et al. by proposing SSFs as a potential alternative to short-selling when it is prohibited. Grundy et al. make an attempt to examine the effect of the 2008 short sale ban on the SSFs market. They find through preliminary examination that there is little evidence that short sellers migrated to the SSFs market. Motivated by the theoretical model developed by Easley et al. and finding by Grundy et al., we conduct a formal analysis of the effect of the 2008 short sale ban on the trading volume of SSFs on banned stocks. Our second research question is: What are the effects of SSFs on price discovery of underlying banned stocks during the ban period? If the SSFs trading volume increased during the ban period, it implies that informed and bearish investors may have migrated to the SSFs market and that there was an increase in information passed through the SSFs market. Fung and Tse (2008) and Kumar and Tse (2008) find that SSFs prices contribute to price discovery of 2

5 underlying stocks by 33 percent in the Hong Kong Exchange and 28 percent in the National Stock Exchange of India, respectively. Shastri et al. (2008) find that contribution of SSFs prices to price discovery is 24 percent in the OneChicago Exchange. Given that bearish investors were kicked out of the stock market and also that trading options was not an option to them, SSFs prices of banned stocks potentially became more informative during the ban period. This motivates us to examine the effects of SSFs on price discovery of underlying banned stocks during the short sale ban period. Our third research question is: What are the effects of SSFs on market quality of underlying banned stocks during the ban period? Boehmer et al. (2013) study the stock market quality during the 2008 short sale ban. They use a difference-in-difference approach to examine the effect of the ban on market quality of stocks for which short selling was banned. They find that market quality of banned stocks deteriorated dramatically during the ban, and the negative effect of the ban on market quality was more distinct in banned stocks in the larger market cap quartiles. As stated earlier, the number of SSFs listings increased by 55.5 percent during the ban period. Of the 291 SSFs that were added at that time, 276 were stocks that were subject to the ban. The number of SSFs listed for banned stocks rose from 64 to 342, a 534 percent increase during the ban period. It can be inferred that the OneChicago Exchange increased the supply of SSFs due to a sudden increase in demand for short positions in underlying banned stocks. It is of our interest to examine whether the presence of SSFs lessen the effect of the ban on market quality of underlying banned stocks. We briefly summarize our findings and answers to the research questions as follows: From the analyses of 64 banned stocks that had SSFs prior to the ban, we find that trading volume of SSFs on these banned stocks increased significantly during the ban period. 3

6 Subsequently, we construct a matched sample of non-banned stocks with similar size, volatility, and liquidity to test whether SSFs trading volume increased significantly larger for banned stocks than for non-banned stocks. We find that relative to their non-banned counterparts, volume of SSFs on banned stocks increased significantly larger during the ban period. Next, we show the results for our second research question that SSFs contribution to price discovery of underlying banned stocks increased dramatically during the ban period. This suggests that during the ban period, prices of SSFs became more informative than during non-ban periods. However, our further investigation shows that SSFs contribution to price discovery increased more for non-banned stocks. We attribute this result to the SEC s Rule 204T, which became effective on September 18, 2008, and made short selling more costly for all U.S. stocks. Lastly, we find that market quality of banned stocks during the ban period was better when there was SSFs trading. We find evidence that the presence of SSFs lessened the negative impact of the short sale ban on stock market quality during the ban period. The rest of the paper is organized as follows: Section 2 describes the sample and data we employ in our analyses. Section 3 discusses our empirical designs. In Section 4, we provide main empirical results. Section 5 concludes. <FIGURE 1 HERE> <TABLE 1 HERE> 2. Data We obtain data from multiple sources. Data on new SSFs listing information are hand collected from the press release page of the OneChicago Exchange. Data on SSFs quotes and trading volume, available at the daily frequency, are from the Bloomberg terminal. Generally 4

7 speaking, each SSFs contract has multiple expiration dates. For a SSFs contract underlying stock i, we choose the one with the shortest time to expiration date. If there are less than 5 trading days until the shortest maturity date, we choose the one with the second shortest time to expiration date. Data on underlying stock price are from the Center for Research in Security Prices (CRSP). Intraday transaction data (trade and quotes) used to calculate the market quality measures are obtained from the Trade and Quote (TAQ) database. The list of stocks that were subject to the 2008 short sale ban is obtained from the NYSE and NASDAQ websites. Most of our analyses cover the sample period from August 1, 2008 to October 31, We divide the sample periods into 3 sub-sample periods: Pre-ban (August 1, 2008 September 18, 2008), ban (September 19, 2008 October 8, 2008), and post-ban period (October 9, October 31, 2008). Our treatment sample consists of 64 banned stocks with SSFs that had SSFs quotes on OneChicago and stock data on CRSP throughout the sample period. In the same manner, we require that control samples to have SSFs quotes and volume and stock data available throughout the sample period to be considered for matching. To match a treatment stock with a control stock, we select control stock i for treatment stock j that has the minimum weighted sum of the absolute differences between the treatment and control matching variables, Size, Volatility, and Liquidity. 3. Methodology First, to examine the effect of the short sale ban on the trading volume of single-stock futures of banned stocks during the short sale ban period, we estimate the following fixed effects model for a daily panel of banned stocks with SSFs trading: 5

8 SSSSSSSS VVVVVVVVVVee ii,tt = αα + ββ 1 BBBBBB pppppppppppp + ββ 2 (BBBBBB pppppppppppp BBBBBBBBBBBB ssssssssss) +ββ 3 PPPPPPPPPPPPPP pppppppppppp + ββ 4 (PPPPPPPPPPPPPP pppppppppppp BBBBBBBBBBBB ssssssssss) + θθθθ + εε ii,tt, (1) where SSSSSSSS VVVVVVVVVVee ii,tt is the daily trading volume of SSFs on underlying stock i on day t. Ban period is an indicator variable that is equal to one during the short sale ban period (September 19 October 8, 2008) for stocks subject to the ban, and zero otherwise. Postban period is an indicator variable that is equal to one after the short sale ban period (i.e., after October 8, 2008), and zero otherwise. Banned stock is an indicator variable that is equal to one for stocks that were subject to the 2008 short sale ban, and zero otherwise. XX is a vector of firm-level characteristics and other macroeconomic variable: Stock return (of underlying stock i), Stock volume (of underlying stock i), and S&P500 Volatility Index (VIX). Second, we follow Hasbrouck (1995) to measure the SSFs market s contribution to price discovery for the underlying security. The Hasbrouck s price discovery methodology is used to examine the price discovery of a single security traded in multiple markets. Prior research on (single-stock) futures markets extend Hasbrouck s methodology to analyze price discovery between stock and SSFs markets (for example, Fung and Tse, 2008, Shastri et al. (2008), and Kumar and Tse, 2009). As in Shastri et al. (2008), consider the case where a single security trades in two markets, stock and SSFs markets. Stock and SSFs prices underlying security i are denoted as S i,t and F i,t, respectively. Let m i,t represent the efficient price of underlying security i. Then stock and SSFs prices share a common efficient price mm ii,tt, such that: 6

9 S i,t = m i,t + ε S,i,t (2) F i,t = m i,t + ε F,i,t, (3) and the common efficient price, m i,t, follows a random walk: m i,t = m i,t 1 + w i,t, (4) where w t ~ i. i. d (0, σ 2 ). We can represent stock price S t as a function of SSFs price F t underlying security i as following:` S i,t = F i,t + Dividend e rt, (5) where t is the time to maturity, r is the risk-free rate, and Dividend is dividends to be paid on the stock before the maturity date. Since we focus SSFs contracts with shortest time to maturity t (for t > 5 trading days), t is small enough such that ee rrrr will be close to 1. With ee rrrr being (almost) 1, Dividend is a constant: S i,t = F i,t + c (6) Denote price vector for security i as: p i,t = S i,t F i,t. If S i,t and F i,t are I(1) and there is a linear combination of the processes S i,t = α + βf i,t + ϵ i,t that is stationary, S i,t and F i,t are said to be co-integrated. Following the general vector error correction model (VECM) specification by Hasbrouck (1995), p i,t can be expressed in terms of an error correction model of order N as: p i,t = φ 1 p i,t 1 + φ 2 p i,t β z i,t 1 b + ϵ i,t, (7) where z i,t 1 b is an error correction term with z i,t 1 = S i,t 1 F i,t 1 and b = E z i,t. If there is a disequilibrium of price relationship S i,t 1 > F i,t 1 S i,t 1 < F i,t 1, coefficient β corrects the error in last 7

10 period t-1 to adjust further towards the equilibrium value. The VECM allows for both short- and long-run dynamics. Equation (7) can be expressed as a VMA model: P i,t = ϵ i,t + ψ 1 ϵ i,t 1 + ψ 2 ϵ t 2 +, (8) where ϵ i,t = ϵ i,s,t ϵ i,f,t and var ϵ i,t = Ω. Since S i,t and F i,t have the same underlying security, εε ii,ss,tt aaaaaa εε ii,ff,ss are likely to be correlated across two markets. Then Ω is not diagonal: VVVVVV ϵ I,S,t Ω = CCCCCC ϵ I,F,t, ϵ I,S,t CCCCCC ϵ I,S,t, ϵ I,F,t, (9) VVVVVV ϵ I,F,t and thus the variance decomposition requires the Cholesky factorization. For covariance matrix Ω, the lower triangle matrix FF provides a Cholesky decomposition such that Ω = FF FF. The total variance of market innovations εε tt = εε 1,tt εε 2,tt is: [ψψ] 1 ΩΩ[ψψ] 1 = [ψψ] 1 FF FF[ψψ] 1, where [ψψ] 1 denotes the first row of ψψ, the row corresponding to εε tt = εε ii,ss,tt εε ii,ff,,tt. The first element of [ψψ] 1 FF is the portion of market innovation from the stock market, whereas the second element of [ψψ] 1 FF is that from the SSFs market. Contribution of market j to price discovery for underlying stock i is: CCCCCCCCCCCCCCCCCCCCCCnn ii,jj = [ψψff ] 2 jj, (10) ψψψψψψ where [ψψff ] jj is the j-th element of the row matrix ψψff for j=s, F. From Equation (10), we define the SSFs market s contribution to price discovery of underlying stock i as: PPDD ii,tt = [ψψff ] 2 FF (11) ψψψψψψ 8

11 Third, to examine the effect of the SSFs market on price discovery for underlying stocks during the ban period, we conduct a differences-in-differences test. First of all, we examine how the contribution by the SSFs market to price discovery changed during the ban period, relative to the pre-ban period. The contribution of the SSFs market to price discovery for underlying stock i during period t is measured by PPDD ii,tt as defined in equation (11). Secondly, we examine how the SSFs market s contribution to price discovery for banned stocks changed relative to stocks for which short-selling was never banned during the ban period. We examine this by performing the following differences-in-differences test: PPDD ii,tt = αα + ββ 1 BBBBBB PPPPPPPPPPPP + ββ 2 BBBBBBBBBBBB SSSSSSSSSS + ββ 3 (BBBBBB BBBBBBBBBBBBBBBBBBBBBB) + εε ii,tt, (12) where PPDD ii,tt is the SSFs market s contribution to price discovery for underlying stock i, measured over time period t for t= pre-ban, ban, and post-ban periods. Dummy variable Ban Period is equal to 1 if an observation is during the ban period. Dummy variable Banned Stock is equal to 1 if underlying stock i was subject to the 2008 short sale ban. We match a group of treatment sample with a group of control stocks by underlying stock s market capitalization, idiosyncratic volatility, and turnover. Market capitalization and idiosyncratic volatility are as of March 1, 2008 and turnover is averaged from January 2008 through July If an underlying stock is listed on NASDAQ, turnover is divided by two. Our treatment group consists of 64 banned stocks that had SSFs listed on the OneChicago exchange throughout the period from March 1, 2008 June 30, We then create a matched control group of 64 unbanned stocks, matched by the variables introduced earlier. Using the model (Equation 12), we make comparison of changes over time in price discovery between treatment (banned) and control (unbanned) groups. 9

12 After testing the difference in price discovery between banned and unbanned stocks during the ban period, we turn our focus to examining how SSFs trading contributes to improving market quality of underlying stocks over time. We follow Boehmer et al. (2013) to test the market quality of the underlying stocks during the ban period. For a treatment group of 64 banned stocks with SSFs (as defined earlier), we create a matched control group of 64 banned stocks that never had SSFs. We use a difference-in-difference approach to examine the effect of SSFs trading on market quality of underlying stocks during the 2008 short sale ban period. Our treatment sample consists of 64 banned stocks with SSFs listed and traded prior to the short sale ban. We match the 64 treatment sample with a group of 64 banned stocks that never had SSFs. Matching made by following the same matching criterion defined earlier in this section. Though our primary focus is on the effect of SSFs trading on market quality of underlying banned stocks during the ban period, we also create two other control groups: The control group 2 that consists of 64 matched non-banned stocks that had SSFs listed and traded before the ban, and the control group 3 that consists of 64 matched non-banned stocks that never had SSFs. We estimate the following fixed effects model of a difference-in-difference test to examine to what extent SSFs trading contributes to improving market quality of underlying banned stocks: YY ii,tt = αα + ββ 1 BBBBBB ppppppiiiiii + ββ 2 (BBBBBB pppppppppppp SSSSSS) + β 3 PPPPPPPPPPPPPP pppppppppppp + ββ 4 (PPPPPPPPPPPPPP pppppppppppp SSSSSS) +θθxx ii,tt + εε ii,tt, (13) where YY ii,tt is the measured market quality of stock i on day t. Dummy variables Ban period and Banned stock are defined same as in Equation 11. SSF is a dummy variable set equal to 1 if stock i had a SSFs listed on day t. XX ii,tt is a vector of firm-level control variables: market capitalization, 10

13 dollar trading volume, intraday value-weighted average price (VWAP), and proportional daily range of stock prices (RVOL) of underlying stock i on day t as in Boehmer et al. (2013). We calculate multiple measurements of market quality as in Boehmer et al. (2013): TThee QQQQQQQQQQQQ SSSSSSSSSSdd ii,tt RRRRSS ii,tt = AAAAkk ii,tt BBBBdd ii,tt MM ii,tt, (14) where AAAAkk ii,tt (BBBBdd ii,tt ) is the ask (bid) price for stock i at trade time t, and MM ii,tt is the price midpoint of the National Best Bid and Offer (NBBO) quotes for stock i at trade time t. TThee EEEEEEEEEEEEEEEEEE SSSSSSSSSSdd ii,tt RRRRSS ii,tt = 2 PP ii,tt MM ii,tt MM ii,tt, (15) where PP ii,tt is the trade price per share for stock i at trade time t. MM ii,tt is the price midpoint of the National Best Bid and Offer (NBBO) quotes for stock i at trade time t. TThee QQQQQQQQQQQQ SSSSSSSSSSdd ii,tt RRRRSS ii,tt = 2 NNNNOO ii,tt NNNNBB ii,tt MM ii,tt, (16) where NNNNOO ii,tt (NNNNBB ii,tt ) is the National Best Offer (Bid) quote for stock i at trade time t. FFFFFFFF MMMMMMMMMMMMMM PPPPPPPPPP IIIIIIIIIItt ii,tt RRRRRR5 ii,tt = DD MM ii,tt+5min MM ii,tt MM ii,tt, (17) where D is an indicator variable that is equal to +1 for buyer-initiated trades based on the Lee and Ready (1991) algorithm. Similarly, D equals to -1 for seller-initiated trades. 11

14 4. Empirical Results 4.1 Did trading volume of SSFs on banned stocks increase during the 2008 short sale ban? In this section, we attempt to answer our first research question: Did trading volume of SSFs on banned stocks increase during the 2008 short sale ban? In doing so, it is worthwhile to take a look at trading activities in the US single-stock futures market around the financial crisis period. Figure 2 shows the daily total dollar volume for the single stock futures market around the financial crisis period. It is clear to see that there is dramatic increase in the amount of trading in the OneChicago market during and immediately following the initiation of the short-sell ban. <FIGURE 2 HERE> Figure 3 exhibits average single-stock futures trading volume per stock (SSFVS) of banned stocks around the financial crisis period in 2008 and 2009, while Figure 4 plots average SSFVS around the 2008 short sale ban period. We compute SSFVS by summing all trades in all SSF contracts on each individual stock on each day (As in Grundy et al., 2012 pg.346). From Figures 3 and 4, SSFs trading volume has increased sharply on September 19, 2008 when the ban becomes effective, and it drops sharply after October 8, 2008 when the ban is lifted. <FIGURE 3 HERE> <FIGURE4 HERE> Figures 3 and 4 also plot 75 th and 90 th percentiles of SSFVS. We see a similar phenomenon that the 75 th and 90 th percentiles of daily trading volume of SSFs on banned stocks increases when the ban becomes effective and declines when it is lifted. In answering our first research question, did trading volume of SSFs on banned stocks increase during the short sale 12

15 ban? we analyze 458 US stocks with SSFs contracts listed and traded during the 2008 short sale ban period. Table 2 summarizes our sample selection criterion. There are 523 stocks that had SSFs listed on the OneChicago exchange prior to September 19, Of 523, 64 are stocks for which short selling was banned during the ban period (September 19 October 8, 2008). First, we analyze 64 banned stocks with SSFs to examine whether trading volume of SSFs increased for banned stocks during the ban period. Second, we analyze a full sample of 64 banned stocks and 394 non-banned stocks to examine whether trading activity in the SSFs market differed for banned stocks than for non-banned stocks during the ban period. Lastly, we create a matched control group of 64 non-banned stocks with similar size, volatility, and liquidity to those of 64 treatment banned stocks, and then conduct a difference-in-difference test to examine whether SSFs trading volume increased significantly more for banned stocks than for their matched nonbanned counterparts. <TABLE 2 HERE> Table 3 reports summary statistics of 458 sample stocks. From the preban period to ban period, logged SSF trading volume for banned stocks increased from to After the ban was lifted, the trading volume decreased to 0.455, suggesting that the 2008 short sale ban induced bearish investors to trade SSFs on banned stocks. A similar phenomenon is observed for non-banned stocks: SSFs trading volume increased from the preban period to ban period, and then decreased after the ban was lifted. <TABLE 3 HERE> Table 4 reports regression results from a daily panel of 64 banned stocks with SSFs traded before the short sale ban. The result from a univariate regression (model 1) suggest that 13

16 the short sale ban significantly increases SSFs trading volume for banned stocks. This effect of the ban on SSFs trading volume holds after we include control variables in models 2, 3, and 4. <TABLE 4 HERE> In Table 5, we include the 394 non-banned stocks sample and then analyze a daily panel of 458 stocks (64 banned and 394 non-banned) with SSFs using panel regressions with firmfixed effects. The short sale ban has a positive and significant effect on SSFs trading volume: When the ban is in effect, it increases SSFs trading volume for both banned and non-banned stocks. Contrary to our initial prediction that SSFs trading volume would increase more for banned stocks than for non-banned stocks during the ban period, however, increment is not significantly larger for banned stocks than for non-banned stocks during the ban period. <TABLE 5 HERE> In Table 6, we compare the 64 banned stocks with SSFs to the 64 matched control (nonbanned) stocks with SSFs by conducting a difference-in-difference test. Though we do not find any significant effect of the ban on SSFs trading volume for banned stocks, it is now of interest to examine whether this finding holds when we compare the 64 treatment sample with its matched control sample with similar size, volatility, and liquidity. In Table 6, we estimate panel regressions with firm-fixed effects. In Model 3, we find that the short sale ban has a significant and positive effect on SSFs trading volume for banned stocks, compared to their matched control sample with similar firm characteristics. <TABLE 6 HERE> From our analyses, we find that single-stock futures trading volume increased for both banned and non-banned stocks during the 2008 short sale ban period, suggesting that the 14

17 presence of SSFs created a viable alternative to short selling. Further analyses are conducted to examine to which extent SSFs trading volume on banned stocks increased relative to stocks that were not subject to the ban. We did not find evidence that trading volume of SSFs on banned stocks increased significantly larger than that for non-banned stocks. This suggests that the 2008 short sale ban had effects on trading of SSFs on both banned and non-banned stocks. We attribute this to not only the short sale ban but also the Rule 204T, which became effective on September 18, 2008 and made short selling more costly for all U.S. stocks. We subsequently use a difference-in-difference approach to compare the effect of the ban on SSFs trading volume for the 64 treatment (banned) stocks to that for the 64 matched non-banned stocks. We find that SSFs trading volume increased significantly more for banned stocks than for the matched nonbanned counterparts during the ban period. 4.2 What are the effects of SSFs on price discovery of underlying banned stocks during the short sale ban? Table 7 reports descriptive statistics for the price discovery variable. Price discovery variable PPDD ii,tt determines how much the SSFs trade contributes to price discovery of underlying stock i, measured over period t. We have a full sample group of 346 banned stocks that had SSFs listing on OneChicago during the ban period. We also create a sub sample of 64 banned stocks that had had SSFs listed before the short sale ban became effective on September 19, Note that we focus more on results on the latter sample group, as trading volume of SSFs on these 64 banned stocks represents almost all of the entire trading volume of SSFs on banned stocks. 15

18 Prior to the short sale ban period, the average contribution of SSFs to price discovery of underlying banned stock is 28.5 percent. The average contribution increased by 14.4 percent, to 43.0 percent, during the ban period for the sub-sample banned stocks, and increased to 44.9 percent for the full sample. After the short sale ban is lifted, the average contribution decreases to 16.4 and 17.3 percent for full- and sub-sample banned stocks, respectively. From this result, we find that SSFs contribution to price discovery of underlying banned stocks increased during the short sale ban. <TABLE7 HERE> Next, it is of interest to examine to what extent contribution of SSFs to price discovery for underlying banned stocks increased relative to non-banned stocks during the short sale ban period. To examine this using a differences-in-differences test, we create a control group of 64 unbanned stocks with SSFs to match with our treatment group of 64 banned stocks with SSFs. Table 8 shows descriptive statistics for matching variables, Market cap, Idiosyncratic volatility, and Turnover of both treatment and control groups. Size reported as the log of market capitalization is very similar for both groups, with a difference in means of Volatility reported as idiosyncratic volatility has essentially no difference between the means. Liquidity reported as turnover is much higher for the treatment group, however this would be expected because the treatment group is comprised of mostly financial firms, which were under a severe financial distress during the financial crisis period. This implies that each treatment sample is well matched with a control sample with similar size, volatility, and liquidity. <TABLE 8 HERE> 16

19 Table 9 makes comparison of the SSFs contribution to price discovery between the two groups. We predict that the effect will be stronger in the treatment group of banned stocks for which shorting was prohibited during the ban period, and hence that SSFs would not contribute to price discovery of unbanned stocks as much as that of banned stocks. Contrary to our prediction, Table 5 shows that SSFs contribution to price discovery increases dramatically for matched unbanned stocks as well. This implies that the 2008 short sale regulations; the short sale ban which was intended to protect financial stocks from falling stock price, and Rule 204T which applied to all stock, had effects on both banned and unbanned stocks, especially in the context of price discovery. <TABLE 9 HERE> What are the effects of the ban on price discovery contribution of single-stock futures when controlling for options? We test to see if our results are robust when taking into account options trading. First we separate the stocks with single-stock futures into two groups those with options trading and those without options trading, we then test for price discovery contribution in each of the groups. We find that when using the full sample of stocks with singles stock futures at the end of the shortsale ban, we have 151 stocks with single-stock futures and options and 198 stocks with singlestock futures and no options that have valid results. In Table 10 we find that the price discovery contribution increases to 44% in both cases during the ban period. <TABLE 10 HERE> 17

20 We then calculate the price discovery contributions of the spot, future and options price together. Chakravarty, Gulen and Mayhew (2004) find the option market contribution to price discovery to be about 17%, in this case we would expect to find similar results. Consistent with our previous findings, in Table 11, we find that the futures price contribution increases from 25% in the pre-ban period to 30% during the ban, while the spot, and option price discovery contribution decrease with the subsample of stocks with single stock futures prior to the ban. <TABLE 11 HERE> 4.3 What are the effects of SSFs on market quality of underlying banned stocks during the short sale ban? Figure 5 shows different measures of market quality from August 1, 2008 to October 31, 2008, for the banned stocks with single-stock futures and each of the three control groups. We see that for almost all measures, prior to the ban, all groups have similar market quality. We also see a spike for most measures at the beginning of the ban, while the non-banned stocks quickly revert, but not quite to pre-ban levels. We also see that banned stocks with single-stock futures have lower relative quoted spreads and relative effective spreads throughout the entire ban period, compared to banned-stocks without single stock futures. While these stocks also have lower 5 minute price impact and realized spreads for most of the ban period, compared to banned-stocks without single stock futures. We also see that all measure, banned stock market quality improves in the post ban period but does not go back to pre-ban levels, while for nonbanned stocks market quality stays at the banned period level, which is worse than the pre-ban 18

21 level. We can also see that non-banned stocks with singles-stock futures have consistently better market quality. <FIGURE 5 Here> Table 12 shows the descriptive statistics of the market quality measures, for banned stocks with single-stock futures and banned stocks without single-stock futures. We can see that in the pre ban period the stocks with single-stock futures have similar market quality measures. However, in general market quality is slightly better for those with single stock futures. During the ban period the market quality gets worse for both sample but is much worse for those without single stock futures. Table 13 reports the difference in difference regression results for banned stocks, with the market quality measures as the dependent variables. The quoted spread for stocks with singlestock futures during the ban period is 0.33 basis point lower than for stocks without futures. The 5 minute price impact for stocks with futures during the ban is basis points lower than for stocks without futures. <TABLE 13 HERE> Due to the illiquidity of the futures market, there are futures contracts that do not trade every day, to adjust for this we use SSF trade as a dummy variable if there is trading volume for the future of a particular stock. We use SSF trade instead of SSF and rerun the previous regression. Table 14 reports the difference in difference results with stock market quality measures as the dependent variables and using SSF trade as one of the independent variables. Table 14 shows that the interaction term SSF trade*ban Period is negative and significant for all market quality measures. We can see that the market quality of the underlying banned stocks is 19

22 better when there is single-stock futures trading compared to banned stocks without single-stock futures. Although market quality decreased for all stocks during the short-sale ban, stocks with single-stock futures had better market quality during the ban than those without single-stock future. <TABLE 14 HERE> 5. Conclusion We investigate the impact that the 2008 short-sale ban had on the single stock-futures market, and what effect single-stock futures had on the underlying equities during the ban period. We suggest that SSFs are a viable alternative for bearish investors when short-selling is restricted. We report a 55.5% increase in the number of stocks with single stock futures available to trade on the OneChicago Exchange. We also find a significant increase in volume in the single stock futures market, implying that the SSFs market provided a potential venue for investors to trade when short selling of designated financial and non-financial stocks was prohibited. We also find that SSF s contribution to price discovery of the underlying security increased across all stocks, although the price discovery increased more for non-banned stocks than for banned stocks. Rule 204T, restricting naked short-selling, was announced by the SEC one day before the short-sale ban on financial stocks was initiated, this increased the importance of SSFs role in information transfer for all stocks, however this importance was augmented for those stocks affected by the short-sale ban. Finally, the presence of single-stock futures lessened the negative impact of the short-sale ban on stock market quality reported by Boehmer et al. (2013). Although the single-stock futures 20

23 market is much smaller in comparison to options and stock markets, they are an essential part of financial markets, especially when short selling is constrained. 21

24 Bibliography Ang, A., Hodrick, R.J., Xing, Y. and Zhang, X., High idiosyncratic volatility and low returns: International and further US evidence. Journal of Financial Economics, 91(1), pp Battalio, R. and Schultz, P., Regulatory uncertainty and market liquidity: The 2008 short sale ban's impact on equity option markets. The Journal of Finance, 66(6), pp Benzennou, B., ap Gwilym, O. and Williams, G., Are single stock futures used as an alternative during a short selling ban?. Journal of Futures Markets. Blau, B.M. and Brough, T.J., Short sales and option listing decisions. Financial Management, 43(3), pp Boehmer, E., Jones, C.M. and Zhang, X., Shackling short sellers: The 2008 shorting ban. The Review of Financial Studies, 26(6), pp Chakravarty, S., Gulen, H. and Mayhew, S., Informed trading in stock and option markets. The Journal of Finance, 59(3), pp Danielsen, B.R., Van Ness, R.A. and Warr, R.S., Single stock futures as a substitute for short sales: Evidence from microstructure data. Journal of Business Finance & Accounting, 36(9 10), pp Easley, D., O'hara, M. and Srinivas, P.S., Option volume and stock prices: Evidence on where informed traders trade. The Journal of Finance, 53(2), pp Fung, J.K. and Tse, Y., Efficiency of single stock futures: An intraday analysis. Journal of Futures Markets, 28(6), pp

25 Grundy, B.D., Lim, B. and Verwijmeren, P., Do option markets undo restrictions on short sales? Evidence from the 2008 short-sale ban. Journal of Financial Economics, 106(2), pp Hasbrouck, J., One security, many markets: Determining the contributions to price discovery. The Journal of Finance, 50(4), pp Kumar, U. and Tse, Y., Single-stock futures: Evidence from the Indian securities market. Global Finance Journal, 20(3), pp Lee, C. and Ready, M.J., Inferring trade direction from intraday data. The Journal of Finance, 46(2), pp Miller, E.M., Risk, uncertainty, and divergence of opinion. The Journal of Finance, 32(4), pp Newey, W.K. and West, K.D., Hypothesis testing with efficient method of moments estimation. International Economic Review, pp Shastri, K., Thirumalai, R.S. and Zutter, C.J., Information revelation in the futures market: Evidence from single stock futures. Journal of Futures Markets, 28(4), pp

26 Tables and Figures

27 Figure 1: New single-stock futures listings on the OneChicago Exchange around the 2008 short sale ban This figure plots the number of new single-stock futures (SSFs) listings on the OneChicago Exchange by listing month (Panel A) and listing date (Panel B). The All stocks group consists of all US stocks whose SSFs contracts were introduced in The Banned stocks group consists of financial and non-financial stocks that were subject to the 2008 short sale ban. Panel A: New single-stock futures listings on the OneChicago Exchange in 2008 (by listing month) Before Feb-08 Mar-08 Apr-08 May-08 Jun-08 Jul-08 Aug-08 Sep-08 Oct-08 All stocks Banned stocks Panel B: New single-stock futures listings on the OneChicago Exchange around the 2008 short sale ban (by listing date) Before 9/19/ /26/2008 9/29/ /1/ /2/ /8/2008 All stocks Banned stocks

28 Figure 2: Daily dollar trading volume of single-stock futures on the OneChicago Exchange around the financial crisis period This figure plots daily total dollar trading volume of SSFs contracts on the OneChicago Exchange during the 01/01/ /01/2009 period. The shaded area is the 2008 short sale ban period: September 18, 2008 October 8, Total dollar volume is reported in millions.

29 Figure 3: Average daily single-stock futures trading volume per stock (SSFVS) during the period This figure plots the daily single-stock futures trading volume per stock (SSFVS) from 03/01/2008 to 06/30/2009. We follow Grundy et al. (2012) to compute SSFVS by summing all trades in all SSF contracts on each individual stock on each day. Panel A plots the average daily SSFVS, Panel B plots the third quartile SSFVS, and Panel C plots the 90 th percentile SSFVS. Each panel uses a different scale on the vertical axis. The solid (dashed) line represents banned (nonbanned) stocks. Panel A: Daily single-stock futures volume per stock (average)

30 Figure 3 (cont.): Panel B: Daily single-stock futures volume per stock (3 rd quartile)

31 Figure 3 (cont.): Panel C: Daily single-stock futures volume per stock (90 th percentile)

32 Figure 4: Average daily single-stock futures trading volume per stock (SSFVS) around the 2008 short sale ban This figure plots the daily single-stock futures trading volume per stock (SSFVS) around the 2008 short sale ban period. The ban became effective on 09/18/2008 and was lifted on 10/08/2008. We follow Grundy et al. (2012) to compute SSFVS by summing all trades in all SSF contracts on each individual stock on each day. Panel A plots the average daily SSFVS, Panel B plots the third quartile SSFVS, and Panel C plots the 90 th percentile SSFVS. Each panel uses a different scale on the vertical axis. The solid (dashed) line represents banned (nonbanned) stocks. Panel A: Daily single-stock futures volume per stock (average) around the 2008 short sale ban

33 Figure 4: (cont.) Panel B: Daily single-stock futures volume per stock (3 rd quartile) around the 2008 short sale ban

34 Figure 4: (cont.) Panel C: Daily single-stock futures volume per stock (90 th percentile) around the 2008 short sale ban

35 Figure 5: Average market quality measures around the 2008 short sale ban This figure plots average market quality measures around the 2008 short sale ban. The sample period covers from 08/01/ /31/2008. Panel A reports the average relative quoted spread, Panel B reports the average relative effective spread, Panel C reports average realized spread, and Panel D reports average 5-minute price impact. The Treatment line represents the average market quality measure of 64 banned-stocks that had SSFs listed prior to the ban. The Control 1 line represents the average market quality of 64 banned-stocks that never had SSFs listed during the sample period. The Control 2 (Control 3) lines represent the average market quality of 64 non-banned stocks that had (never had) SSFs listed during the sample period. Relative quoted spread is time-weighted, and the other market quality measures are equal-weighted. Each panel uses a different scale on the vertical axis. Panel A: Average relative quoted spread

36 Figure 5: (cont.) Panel B: Average relative effective spread

37 Figure 5: (cont.) Panel C: Average relative realized spread

38 Panel 5: (cont.) Panel D: Average 5-minute price impact

39 Table 1: New single-stock futures listings on the OneChicago Exchange around the 2008 financial crisis time This table exhibits the number of new single-stock futures (SSFs) listings around the 2008 financial crisis time. Prior to 2008, there were a total of 193 SSFs listings. Panel A reports the number of new SSFs listings in 2008 by month. Panel B reports the number of new SSFs listings during the 2008 short sale ban period (09/19/ /08/2008). Column All stocks reports new listings of SSFs on both banned and non-banned stocks, whereas column Banned stocks reports those on stocks that were subject to the ban. Month Panel A: New single-stock futures listings in 2008 by month All stocks New SSFs Listings Cumulative SSFs listings Month Banned stocks New SSFs Listings Cumulative SSFs listing Before Before Feb Feb Mar Mar Apr Apr May May Jun Jun Jul Jul Aug Aug Sep Sep Oct Oct Panel B: New single-stock futures listings around the 2008 short sale ban period Day All stocks New SSFs Listings Cumulative SSFs listings Month Banned stocks New SSFs Listings Cumulative SSFs listing Before 9/19/ Before 9/19/ /26/ /26/ /29/ /29/ /1/ /1/ /2/ /2/ /8/ /8/

40 Table 2: Sample selection and filtering The sample period is from August 1, 2008 to October 31, The 2008 short sale ban was initiated on September 19, 2008, and was lifted on October 8, To be included in our sample, stocks must have singlestock futures (SSFs) listed and traded on the OneChicago Exchange prior to September 19, 2008 and have SSFs data available on the Bloomberg terminal throughout the sample period. We also exclude stocks with data unavailable on CRSP from our analyses. We obtained a list of SSFs listings from the press release page of the OneChicago Exchange ( Sample selection Full sample Banned + non-banned stocks Sub sample Nonbanned Banned stocks stocks Number of stocks with SSFs traded prior to September 19, (Less) Number of stocks for which SSFs data are not available on the Bloomberg terminal (Less) Number of stocks for which data are not available on CRSP Total number of sample stocks in our sample

41 Table 3: Descriptive statistics Panel A of this table reports summary statistics of SSF trading volume, Stock return, Stock volume, and VIX. The sample consists of 458 U.S. stocks that had had single-stock futures (SSFs) listed before the 2008 short sale ban was initiated on September 19, stocks are banned stocks that had SSFs listed prior to the ban. 394 stocks are stocks for which short selling was never banned and that had SSFs listed before the ban. The sample period is from August 1, 2008 to October 31, The preban period is August 1, 2008 September 18, 2008; the ban period is September 19, 2008 October 8, 2008; and the postban period is October 9, 2008 October 31, SSF Trading volume is logged trading volume of SSFs contract on underlying stock i on day t. Stock return is daily return of underlying stock i on day t. Stock volume is daily trading volume of underlying stock i on day t. VIX is the S&P500 Volatility Index. Panel B reports the time-series average of the cross-sectional correlations between variables. Panel A: Summary statistics Full sample Sub sample Banned + non-banned stocks Banned stocks Non-banned stocks Preban Ban Postban Preban Ban Postban Preban Ban Postban SSF trading volume (logged) Stock return Stock volume (in Million) VIX Number of stocks Panel B: Correlations between variables SSF trading volume Ban period Post ban period Stock return Stock volume SSF trading volume Ban period 0.023* Post ban period * Stock return * 0.036* Stock volume 0.194* 0.018* 0.050* * VIX * 0.861* * 0.080* * indicates that significance at the 5 percent. VIX

42 Table 4: The effect of the 2008 short sale ban on trading volume of SSFs on banned stocks This table reports firm-fixed effects regression results from a daily panel of 64 banned stocks that had SSFs listed prior to the 2008 short sale ban. The dependent variable is logged daily SSFs trading volume of stock i. Firm-level control variables are Stock return and Stock volume. VIX is the S&P500 Volatility Index. Ban period is an indicator variable that is equal to one if an observation is during the ban period (September 19, 2008 October 8, 2008), and zero otherwise. Postban period is an indicator variable that is equal to one if an observation is after the ban period (after October 8, 2009), and zero otherwise. Significance is computed using Newey-West (1987) standard errors. Standard errors are reported in parentheses. ***, **, and * indicate p<0.01, p<0.05, and p<0.1, respectively. Dependent variable: SSFs trading volume (1) (2) (3) (4) Ban period 0.107** ** * ** (0.0457) (0.0479) (0.0479) (0.0460) Post ban period (0.0426) (0.0416) Stock return (0.178) (0.180) Stock volume *** *** ( ) ( ) VIX ( ) Observations 4,112 4,112 4,112 4,112 Adjusted R-squared

43 Table 5: The effect of the 2008 short sale ban on trading volume of SSFs on banned and non-banned stocks This table reports firm-fixed effects regression results from a daily panel of 458 stocks- 64 banned and 394 nonbanned stocks that had SSFs listed prior to the 2008 short sale ban. The dependent variable is logged daily SSFs trading volume of stock i. Banned stock is an indicator variable that is equal to one if a stock was subject to the ban, and zero otherwise. Firm-level control variables are Stock return and Stock volume. VIX is the S&P500 Volatility Index. Ban period is an indicator variable that is equal to one if an observation is during the ban period (September 19, 2008 October 8, 2008), and zero otherwise. Postban period is an indicator variable that is equal to one if an observation is after the ban period (after October 8, 2009), and zero otherwise. Significance is computed using Newey-West (1987) standard errors. Standard errors are reported in parentheses. ***, **, and * indicate p<0.01, p<0.05, and p<0.1, respectively. The coefficient on VIX is multiplied by Dependent variable: SSFs trading volume (1) (2) (3) (4) Ban period *** *** ** *** (0.0155) (0.0162) (0.0174) (0.0166) Ban period*banned stock (0.0505) (0.0484) Post ban period (0.0162) (0.0182) Post ban period*banned stock (0.0446) Stock return (0.0880) (0.0892) Stock volume *** *** ( ) ( ) VIX ( ) Observations 29,491 29,491 29,480 29,480 Adjusted R-squared