Trading Puts and CDS on Stocks with Short Sale Ban

Transcription

1 Trading Puts and CDS on Stocks with Short Sale Ban Sophie Xiaoyan Ni and Jun Pan September 14, 2010 Abstract We focus on the short sale ban of 2008 to examine the interaction between price discovery in banned stocks and the trading of options and CDS. Within the sample of banned stocks with exchange traded options, stocks whose put-call ratios are in the top quintile underperform the middle group by 1.56% and 2.84%, respectively, over the next two- and five-day returns. By contrast, the bottom quintile does not perform differently from the middle group. Within the sample of banned stocks with CDS traded and using their one-day percentage change in CDS spreads as a signal, we find cross-sectional predictability CDS signal for future stock returns. Again, the predictability is asymmetric, driven mostly by stocks with more positive percentage change in CDS spreads, and therefore more negative information according to the CDS market. Overall, our results confirm that in the presence of short sale ban, it takes time for the negative information contained in either the options market or the CDS market to get incorporated into stock prices. Sophie is from Hong Kong University of Science and Technology (HKUST) (sophieni@ust.hk); Pan is from MIT Sloan School of Management and NBER (junpan@mit.edu) 1

2 1 Introduction Shortly after the bankruptcy filing of Lehman and the government $85 billion bailout of AIG, the Securities and Exchange Commission issued an emergency ban on short selling for all financial stocks. While the original motivation for SEC to institute such a ban was to prevent further price disruptions and manage the wide-spread crisis of confidence in financial institutions, its effectiveness is questionable and the unintended consequences were many. 1 In this paper, we focus on the trading behavior in the options market during the short sale ban (SSB). As long as the cost of options trading is not prohibitively expensive, one could think of buying a put on the SSB stock as an indirect, but open channel, through which any potential negative information about the underlying stock could get in. In other words, although the more direct channel is closed, as long as the door is not completely shut, the information will find its way to securities prices. One interesting implication of this hardly surprising observation is that it allows us to follow the flow of information: from options market to the underlying stock market. Indeed, our first main result shows that in the presence of short sale ban, it takes time for the negative information expressed in the options market to get incorporated into stock prices. Our analysis focuses on the roughly 225 banned financial stocks with exchange traded options over the 14-day period between September 19 through October 8. 2 Using data from Chicago Board of Options Exchange, we construct put-call ratios from put and call option volumes initiated by customer buyers to open new option positions on the SSB stocks. For each of the 14 days, we sort the SSB stocks by their put-call ratios into three groups: the top and bottom 20% and the middle 60%. The stocks in the top 20% group are the ones that options investors express the most negative information, and they are also the most interesting stocks for us to study since short-sale ban has the potential to impede the incorporation of negative information into stock prices. We find that during the short sale ban, this top 20% group underperforms the middle 60% group by 1.56% over the next two-day return and by 2.84% over the next five-day return. The differences are not only economically large, but also statistically significant, even though this test was done over a 14-day window. By contrast, the performance of the bottom 20% group, which relatively speaking options investors are most positive about, is not economically or statistically different from that of the middle 60% group. In other words, while short sale ban does impede the incorporation of negative information into stock prices, it does not have an impact on the incorporation of 1 See, for example, the New York Times opinion piece by Arturo Bris on September 29, Over this time period, there were additions and removals from the original list, which are incorporated in our analysis. 2

3 positive information. By performing our analysis within the group of stocks with the same short sale ban and examine the predictability cross-sectionally within this group, we have perhaps provided one of the cleanest results in the literature to document the effect of short sale constraint on the incorporation of negative information. The cross-sectional aspect is particularly relevant given the volatile nature of the market at the time. It is, however, useful to perform a few more robustness tests. Matching each banned stock with an unbanned stock that also has exchange traded options, we perform the same analysis on this pool of control stocks over the same 14-day period of short sale ban and find no predictive result in either the top or bottom 20% group. For the SSB stocks, we also perform the same analysis over the 14-day periods immediately before and after the short sale ban, and find no result. The second main result of our paper focuses on the interaction between credit default swaps and stocks during the short sale ban. Given that CDS is an insurance on the potential default of a firm, it is naturally more sensitive to negative information, making it a good candidate for our analysis. It is, however, difficult to obtain trading information on CDS since it is traded over the counter. Moreover, our CDS sample is markedly smaller than the option sample: there are only 60 banned stocks with active CDS quotes. Nevertheless, our result finds that, in the presence of short sale ban, it takes time for the negative information already incorporated in CDS prices to get into stock prices. For each banned stock, we construct its daily percentage change in CDS spreads, and use it as a CDS signal. Cross-sectionally, a firm with more positive change in CDS is the one with more negative information, as far as the CDS market is concerned. Connecting this signal to stocks, we find that over the 14-day short ban period, if the CDS signal of one banned stock is one standard deviation higher than that of another banned stock, this stock on average underperforms the other by 1.29% over the next two-day return, and 2.69% over the next five-day return. Moreover, this predictability is asymmetric, driven mostly by stocks with more positive percentage change in CDS. Performing the same analysis over the 14-day periods immediately before and after the short sale ban, we do not find predictability. Performing the same analysis on 60 unbanned control stocks with CDS traded, we do not find any predictability. In other words, the short sale ban is an important ingredient in generating the CDS predictability, which is consistent with our main result from the options market. Conceptually, our paper follows the theoretical work of Diamond and Verrecchia (1987), who models effects of short-sale constraints on the speed of adjustment to private information of security prices. In particular, their model predicts that the effect of prohibiting short-sales 3

4 on the impoundment of private information into price is relatively more pronounced in the case of bad news versus good news. The asymmetric nature of our predictability results, for both options and CDS, are very much a confirmation of their prediction. 3 The impact of the 2008 short ban has been studied by a number of recent papers. Boehmer, Jones, and Zhang (2009) are among the first and they find that banned stocks suffered a severe degradation in quality, as measured by spreads, price impacts, and intraday volatility. Battalio and Schultz (2009) examine the impact of the ban on equity options. They find no evidence of migration of stock investors to the option market to avoid the ban, but they do find that the option market liquidity was severely affected by the short sale ban and synthetic share prices for banned stocks were lower than actual share prices. Kolasinksi, Reed, and Thornock (2009) examine both the ban on naked short sale on 19 stocks and the ban on all financial stocks, they find that the ban decreased market liquidity and increased the informativeness of short sales, and that both these changes were especially strong for stocks with listed options. Gagnon and Witmer (2009) document pricing differences between the U.S. and Canadian shares and trading volume migrations from U.S to Canada among banking stocks during the ban. Harris, Namvar, and Phillips (2009) use a factor-analytic model to show that the ban led to substantial price inflation in the SSB stocks. Our paper is also related to Sorescu (2000), Danielsen and Sorescu (2001), and Mayhew and Mihov (2005), who focus on option introductions and investigate whether options help relax binding short-sale constraints. Given that option introductions are generally endogenous events, it is difficult to learn conclusively about how the presence of options market might affect the underlying stock market. Our paper takes advantage of an event that was also highly endogenous, but this event was imposed on a broad enough set of stocks for us to examine the cross-sectional difference among the SSB stocks. Our paper is also related to the growing body of literature that examines the cross-market information transmission between stocks and options. Among others, Easley, O Hara, and Srinivas (1998) provide a formal framework to examine this issue and Pan and Poteshman (2006) present strong evidence that option trading volume contains information about future stock prices. Their result indicates that this information transmission is present during normal times and the predictability is symmetric with respect to positive and negative information. 4 We focus 3 It is, however, important to point out that the information that are used to gauge the speed of adjustment is in fact publicly available: both the put-call ratio and the CDS prices can be obtained in real time with a fee. 4 One important ingredient of their result is that the option volume information used in their analysis was not publicly available during their sample period. More recently and after the publication of their paper, however, the CBOE started to release such information in real time, and the predictability has decreased 4

5 instead on the short sale ban, and our predictability is asymmetric with respect to the nature of information: the predictability is present only for negative information, but not for positive information. The rest of the paper is organized as follows. Section 2 provides some details on the short sale ban and the option and CDS data used in our analysis. Section 3 presents our main result with respect to option trading and future stock returns during the short sale ban, and Section 4 presents the result relating CDS to future stock returns. Section 5 concludes. 2 Short Sale Ban, Data and Summary Statistics 2.1 Short Sale Ban On September 19, the SEC issued a short sale ban (SSB) on financial stocks. Initially, the ban covered 797 financial stocks selected by the SEC. On September 22, the following trading day, the SEC allowed exchanges to determine which firms would be included or excluded from the ban. Eventually, around 990 stocks were elected to be covered by the ban, with some stocks entering and exiting the ban list. The ban was set to expire in 10 days, but could be extended to 30 days at the SEC s discretion. On October 2, 2008, at the end of the initial 10-day effective period, the SEC extended the ban to the earlier of October 17, 2008 or three business days following the $700 billion financial rescue legislation was passed into law. On October 3 immediately after the rescue plan passed the Congress, the SEC announced that the ban would expire at 11:59pm ET Wednesday, October 8, The ban lasts 14 trading days from September 19 to October 8, we choose August 29 to September 18 as the pre-ban period, and October 9 to 28 as the post-ban period, so that both pre- and post- ban periods cover 14 trading days. 2.2 Option Data The data used to compute option trading activity is obtained from the CBOE. The dataset contains daily non-market maker volume for all CBOE-listed options. For each option, the daily trading volume is divided into four types of trades: open-buy, in which non-market markers buy options to open new long positions; close-buy, in which non-market makers buy options to close out existing written option positions; open-sell, in which non-market makers sell options to open new short positions; and close-sell, in which non-market makers markedly. For example, during our very short, but special sample period, we do not find any predictability of put-call ratio for unbanned stocks. 5

6 sell options to close out existing long options positions. Among these four types of option trades, the information content of the open-buy volume is perhaps the highest given that it does not involve any pre-existing option positions or margin requirements. For this reason, our main analysis will focus mostly on option volumes associated with open buy originated by by public customers. 5 During the SSB period, there are 225 banned stocks and 2036 non-banned stocks with options traded on CBOE. For comparison, we match each optionable SSB stock with an optionable control stock, which is selected by minimizing the sum of the squared percentage differences of stock market capitalization, stock volume, and stock price. The matching is done for the period from August 29 to October 28 and no control stock is used twice. Table 1 summarizes the option trading activities of the SSB and control stocks before, during, and after the short-sale ban. Reported are the call and put option volumes, as well as the relative trading volumes between option and the underlying stock. For each variable, we compute the cross-sectional mean, median and standard deviation for each trading day and report their time series averages. Comparing the call and put option volumes between the SSB and control stocks over the different time periods, we find mixed results depending on whether we use mean or median. But when the option volume is measured relative to the underlying stock volume, we see an increased activity in the options market for the SSB stocks. This can be seen in two ways. First, during the SSB period, the option/stock volume of the SSB stocks is higher than that of the control stocks. Second, the option/stock volume of the SSB stocks is higher during the SSB period than before and after the SSB period. Breaking down the option volume to call and put volumes suggests that the high relative option to stock volume mainly comes from put options. For example, during the SSB period, the mean of relative put to stock volume is 207 basis points for the SSB stocks and 136 for the control stocks, while the mean of relative call volume is 182 for the SSB stocks and 171 for control stocks. To further quantify the relative trading between put and call options, we use a measure of put/call ratio, which is the put option volume divided by sum of the put and call option volumes. Table 1 reports the summary statistics of the put/call ratios using put and call option volumes associated with the four types of trades: open buy, open sell, close buy, and 5 For each type of non-market maker trade, the Option Clearing Corporation assigns one of the two origin codes: F for firm proprietary traders, C for public customers. An example of a firm proprietary trader would be an employee of Goldman Sachs trading for the bank s own account. Public customers are the rest of non-market maker investors. The CBOE then subdivides public customer volumes into volumes of small, medium and large trades for orders less than 100 contracts, between 100 and 199 contracts, and larger than 199 contracts, respectively. 6

7 Table 1: Summary Statistics of Stocks with CBOE Options Traded SSB Stocks Control Stocks mean median std mean median std Panel A: Ban Period, Sep 19 - Oct 8 Call volume (contract) Put volume (contract) Option/Stock Volume (bp) Call/Stock Volume (bp) Put/Stock Volume (bp) Open Buy p/(p+c)(%) Open Sell p/(p+c)(%) Close Buy p/(p+c)(%) Close Sell p/(p+c)(%) Panel B: Before Ban Aug 29 - Sep18 Call volume (contract) Put volume (contract) Option/Stock Volume(bp) Call/Stock Volume(bp) Put/Stock Volume(bp) Open Buy p/(p+c)(%) Open Sell p/(p+c)(%) Close Buy p/(p+c)(%) Close Sell p/(p+c)(%) Panel C: After Ban Oct 9 - Oct28 Call volume (contract) Put volume (contract) Option/Stock Volume(bp) Call/Stock Volum(bp) Put/Stock Volume(bp) Open Buy p/(p+c)(%) Open Sell p/(p+c)(%) Close Buy p/(p+c)(%) Close Sell p/(p+c)(%) Reported are the time-series averages of cross-sectional mean, median, and standard devation. Put/call ratios are defined as p/(p+c), where p and c are the put and call option volumes, respectively. 7

8 0.65 OpenBuy P/(P+C) 0.6 ssb control /29 09/19 10/08 10/28 Figure 1: Daily Time-Series of Open-Buy Put/Call Ratios 8

9 close sell. Overall, we see that the SSB stocks have higher open-buy put/call ratios than the control stocks. Figure 1 paints a more detailed picture by plotting the daily time series of the open buy put/call ratios. As we can see, there was a run up in the put/call ratio for the SSB stocks before the ban. But this ratio dropped dramatically on the the first day of the ban, when the option market was severely disrupted. Except for the first day of the ban, the SSB stocks had much higher put/call ratios than the control stocks during the ban. After the ban, the difference became less severe. 2.3 CDS Data The CDS data used in this paper are from Thomson Datastream. During the short ban period, there are 395 stocks with CDS quotes, among which 60 are SSB stocks and 335 are non-ssb stocks. We select 60 control stocks from the 335 non-ssb stocks using the same matching method as before. In particular, control stocks are selected by matching stock market capitalization, stock trading volume and stock price. Table 2: Summary Statistics of SSB and Control Stocks with CDS SSB Control mean median std mean median std Panel A: Ban Period, Sep 19 - Oct 8 CDS Spread (bp) Daily Change (bp) Daily Percentage Change (%) Panel B: Before Ban, Aug 29 - Sep18 CDS Spread (bp) Daily Change (bp) Daily Percentage Change (%) Panel C: After Ban, Oct 9 - Oct 28 CDS Spread (bp) Daily Change (bp) Daily Percentage Change (%) Reported are the time-series averages of cross-sectional mean, median, and standard devation. Table 2 summarizes the CDS data during, before and after the ban. During the ban, the SSB stocks on average have higher CDS spread than the control stocks, and this is reflected in both the cross-sectional mean and median. For example, the average CDS spread was bps for the SSB stocks, compared with that of bps for the control stocks. Before and 9

10 after the ban, however, this relation is somewhat mixed, depending on whether mean or median is used in the comparison. We also find that the SSB sample is more dispersed with a higher cross-sectional standard deviation. Overall, there is a general pattern of increasing CDS spreads for both the SSB and control stocks over the sample period as the broad economic condition worsens during our sample period. 3 Option trading and future stock returns 3.1 Main Results This section investigates whether the option trading contains negative information about banned stocks. As predicted by the model of Diamond and Verrecchia (1987), in the presence of short sale prohibition, it takes longer for the stock prices to incorporate negative information relative to positive information. With access to the options market, however, investors can take advantage of the ban by taking an open buy position on put options. If stock prices incorporate negative information with a delay, we would expect that the stocks with high put/call ratios during the ban have lower future returns, while stocks with low put/call ratios may not have high future returns. Our main empirical investigation is to test this asymmetric predictability for the SSB stocks during the SSB period. Moreover, we expect this particular predictability to be less obvious for control stocks during the ban or the SSB stocks right before or after the ban. We focus first on the sample of SSB stocks during the ban period. On each date t during the ban period, we sort SSB stocks by their put/call ratios into quintiles and focus on the stocks in the top and bottom quintiles. Specifically, if stock i is in the top (highest) quintile on date t, we assign Ht i to be 1, and zero otherwise. Likewise, if stock i is in the bottom (lowest) quintile on date t, we assign L i t to be 1 and zero otherwise. We then test the asymmetric predictability using the following specification: Rt,t+τ i = a + b H Ht i + b L L i t + c ln(size i ) + d (opt/stk vol) i t + f Rt 5,t 1 i + ϵ i t, (1) where Rt,t+τ i with τ = 1, 2,..., 5 is the future τ-day cumulative excess return for stock i adjusted using the Fama and French (1993) three-factor model. In addition to the H and L dummies, which are the key components of our specification, we also include a few control variables: size is the market capitalization of the underlying stock, (opt/stk vol) i t is the ratio of option to stock volume on date t for stock i, and Rt 5,t 1 i is stock i s cumulative excess return from date t 5 to date t 1. 10

11 The most interesting aspect of our test relates to the difference between the regression coefficients b H and b L in Equation (1). If short sale prohibition does slow down the adjustment of stock prices to negative information and if this negative information is indeed contained in the trading of put options, we would expect the coefficient b H to be negative and significant for the sample of SSB stocks during the ban period. By contrast, the coefficient b L picks up the positive side of the information content and is not affected at all by the presence of short-sale ban. Moreover, we would expect this asymmetry between b H and b L be to present only during the ban period for the sample of SSB stocks. In particular, we do not expect to observe the same asymmetric predictability for the sample of control stocks during the ban, or for the SSB sample before or after the ban period. Our results are summarized in Table 3. Indeed, as reported in Panel A of Table 3, we find that for the sample of SSB stocks during the ban, the slope coefficient b H on the high put/call ratio dummy H is negative and statistically significant while the slope coefficient b L on the low put/call ratio dummy L is not significant. When the dependent variable is future two-day return, the coefficient estimate on H is bps with a t-stat of 2.30, while that on L is bps with a t-stat of When the dependent is future five-day return, the coefficient estimate on H is with a t-statistics of 4.48, while the estimate on L is with a t-stat of These numbers indicate that relative to the average SSB stocks, those SSB stocks in the top put/call ratio quintile underperforms by 1.56% and 2.84%, respectively, over the next two and five business days. By contrast, the future returns of the SSB stocks with below 20th percentile put-call ratios are not different from the average SSB stocks during the ban. As a comparison, we also perform the same regression specification for control stocks during the ban as well as for SSB stocks over the non-ban periods. Panel B of Table 3 reports the result for the control sample during the ban period. As expected, we do not find any predictability of option trading volume for future stock returns: neither the coefficient on the high put/call ratio dummy H nor that on the low put/call ratio dummy L is statistically significant. Panel C of Table 3 reports the estimates of coefficients on H and L for the SSB sample 14 trading days before the ban (August 29 to September 18), 14 trading days after the ban (October 9 to October 28), and during regular period (January to July 2008). Again, we do not find any predictability of option trading volume for future stock returns. Overall, the asymmetric predictability documented in our paper is specific only for the SSB stocks during the ban period, which is consistent with our hypothesis. For the coefficient estimates on control variables, the estimates on size are negative for both the SSB and control stocks during the ban, and statistical significant for the control 11

12 Table 3: Future Stock Returns and Open Buy Put Call Ratio τ const Ht i L i t size opt/stk Rt 5,t 1 i Panel A: SSB stocks during the ban [0.26] [-1.74] [1.30] [-0.32] [-1.13] [-2.53] [0.46] [-2.30] [0.48] [-0.54] [-1.26] [-2.78] [1.00] [-1.49] [0.28] [-1.08] [-1.39] [-2.84] [1.65] [-2.39] [-1.79] [-1.64] [-1.55] [-3.53] [1.04] [-2.44] [-0.49] [-1.00] [-1.14] [-3.47] Panel B: Control stocks during the ban [-0.36] [0.70] [1.25] [-0.02] [-1.59] [-1.02] [1.08] [0.35] [0.95] [-1.75] [-1.73] [-0.55] [1.95] [-0.54] [0.00] [-2.45] [-3.35] [-0.89] [2.67] [-0.89] [-0.14] [-3.13] [-2.46] [-1.72] [3.70] [-0.97] [-0.58] [-3.85] [-2.36] [-1.35] Panel C: SSB stocks in non-ban periods Ht i L i t Ht i L i t Ht i L i t [-0.64] [0.27] [0.19] [-0.09] [-0.64] [1.84] [-1.18] [0.47] [-0.04] [0.41] [-0.64] [0.80] [-1.69] [0.26] [0.11] [-0.58] [-0.72] [1.14] [-1.54] [0.81] [0.19] [-0.59] [-1.29] [1.11] [-1.14] [0.69] [0.47] [0.15] [-1.41] [1.10] Daily cross-sectional regression of cumulative excess returns Rt,t+τ i on the high and low put/call dummies Ht i and L i t. On each day t, stocks within the respective sample are sorted by their open buy put/call ratio and those ranked above 80% have Ht i = 1 and those below 20% have L i t = 1. Additional controls include opt/stk, the ratio of option and stock volumes; Rt 5,t 1, i lagged returns; and the log of equity size. Fama-MacBeth T-stats are reporetd in squared brackets. 12

13 stocks, suggesting large stocks earn lower excess returns than small stocks. The estimates on relative option to stock volume are negative for the SSB and control stocks but with little statistical significance. The estimates on previous five trading day returns are negative for both the SSB and the control stocks and statistically significant for the SSB stocks. 3.2 Daily Coefficients Given that the short-sale ban lasts only for 14 trading days, it is possible that our asymmetric predictability result of the SSB stocks is driven by a few extreme observations. Also, it s worthwhile to check whether or not the predicted returns reverse in a longer time period. Figure 2 plots the daily coefficient estimates on H when the dependent variables are future one-day and ten-day returns, respectively. We find that the high put/call ratio dummy H can predict future banned stock returns on most of the trading days during the ban: when the dependent variable is the future one-day return, 12 out of 14 coefficients on H are negative. In other words, our result is not driven by a few extreme observations. Moreover, the predicted returns remain negative for longer time period: 11 coefficients are negative when the dependent is the ten-day future return. On September 19th, the first day of the ban, SEC only granted option market maker to sell short when selling short as part of bona fide market making and hedging activities related directly to bona fide market making in derivatives on the SSB stocks until 11:59 pm on September 19th. This suggests that option market makers would be unable to sell short for any reason during the remainder of ban. While on early morning of Monday September 22nd, the second trading day of the ban, the SEC confirmed that option market maker can short sell during the remainder of the ban. Battalio and Schultz (2009) document that on the first day of the ban, when market makers expected that they would be prohibited from short sale, the transaction cost for options on banned stocks are extremely high; the marginal effect of the ban on bid ask spread of options on the SSB stocks is twice as large as on the other ban days. Figure 2 shows that on the first day of the ban, high put/call ratio dummy, H, has no power in predicting future stock returns. The coefficients on H are both positive when the dependent is one day and ten day returns. Also as shown in Figure 1, the put/call ratios are exceptionally low on the same day. These findings imply that when option market makers are prohibited from short sale, they are unwilling to sell puts to retail investors, and option trading does not contain negative information. 13

14 1000 Daily Coefficients 0 Basis Point day 10 days 19/09 26/09 03/10 08/10 Figure 2: Daily Coefficient on High Put/Call Ratio Dummy H 14

15 3.3 Robustness Test It is possible that the predictability of option volumes of the SSB stocks is from those illiquid options, of which the put/call ratios are erratic and unlikely to contain information. Then the observed link between high put/call ratios and low future returns of the SSB stocks during the ban may be just a co-incidence. For example, a stock with 10 open buy put contracts and 1 open buy call contract experiences low return in following days. To exclude this possibility, we redo the test specified in equation (1) and restrict the sample to SSB stocks with at least 150 daily average open purchased option contracts from January to July Also samll sample problem may be the reason that there s no predictability in option volumes of control stocks or option volumes of banned stocks pre- or post ban. To examine this alternative explanation, we re-estimate the coefficients in equation (1) use all non-ssb stocks as control, and expand pre- or post ban period from 14 trading days to two months. Table 4 reports the average daily coefficient estimates on high and low put/call ratio dummies for the robustness test. To save space, we omit the estimates on the constant and the control variables. After we restrict SSB stocks to those with liquid options, the coefficients on high put/call ratio dummy (H) become more negative. For example, when the dependent variable is future five day returns, the coefficient on H is (t-statistics: -3.42) for the SSB stocks with liquid options, and only (t-statistics: 2.44) for the whole sample of the SSB stocks. This result suggests that the predictability of high put/call ratios of the SSB stocks during the ban comes mainly from the stocks with liquid options. Table 4 also shows that insufficient number of observations is not the reason for the nonpredictability of put /call ratios of control stocks during the ban or SSB stocks pre- or post the ban. The coefficients on H are not statistically different from zero either for all non-ssb stocks during the ban or for the SSB stocks two months pre- or post the ban. 3.4 Different Trade Sizes and Trade Volumes We now examine the information content of options on the SSB stock by breaking down option volumes into different trade sizes. CBOE divides public customer volumes into volumes of small, medium and large trades. The small trades are orders less than 100 contracts, the larger trades are orders larger than 199 contracts, and the medium are in between. As specified in equation (1), we regress the future risk-adjusted returns on the high and the low put-call ratio dummies constructed from the open buy volumes of different contracts is the median of daily average option volumes for the SSB stocks during the period from January to July

16 Table 4: Future Stock Returns and Put Call Ratios, Robustness Test τ Ht i L i t Ht i L i t Panel A: SSB and Non-SSB Stocks during the Ban SSB Stocks with Liquid Options All non-ssb Stocks [-1.59] [0.29] [-0.48] [ [-1.81] [0.38] [-0.91] [-0.83] [-1.84] [0.10] [-0.35] [-1.03] [-2.43] [-1.69] [-0.42] [-0.76] [-3.42] [-0.88] [0.29] [-0.51] Panel B: SSB Stocks 2 Months Before and After the Ban 2 Months Before 2 Months After [-1.70] [1.65] [-0.72] [1.09] [-1.25] [1.39] [-1.30] [1.73] [-1.38] [1.80] [-1.00] [1.37] [-1.67] [1.50] [-1.42] [1.39] [-1.60] [1.22] [-1.36] [1.90] Daily cross-sectional regression of cumulative excess returns R i t+1,t+τ on the high and low put/call dummies H i t and L i t. On each day t, stocks within the respective sample are sorted by their open buy put/call ratio and those ranked above 80% have H i t = 1 and those below 20% have L i t = 1. Controls include opt/stk, the ratio of option and stock volumes; R i t 5,t 1, lagged returns; and the log of equity size. Fama-MacBeth T-stats are reporetd in squared brackets. Coefficients on controls are not reported. 16

17 trade sizes. The coefficients are reported in Table 5 Panel A. Again, to save space, we omit coefficients on the constant and conrol variables. The open-buy volumes from small and medium trades provide stronger predictive power than those from large trades in both magnitude and statistical significance. Similar to the main results, the prediction only occurs on stocks with high put-call ratios but not low put-call ratios. The different predictabilities of small and large trades may arise from increased trading cost. Battalio and Schultz (2009) document that market makers increased bid ask spread of the option prices especially for the SSB stocks during the ban. Confronting with enlarged bid ask spread, investors may be more likely to use small trades to reduce their trading impact on option prices. Figure 3 plots option volumes from small, medium and large trades and shows that indeed during the ban volumes from small trades are higher than those from large trades for the SSB stocks, but not for the control stocks. While during regular trading period before the ban, the trading volumes from large trades are higher than those from small trades for both the SSB and control stocks SSB stocks Contracts (000) small medium large 0 01/01 06/01 09/1910/08 11/ Control stocks Contracts (000) small medium large 0 01/01 06/01 09/1910/08 11/28 Figure 3: Option Volumes of Different Trade Sizes 17

18 Table 5: Future Stock Returns and Put Call Ratios from Different Trade Sizes and Option Volumes Panel A: Put Call Ratios of Different Trade Size Small Medium Large τ Ht i L i t Ht i L i t Ht i L i t [-2.38] [0.79] [-2.12] [-1.71] [-1.41] [0.12] [-2.55] [1.14] [-2.92] [0.27] [-0.91] [0.65] [-2.87] [0.81] [-2.18] [0.00] [-0.80] [-0.27] [-3.14] [0.16] [-2.25] [-0.15] [-0.44] [-0.20] [-3.35] [1.48] [-0.80] [0.17] [-0.05] [-0.96] Panel B: Put Call Ratios of Different Trade Volumes τ Open Sell Close Buy Close Sell Ht i L i t Ht i L i t Ht i L i t [-0.30] [-0.90] [-2.16] [-0.71] [-2.18] [-0.73] [-0.41] [-1.15] [-2.13] [-0.40] [-1.81] [-0.05] [-0.89] [-1.14] [-1.77] [-0.39] [-0.72] [0.80] [0.11] [-1.55] [-2.15] [-0.98] [-0.83] [0.38] [0.01] [-0.95] [-3.69] [-1.37] [-1.39] [-1.03] Daily cross-sectional regression of SSB cumulative excess returns R i t,t+τ on the high and low put/call dummies H i t and L i t calculated from various option volume duing the ban. On each day t, SSB stocks are sorted by their put/call ratio co and those ranked above 80% have H i t = 1 and those below 20% have L i t = 1. Additional controls include opt/stk, the ratio of option and stock volumes; R i t 5,t 1, lagged returns; and the log of equity size. Coefficient estimates on contols are omitted. Fama- MacBeth T-stats are reporetd in squared brackets. 18

19 We now examine the information content of the other option volume types: open sell, close buy, and close sell. When in possession of a negative private signal about an underlying stock, an investor can buy fresh (open buy) put options or sell fresh (open sell) call options. If informed traders bring negative information to the open sell volume, we would expect positive slopes on the high put-call ratio dummy in the predictive regression. The results reported in Table 5, however, indicate that investors do not bring negative information to the open sell volumes of banned stocks during the ban period. This may be because sale of fresh put has margin requirement and less appeal to informed option investors. Informed investors can also close their existing option positions and thereby bring information to the close buy and close sell option volumes. Pan and Poteshman (2006) document that the information content from closing trades is usually lower than that from open trades because traders can only use information to close positions if they happen to have appropriated positions open at the time they become informed. However when SEC granted option market maker to short sell, it added a provision that market makers could not short if market makers knew the counter party was increasing an economic net short position in the shares of that stocks. This provision implicitly assumes that the market maker can distinguish whether a trade is initiated to open a new option position or to close an existing option position.ni, Pan, and Poteshman (2008) document that market makers can somehow distinguish close and open trades by showing that the open trades have higher price impact than close trades. If market makers observe SEC s provision, the open trades would be more difficult to be executed than the close trades during the ban period. Hence the predictability of closing trades may not be lower than that of open trades on the SSB stocks. As expected, Table 5 Panel B shows that the predictability of close buy volume is similar to that of open buy volume; for example, when the dependent variable is future five day returns, the coefficient on open buy high put/call ratio dummy (H) is -284, and the coefficient on close buy H is CDS and future stock returns Another channel for investors to profit from negative information is to trade CDS. Given that CDS is an insurance on the default of a firm, it is naturally more sensitive to negative information, making it a good candidate for our analysis. Cross-sectionally, a firm with more positive change in CDS is the one with more negative information.if short sale ban slows down the stock price adjustment to negative information, we would expect an increase in CDS predicts a low future stock return. To test this predicatability we use the following 19

20 regression model: R i t,t+τ = a + b Y i t + c ln(size i t) + d R i t 6,t 1 + ϵ i t, (2) where Y i t = 100% x (CDS i t CDS i t 1)/CDS i t 1, (3) where Rt,t+τ i is the cumulative return of stock i from trade day t to t+τ, Yt i is the percentage change of CDS from date t 1 to date t, Rt 6,t 1 i is past five day return. If short sale ban slows down the stock price adjustment to negative information, the coefficient on Yt i will be negative for the SSB stocks during the ban, but not different from zero for the control stocks during the ban or SSB stocks during the non-ban periods. The result presents in Table 6 shows that the coefficient estimates on Y are negative and statistically significant only for the SSB stocks during the ban. When the dependent variable is two (five) day future returns, the coefficient estimate is (24.07), implying a one standard deviation increase in Y is associated with 1.29% (2.69%) return decrease in next two (five) days. While for the control stocks, the coefficient estimates on Y are not significantly different from zero. Panel B of Table 6 reports the coefficient estimates on Y for the SSB stocks during the non-ban periods. To save space we omit the estimates on the constant and control variables. As shown in Panel B, the CDS has no predictability during non-ban periods. Right before the ban and during the normal time period from January to July 2008, the coefficients on Y are negative without statistical significance, and during the post ban period, the coefficients are mixed on sign and not significant either. If short sale ban impedes stock prices from incorporating negative information but not positive information, we would expect only increases of CDS during the ban can predict low stock returns, but decreases of CDS cannot predict high stock returns. asymmetric predictability hypothesis we use following specification: To test this R i t,t+τ = a + b Y i t + c Y i t Y H i t + d Y H i t + f ln(size i t) + f R i t 6,t 1 + ϵ i t, (4) where Y Ht i is the high CDS change dummy; it is one for company i on day t if its Yt i is above median on day t. If the predictability of Y is asymmetic for the SSB stocks during the ban, the coefficient on the interaction term between Y and Y H will be negative, and the predictability of Y in equation (4) will be weaker than that in equation (2). As shown in Table 7, the coefficient estimates on the interaction term between Y and Y H are all negative and statistically significant. After adding the interaction term, the CDS change (Y ) no longer has predictability for the SSB stock returns. For example, when the dependent variable is two day future returns, the coefficient estimate on the interaction term is (t-statistics: -2.09), while the estimate on Y is (t-statistics: 1.29). These 20

21 Table 6: Future Stock Returns and Percentage Change of CDS Panel A. The SSB and the control stocks during the ban SSB stocks Control stocks τ const Yt i size Rt 5,t 1 i const Yt i size Rt 5,t 1 i [-2.59] [-1.89] [2.23] [-2.25] [-4.48] [1.48] [3.35] [0.25]] [-3.86] [-2.61] [2.75] [-2.18] [-3.26] [0.90] [1.84] [0.92] [-5.40] [-1.96] [3.97] [-2.47] [-3.81] [1.09] [2.41] [1.29] [-4.57] [-1.91] [4.75] [-2.04] [-4.37] [0.49] [3.06] [0.70] [-5.39] [-2.01] [4.42] [-1.44] [-4.92] [0.56] [3.25] [0.98] τ Panel B. The SSB stocks in non-ban periods Y i t [-1.64] [0.10] [-1.29] [-1.49] [0.92] [-1.40] [-0.57] [1.10] [-0.69] [-0.87] [0.75] [-0.96] Daily cross-sectional regression of cumulative returns Rt,t+τ i on CDS percentage change Yt i in basis points. Controls include the log of equity size and lagged returns, Rt 5,t 1. i Coefficient estimates on controls are omitted in Panel B. Fama-MacBeth t-stats are reporetd in squared brackets. Y i t Y i t 21

22 Table 7: Future Stock Returns and High Percentage Change of CDS τ const Y i t The SSB Stocks during the Ban Y i t H i t H i t size R i t 5,t [-2.33] [1.67] [-2.84] [0.83] [1.91] [-2.11] [-3.43] [1.29] [-2.06] [-0.18] [2.52] [-1.94] [-4.81] [1.16] [-3.59] [1.01] [3.59] [-2.42] [-4.50] [0.74] [-2.14] [0.64] [5.30] [-2.07] [-4.52] [0.52] [-1.93] [-0.48] [4.11] [-1.40] Daily cross-sectional regression of cumulative returns R i t,t+τ on CDS percentage change, Y i t, in basis points, and its intersection with high CDS change dummy, H i t. Controls include the log of equity size and lagged returns, R i t 5,t 1. Fama-MacBeth t-stats are reporetd in squared brackets. numbers indicate that the predictability of CDS during the ban is driven by the SSB stocks with high increase in CDS but not those with low or negative increase in CDS. This result is consistant with that short sale ban reduces the speed of stock price adjustment to negative information. 5 Conclusions In this paper, we focused on the short sale ban of 2008 and examined the interaction between price discovery in banned stocks and the trading of options and CDS during the 14-day period. We found that in the presence of short sale ban, it takes time for the negative information contained in either the options market or the CDS market to get incorporated into stock prices. Given the volatile nature of the markets during the short sale ban, it is important for us to minimize the influence of the overall market movement on our predictive results. For this reason, we structured our test to focus within the group of stocks with the same short sale ban and examine the predictability cross-sectionally within this group. In other words, our finding of sluggishness in price impoundment is a measure of one banned stock relative to another banned stock. The only thing differentiating one from another is the fact that it has some negative information according to the options market or the CDS market. 22

23 In extracting the information from options market, we chose to use the trading information instead of prices, since we would like to capture more of an action variable. Given the volatile nature of the time period and the large bid/ask spreads in the options market, this action variable could perhaps capture option information in a cleaner way. In extracting the information from the CDS market, we could only use the CDS price since trading information was not available. This differing information content has some interesting implications. While our result shows that it takes time for the negative information expressed in the options market to get incorporated into stock prices, our paper is silent on whether or not this information has already been incorporated in the option prices. By using the CDS price data, however, our paper does show that, in the presence of short sale ban, it takes time for the negative information already incorporated in CDS prices to get into stock prices. Finally, although our original idea was as simple as following the flow of information, our result is a confirmation of the theory prediction of Diamond and Verrecchia (1987). In particular, our result shows that the sluggishness in price impoundment is asymmetric: the information collected from either the options market or the CDS market has predictability only on the negative side, not on the positive side. Moreover, this predictability was not present for the periods immediately before and after the short sale ban, nor was it present for a matching sample of stocks that were not banned. 23

24 References Battalio, R. and P. Schultz (2009). Regulatory Uncertainty and Market Liquidity: The 2008 Short Sale Ban s Impact on Equity Option Markets. Working Paper, University of Notre Dame. Boehmer, E., C. Jones, and X. Zhang (2009). Shackling Short Sellers: The 2008 Shorting Ban. Working Paper, University of Oregon and Columbia Business School and Cornell University. Danielsen, B. and S. Sorescu (2001). Why Do Option Introductions Depress Stock Prices? An Empirical Study of Diminishing Short-Sale Constraints. Journal of Financial and Quantitative Analysis 36, Diamond, D. and D. Verrecchia (1987). Constraints on Short-Selling and Asset Price Adjustment to Private Information. Journal of Financial Economics 18, Easley, D., M. O Hara, and P. Srinivas (1998). Option Volume and Stock Prices: Evidence on Where Informed Traders Trade. Journal of Finance 53, Fama, E. and K. French (1993). Common Risk Factors in the Returns on Stocks and Bonds. Journal of Financial Economics 33, Gagnon, L. and J. Witmer (2009). Short Changed? The Market s Reaction to the Short Sale Ban of Working Paper, Queen s University. Harris, L., E. Namvar, and B. Phillips (2009). Price Inflation and Wealth Transfer During the 2008 SEC Short-Sale Ban. Working Paper, University of Southern California. Kolasinksi, A., A. Reed, and J. Thornock (2009). Prohibitions versus Constraints: The 2008 Short Sales Regulations. Working Paper, University of North Carolina. Mayhew, S. and V. Mihov (2005). Short Sale Constraints, Overvaluation, and the Introduction of Options. Working Paper, SEC and Texas Christian University. Ni, S. X., J. Pan, and A. Poteshman (2008). the Information in Option Volume for Future Stock Volatility. Journal of Finance 63, Pan, J. and A. Poteshman (2006). The Information in Option Volume for Future Stock Prices. Review of Financial Studies 19, Sorescu, S. (2000). The Effect of Options on Stock Prices: Finance 55, to Journal of 24