Informed Options Trading prior to M&A Announcements Preliminary Draft - Do not quote without permission

Transcription

1 Informed Options Trading prior to M&A Announcements Preliminary Draft - Do not quote without permission Patrick Augustin Menachem Brenner Marti G. Subrahmanyam McGill University, Desautels New York University, Stern New York University, Stern November 17, 2013 Abstract We investigate the possibility of insider trading in equity options around major informational events such as corporate mergers and acquisitions (M&A) for a sample of 1,859 US events, focusing on both the target and the acquirer. Our goal is to quantify the likelihood of informed trading and information leakage based on trading volume and implied volatility attributes. We further focus our attention on options strategies which, in anticipation of public news announcements and the presence of private information, should result in abnormal returns to insiders. We find a highly positive abnormal trading volume and excess implied volatility in anticipation of M&A announcements, with stronger effects for OTM call options on the target. We further document a decrease in the term structure of implied volatility, an increase in the implied volatility spread and an average rise in percentage bid-ask spreads of approximately 20 percentage points prior to the announcements. The latter effect arises from short-dated and OTM options. Keywords: Asymmetric Information, Insider Trading, Mergers and Acquisitions, Market Microstructure, Equity Options JEL Classification: G13, G14, G34, G38 We thank participants in the OptionMetrics Research Conference, New York, and a seminar at NYU Stern for comments on a prior draft. McGill University - Desautels Faculty of Management, 1001 Sherbrooke St. West, Montreal, Quebec H3A 1G5, Canada. Patrick.Augustin@mcgill.ca. New York University - Leonard N. Stern School of Business, 44 West 4th St., NY New York, USA. mbrenner@stern.nyu.edu. New York University - Leonard N. Stern School of Business, 44 West 4th St., NY New York, USA. msubrahm@stern.nyu.edu.

2 1 Introduction The recent announcement of the leveraged buyout of H.J. Heinz by an investor group made up of Warren Buffett s Berkshire Hathaway Inc. and Brazilian private-equity firm 3G Capital has sparked concerns about the unusual option activity prior to the deal announcement. Was this abnormal volume in the options of Heinz an indication of trading based on insider information? Apparently, the Securities and Exchange Commission (SEC) thought so, alleging that a brokerage account in Switzerland was used for illegal insider trading. Another noteworthy case from an earlier period is the merger of Bank One with JP Morgan (JPM) Chase in July of 2004, where one investor was alleged to have bought deep out-of-the-money (DOTM) calls just (hours) before the announcement. While these cases received considerable publicity, they are by no means isolated cases of such activity. Indeed, while the SEC has taken action in several cases where the evidence was overwhelming, one can assume that there are many more cases that go undetected, or where the evidence is not as clear-cut, in a legal/regulatory sense. 1 Academic research on the role of informed trading in equity options around major news events, and in particular, the announcements of mergers and acquisitions (M&A), has been scanty. 2 We aim to fill this gap in the research presented in this paper. The objective of our study is to investigate the pervasiveness of trading on inside information in the context of M&A activity. We conduct a forensic analysis of volume and implied volatility over the 30 days preceding a formal announcement. In addition, we examine alternative strategies that may yield abnormal returns to informed traders. The focus is on options strategies, although some of these may also involve trading in the underlying stocks. We examine companies on both sides in an M&A transaction: the target and the acquirer. More specifically, we examine option trading volume (and prices) prior to M&A announcements in the U.S. from January through December We attempt to quantify the likelihood that a sudden and significant spike in put or call trading volume prior to a major informational event is based on insider trading activity, rather than being random. We also analyze the microstructure effects in the equity options market arising from the arrival of informed trading prior to the M&A announcement dates, as a consequence 1 See, for example, Options Activity Questioned Again in the Wall Street Journal, 18 February The related issue of insider trading activity prior to earnings announcements and other important corporate announcement has also received very little attention. 1

3 of leakage of information due to the trading activity of insiders or some other means. We find evidence of statistically significant average abnormal trading volume in equity options written on the target firm over the 30 days preceding M&A announcements. Approximately 25% of all cases have positive abnormal volume that is significant at the 5% level. The proportion of cases with positive abnormal volume is relatively higher for call options (25%) than for put options (15%). Stratifying the results by depth-in-moneyness, we find that there is significantly higher abnormal trading volume (both in levels and frequencies) in OTM call options compared to ATM and ITM calls. In addition, the ratio of average abnormal trading volume in OTM call options relative to OTM put options, about 2.06, is significantly higher than the ratio of average abnormal trading volume in ITM call options relative to ITM put options, which is lower at around This is preliminary evidence that insiders may not only engage in OTM call transactions, but potentially also sell ITM puts. In contrast, we find that there is no statistically significant average cumulative abnormal trading volume for equity options written on the acquirer firms over the 30 pre-announcement days, while it is positive and significantly different from zero for targets ( 15,267 contracts). In addition to evidence of abnormal trading volume in anticipation of M&A announcements, we provide statistical evidence that the two-dimensional volume-moneyness distribution significantly shifts, to options with higher strike prices, over 30 days as the announcement day approaches. Such evidence is less convincing for trading volume for equity options on the acquirer firms. We further provide statistical tests of positive (no) excess implied volatility for target (acquirer) firms in the pre-event window. Thus, informed trading has an impact on equity option prices and we show that this leads to an attenuation of the term structure of implied volatility for target, but not for acquirer firms. Similarly, the relative higher abnormal volume in OTM call options for targets translates on average into a steepening of the implied volatility spread prior to the announcement day. The evidence also suggests that private information, potentially arising from insider trading, about future takeovers leaks to the market. The percentage bid-ask spread on target equity options rises from an average of 35% to 55% over the 30 days preceding the announcement. This effect is limited to DOTM and OTM, as well as short- to medium-dated options. Similar evidence for acquirer equity options is weak. 2

4 This paper provides a preliminary forensic analysis of trading volume and implied volatility for equity options on both targets and acquirers involved in M&A announcements in the pre-event window. It naturally suggests a classification scheme based on volume and price attributes that may help regulators and prosecutors to detect illicit insider trading. Our evidence suggest that informed trading is more pervasive than what one would expect based on the modest number of prosecuted cases announced in the financial press. The outline of the paper is as follows. In Section 2, we provide a review of the relevant literature. We describe the data selection process and review the basic summary statistics in Section 3. The main hypotheses and methodology are presented in Section 5. We analyze the results in the subsections of Section 6. We end with a summary and preliminary conclusions in Section 7. 2 Literature Review Our work relates more generally to the theoretical literature emphasizing the options market as a preferred venue for informed traders, such as for example Easley et al. (1998). 3 While the bulk of the empirical research on options markets focuses on index options, there are fewer studies using equity options (i.e., options on individual stocks), although they were trading for almost a decade prior to the introduction of index options in the U.S. 4 There are even fewer studies relating to informed trading around major informational events, such as mergers and acquisitions, using options strategies, and these are typically based on relatively small data sets. Cao et al. (2005), for example, focus on the target firms in M&A transactions and study 78 U.S. merger or takeover firms between 1986 and They find evidence that the options market displaces the stock market for information-based trading during periods immediately preceding takeover announcements. More specifically, buyer-seller initiated call-volume imbalances, but not stock imbalances, are associated with higher next-day stock returns. However, during periods of normal trading activity, only buyer-seller initiated stock-volume imbalances exhibit predictability, while option volume is uninformative. Option volume imbalances before M&A transactions are 3 John et al. (2003) illustrate that the choice of informed trading in the derivative market depends on frictions such as margin requirements and wealth constraints. 4 The main constraint in the earlier period was the availability of reliable data, which has changed dramatically with the advent of OptionMetrics as a reliable source for academic research in this area. 3

5 concentrated in firms that eventually have successful takeovers and cannot be explained by target firm characteristics. Chan et al. (2013), on the other hand, focus on the acquiring firms, using a sample of 5,099 events for 1,754 acquirers, over the period 1996 to The one-day pre-event implied volatility spread and the implied volatility skew, two proxies for informed option trading, are respectively positively and negatively associated with acquirer cumulative abnormal returns. 5 The predictive power of both measures increases if the liquidity of options is high relative to that of the underlying stocks. Barraclough et al. (2012) exploit the joint information set of stock and option prices to disentangle true synergies from news in merger and acquisition transaction announcements. A sample of 167 takeover announcements from April 1996 through December 2008 suggests that the increase in trading volume from the pre-announcement period to the announcement day is most dramatic for call options with an increase of 212.3% for bidder call options, and an increase of 1,619.8% for target call options. Acharya and Johnson (2010) study a sample of leveraged buyout deals from and document that the presence of more insiders lead to greater levels of insider activity, in the sense that a larger number of equity participants in the syndicate is associated with suspicious stock and options activity. Podolski et al. (2013) also provide some indirect evidence that the option-to-stock volume ratio increases in the pre-takeover period, and increases relatively more for small deals that are less likely to be detected. Finally, Nicolau (2010) studies the behavior of implied volatility around merger announcements. Several other papers are peripherally related to the specific issue studied in this paper. Spyrou et al. (2011) provide evidence of informed trading and the role of options markets in information revelation around M&A announcements from the UK equity and options market. Keown and Pinkerton (1981) studies excess stock returns earned by insider trading in the presence of merger announcements, but does not investigate equity option activity. Meulbroek (1992) studies a sample of illegal insider trading cases detected and prosecuted by the SEC from 1980 to 1989 and likewise does not focus on option trading. Roll et al. (2010), on the other hand, study the relationship 5 The implied volatility spread is calculated as the average difference between implied volatilities of same security/strike/maturity calls and puts). The implied volatility skew is calculated as the difference between implied volatilities of OTM puts and ATM calls. 4

6 between the option-to-stock trading volume and post-earnings announcement returns. Bester et al. (2011) and Subramanian (2004) develop theoretical option pricing models for the target in the case of cash and stock-for-stock mergers respectively. Dubinsky and Johannes (2006) develop estimators of the anticipated uncertainty of fundamental information released on earnings announcement dates using equity option prices. Easley et al. (1998), Pan and Poteshman (2006), Xing et al. (2010), Cremers and Weinbaum (2010), Johnson and So (2011), Driessen et al. (2012), Jin et al. (2012) and Hu (2013) more generally relate various information-based measures derived from option trading volume and prices to the predictability of stock returns. Finally, while Bollen and Whaley (2004) and Garleanu et al. (2009) show how price pressure affects option prices, Cao and Ou-Yang (2009) theoretically studies how differences in beliefs generate trades in options relative to stocks. While the above studies tend to focus on either the target or the acquirer, we study the trading patterns of equity options of both the target and acquirer, using data on both trading volume and prices. More specifically, we focus on the behavior of the entire volume distribution and optionimplied volatility across the depth-in-the-money dimension, prior to takeover announcements. Importantly, while some papers in the previous literature have investigated the informational content of option trading volumes for post-announcement stock returns, none of them has focused on the role of options strategies in illegal insider trading. Moreover, in contrast to the above studies, which focus on various aspects of the M&A announcements using options data, our study focuses on the extent to which informed trading, some of it illegal, can be detected by analyzing various option strategies, using both puts and calls, of the target company and the acquirer. The likelihood of informed trading in these cases will be quantified in our analysis. Our study is also more comprehensive in scope than the above prior studies, based on a much larger sample, with more rigorous statistical tests. Through this contribution, we hope to provide tangible policy guidance regarding the detection of this type of illegal activity, which would aid regulatory action. 3 Data Selection and Summary Statistics The data for our study are obtained from three primary sources, Thomson Reuters SDC Platinum Database, Center for Research in Securities Prices (CRSP) and OptionMetrics. The start date of our 5

7 sample period is dictated by the availability of option information in OptionMetrics, which initiated its reporting on 1 January Our final sample consists of 1,859 corporate transactions, for which we could identify matching stock and option information for the target. These deals are undertaken by 1,279 unique acquirers on 1,669 unique targets. 6 For a sub-sample of 792 transactions, option information is available for both the target and the acquirer. The starting point of our sample selection is the full domestic Mergers and Acquisitions sample for US targets from the Securities Data Company Platinum database over the time period January 1996 through December We restrict the sample to deals with the intent of effecting change of control, i.e., to be included in our sample, the acquirer needs to own less than 50% of the target s stock before the transaction, and is seeking to own more than 50% post transaction. Hence, we include only mergers, acquisitions and acquisitions of majority interest in our sample, thereby excluding all deals that are acquisitions of partial interest/minority stake purchases, acquisitions of remaining interest, acquisitions of assets, acquisitions of certain assets, recapitalizations, buybacks/repurchases/self-tender offers and exchange offers. In addition, we exclude deals for which the status is pending or unknown, i.e., we only include completed, tentative or withdrawn deals. Next, we require available information on the deal value and eliminate all deals with a transaction value below 1 million USD. Finally, we match the information from SDC Platinum with price and volume information for the target in both CRSP and OptionMetrics. We require a minimum of 90 days of valid stock and option price and volume information on the target prior to and including the announcement date. We further retain short-dated options expiring before the announcement date, as long as they are at-the-money. All matches between SDC and CRSP/OptionMetrics are manually checked for consistency based on the company name. 7 Panel A in Table 1 reports the basic characteristics for the full sample, in which we require option information availability only for the target. Pure cash offers make up 48.6% of the sample, followed by hybrid financing offers with 22.3% and share offers with 21.7%. 82.9% of all transactions are completed, and mergers are mostly cross-industry with 89.2% of all deals being undertaken with 6 Thus, 190 of the targets were involved in an unsuccessful merger or acquisition that was ultimately withdrawn. However, we include these cases in our sample, since the withdrawal occurred after the takeover announcement. 7 Overall, we have up to a maximum of one year of stock and option price information before and after the announcement date. The cut-off of one year is arbitrary, but follows from the trade-off of the following two objectives: having of a sufficiently long time series before the announcement day to conduct the event study analysis, and keeping the size of the dataset manageable to minimize computational complexity. 6

8 a company not in the same industry based on the 4-digit SIC code. 90.2% of all deals are considered to be friendly and only 3.4% are hostile, while 11.6% of all transactions are challenged. For a small sub-sample of 6.5%, contracts contain a collar structure, 76.5% of all deals contain a termination fee, and in only 3.5% of the transactions did the bidder already have a toehold in the target company. Panel B shows that the average deal size is 3.8 billion USD, with cash-only deals being, on average, small (2.2 billion USD), in comparison to stock-only transactions (5.4 billion USD). The average one-day offer premium, defined as the offer price to the target s closing stock price 1 day before the announcement date, is 31%. Statistics for the sub-sample for which we have option information on both the target and acquirer are qualitatively similar. In Figure 1, we plot the average option trading volume in calls and puts for both the target and acquirer, 60 days before, and after the announcement date. The run-up in volumes is a first indication of information leakage prior to the public news announcements. However, these simple averages mask significant cross-sectional differences across firms and options in abnormal trading volumes. A more detailed analysis is provided in Section 6, the empirical section that follows the discussion of our hypotheses. 4 Research Questions and Hypotheses We attempt to quantify the likelihood of insider trading by investigating various trading strategies that, in the presence of private information, should result in abnormal trading volumes and returns in equity options in anticipation of public M&A announcements. 8 We investigate several hypotheses to test for such activity. 9 The underlying assumption for all these hypotheses is that insiders are capital constrained. 8 In this version, we do not present results for hypothesis H3. Additional evidence supporting the other hypotheses as well as robustness checks will be presented in the next version of the paper. 9 We write these hypotheses as statements of what we expect to find in the data, rather than as null hypotheses that would be rejected. 7

9 5 Research Questions and Hypotheses We attempt to quantify the likelihood of insider trading by investigating various trading strategies that, in the presence of private information, should result in abnormal trading volumes and returns in equity options in anticipation of public M&A announcements. 10 We investigate several hypotheses to test for such activity. 11 The underlying assumption for all these hypotheses is that insiders are capital constrained. 5.1 Insider Activity without Leakage of Information H1: There is evidence of positive abnormal trading volume in equity options, written on the target firms, prior to M&A announcements. If informed trading is present, without leakage of information, informed traders should benefit relatively more from strategies that use options due to the leverage they obtain, if they are capital constrained. A takeover announcement tends to be generally associated with a stock price increase for the target (see Schwert 1996). An insider engaging in illegal trading is, therefore, likely to engage in directional trading strategies to maximize his profits on his private information, given his capital constraints. In that case, we would expect to see significant positive abnormal trading volume in options for the target firms, in anticipation of major corporate takeover announcements. H2: There is a higher ratio of the abnormal trading volume in (a) OTM call options compared to ATM and ITM call options, and (b) ITM put options compared to ATM and OTM put options, written on the target firms, prior to M&A announcements. In the presence of informed trading, an option strategy involving the purchase of OTM call options should generate significantly higher abnormal returns, as a consequence of the higher leverage. Hence, we expect a relatively larger increase in abnormal trading volume for OTM calls relative to ATM and ITM calls, in the presence of such activity. 12 Moreover, an in- 10 In this version, we do not present results for hypothesis H3. Additional evidence supporting the other hypotheses as well as robustness checks will be presented in the next version of the paper. 11 We write these hypotheses as statements of what we expect to find in the data, rather than as null hypotheses that would be rejected. 12 This case corresponds to the case study of JPM-Chase merging with Bank One, which exhibits such a pattern. 8

10 sider taking advantage of his privileged knowledge of the direction of the target s stock price evolution is also likely to increase the trading volume through the sale of ITM puts. H3: In anticipation of major news events, there should be a volume increase in long-vega trading strategies for the acquirer firm, prior to M&A announcements. Unlike the target firm, where the stock price almost always goes up after the announcement, there is generally more uncertainty associated with the post-announcement direction of the stock price of the acquiring firm. For the acquirer, we therefore anticipate an increase in trading volume of option pairs which have high vega exposure, such as straddle strategies for example. 5.2 Insider Activity with Leakage of Information H4: There is positive (no) excess implied volatility for equity options, written on the target (acquiring) firms, prior to M&A announcements. Information about anticipated takeover targets may leak to financial markets and give rise to informed trading. Jarrell and Poulsen (1989) argue that one such possible source of legitimate information available to market participants is 13D filings when investors acquire more than 5% of the target firms stock. Thus, anticipating a price increase for target but not for acquiring companies, we would expect to observe positive excess implied volatility on average for takeover targets, and weak or no excess implied volatility for acquirers. H5: : The percentage bid-ask spread for options written on target firms should widen, prior to M&A announcements. Similar to H4, there should be no pattern in the bid-ask spread in the options on the target firms as the announcement date approaches, absent insider activity with leakage of information. An increase in the percentage bid-ask spread conditional on positive abnormal trading volume would be evidence that information about a potential merger has leaked to the market makers. H6: The convexity of the option smile, for target firms, should increase for call options and decrease for put options, prior to M&A announcements. 9

11 According to H2, we expect abnormal trading volume in OTM call options and ITM put options. Higher demand for OTM call options and ITM put options from the insider could increase the relative expensiveness of OTM Call options (ITM puts) and hence, increase the convexity of the option smile., in the presence of insider trading when the trading causes a leakage of information to the market. Both real insider trading and higher trading activity due to leaked information are associated with positive abnormal trading volume. Conditional on positive abnormal trading volume, leakage should be identified through an increase in both volume and prices, i.e., the convexity of the option smile should increase.. However, if we identify abnormal trading volume without a statistically significant shift in the slope of the call option smile, then the increasing trading activity is more likely based on illegal insider information, which has not yet leaked to the market. H7: The term structure of implied volatility should decrease for options on the target firms before takeover announcements, and remain flat for options on the acquiring firms. Informed traders obtain the highest leverage by investing into short-dated OTM call options. Demand pressure on short-dated options should lead to a relative price increase in options with a short time to expiration compared to long-dated options. Thus, the term structure of implied volatility should decrease for call options written on target firms and remain flat for call options written on acquirer firms. 6 Analysis We investigate the above hypotheses along two dimensions: volume and price. We begin by looking into the behavior of volume prior to the M&A announcement dates. 6.1 Abnormal Volume In order to address hypotheses H1 to H2, we conduct a forensic analysis of trading volume in equity options during the 30 days preceding takeover announcements. We first summarize the trading volume in our sample. We next look at specific trades that are most susceptible of insider trading, the strongly unusual trading sample, and compare it to a matched random sample. We 10

12 further test for the presence of positive abnormal volume in call and put options across moneyness categories, using the event study methodology. Next, we formally test, using an approximation to the bivariate Kolmogorov-Smirnov test, whether the entire volume-moneyness distribution shifts in anticipation of takeover news releases. Finally we show some evidence on the behavior of conditional trading volume Statistics on Equity Option Trading Volume We start by reporting basic summary statistics on option trading volume, stratified by time to expiration and depth-in-money in Table We classify three groups of time to expiration: less or equal than 30 days, bigger than 30 but less or equal than 60 days and more than 60 days. In addition, we assign five groups for depth-in-moneyness, where depth-in-moneyness is defined as S/K, the ratio of the stock price S over the strike price K. Deep out-of-the-money (DOTM) corresponds to S/K [0, 0.80] for calls ([1.20, ] for puts), out-of-the-money (OTM) corresponds to S/K ]0.80, 0.95] for calls ([1.05, 1.20[ for puts), at-the-money (ATM) corresponds to S/K ]0.95, 1.05[ for calls (]0.95, 1.05[ for puts), in-the-money (ITM) corresponds to S/K [1.05, 1.20[ for calls (]0.80, 0.95] for puts), and deep in-the-money (DITM) corresponds to S/K [1.20, [ for calls ([0, 0.80] for puts). Panels A to C report statistics on all options in the sample, while Panels D to F and G to I report the numbers separately for calls and puts. A first observation is that, regardless of depth-in-money, the level of trading volume, as indicated by the mean volume statistics, is significantly higher for short and medium-dated options compared to long-dated options. For example, the average number of traded contracts in OTM options, for the target (acquirer), is 370 and 285 (497 and 384) contracts for maturities of respectively less than 30 and 60 days, while it is 130 (193) contracts for options with more than 60 days to maturity. This difference is more pronounced for call options than for put options. Second, the average trading volume is higher for options on the acquirer firms (547 contracts) than on targets (283 contracts). Third, the distribution of volume as a function of depth-in-moneyness exhibits a humpshaped pattern for acquirers, irrespective of whether the options are short- or long dated. Hence, trading volume tends to be highest ATM and decreases as S/K moves further into or OTM. In the 13 Data on equity options volume have the particular feature of multiple zero-volume observations. These data points are omitted for the purpose of the basic summary statistics. 11

13 entire universe, for instance, the average volume is 1,084 contracts ATM, 497 and 398 contracts respectively OTM and ITM, and 127 and 214 contracts respectively for DOTM and DITM options. On the other hand, for targets, the highest average trading volume tends to be associated with OTM options. Finally, we observe that the standard deviation of trading volume decreases with time to expiration Strongly Unusual Trading Volume and Matched Random Sample Our goal is to distinguish insider trading from random speculative bets. We are looking for unusual trading patterns that are considerably different from patterns exhibited by randomly selected samples. Evidence of non-random trading would point to the existence of illicit option trades, i.e., trades based on insider information. Thus, we start with cases that potentially are the most likely ones to reflect insider trading. We define as strongly unusual trading (SUT) observations corresponding to the following four criteria for individual options: 14 (1) the daily best recorded bid is zero. This corresponds normally to DOTM options where the market maker, through his zero bid, signals his unwillingness to buy but is willing to sell at a non-zero ask price. (2) The option should expire on or after the announcement day, but is the first one to expire thereafter (the so called front month option). Obviously, an insider would buy options that expire after the announcement. In order to get the biggest bang for the buck, he would try to buy the cheapest ones that are most likely to end up in the money. Short-dated OTM options tend to be cheaper and provide the greatest leverage. (3) The option must have strictly positive trading volume. As many individual equity options, especially OTM, have zero trading volume (all options have quotes in a market making system) we focus on those that have positive volume assuming that the insiders pick those OTM options that are the most likely to be exercised. (4) Finally, the transaction must take place within 30 days of the event-date, defined as the 0 date (i.e., from event date -29 to 0). An informed trader faces the tradeoff of leveraging on his private information prior to the event, while likely not trading too close to the event, which may entail a higher risk of getting caught An observation corresponds to an option-day pair, i.e., the end-of-day volume for a given option on the target or acquirer. 15 An additional aspect that we do not explicitly consider is the number of insiders involved and their connections with each other. Such information may reveal whether the information was shared by many players and potentially leaked to them. 12

14 Table 3 presents the sample statistics for the SUT sample. From the entire dataset, we identify 2,042 (1,847) option-day observations for the target (acquirer), passing the SUT selection criteria. 16 The share of calls is slightly more than half, with 1,106 (954) observations for the target (acquirer). The average trading volume is 124 (110) contracts for targets (acquirers) and the average trading volume for calls and puts is respectively 137 and 108 (126 and 94) for the target (acquirer). Median trading volume is somewhat more stable, with a value of 20 contracts for options written on the target, and around 15 contracts for options written on the acquirer. We observe that, in the sample most susceptible of insider trading, the volume of the target firm, at each percentile of the distribution, is significantly above that of the acquirer. The basic summary statistics in Table 2 showed exactly the opposite. We compare the statistics from the SUT sample to those from a randomly selected sample. The sampling procedure is as follows: For each of the 1,859 (792) events with traded options on the target (acquirer), we randomly select a pseudo-event date. We treat the pseudo-event date as a hypothetical announcement date, and then apply the SUT selection criteria, i.e., we keep optionday observations with a zero bid price, non-zero trading volume, that are within 30 days of the pseudo-event date and which have an expiry date after the pseudo-event date. The SUT sample statistics are compared to the Random Sample Trading (RST) statistics in Panel B of Table The number of observations, deals and options are somewhat higher in the RST sample than in the SUT sample. The ratios of number of observations, deals and options in the RST sample to those in the SUT sample range approximately between 1.4 and 1.8. However, the average and median trading volume in the SUT sample is more than double that of the RST sample. The maximum observed trading volumes are significantly higher in the SUT sample than in the RST sample. However, the distributional statistics illustrate that this effect arises not because of outliers. In the RST sample, around the 50th percentile of the distribution and upwards, volumes are consistently less than half the trading volumes observed in the SUT sample at comparable cutoffs of the volume distribution. Another interesting feature is that the distance between the median 16 Note that the full sample has approximately 12,000,000 observations. For each event, the event time spans from roughly 1 year before to 1 year after the announcement date. 17 As we have confined our study to a limited period and due to the fact that the variance may be large, we have double checked our results using 100 random samples of 1,859 (792) pseudo-events for the target (acquirer) in order to minimize the standard error of our estimates. The results from this robustness check were very similar. 13

15 and the mean is roughly constant around 100 traded contracts in the SUT sample. Finally, while a similar pattern emerges for options written on the acquirer, the effect is overall weaker. In particular statistics for put options are statistically similar across both samples. The significant difference in average and median trading volumes between the SUT and RST samples for target firms is a first indication of a shift in the volume distribution in anticipation of M&A announcements. We point out that the difference between the two samples is likely to be conservative. For each event, we have a maximum of 1 year of data before and after the event, rather than the whole universe of traded options. Under normal circumstances, we should be using for each event all observations going as far back as January 1996 until today. In such a situation, the difference is likely to be even stronger. To summarize, the entire distribution in trading volume differs significantly between the SUT and RST samples for target firms, but not for acquirer firms. In particular, we observe that an average trading volume above 100 contracts, with the mean to median distance of 100 contracts, can be considered to be strongly unusual and non-random, when the transactions occurs at a zero-bid within 30 days of the announcement date on options expiring after the announcement. This first pass test is preliminary evidence in favor of the hypothesis H1 by showing that there is a non-random increase in trading volume on targets, but on acquirers, prior to a public M&A announcement Shifts in Option Trading Volume Density The previous section has illustrated that the 30 days prior to M&A announcement dates exhibit strongly unusual trading patterns, in particular for targets The question is whether there is a monotonic shift in the option trading volume distribution as the announcement date approaches. We formally test for a shift of the bivariate volume-moneyness distribution across time in anticipation of the announcement dates. Figure 2 visually illustrates the shift in volume distribution for calls and puts written on the target as we approach the announcement date. Each individual line reflects a local polynomial function fitted to the volume-moneyness pairs. It is striking to see how the volume distribution, for target call options, shifts to the tails and puts more weight on DITM and DOTM categories as 14

16 we approach the announcement date. In addition the volume level keeps increasing, in particular in the event window [ 4, 1]. The last event window [0, 0] incorporates the announcement effect, whereby the overall average trading level is lifted upwards, and the distribution shifts to ITM call options and OTM puts. The lower two panels illustrate the same graphs for the acquirer firms. While it is possible to see a small shift in the volume distribution across time, the magnitudes of the shifts are small. It is mainly the announcement effect that is clearly distinguishable for acquirers. Another way to visualize the change in the distribution is given in Figure 3, although this graph doesn t reflect the bivariate dimension of the distribution. The dashed blue line and the solid green line in each plot represent the 90th and 95th percentile of the distribution, whereas the dotted red lines reflect the interquartile range. The percentage increase in the percentiles of the volume distribution is clearly much more pronounced for the targets than for the acquirers. For example, the interquartile range for target call options increases from a level below 50 contracts to approximately 2000 contracts on the announcement day. For acquirer call options, on the other hand, the interquartile range increase from approximately 200 to about 1,200 contacts on the announcement day. To summarize, there is a significant shift in both the mean and median trading volume in anticipation of U.S. M&A transactions for target firms. This shift is more pronounced for DOTM and OTM call options, compared to ITM and DITM options. On the other hand, the increase in the level of trading volume for acquirers is more moderate and doesn t seem to be much different for ITM options relative to OTM calls and puts. This confirms our hypothesis H2 that there is a higher abnormal trading volume in deep OTM call options, compared to ATM and ITM call options, but that abnormal trading volume, if any, should be equal across moneyness categories for acquirers. In what follows, we apply a formal statistical test for the shift in volume distribution. In order to test whether the bivariate volume-moneyness distribution shifts across time prior to announcement dates, we use a two-sample bivariate Kolmogorov-Smirnov test. The two-sample Kolmogorov-Smirnov (KS) test is a non-parametric test for the equality of two continuous distribution functions. Essentially, the KS statistic quantifies the distance between the two empirical cumulative distribution functions. While the test statistic is straightforward to compute in the univariate setting with distribution-free properties, the computation in the multivariate setting can 15

17 become burdensome, in particular, when the sample size is large. The reason for this is because in the univariate setting, the empirical cumulative distribution function diverges only at its observed points, while it diverges at an infinite number of points in the multivariate setting. To see this, remember that in a multivariate setting, there is more than one definition of a cumulative distribution function. In particular, in the bivariate setting, the four regions of interest are H (1) (x, y) = P [X x, Y y], H (1) (x, y) = P [X x, Y y] (1) H (1) (x, y) = P [X x, Y y], H (1) (x, y) = P [X x, Y y], (2) and we need to evaluate the empirical cumulative distribution function in all possible regions. To reduce computational complexity, we rely on the Fasano and Franceschini generalization of the ( ) ( ) two-sample bivariate KS test. Define the two sample sizes { x 1 j, y1 j : 1 j n} and { x 2 j, y2 j : 1 j m}, with their corresponding empirical cumulative distribution functions H (k) n and H (k) m regions k = 1, 2, 3, 4. The Fasano-Franceschini (Fasano and Franceschini (1987)) FF-test statistic is then defined as for Z n,m = max{t (1) n,m, T (2) n,m, T (3) n,m, T (4) n,m}, (3) where T (k) nm n,m = sup (x,y) R 2 n + m H n (k) (x, y) H m (k) (x, y). (4) Although the analytic distribution of the test statistic is unknown, its p-values can be estimated using an approximation, based on Press et al. (1992), to Fasano and Franceschini s (FF) Monte Carlo simulations. Our prior is that the FF-statistic, which reflects the distance between the two bivariate empirical distribution functions (EDF), should monotonically increase for target firms as we get closer to the announcement date. 18 In contrast, there should be no evidence for such an increase, or at worst 18 One can think of the FF-statistic as a KS-statistic in the multivariate setting. The FF-statistic is computationally less intensive in the multivariate case, but is consistent and does not compromise power in large sample sizes. See Simon L. Greenberg, Bivariate Goodness-of-fit tests based on Kolmogorov-Smirnov Type Statistics. 16

18 weaker evidence, for acquirer firms. Essentially, the difference in EDFs should be larger between event windows [ 29, 25] and [ 24, 20], than between [ 29, 25] [ 19, 15]), etc. In addition, the FF-statistics should increase relatively more for short-dated options, which are closer to the announcement date. These predictions are clearly confirmed by the results in Tables 4 and 5. The FF test reveals statistically significant differences in the bivariate volume-moneyness distributions, as we move closer to the announcement date. We compare the distributions in event window blocks of 5 days. A glance at the table reveals that the test is statistically significant, at the 1% level, for almost all comparisons. In addition, the magnitude of the statistic is monotonically increasing as we move from the left to the right, and as we move from the bottom to the top. Panel A and B in Table 4 report results for calls and puts respectively for the target. For example, the first row shows that the bivariate distribution significantly shifts from event window [ 29, 25] to [ 24, 20], with an FF statistic of The test statistic increases to , if we compare event windows [ 29, 25] and [ 4, 1], and to , for the event windows [ 29, 25] and [0, 0]. Overall, the largest test statistics seem to be associated with comparisons between the announcement date ([0, 0]) and the event window immediately preceding it ([ 4, 1]). However, for short-dated options with time to expiration less than 30 days, the statistic for the difference in distributions for the shift from event window [ 29, 25] to [ 4, 1], excluding the announcement effect, is with a value of (0.34) for call (put) options higher than the announcement effect going from the event window [ 4, 1] to the announcement date. Changes in the bivariate distributions are statistically significant at the 1% level for almost all event windows. Results for the acquirer are reported in Table 5 and confirm our prior. The shifts in the distributions, as indicated by the FF test statistics, are generally significantly smaller than for the target. For instance, comparing the period [ 29, 25] to [ 4, 1] for call options in the entire sample, the test-statistic indicates a value of for the acquirer and for the target. The average ratio across comparisons is about 2. Even the shifts on the announcement dates are less acute than for the target companies. Finally, statistical significance is much weaker for adjacent event windows, and is mostly restricted to short-dated options just prior to the announcement date. 17

19 6.1.4 Abnormal Trading Volume - Event Study Hypothesis H1 asserts that there is positive abnormal trading volume in call equity options written on the target prior to a public M&A announcement. We test this by running a classical event study. For each of the 1,859 deals in the sample, we source the aggregated total option volume on the target s stock, as well as the total aggregated volume traded in calls and puts. Option information for the acquirer is limited to a subsample of 792 deals. To compute abnormal trading volume, we use as a benchmark a constant-mean-trading-volume model, as well as two different volume-based versions of the market models. We define the market trading volume as the median (mean) total call and put trading volume across all options in the OptionMetrics database. As we are interested in the abnormal trading volume in anticipation of the event, we use as the estimation window the period starting 90 days before the announcement date until 30 days before the event. The event window in our case stretches from 30 days before until one day prior to the event. To account for the possibility of clustered event dates, we correct all standard errors for cross-sectional dependence. The results are reported in Table 6. The average cumulative abnormal trading volume is positive and statistically significant for the target across all model specifications. The magnitude of the average cumulative abnormal volume over the 30 pre-event days is also economically meaningful with an estimate of 11,969 contracts for call options using the median market model. For put options on the target, average cumulative abnormal volume is also positive and highly statistically significant, but the average cumulative abnormal volume over the 30 pre-event days at 3,471 contracts is economically much smaller. The evolution of average abnormal and cumulative abnormal trading volume for targets is illustrated in the upper two panels in Figure 4. It is apparent that the average cumulative abnormal trading volume in put options is quantitatively less important than the average cumulative abnormal trading volume in call options, which is primarily driving the results for the overall sample. The daily average abnormal volume for call options is positive and steadily increasing to a level of approximately 1,500 contracts the day before the announcement. Individually, the number of deals with positive abnormal trading volume at the 5% significance level ranges from 472 to 492 for calls, and from 271 to 319 for puts, corresponding to respectively 18

20 25% and 15% of the entire sample. 19 These results confirm the null hypothesis H1, that there is positive abnormal trading volume in call and put equity options written on the target prior to a public M&A announcement. Focusing on the acquirer in Table 6, the average cumulative abnormal trading volume over the 30 pre-event days is statistically indistinguishable from zero and fluctuates between positive and negative values depending on the model specification. Also, the number of deals for which there is positive and statistically significant abnormal volume on the acquirer is lower than for the targets. Depending on the model specification, there about 54 to 56 such cases for call options and 45 to 51 for put options. This corresponds to about 7% and 6% of all deals for calls and puts respectively. The bottom two panels in Figure 4 illustrate the evolution of the average abnormal and cumulative abnormal trading volume for acquirers. In line with our prior, there is no clear evidence of positive abnormal trading volume for acquirers, whether for calls or puts. Daily average abnormal volume resembles more a random walk rather than any clear pattern. In addition to the aggregated results, we stratify our sample by moneyness and conduct an event study for each category. We find that there is significantly higher abnormal trading volume for targets in OTM and ITM call options, compared to ATM and ITM calls, both in terms of volume levels and frequencies. Using the median market model, for instance, the average cumulative abnormal volume is 3,797 (1,860) contracts for OTM calls (puts) and 1,702 (1,110) contracts for ITM calls (puts), while it is 1,059 (188) for ATM calls (puts). These values correspond to 383 (300, 448) deals or 21% (16%, 24%) for OTM (ATM, ITM) calls and 387 (254, 316) deals or 21% (14%, 17%), for OTM (ATM, ITM) puts, respectively. In addition, while we find that the average cumulative abnormal volume is positive and statistically significant for both OTM and ITM calls and puts, it is only statistically significant at the 5% level for ATM call options, but not for put options. Panel B reports the results from paired t-tests for differences in means of the cumulative average abnormal volumes across different depths. Consistent with our hypothesis H2, these results emphasize that there is a higher ratio of abnormal trading volumes for OTM call options compared to ATM and ITM call options. The difference in means, using the median market model, for 19 Unreported results indicate that, at the 1% significance level, the number of deals with positive abnormal trading volume for the entire sample range from 278 to 292 for calls, and from 138 to 195 for puts, corresponding to a frequency of respectively 16% and 8%. 19

21 OTM calls relative to ATM and ITM calls is 2,738 and 2,096 respectively, which is positive and statistically different from zero. On the other hand, the difference in means between ATM and ITM calls is slightly negative (-643), but not statistically different from zero. Even though we do confirm that the ratio of the average cumulative abnormal volume for ITM put options is higher than for ATM put options, we also find that there is a higher expected abnormal volume for OTM compared to ITM put options, even though the difference of 750 contracts is small in magnitude, given that it is a cumulative measure over 30 days. This is some preliminary evidence that insiders may not only engage in OTM call transactions, but potentially also sell ITM puts. The corresponding statistics for the acquirer stratified by moneyness do not show any clear pattern. To summarize, the event study further supports hypotheses H1 and H2. In other words, there is ample evidence of positive abnormal volume in equity options for the targets, but not for the acquirers in M&A transactions, prior to the announcement date. In addition, we document that, for the targets, there is a significantly larger amount of abnormal trading volume in OTM call options compared to ATM and ITM call options. However, the evidence that insiders may also engage in writing ITM put options is weak. The results of the event study are supported by formal statistical evidence that the two-dimensional volume-moneyness distribution significantly shifts, over both time and moneyness, over the 30 days preceding the announcement day. Hence, the level of the volume distribution increases with a higher frequency of trades occurring in both OTM and ITM options. Moreover, this shift in the bivariate distribution is not random, as we illustrate by comparing the volume distribution from a suspicious trading sample to that of a randomly matched sample Non-parametric Volume at Risk. To supplement our forensic analysis on the behavior of options volume before takeover announcements, we analyze trading volume patterns conditional on no trading volume for one to five earlier days. This analysis may be insightful, given the significant amount of zero volume observations that characterize the data for equity options. To do so, we calculate, non-parametrically, the probability of spikes in trading volume, conditional on zero trading volume between one and five days prior to the spike. We define as the conditionalvolume-at-risk (VolaR), at a confidence level α (0, 1), the 20

22 smallest trading volume ν, such that the probability of the volume V exceeding ν is less than or equal to (1 α). Formally, V olar is given by: i V olarα i = inf{ν R : P (V > ν) 1 α, V t j = 0}. (5) The statistics in Panel A of Table 9 show the VolaR statistics for the 30 days preceding the announcement dates. We see that for target companies, if there is a day with zero option trading volume on a given company, the probability that the volume exceeds 100 (200; 1,000) contracts j=1 during the next day is less than 10% (5%, 1%). When we look at the options written on the acquiring firm, the numbers look very similar. The probability of observing more than 96 (191; 1,000) traded contracts on a company when there was no option trading volume on its stock during the preceding day, is less than 10% (5%, 1%). Naturally, the V olar i α levels decrease for the more restrictive cases with several preceding days with zero trading volume (i > 1). However, the values are close to the benchmark case (i = 1) of zero trading volume for one preceding day only. In the case with five earlier days with no recorded option trading, the probabilities of exceeding respectively more than 65, 126 or 710 traded contracts are less than 10%, 5% and 1%. Statistics for the acquirer sample are again very similar, with V olarα 5 values of 70, 143 and 700 contracts at the confidence levels 90, 95 and 99. We compare these critical levels of trading volumes to the sample statistics from the SUT selection, which correspond to the same 30 days before the announcement dates. We have argued that the unconditional probability of exceeding more than 100 traded contracts on any given company, when there was no trading volume on the preceding day, is less than 10%. This volume of 100 contracts seems small in comparison to the average trading volume in call options, written on the target, from the SUT sample. The average trading volume of option trades characterized as strongly unusual, which occurred during the 30 days preceding a M&A announcement, is 124 contracts, with the volume for call options being slightly higher (137 contracts). Thus, the VolaR statistics further emphasize the strongly unusual trading patterns in anticipation of these publicly unexpected events. In addition, we compare the VolaR statistics to those of a randomly matched sample. Similarly 21

23 to the earlier exercise, we randomly choose 1,859 (792) pseudo-event dates for the target (acquirer) companies, and then investigate the VolaR statistics over the 30 days preceding the pseudo-event dates. Given that trading picks up during the pre-announcement period, one would expect to see fewer days with zero trading volume for individual equity options, and as a consequence, the VolaR statistics should be lower in sample. Panel B in Figure 9 compares our prior. The VolaR levels are higher in the randomly matched sample. For instance, the 1% VolaR level for target (acquirer) companies is 1,525 (1,492) option contracts in the random sample compared to 1,000 (986) in the real sample. The forensic analysis of the trading volume behavior in equity options confirms our priors stated in hypotheses H1 and H2. The next step is to investigate hypotheses H4 to H7 by focusing on the information embedded in equity options prices, based on their implied volatilities. 6.2 Implied Volatiliy The price behavior of options is embedded in the summary statistic of implied volatility. As a complement to the volume results, we conduct a forensic analysis on implied volatility over the 30 days preceding the announcement date. We first conduct an event study to test for the presence of positive excess implied volatility relative to a market benchmark. Second, we study the behavior of the convexity of the option smile, in anticipation of news releases, followed by investigations of the bid-ask spread. Finally, we address the hypothesis related to the term structure of implied volatility Excess Implied Volatility - Event Study As informed traders benefit from their private information only to the extent that the outcome of the announcement on stock prices is certain, we expect to find positive excess implied volatility for equity options on target, but not acquirer companies. We use the interpolated volatility surface in OptionMetrics for this exercise. To analyze the behavior of ATM implied volatility, we use the 50 delta (or 0.50) options, and we reference the 80 and 20 delta options for ITM and OTM options respectively. We test two different model specifications for our results: a simple constant mean volatility model and a market model, where we use the S&P 500 VIX index as the market s 22

24 benchmark for implied volatility. The estimation window runs from 90 to 31 days, before to the announcement date, while our event window relates to the 30 days before the event, excluding the announcement itself. All standard errors are clustered by time to account for the clustering of events. Panel A in Table 7 documents that excess implied volatility is quite pervasive in our sample. At the 5% significance level, using the market model, there are about 812 cases (44% of the 1,859 deals) with positive excess implied volatility for call options and about 798 cases (43% of the 1,859 deals) with positive excess implied volatility for put options. The frequencies are similar for OTM implied volatilities and slightly lower for ITM implied volatilities, where positive excess implied volatility is documented for 39% (calls) and 41% (puts) of all cases. In addition, cumulative excess implied volatilities are consistently positive and statistically significant for the target companies. The average cumulative excess implied volatility over the 30 pre-announcement days is approximately 50% for calls and 40% for puts. In contrast to the results for targets, average cumulative excess implied volatilities fluctuate around zero and are statistically indistinguishable from zero. However, on a case by case basis, 39% (33%, 38%) of the deals exhibit positive and statistically significant positive excess implied volatility for ATM (ITM, OTM) call options, while 38% (36%, 32%) of the deals exhibit positive and statistically significant positive excess implied volatility for ATM (ITM, OTM) put options. To summarize, the event study confirms our hypothesis H4, which asserts that there should be on average positive cumulative excess implied volatility for the target companies, but not for the acquirer firms. These results are graphically presented in Figure 5 for ATM implied volatilities. The daily average excess ATM implied volatility starts increasing about 18 days before the announcement date and then steadily rises. In contrast, for acquirers, the daily average ATM excess implied volatility fluctuates around zero and behaves much more like a random walk Information Dispersion and Bid-Ask Spreads To address hypothesis H5, we study the evolution of the bid-ask spread in anticipation of the M&A announcement. The prediction of the null hypothesis H5 is that the percentage bid-ask spread in option premia should widen prior to the announcement. Strong evidence in favor of this hypothesis 23

25 would indicate that the information about a potential merger has leaked to the market makers and that unusual trading activity is more likely to be associated with illicit trading. Figure 6 plots the evolution of the average percentage bid-ask spread from 90 days before the announcement date to 90 days after the event. The average percentage bid-ask spread rises from about 35% to 55% and then jumps up to approximately 80% because of the announcement effect. Interestingly, this rise in bid-ask spreads is restricted to DOTM and OTM options. Figure 6c shows that the evolution of the percentage bid-ask spread before the announcement for ITM and DITM options is flat and jumps only at the announcement, while there is a strong increase for both DOTM and OTM options. If we separate the sample into different buckets of time-to-expiration, it is also easy to see in Figure 6e, that the widening of bid-ask spreads is restricted to options with a maturity of less than 60 days. The right hand side panels in Figures 6d, 6d and?? plot the equivalent graphs for the subsample for which we have information on options on the acquirer. While the level of the average percentage bid-ask spread is also increasing as we approach the announcement date, the increase is economically less significant than for the options on the target, as it rises from approximately 24% 90 days before the announcement date to 39%, and then stabilizes at event time zero. The stratified samples document that there is hardly any change in the cross-sectional percentage bid-ask spreads for different time to expirations, and the rise can almost entirely be explained by DOTM options. Overall, the results are suggestive of the fact the information leakage, if present, is much more centered on the target companies and less on the acquirers. To summarize, we find evidence that the percentage bid-ask spread in targets option premia increases prior to the announcement, which is evidence that information about the future transaction leaked to the market. The effect is restricted to the DOTM and OTM options with expirations of less than 60 days Volatility Smile and the Term Structure of Implied Volatility Hypotheses H6 predicts that the convexity of the option smile, for target firms, should increase for call options and decrease for put options, prior to M&A announcements. We investigate this question by plotting in Tables 7 and 8 various measures relating to the convexity of the option 24

26 smile. 7a and 8a illustrate two documented measures of the implied volatility skewness. The first measure, on the left scale, is measured as the difference between OTM call and put implied volatility, standardized by ATM implied volatility. The second measure, on the right axis, is measured as the difference between OTM implied put and ATM implied call volatility. To our surprise, both measures seem to remain flat prior to the announcement date, whether for target or for acquiring companies. We therefore look at the implied volatility spread, which is supposed to capture informed trading in the options market (see Cremers and Weinbaum (2010)). Prior literature shows that the implied volatility spread between call and put options is a bullish signal for future returns on the underlying stocks. A common interpretation is that a high call-put implied volatility spread indicates favorable private information revealed by informed option investors. We compute the volatility spread as the difference between ATM implied call and put volatility. This measure, reported on the left axis in Figure 7c, clearly suggests that call options become relatively more expensive than put options before the news announcement, which is suggestive of information leakage and informed trading. However, consistent with our prior, there is a flat evolution for the same measure applied to acquirer firms, as illustrated in Figure 8c. Finally, if abnormal trading volume is relatively more pronounced for OTM than for ITM put options, then we should also see a steeping in the difference between OTM implied call and ITM implied put volatility. This is what we observe on the right axis of the same graph. Again, there is no such evidence for the acquirer. Finally, hypothesis H7 states that the term structure of implied volatility should decrease for options on the target firms before takeover announcements, and remain flat for options on the acquiring firms. The justification for this hypothesis is that informed traders obtain the highest leverage by investing into short-dated OTM call options. Hence, demand pressure on short-dated options should lead to a relative price increase in options with a short time to expiration compared to long-dated options. Thus, a confirmation of our hypothesis would be supportive of the fact that, on average, activity in the options market before major takeover announcements is partially influenced by investors with inside information. Figure 7b documents that the average term structure of implied volatility, calculated by the difference between the implied volatility of 3-month and 1- month options, is decreasing from -1.8% by about 2.5 percentage points to approximately -4.3% over the 30 days before the announcement date. This result obtains for both call and put options. 25

27 However, the evidence is much weaker for acquirers and is restricted, if any, to call options. Figure 8b illustrates that the level of the implied volatility term structure decreases from about -0.95% to-1.45% for call options, and from roughly -1.2% to -1.45% for put options. These decreases are statistically insignificant. In a nutshell, we find supportive evidence of the fact that the average implied volatility spread significantly increase for target firms, but not for acquirer firms, prior to M&A announcements. In addition, the term structure of implied volatility becomes more negative for targets and remains roughly flat for acquirers, as we approach the announcement. 6.3 Joint evolution of abnormal volume and implied volatility We document that abnormal volume and excess implied volatility is more pervasive ahead of merger and acquisitions than what would be expected based on the number of insider trading litigations. The ultimate goal of our forensic study is to provide regulators with a clear guidance on which suspicious trades are worth pursuing. In the future, we therefore plan to double sort all cases based on their abnormal volume and excess implied volatility levels. This should in theory lead to a classification scheme to those cases that are most susceptible of insider trading. An additional filter given by the SUT selection criteria should further sharpen the monitoring results. This analysis will follow in our future research. 7 Conclusion Research on trading in individual equity options has been scanty, and even more so, centered on major informational events, such as M&As. In light of recent investigations into insider trading based on unusual abnormal trading volume in anticipation of major corporate acquisitions, we investigate the presence of informed options trading around such unexpected public announcements. We focus on equity options written on both the target and the acquirer. Our goal is to quantify the likelihood of informed trading by investigating various options trading strategies, which should a priori lead to unusual abnormal trading volume and returns in the presence of private information. A forensic analysis of trading volume and implied volatility over the 30 days preceding formal 26

28 takeover announcements suggests that that informed trading, whether based on inside information or leakage, is more pervasive than what would be expected based on the actual number of prosecuted cases. We find that there is statistically significant abnormal trading volume in call options written on the target, prior to M&A announcements, with particular pronounced effects for OTM calls. We provide formal tests of shifts in the bivariate volume-moneyness distribution and illustrate that the levels of unusual volume in options trading cannot be replicated in a randomly matched sample. We further find strong support for positive excess implied volatility for the target companies, but not for the acquirers. In addition, for targets, the term structure of implied volatility becomes more negative and the implied volatility spread widens, but there is no such effect for acquirers. Finally, the information leakage leads to wider percentage bid-ask spreads of approximately 20 percentage points over the 30 days preceding an announcement, but mainly for short-dated and OTM options. Future analysis, based on the attributes of abnormal volume and excess implied volatility, will lead to a classification that should ultimately be reflective of those cases that are most likely of insider trading. We thereby hope to provide some guidance to regulators on where to start their investigations into illicit options trading. While some insider cases are obvious, detecting others is like finding the proverbial needle in a hay stack. References Acharya, V. V. and Johnson, T. C. (2010). More insiders, more insider trading: Evidence from private-equity buyouts, Journal of Financial Economics 98(3): Barraclough, K., Robinson, D. T., Smith, T. and Whaley, R. E. (2012). Using option prices to infer overpayments and synergies in m&a transactions, Review of Financial Studies. Bester, A., Martinez, V. H. and Rosu, I. (2011). Option pricing and the probability of success of cash mergers, Working Paper. Bollen, N. P. B. and Whaley, R. E. (2004). Does net buying pressure affect the shape of implied volatility functions?, The Journal of Finance 59(2): Cao, C., Chen, Z. and Griffin, J. M. (2005). Informational content of option volume prior to takeovers, The Journal of Business 78(3): pp Cao, H. H. and Ou-Yang, H. (2009). Differences of opinion of public information and speculative trading in stocks and options, Review of Financial Studies 22(1): Chan, K., Ge, L. and Lin, T.-C. (2013). Informational content of option trading on acquirer announcement return, forthcoming in the Journal of Financial and Quantitative Analysis. 27

29 Cremers, M. and Weinbaum, D. (2010). Deviations from put-call parity and stock return predictability, The Journal of Financial and Quantitative Analysis 45(2): pp Driessen, J., Tse-Chun, L. and Xiaolong, L. (2012). Why do option prices predict stock returns?, Working Paper. Dubinsky, A. and Johannes, M. (2006). Fundamental uncertainty, earning announcements and equity options, Working Paper Columbia University. Easley, D., O Hara, M. and Srinivas, P. S. (1998). Option volume and stock prices: Evidence on where informed traders trade, The Journal of Finance 53(2): pp Fasano, G. and Franceschini, A. (1987). A multidimensional version of the kolmorogov-smirnov test., Monthly Notices of the Royal Astronomical Society 225: Garleanu, N., Pedersen, L. H. and Poteshman, A. M. (2009). Demand-based option pricing, Review of Financial Studies 22(10): Hu, J. (2013). Does option trading convey stock price information?, Journal of Financial Economics, forthcoming. Jarrell, G. A. and Poulsen, A. B. (1989). Stock trading before the announcement of tender offers: Insider trading or market anticipation?, Journal of Law, Economics, & Organization 5(2): pp Jin, W., Livnat, J. and Zhang, Y. (2012). Option prices leading equity prices: Do option traders have an information advantage?., Journal of Accounting Research 50(2): John, K., Koticha, A., Narayanan, R. and Subrahmanyam, M. G. (2003). Margin rules, informed trading in derivatives and price dynamics, Working Paper New York University, Stern School of Business. Johnson, T. L. and So, E. C. (2011). The option to stock volume ratio and future returns, forthcoming in the Journal of Financial Economics. Keown, A. J. and Pinkerton, J. M. (1981). Merger announcements and insider trading activity: An empirical investigation, The Journal of Finance 36(4): pp Meulbroek, L. K. (1992). An empirical analysis of illegal insider trading, The Journal of Finance 47(5): pp Nicolau, A. A. (2010). An examination of the behavior of implied volatility around merger announcements, Bachelor of Science Thesis, New York University, Leonard N. Stern School of Business. Pan, J. and Poteshman, A. M. (2006). The information in option volume for future stock prices, The Review of Financial Studies 19(3): pp Podolski, E. J., Truong, C. and Veeraraghavan, M. (2013). Informed options trading prior to takeovers does the regulatory environment matter?, Journal of International Financial Markets, Institutions & Money. 28

30 Roll, R., Schwartz, E. and Subrahmanyam, A. (2010). O/s: The relative trading activity in options and stock, Journal of Financial Economics 96(1): Schwert, G. (1996). Markup pricing in mergers and acquisitions, Journal of Financial Economics 41(2): Spyrou, S., Tsekrekos, A. and Siougle, G. (2011). Informed trading around merger and acquisition announcements: Evidence from the uk equity and options markets, Journal of Futures Markets 31(8): Subramanian, A. (2004). Option pricing on stocks in mergers and acquisitions, The Journal of Finance 59(2): pp Xing, Y., Zhang, X. and Zhao, R. (2010). What does the individual option volatility smirk tell us about future equity returns?., Journal of Financial & Quantitative Analysis 45(3):

31 Table 1: Descriptive and Financial overview of M&A Sample - Full Sample Panel A in this table provides an overview of the M&A deal characteristics for all US domestic mergers and acquisitions in the Thomson Reuters SDC Platinum database over the time period January 1996 through 31 December 2012, where a match was found for the target in the CRSP master file and in OptionMetrics based on the 6-digit CUSIP. The sample excludes deals with an unknown or pending deal status, includes only deals with available deal information, where the deal value is above 1 million USD and where an effective change of control was intended. In addition, we require valid price and volume information in both CRSP and OptionMetrics for the target for at least 90 days prior to and on the announcement day. Panel B illustrates the financial statistics of the deals. P1d (P1w, P4w) refers to the Premium 1 Day (1 week, 4 weeks) prior to Announcement Date in percentage terms. Explanations: The deal value is the total value of consideration paid by the Acquirer, excluding fees and expenses. The dollar value includes the amount paid for all common stock, common stock equivalents, preferred stock, debt, options, assets, warrants, and stake purchases made within six months of the announcement date of the transaction. Liabilities assumed are included in the value if they are publicly disclosed. Preferred stock is only included if it is being acquired as part of a 100% acquisition. If a portion of the consideration paid by the Acquirer is common stock, the stock is valued using the closing price on the last full trading day prior to the announcement of the terms of the stock swap. If the exchange ratio of shares offered changes, the stock is valued based on its closing price on the last full trading date prior to the date of the exchange ratio change. For public target 100% acquisitions, the number of shares at date of announcement (CACT) is used. The premium paid is defined as the Premium of offer price to target closing stock price 1 day (1 week, 4 weeks) prior to the original announcement date, expressed as a percentage. Source: Thomson Reuters SDC Platinum. Panel A: Deal Information Offer Structure Cash Only Hybrid Other Shares Unknown Total Description No. % of Tot. No. % of Tot. No. % of Tot. No. % of Tot. No. % of Tot. No. % of Tot. Nbr. of Deals % % % % % 1, % Completed Deals % % % % % 1, % Hostile Deals % % 3 0.2% 7 0.4% 4 0.2% % Same Industry Deals % % 6 0.3% % 5 0.3% % Challenged Deals % % 7 0.4% % % % Competing Bidder % % 3 0.2% % 4 0.2% % Collar Deal 4 0.2% % 3 0.2% % 7 0.4% % Termination Fee % % % % % 1, % Bidder has a toehold % % 2 0.1% 7 0.4% 3 0.2% % Panel B: Deal Financials Offer Structure Cash Only Hybrid Other Shares Unknown Total Description Mean Sd Mean Sd Mean Sd Mean Sd Mean Sd Mean Sd DVal (mil) $2,242.0 $4,147.2 $5,880.9 $10,071.5 $5,074.2 $10,387.7 $5,429.8 $15,158.5 $1,635.7 $2,503.7 $3,848.4 $9,401.3 P1d 33.6% 31.7% 28.5% 27.5% 25.1% 40.5% 28.3% 39.5% 33.3% 29.6% 31.0% 33.1% P1w 36.6% 31.0% 32.4% 29.1% 29.5% 42.5% 33.6% 61.5% 33.4% 29.8% 34.7% 39.8% P4w 41.1% 35.6% 35.0% 32.4% 31.2% 46.1% 36.7% 45.3% 38.0% 33.6% 38.3% 37.7% 30

32 Table 2: Summary Statistics - Option Trading Volume (Without Zero Volume Observations) This table presents basic summary statistics on option trading volume, excluding zero-volume observations, stratified by time to expiration and depth-inmoneyness. We classify three groups of time to expiration: less or equal than 30 days, bigger than 30 but less or equal than 60 days and more than 60 days. We assign 5 groups for depth-in-moneyness, where depth-in-moneyness is defined as S/K, the ratio of the stock price S over the strike price K. Deep out-of-themoney (DOTM) corresponds to S/K [0, 0.80] for calls ( [1.20, ] for puts), out-of-the-money (OTM) corresponds to S/K ]0.80, 0.95] for calls ([1.05, 1.20[ for puts), at-the-money (ATM) corresponds to S/K ]0.95, 1.05[ for calls ( ]0.95, 1.05[ for puts), in-the-money (ITM) corresponds to S/K [1.05, 1.20[ for calls (]0.80, 0.95] for puts), and deep in-the-money (DITM) corresponds to S/K [1.20, [ for calls ([0, 0.80] for puts). Source: OptionMetrics Target (N = 2,214,260) Acquirer (N = 3,582,394) DITM Mean SD Min Med p75 p90 Max Panel A: All options, TTE = [0,30] DOTM (3%) 246 1, ,177 OTM (5%) 370 1, ,086 ATM (79%) 273 1, ,204 ITM (5%) 356 6, ,482 DITM (5%) 275 3, ,000 Total (100%) 283 2, ,482 Panel B: All options, TTE = ]30,60] DOTM (6%) ,045 OTM (9%) 285 1, ,222 ATM (71%) ,822 ITM (6%) 190 3, ,513 DITM (6%) 208 5, ,053 Total (100%) 194 1, ,053 Panel C: All options, TTE = ]60,...] DOTM (25%) 117 1, ,751 OTM (24%) ,885 ATM (20%) ,416 ITM (14%) ,647 DITM (15%) 83 1, ,804 Total (100%) ,751 Panel D: Call options, TTE = [0,30] DOTM (2%) 285 1, ,937 OTM (4%) 438 2, ,637 ATM (78%) 302 1, ,204 ITM (7%) 446 7, ,482 DITM (7%) 220 3, ,000 Total (100%) 311 2, ,482 Panel E: Call options, TTE = ]30,60] DOTM (4%) ,000 OTM (8%) 313 1, ,955 ATM (70%) ,208 ITM (7%) 213 3, ,513 DITM (8%) 213 5, ,053 Total (100%) 212 2, ,053 Panel F: Call options, TTE = ]60,...] DOTM (23%) 108 1, ,751 OTM (26%) ,885 ATM (20%) ,416 ITM (13%) 108 1, ,647 DITM (16%) 82 1, ,804 Total (100%) 114 1, ,751 Panel G: Put options, TTE = [0,30] DOTM (4%) 220 2, ,177 OTM (6%) 306 1, ,086 ATM (81%) 234 1, ,819 ITM (4%) ,708 DITM (2%) 485 3, ,010 Total (100%) 242 1, ,010 Panel H: Put options, TTE = ]30,60] DOTM (9%) ,045 OTM (10%) 253 1, ,222 ATM (71%) ,822 ITM (5%) ,177 DITM (3%) 192 1, ,004 Total (100%) ,822 Panel I: Put options, TTE = ]60,...] DOTM (29%) ,123 OTM (22%) ,066 ATM (21%) ,000 ITM (14%) ,906 DITM (12%) ,014 Total (100%) ,066 DITM Mean SD Min Med p75 p90 Max Panel A: All options, TTE = [0,30] DOTM (10%) ,377 OTM (22%) 497 1, ,207 55,167 ATM (26%) 1,084 3, , ,146 ITM (23%) 398 5, ,620 DITM (16%) 214 3, ,841 Total (100%) 547 3, , ,620 Panel B: All options, TTE = ]30,60] DOTM (14%) ,000 OTM (27%) 384 1, ,552 ATM (25%) 551 1, ,299 90,497 ITM (20%) 236 3, ,019 DITM (12%) , ,609,002 Total (100%) 354 4, ,609,002 Panel C: All options, TTE = ]60,...] DOTM (24%) ,430 OTM (25%) 193 1, ,507 ATM (18%) ,131 ITM (15%) ,027 DITM (15%) 80 1, ,500 Total (100%) 145 1, ,500 Panel D: Call options, TTE = [0,30] DOTM (6%) ,553 OTM (21%) 523 1, ,281 55,167 ATM (25%) 1,285 3, ,106 3, ,146 ITM (24%) 499 6, ,620 DITM (23%) 192 3, ,841 Total (100%) 603 4, , ,620 Panel E: Call options, TTE = ]30,60] DOTM (9%) ,000 OTM (27%) 425 1, ,060 ATM (24%) 657 1, ,593 90,497 ITM (21%) 297 4, ,019 DITM (17%) , ,609,002 Total (100%) 412 6, ,609,002 Panel F: Call options, TTE = ]60,...] DOTM (19%) ,430 OTM (27%) 199 1, ,507 ATM (18%) ,131 ITM (15%) ,027 DITM (20%) 75 1, ,500 Total (100%) 147 1, ,500 Panel G: Put options, TTE = [0,30] DOTM (14%) ,377 OTM (25%) 468 1, ,128 40,432 ATM (29%) 855 2, ,185 77,874 ITM (21%) 249 1, ,584 DITM (8%) 294 2, ,004 Total (100%) 471 1, , ,584 Panel H: Put options, TTE = ]30,60] DOTM (21%) ,195 OTM (27%) 332 1, ,552 ATM (26%) 424 1, ,010 32,239 ITM (18%) ,470 DITM (6%) 281 1, ,401 Total (100%) 280 1, ,552 Panel I: Put options, TTE = ]60,...] DOTM (31%) ,103 OTM (23%) ,492 ATM (19%) ,516 ITM (16%) ,420 DITM (9%) ,051 Total (100%) ,103 31

33 Table 3: Strongly Unusual Trading (SUT) Sample and Matched Random Sample Panel A presents sample statistics for the Strongly Unusual Trading (SUT) sample, reflecting four selection criteria: (1) the best bid price of the day is zero, (2) non-zero volume, (3) option expiration after the announcement date and (4) transaction within 30 days prior to the announcement date. Panel B presents comparative statistics for a randomly selected sample from the entire data, where for each event we choose a pseudo event date and then apply the same selection criteria as for the SUT sample. Source: OptionMetrics Panel A: SUT Selection with the historical 1,859 (792) event dates for the target (acquirer) Target Obs # Deals # Options Mean Vol Med Vol Min Vol 1st pctile 5th pctile 25th pctile 75th pctile 95th pctile 99th pctile Max Vol All 2, , ,076 13,478 Calls 1, ,517 6,161 Puts ,494 13,478 Acquirer Obs # Deals # Options Mean Vol Med Vol Min Vol 1st pctile 5th pctile 25th pctile 75th pctile 95th pctile 99th pctile Max Vol All 1, ,829 16,837 Calls ,899 16,837 Puts ,829 5,004 Panel B: 1 random sample of 1,859 (792) pseudo event dates for the target (acquirer) Target Obs # Deals # Options Mean Vol Med Vol Min Vol 1st pctile 5th pctile 25th pctile 75th pctile 95th pctile 99th pctile Max Vol All 3, , ,000 Calls 1, ,000 Puts 1, ,000 Acquirer Obs # Deals # Options Mean Vol Med Vol Min Vol 1st pctile 5th pctile 25th pctile 75th pctile 95th pctile 99th pctile Max Vol All 2, , ,102 5,938 Calls 1, ,000 4,287 Puts 1, ,540 5,938 32

34 Table 4: Bivariate Kolmogorov-Smirnov Tests - Target The table reports the test statistics from a generalization of the bivariate two-sample Kolmogorov Smirnov test based on Fasano and Franceschini (Fasano and Franceschini (1987)). The null hypothesis of the test is that two bi-variate samples come from the same empirical distribution function., and denote statistical significance at the 1%, 5% and 10% level respectively. Panel A: Calls Full Sample Event Window [ 29, 25] [ 24, 20] [ 19, 15] [ 14, 10] [ 9, 5] [ 4, 1] [0, 0] [ 29, 25] [ 24, 20] [ 19, 15] [ 14, 10] [ 9, 5] [ 4, 1] [0, 0] TTE = [0,30] Event Window [ 29, 25] [ 24, 20] [ 19, 15] [ 14, 10] [ 9, 5] [ 4, 1] [0, 0] [ 29, 25] [ 24, 20] [ 19, 15] [ 14, 10] [ 9, 5] [ 4, 1] [0, 0] TTE = ]30,60] Event Window [ 29, 25] [ 24, 20] [ 19, 15] [ 14, 10] [ 9, 5] [ 4, 1] [0, 0] [ 29, 25] [ 24, 20] [ 19, 15] [ 14, 10] [ 9, 5] [ 4, 1] [0, 0] TTE = [60,...] Event Window [ 29, 25] [ 24, 20] [ 19, 15] [ 14, 10] [ 9, 5] [ 4, 1] [0, 0] [ 29, 25] [ 24, 20] [ 19, 15] [ 14, 10] [ 9, 5] [ 4, 1] [0, 0] Panel B: Puts Full Sample Event Window [ 29, 25] [ 24, 20] [ 19, 15] [ 14, 10] [ 9, 5] [ 4, 1] [0, 0] [ 29, 25] [ 24, 20] [ 19, 15] [ 14, 10] [ 9, 5] [ 4, 1] [0, 0] TTE = [0,30] Event Window [ 29, 25] [ 24, 20] [ 19, 15] [ 14, 10] [ 9, 5] [ 4, 1] [0, 0] [ 29, 25] [ 24, 20] [ 19, 15] [ 14, 10] [ 9, 5] [ 4, 1] [0, 0] TTE = ]30,60] Event Window [ 29, 25] [ 24, 20] [ 19, 15] [ 14, 10] [ 9, 5] [ 4, 1] [0, 0] [ 29, 25] [ 24, 20] [ 19, 15] [ 14, 10] [ 9, 5] [ 4, 1] [0, 0] TTE = ]60,...] Event Window [ 29, 25] [ 24, 20] [ 19, 15] [ 14, 10] [ 9, 5] [ 4, 1] [0, 0] [ 29, 25] [ 24, 20] [ 19, 15] [ 14, 10] [ 9, 5] [ 4, 1] [0, 0]

35 Table 5: Bivariate Kolmogorov-Smirnov Tests - Acquirer The table reports the test statistics from a generalization of the bivariate two-sample Kolmogorov Smirnov test based on Fasano and Franceschini (Fasano and Franceschini (1987)). The null hypothesis of the test is that two bi-variate samples come from the same empirical distribution function., and denote statistical significance at the 1%, 5% and 10% level respectively. Panel A: Calls Full Sample Event Window [ 29, 25] [ 24, 20] [ 19, 15] [ 14, 10] [ 9, 5] [ 4, 1] [0, 0] [ 29, 25] [ 24, 20] [ 19, 15] [ 14, 10] [ 9, 5] [ 4, 1] [0, 0] TTE = [0,30] Event Window [ 29, 25] [ 24, 20] [ 19, 15] [ 14, 10] [ 9, 5] [ 4, 1] [0, 0] [ 29, 25] [ 24, 20] [ 19, 15] [ 14, 10] [ 9, 5] [ 4, 1] [0, 0] TTE = ]30,60] Event Window [ 29, 25] [ 24, 20] [ 19, 15] [ 14, 10] [ 9, 5] [ 4, 1] [0, 0] [ 29, 25] [ 24, 20] [ 19, 15] [ 14, 10] [ 9, 5] [ 4, 1] [0, 0] TTE = [60,...] Event Window [ 29, 25] [ 24, 20] [ 19, 15] [ 14, 10] [ 9, 5] [ 4, 1] [0, 0] [ 29, 25] [ 24, 20] [ 19, 15] [ 14, 10] [ 9, 5] [ 4, 1] [0, 0] Panel B: Puts Full Sample Event Window [ 29, 25] [ 24, 20] [ 19, 15] [ 14, 10] [ 9, 5] [ 4, 1] [0, 0] [ 29, 25] [ 24, 20] [ 19, 15] [ 14, 10] [ 9, 5] [ 4, 1] [0, 0] TTE = [0,30] Event Window [ 29, 25] [ 24, 20] [ 19, 15] [ 14, 10] [ 9, 5] [ 4, 1] [0, 0] [ 29, 25] [ 24, 20] [ 19, 15] [ 14, 10] [ 9, 5] [ 4, 1] [0, 0] TTE = ]30,60] Event Window [ 29, 25] [ 24, 20] [ 19, 15] [ 14, 10] [ 9, 5] [ 4, 1] [0, 0] [ 29, 25] [ 24, 20] [ 19, 15] [ 14, 10] [ 9, 5] [ 4, 1] [0, 0] TTE = ]60,...] Event Window [ 29, 25] [ 24, 20] [ 19, 15] [ 14, 10] [ 9, 5] [ 4, 1] [0, 0] [ 29, 25] [ 24, 20] [ 19, 15] [ 14, 10] [ 9, 5] [ 4, 1] [0, 0]

36 Table 6: Positive Abnormal Trading Volume Panel A reports the number and frequency of events with positive cumulative abnormal volume, as well as the t- statistic on the average cumulative abnormal volume. Panel B reports the t-test statistics for the differences in the average cumulative abnormal volumes across moneyness categories. We report heteroscedasticity-robust standard errors. For the market model, the market option volume is defined as either the mean or the median of the total daily trading volume across all options (respectively calls or puts) in the OptionMetrics database. The estimation window starts 90 days before the announcement date and runs until 30 days before the event. The event window stretches from 30 days before until one day prior to the event. Panel A Market Model (Median) Market Model (Mean) Constant Mean Model Option Type All Calls Puts All Calls Puts All Calls Puts All Options - Target Sign.t-stat 5% (#) Sign.t-stat 5% (freq.) E [CAV ] tcav OTM Options - Target Sign.t-stat 5% (#) Sign.t-stat 5% (freq.) E [CAV ] tcav ATM Options - Target Sign.t-stat 5% (#) Sign.t-stat 5% (freq.) E [CAV ] tcav ITM Options - Target Sign.t-stat 5% (#) Sign.t-stat 5% (freq.) E [CAV ] tcav All Options - Acquirer Sign.t-stat 5% (#) Sign.t-stat 5% (freq.) E [CAV ] tcav OTM Options - Acquirer Sign.t-stat 5% (#) Sign.t-stat 5% (freq.) E [CAV ] tcav ATM Options - Acquirer Sign.t-stat 5% (#) Sign.t-stat 5% (freq.) E [CAV ] tcav ITM Options - Acquirer Sign.t-stat 5% (#) Sign.t-stat 5% (freq.) E [CAV ] tcav Panel B Statistics Diff s.e. p-val Diff s.e. p-val Diff s.e. p-val All Options - Target OTM-ATM OTM-ITM ATM-ITM Call Options - Target OTM-ATM OTM-ITM ATM-ITM Put Options - Target OTM-ATM OTM-ITM ATM-ITM All Options - Acquirer OTM-ATM OTM-ITM ATM-ITM Call Options - Acquirer OTM-ATM OTM-ITM ATM-ITM Put Options - Acquirer OTM-ATM OTM-ITM ATM-ITM

37 Table 7: Positive Excess Implied Volatility This table reports the results from a classical event study where we test whether there was statistically significant positive excess implied volatility in anticipation of the M&A announcement. Two different models are used: excess implied volatility relative to a constant-mean-volatility model, as well as a market model, where we use as the market-implied volatility the CBOE S&P500 Volatility Index (VIX). The estimation window starts 90 days before the announcement date and runs until 30 days before the event. The event window stretches from 30 days before until one day prior to the event. Panel A reports the number and frequency of events with positive excess implied volatility, as well as the t-statistic on the average cumulative excess implied volatility. Panel B reports the t-test statistics for the differences in means of the average cumulative excess implied volatilities across moneyness categories. We report heteroscedasticity-robust standard errors. Panel A Market Model (VIX) Constant Mean Model Option Type Calls Puts Calls Puts 30-day ATM Implied Volatility ( δ = 50) - Target Sign.t-stat 5% (#) Sign.t-stat 5% (freq.) E [CEIV ] tceiv day ITM Implied Volatility ( δ = 80) - Target Sign.t-stat 5% (#) Sign.t-stat 5% (freq.) E [CEIV ] tceiv day OTM Implied Volatility ( δ = 20) - Target Sign.t-stat 5% (#) Sign.t-stat 5% (freq.) E [CEIV ] tceiv day ATM Implied Volatility ( δ = 50) - Acquirer Sign.t-stat 5% (#) Sign.t-stat 5% (freq.) E [CEIV ] tceiv day ITM Implied Volatility ( δ = 80) - Acquirer Sign.t-stat 5% (#) Sign.t-stat 5% (freq.) E [CEIV ] tceiv day OTM Implied Volatility ( δ = 20) - Acquirer Sign.t-stat 5% (#) Sign.t-stat 5% (freq.) E [CEIV ] tceiv Panel B Market Model (VIX) Constant Mean Model Statistics Diff s.e. p-val Diff s.e. p-val Call Options - Target ITM OTM Put Options - Target ITM OTM Call Options vs. Put Options - Target C IT M P IT M C IT M P OT M C OT M P IT M C OT M P OT M Call Options - Acquirer ITM OTM Put Options - Acquirer ITM OTM Call Options vs. Put Options - Acquirer C IT M P IT M C IT M P OT M C OT M P IT M C OT M P OT M

38 Table 8: Non-parametric Volume-at-Risk Panel A reports statistics (Number of observations, frequency, mean, median, minimum, maximum) on non-zero trading volume observations conditional on no trading volume during the one to 5 preceding days. Source: OptionMetrics Target Acquirer Conditional Volume Statistics variable N Freq. mean p50 sd min max N Freq. mean p50 sd min max Panel A1: Vt > 0 Vt 1 = 0 All Options 900, ,050 1,231, , ,609,002 Calls 519, , , , , ,609,002 Puts 380, , , ,103 Panel B1: Vt > 0 Vt 1 = 0, Vt 2 = 0 All Options 619, , , , ,609,002 Calls 346, , , , ,609,002 Puts 272, , , ,103 Panel C1: Vt > 0 Vt 1 = 0, Vt 2 = 0, Vt 3 = 0 All Options 474, , , , ,609,002 Calls 260, , , , ,609,002 Puts 214, , , ,001 Panel D1: Vt > 0 Vt 1 = 0, Vt 2 = 0, Vt 3 = 0, Vt 4 = 0 All Options 383, , , , ,609,002 Calls , , , ,609,002 Puts , , ,001 Panel E1: Vt > 0 Vt 1 = 0, Vt 2 = 0, Vt 3 = 0, Vt 4 = 0, Vt 5 = 0 All Options 320, , , , ,609,002 Calls 169, , , , ,609,002 Puts 150, , , ,001 37

39 Table 9: Non-parametric Volume-at-Risk This table reports the trading levels ν such that the probability of exceeding a trading volume of ν contracts, conditional on no earlier trading volume during 1, 2, and up to 5 trading days, is less than or equal to x%. Panel A compares results from the event window (30 days before the announcement date t [ 29, 1] to statistics from a random sample of 1859 pseudo event dates in Panel B. The statistics are based on the empirical distribution functions of volume, defined as the number of contracts traded.) Panel A: Event Window t [ 30, 1] Target Acquirer Panel A.1: P (Vt ν n i V t i = 0; i = 1) x% ν ν x x Panel A.2: P (Vt ν n i V t i = 0; i = 2) x% ν x ν x ν x ν x Panel A.3: P (Vt ν n i V t i = 0; i = 3) x% Panel A.4: P (Vt ν n i V t i = 0; i = 4) x% Panel A.5: P (Vt ν n i V t i = 0; i = 5) x% ν x ν x ν x ν x (Vt (Vt Panel B: Random Sample Target Panel B.1: P ν n Acquirer i V t i = 0; i = 1) x% ν ν x Panel B.2: P ν n x i V t i = 0; i = 2) x% ν x ν x ν x Panel B.3: P (Vt ν n i V t i = 0; i = 3) x% Panel B.4: P (Vt ν n i V t i = 0; i = 4) x% n Panel B.5: P (Vt ν i V t i = 0; i = 5) x% ν x ν x ν x ν x ν x

40 Figure 1: Trading Volumes around Announcement Dates These graphs plot the average option trading volume (left axis) and the average number of options used in the calculation (right axis) 60 days before and after the announcement date. Graphs (1a) and (1b) plot the average call trading volume for respectively the Acquirer and target. Graphs (1c) and (1d) plot the average put trading volume for respectively the Acquirer and target. For each deal, the total trading volume on each day across calls and puts is calculated. The bars represent the average daily total trading volume across all deals. Volume is defined as the number of option contracts. (a) (b) (c) (d) 39