Journal of High Technology Management Research 25 (2014) 36–53

Contents lists available at ScienceDirect

Journal of High Technology Management Research

Private information leakages and informed trading returns oftech target firms

Jeff Madura a, Thanh Ngo b,⁎a Florida Atlantic University, United Statesb East Carolina University, United States

a r t i c l e i n f o

⁎ Corresponding author.E-mail addresses: jeffmadura@bellsouth.net (J. Ma

1 See http://www.cbsnews.com/8301-505125_162-2 See http://www.reuters.com/article/2010/12/17/u3 See http://www.americancriminallawreview.com/4 See http://www.insideinvestorrelations.com/artic

1047-8310/$ – see front matter. Published by Elsevierhttp://dx.doi.org/10.1016/j.hitech.2013.12.002

a b s t r a c t

Available online xxxx

Wemeasure private information leakages about target tech firms in mergers. We find that techtarget firms with a higher level of asymmetric information are more exposed to mispricing,which allows for larger stock price gains from using expert networks or other means to obtainprivate information about impending mergers. We also find that the level of informationleakages is reduced since the Sarbanes–Oxley Act and Galleon case. However, the reduction inthe information leakage prior to tech merger announcements has been offset by the increasedshare price responses of tech firms to the merger announcements. Therefore, the potentialrewards from using expert networks or other means to retrieve private information about techtarget firms are still substantial for informed traders.

Published by Elsevier Inc.

Keywords:Insider tradingHigh-tech firmsMergers and acquisitions

1. Introduction

Social networking has taken on a newmeaning in the tech sector.With the high degree of asymmetric information surrounding techfirms, institutional investors are relying more heavily on expert networks for private information. Tech firms face a vicious cycle ofseeking continual investment to support innovation, which drives their growth. They cannot rely on a narrow product line like somemanufacturing firms. Because theymust hide their innovations in progress from their competitors, their public disclosure of the researchand development is very limited. Consequently, they exhibit a high degree of asymmetric information, which causes uncertainty abouttheir true valuation. A single piece of technology can have a major impact on a tech firm's valuation, and investors have an incentive toobtain the private information about tech firms that other investors do not have.

Before 2000, tech firms commonly accommodated their favored analysts and investors by providing private information.1 Thiswas explicitly outlawed by Regulation FD, which was implemented on October 23rd, 2000. Analysts still attempt to obtain privateinformation about tech firms, but have to work harder at obtaining it. In an article on inside information in the tech sector, a techanalyst states “there's just a huge amount of invested dollars focused on tech, and financial analysts chase each other for who hasthe best information.”2

Hedge funds have grown over time, and many of them invest in technology firms. They seek alternative sources of privateinformation (see “Unintended consequence of Reg. FD” by Zuckerman & Pulliam, 2010). The growth and competition for materialinformation has resulted in the development of expert networks.3 In 2006, there were about 30 expert networks that had annualrevenues exceeding $300 million. By 2009, there were more than 45 expert networks with annual revenues exceeding$400 million.4 The connection between the experts and the investors has facilitated by the development and growth of expertnetwork firms. For example, the Gerson Lehrman Group has about 200,000 experts. The expert networks are especially popular in

dura), ngotn123@gmail.com (T. Ngo).28243043/is-insider-trading-still-rampant-on-wall-streets-insidertrading-culture-idUSTRE6BF5WP20101217Drupal/blogs/blog-entry/expert-networks-crossing-insider-trading-line-10-03-2011les/sell-side/15351/rise-expert-networks

Inc.

37J. Madura, T. Ngo / Journal of High Technology Management Research 25 (2014) 36–53

the tech sector, as retired executives and existing employees of tech firms are paid large compensation to provide information tohedge funds, venture capitalists, and other investors. The investors commonly want information about a firm's supply chain, productinnovations, and large purchase orders, which could indicate future movements in earnings. Experts can provide information thatoffers valuable insight about a firm or industry without violating Regulation FD. However, some investors that pursue expertnetworks are in search of material non-public information.5 This may explain why some individual experts have been paid $150 to$1000per hour, ormore than $100,000 over time to participate in phone calls with clients.6 AMcKinsey consultantwas paid $500,000per year by the Galleon hedge fund for several years to provide his knowledge.

Many of the insider trading charges by the Securities and Exchange Commission (SEC) cite expert network participants.MarketWatch reported that regulators were putting a “spotlight on a trend of so-called expert networks.” It quoted John Coffee, aColumbia Law School professor: “The expert network says there shall be no exchange of information, but why is the hedge fundpaying $30,000 to $40,000 to meet these people.”7

Tech firms are more susceptible to insider trading because there are more corporate transactions, there is more focus onmergers, and there is more focus on earnings announcements.8 Since these types of events can have a major impact on shareprices of tech firms, investors who possess the information in advance can generate large trading returns. In one of the mostpublicized cases, the SEC alleged on November 20, 2012 that hedge funds or their investment advisors achieved profits or avoidedlosses totaling $276 million from trading ahead of news in July 2008 about an Alzheimer's drug. A doctor who was the chairmanof a safety monitoring committee that was overseeing the clinical trial was also paid more than $1000 per hour for his expertise inan expert network. He received about $108,000 for his consultations with investors.9

In some cases, investors hiring the experts have indicated that the phone calls will not be recorded, which could be intended toencourage experts to divulge inside information.10 The establishment and growth of expert networks have resulted in concerns assummarized in a CBS news article: “As long as there's a stock market, there will be investment professionals looking for acompetitive edge and a percentage of those willing to cross the line.”11

Stock price movements prior to public announcements can reveal the level of private information, and mergers are especiallyappealing to informed investors, because publicly traded target firms tend to experience a very large jump in their stock pricewhen they a merger bid is announced. Some studies, including those by Keown and Pinkerton (1981), Jarrell and Poulsen (1989),Schwert (1996), and Chira and Madura (forthcoming) have measured the degree of information leakages experienced by publictargets based in the U.S. prior to merger bid announcements in order to offer inferences about insider trading. However, no studyto our knowledge has attempted to measure the effects on tech firms that are subject to takeover bids, or to explain why the levelof information leakage may be higher for some types of tech target firms than others. Yet, when investigating insider trading, techtarget firms deserve special attention because of the established expert networks within the technology sector. Our objective is toinvestigate the private information surrounding tech firms that reaches investors.

We focus onmerger announcements involving publicly-traded tech targets, so that we canmeasure the level of private informationrevealed about the tech targets before the announcements. A negative stock price runup prior to a merger announcement implies lessthan zero private information and could be misleading. We therefore apply a transformation to any tech targets that experience anegative runup prior to their merger announcement. We also apply an alternative method in which we delete these targets from oursample.

Our analysis shows much variation in the level of private information among tech targets. In fact, more than one-third of thetech targets experience a negative runup. When excluding these firms, the mean level of private information prior to mergerannouncements is substantial, along with potential trading returns for investors who trade on that information.

We apply a multivariate analysis to explain the variation in the level of private information and trading returns among techfirms. In general, tech firms that have less investment in research and development, larger capital expenditures and more risk aresubject to higher levels of private information leakage. These firms may be subject to more asymmetric information and moremispricing, which allows for larger stock price gains from using expert networks or other means to obtain private informationabout the impending mergers. In addition, tech target firms with higher trading volume experience higher levels of privateinformation leakage, as informed traders (which we define as traders using private information) may be able to more easily hidetheir trades of those targets' shares.

Our multivariate analysis also confirms a reduction in private information leakage following the Sarbanes–Oxley Act and theGalleon case while accounting for other characteristics. However, the returns earned by informed traders from trading on privateinformation are not reduced since these government initiatives. In essence, the reduction in leakage of private information hasbeen offset by higher share price responses of tech targets to merger announcements after the Sarbanes–Oxley Act and Galleoncase. Consequently, informed traders are still able to extract large profits from illegally capitalizing on private information,although the risk of prosecution and criminal or civil penalties from such trading is now higher.

5 See http://www.edn.com/electronics-news/4369622/Insider-trading-sting-reverberates-throughout-the-tech-industry6 See http://dealbook.nytimes.com/2010/12/16/four-arrested-in-insider-trading-investigation/7 See http://www.marketwatch.com/story/expert-networks-key-to-sec-insider-trading-cases-2012-11-218 See http://www.bloomberg.com/news/2011-04-07/raissi-says-tech-industry-more-prone-to-insider-trading-video.html9 See http://online.wsj.com/article/SB10001424127887323713104578130930796204500.html.

10 See http://www.integrity-research.com/cms/2011/02/07/sec-files-charges-in-insider-trading-probe/11 See http://www.cbsnews.com/8301-505125_162-28243043/is-insider-trading-still-rampant-on-wall-street/

38 J. Madura, T. Ngo / Journal of High Technology Management Research 25 (2014) 36–53

2. Related literature

Many studies have established the potential impact of asymmetric information between investors and firms. Studies by Lelandand Pyle (1977), Grossman and Hart (1981), Bhattacharya and Ritter (1983), Amihud and Mendelson (1986), and Easley andO'Hara (2004) demonstrate that information asymmetry can disrupt managerial behavior and raise the firm's cost of capital. Techfirms are subject to a high degree of asymmetric information, based on common proxies such as research and developmentexpenses (see Helwege & Liang, 1996), analyst forecast dispersion (see Barron, Byard, Kile, & Riedl, 2002), and idiosyncraticvolatility (see Moeller, Schlingemann and Stulz 2006). While tech firms generally have high levels of asymmetric information, alltech firms are not alike. Barron et al. (2002) find that tech firms with relatively high levels of intangible assets are subject to ahigher degree of uncertainty.

Several studies attempt to determine the causes and/or effects of technology-related mergers. Higgins and Rodriguez (2006)find that a firm's declining R&D productivity may be the underlying motivation to acquire R&D intensive firms. Kohers and Kohers(2000) suggest that tech firms are unique because of their business model, growth prospects, and the uncertainty surroundingtheir valuation. They find that high technology mergers elicit a favorable market response. However, Kohers and Kohers (2001)find that long-term stock price performance of tech firms that engage in takeovers is weak. Grimpe and Hussinger (2008) findthat technology mergers can generate innovation advantages over competitors. Ahuja and Katila (2001), Cassiman, Colomboc,Garrone, and Veugelers (2005), Makri, Hitt, and Lane (2010), Ornaghi (2009), Stahl (2010), and Sevilir and Tian (2011) assesshow bidders refine their R&D efforts following takeovers.

While these studies offer insight on the causes and effects of tech mergers from the corporate perspective, they do not attemptto detect whether investors obtain private information in order to anticipate tech mergers, or measure the potential profit frominformed trading of tech target shares. Our goal is to plug this research gap by measuring the level of private information leakagefor tech targets prior to the announced takeover bids and to identify technology characteristics and government initiatives thatexplain the variation in private information among tech targets. We also attempt to explain the variation in returns earned frominformed trading of tech target shares.

3. Hypotheses

When a tech firm possesses much private information that has not been publicly disclosed, there is greater potential for informedtraders to benefit from accessing that private information. Characteristics that cause a greater level of private information may offerlarger potential gains from informed trading. We develop hypotheses for characteristics that can influence the leakage of privateinformation and potential returns on informed trading of tech target shares.

3.1. Technology characteristics that can influence the leak of private information

3.1.1. Firm size (SIZE)Relatively small tech firms (compared to others in the same industry) should exhibit a higher degree of asymmetric information,

because the public disclosure by these firms and the analyst coverage of these firms are limited.Wemeasure firm size as the natural logof the target firms at the end of the year preceding the merger announcement.

Hypothesis 1. Smaller target firms should be exposed to more pronounced information leakages.

3.1.2. Research and development (RD)Many tech firms rely heavily on research and development. Tsai and Wang (2005), Andres-Alonso, Azofra-Palenzuela, and

Fuente-Herrero (2006), Ho, Tjahjapranata, and Yap (2006), Coad and Rao (2008), and Oriani and Sobrero (2008) find that R&Dcan enhance sales growth options and firm valuation. However, tech firms that rely heavily on R&D may also be subject to largelosses if their innovations fail. There is limited disclosure and therefore much uncertainty surrounding the progress of R&D. Chan,Lakonishok, and Sougiannis (2001) and Eberhart, Maxwell, and Siddique (2004) find that firms with heavy R&D are mispriced,which may attract more investors who attempt to obtain private information about the firm. Desyllas and Hughes (2008, 2009)suggest that R&D can enhance acquisition likelihood, which may attract more investors who attempt to obtain privateinformation about the firm. We measure R&D in proportion to the tech target's sales at the end of the year preceding the mergerannouncement.

Hypothesis 2. Target firms with more R&D should be exposed to more pronounced information leakages.

3.1.3. Intangible assets (INTANG)Along with heavy R&D, tech firms are known for their reliance on intangible assets. Barron et al. (2002) find that firms with

relatively high levels of intangible assets are subject to a higher degree of uncertainty. These firms may attract informed traderswho can access private information through expert networks or other means to capitalize on their information advantages. Wemeasure intangible assets in proportion to the total assets of the tech target firms at the end of the year preceding the mergerannouncement.

39J. Madura, T. Ngo / Journal of High Technology Management Research 25 (2014) 36–53

Hypothesis 3. Target firms with more intangible assets should be exposed to more pronounced information leakages.

3.1.4. Capital intensity (CAPX)Capital investment can allow tech firms to pursue growth opportunities. It enables the firms to expand or restructure their

existing assets (Trigeorgis, 1993). However, tech firms with heavy capital investment may be challenged to recapture theirinvestment (see Anderson and Garcia-Feijoo (2006) and Pindado, Queiroz, and Torre (2010)). Tech firms engaging in heavycapital spending may rely more heavily on innovations to recapture their investment. We expect that these firms may have muchprivate information, and therefore investors may use expert networks or other means to obtain the private information about thefirm. We measure capital intensity as the ratio of capital expenditures to sales for tech target firms at the end of the yearpreceding the merger announcement.

Hypothesis 4. Target firms with more capital intensity should be exposed to more pronounced information leakages.

When assessing tech-specific characteristics such as R&D, intangible assets, and capital expenditures, we consider a counterhypothesis. Tech firms with much R&D, intangible assets, or capital expenditures have substantial growth options that may already bepriced, and therefore might not attract as much informed trading.

3.1.5. Idiosyncratic risk (RISK)Tech firms that are subject to a high level of uncertainty relative to others in their corresponding industry exhibit a high degree

of asymmetric information. These firms should be subject to a higher degree of mispricing and therefore may attract informedtraders who want to use expert networks or other means to capitalize on the private information. Therefore, we expect that targetfirms with relatively high levels of idiosyncratic risk will exhibit a higher level of informed trading prior to the mergerannouncement. While risk could be partially influenced by the firm's size or technology characteristics, it could attract informedtrading beyond the influences of the other characteristics. Wemeasure a firm's risk as the standard deviation of the residuals fromthe CAPM model (based on a CRSP equally-weighted index) applied to the (−270, −65) day window.

Hypothesis 5. Target firms with more idiosyncratic risk should be exposed to more pronounced information leakages.

3.1.6. Debt ratio (DEBT)Tech firms that carry a relatively high level of debt compared to their corresponding industry may be more likely to engage in

major restructuring changes to improve their position within the industry. We expect that these firms would receive closerscrutiny from outsiders who are attempting to obtain private information (through expert networks or other means). Therefore,target firms with relatively high debt ratios may exhibit a higher level of informed trading prior to the merger announcement. Weobtain the total debt-to-total asset ratio of each firm at the end of the fiscal year preceding the merger announcement.

Hypothesis 6. Target firms with relatively high debt ratios should be more exposed to information leakages.

3.1.7. Location of firm (CALIFORNIA)We consider whether location has an influence on the level of private information leaked prior to a merger announcement. An

article on inside information in the tech sector describes the Silicon Valley as a “highly refined gossip mill.”12 We use a dummyvariable to designate tech target firms located in California.

Hypothesis 7. Target firms based in California should be exposed to more pronounced information leakages.

3.1.8. Reliance on financial intermediaries (FIN)According to Acharya and Johnson (2010), when tech mergers require the participation of financial intermediaries, more

private information may be leaked. When bidders finance acquisitions, they disclose information to lenders about their reason forfinancing. We control for bidder financing with the variable FIN, set equal to 1 when the bidder finances the acquisition and zerootherwise.

Hypothesis 8. Target firms that are to be acquired by bidder firms that rely on financing should be more exposed to informationleakages.

3.2. Characteristics that can conceal knowledge of private information

When informed traders capitalize on private information, they may attempt to hide their trades so that they do not revealtheir actions to other traders or to regulators (see Barclay & Warner, 1993). Tech firms may attract more informed traders if their

12 See http://www.reuters.com/article/2010/12/17/us-insidertrading-culture-idUSTRE6BF5WP20101217

40 J. Madura, T. Ngo / Journal of High Technology Management Research 25 (2014) 36–53

trading characteristics allow informed traders to more easily conceal their trades, and therefore conceal their knowledge of privateinformation from the SEC or other traders. The following characteristics may allow informed traders to conceal their trades.

3.2.1. Trading volume (VOLUME)Tech target firms with higher trading volume may be subject to higher degree of informed trading prior to merger announcements

because traders canmore easily hide their trades of shareswithout theprevailing trading activity (seeAdmati andPfleiderer (1988)). Thisshould allow informed traders to accumulatemore shares at lower prices before other traders detect from trading behavior that the firmhas become a takeover target.Wemeasure trading volume as the ratio of trading volume to total shares outstanding in the (−270,−65)day window relative to the merger announcement date.

3.2.2. Listed stock options (OPTION)Tech targets that have traded stock options may be subject to a higher degree of informed trading, because informed traders

may be able to conceal their trading and can lever their information to achieve higher profits for a given investment. Jayaraman,Frye, and Sabherwal (2001) show evidence of significant increases in trading of both call and put options in targets shortly beforea takeover. They also find evidence that the trading in the options market leads the trading in the underlying stock. We use adummy variable set to 1 for targets that have traded stock options.

3.2.3. Merger activity (M&A ACTIVITY)Studies by Harford (2005), Powell and Yawson (2005), and Dittmar and Dittmar (2008) show how merger activity changes in

response to economic conditions or shocks. A higher level of merger activity may encourage more investor efforts to obtain privateinformation about tech firms, in order to anticipate subsequent acquisitions. In addition, informed traders may be able to more easilyhide their trading activity from other traders or the SEC when there is more merger activity. Wemeasure merger activity as the naturallogarithm of the total deal values of all mergers taken place in the same industry and same year as the merger in consideration.

3.3. Effects of government initiatives on the leakage of private information

We hypothesize that the level of private information that occurs prior to tech merger announcements is influenced by threegovernment initiatives, as explained next.

3.3.1. Effects of regulation FD (FD)The information environment was the least restrictive up until Regulation Fair Disclosure (FD) was implemented (October,

2000). Studies have clearly documented information leakages regarding earnings information in this period (see Libby and Tan(1999), Anilowski, Feng, and Skinner (2007), and Sinha and Gadarowski (2010)). We expect leakage of private information abouttech target firms to decline since Regulation FD. We employ a dummy variable FD to capture the effect of regulation FD.

3.3.2. Effects of Sarbanes–Oxley Act (SOX)The Sarbanes–Oxley Act requires that firms develop internal control processes for financial reporting. It also forces financial

managers of a firm to be more accountable in their reporting. Therefore, it provides additional forces beyond the Regulation FDguidelines to prevent insiders from leaking information to outsiders. While the Sarbanes–Oxley is not directly intended to reduceillegal leakages of information, it may have indirectly led to this effect by requiring executives to be more accountable for theinformation that is disclosed by their firm. Thus, we expect that the leakage of private information about tech targets declinedsince the Act. We employ a dummy variable SOX to capture the effect of the Act.

3.3.3. Effects of Galleon case (GALLEON)On October 16, 2009, the U.S. government announced its charges against Galleon Group LLC. This announcement could have

sent a signal to discourage inside trading in the tech firms, because it reflected a more dedicated government role in regulatinginsider trading. The government announced its plans to pursue individuals from expert networks that were divulging nonpublicmaterial information to investors. It also announced its plans to use wiretap evidence and ultimately demonstrated its ability touse that evidence. The wiretaps can be especially effective within networks because they may enable experts who are charged toengage in recording conversations that could implicate others.

The Galleon case could reduce private information environment by forcing participants to be more vigilant. Expert networkfirms that connect experts with investors may more strongly emphasize to their clients that the information provided by expertsmust not violate existing laws regarding the disclosure of material non-public information.

As a result of the potential government enforcement signaled by the Galleon case, hedge funds or venture capital firms that arenot seeking illegal inside information from experts may warn experts to only offer information that does not violate the existingguidelines for information disclosure.13 Most importantly, the Galleon case may signal to the experts that they could be subject togovernment prosecution from leaking inside information. The signal of tougher enforcement relayed by the Galleon case has beenreinforced by the SEC's efforts from that point forward. Since it publicly announced charges against Galleon in October, 2009, the

13 [see http://online.wsj.com/article/SB10001424052748703864204576319122050566588.html]

41J. Madura, T. Ngo / Journal of High Technology Management Research 25 (2014) 36–53

SEC has filed 170 insider trading actions that involve more than 400 defendants (individuals or entities). The alleged illicit gains(composed of either trading profits or avoidance of losses) reflected within these actions total $870 million.14

We hypothesize that the Galleon case either restricts the flow of private information about tech firms, or discourages the use ofprivate information by investors. We expect that the tech targets experience a lower leakage of private information since theinception of the Galleon case, and that profits from informed trading of tech shares are lower since then.

Research by Brigida and Madura (2012) assesses how informed trading of publicly-traded firms is associated with variouscharacteristics, and find that informed trading has declined since the Sarbanes–Oxley Act. In addition, Chira and Madura (2013)find that the Galleon case results in lower information leakages. However, these studies focus on all publicly traded targets inaggregate. Our focus is on tech targets, which are more exposed to information leakages due to expert networks. Given our focus,we give attention to the influence of tech-specific characteristics that have not been considered in any other studies on informedtrading.

Unlike the aforementioned studies, we also adjust our sample or measurement of runup for targets that experience negativeinformation leakages. Furthermore, we estimate the level of informed trading returns that can be earned from capitalizing oninformation leakages, and explain the variation in informed trading returns among tech targets.

4. Data and method

To conduct our analysis on private information and informed trading of tech targets, we compile a set ofmerger announcementsinvolving publicly-traded tech target firms over the 1980–2011 period. We obtain merger announcements from Thompson's SDCMergers and Acquisitions database (SDC). In order for a deal to be included in our final sample, we require that the target stockprices should be available in CRSP for the window from day −270 to day +1 relative to the merger announcement date. We alsoremove deals for which deal value information is not available from SDC. This screening process yields an initial sample of 1075merger announcements.

We only consider mergers in which the cumulative abnormal return in the (0, +1) daywindow at the time of the announcement isat least 5%. This process is intended to ensure that allmerger announcements in our sample represent a significant surprise to themarket.Since our theme is on investigating the private information leakage prior to a public announcement that should elicit a strong favorablemarket response, we wish to exclude any merger announcements in which the news may have been publicly disseminated before theofficial announcement date. This screening results in a final sample of 833 merger announcements.

4.1. Measuring private information revealed before merger announcements

Tomeasure the private information revealed before eachmerger announcement involving a tech target firm,wemeasure the “runup”of each tech target, which we define as the abnormal stock price movement during a window prior to the merger announcement. Weapply the market model in the (−270,−65) period in event time prior to the merger announcement in order to derive expected dailyreturns of each tech target under normal conditions. The accumulated difference between actual returns and expected returns of thetarget during the window prior to the merger announcement represents the runup and serves as our measure of private informationleakage. Studies by Keown and Pinkerton (1981), Gupta and Misra (1988), Jarrell and Poulsen (1989), Meulbroek (1992), Schwert(1996), Jabbour, Jalilvand, and Switzer (2000), King (2009), and Chira and Madura (forthcoming) measure the runup of targets in asimilar manner and document that the mean runup is positive and significant for publicly traded target firms in general. As a test ofrobustness,we repeat all of our analyses by applying the Fama–Frenchmodel to estimate the expected daily returns in order to derive therunup.

We report the sample distribution by year in Table 1 for the full sample of 833 publicly-traded tech target firms. The largestproportion of merger announcements occurs during the 1997–2001 period and 2008–2010 period. Descriptive statistics of thetech firms are provided in Table 2. On average, bidder firms are approximately 20 times bigger than target firms (in terms ofmarket capitalization). The target firms have negative cash flows and return on assets on average. Target firms also have high R&Dexpenses as a proportion of sales (the mean R&D expense ratio is 60% and the median is 11%).

5. Level of private information leaked before mergers

The level of private information revealed (as measured by runup prior to the merger announcements) about tech targets isdisplayed in Panel A of Table 3 for the full sample of 833 observations. The mean abnormal stock price runup of tech targets in the(−10, −1) day window is 5.625% based on the market model, and 5.58% based on the Fama–French model. The mean runup forthe window (−20, −1) is 7.548% based on the market model, and 7.514% based on the Fama–French model, and increasesslightly as the window is extended backward in event time. For the window (−50, −1), the mean runup of tech targets is10.841% based on the market model, and 10.748% based on the Fama–French model. We report the one-sample mean t-statisticsand the M-sign statistics for the significance level of the mean and the median of the runup, respectively.

The runup levels off at about 50 days prior to the publicized merger announcement date. While our sample is distinguishedfrom most other studies that investigate informed trading before mergers in that it is focused on tech firms only and includes a

14 See http://www.sec.gov/news/press/2012/2012-237.htm]. The Galleon case was also reinforced by the Dodd–Frank Act (passed in 2010), which increasedpenalties for some forms of illegal insider trading.

Table 1Sample distribution.

Year N Percent of sample

1980 2 0.241981 1 0.121982 1 0.121983 3 0.361984 4 0.481985 9 1.081986 3 0.361987 8 0.961988 9 1.081989 7 0.841990 5 0.61991 11 1.321992 8 0.961993 10 1.21994 30 3.61995 26 3.121996 37 4.441997 47 5.641998 57 6.841999 65 7.82000 49 5.882001 57 6.842002 39 4.682003 45 5.42004 31 3.722005 38 4.562006 38 4.562007 36 4.322008 54 6.482009 45 5.42010 47 5.642011 11 1.32Total 833 100

This table reports the sample distribution by year.

42 J. Madura, T. Ngo / Journal of High Technology Management Research 25 (2014) 36–53

more recent period (through 2011), our general results for the mean runup are somewhat similar to those reported by otherrelated studies, including those by Keown and Pinkerton (1981), Jarrell and Poulsen (1989), Meulbroek (1992), and Schwert(1996).

5.1. Distribution of runup among tech targets

Related studies have focused on the mean runup effect without giving much attention to the dispersion of the runup for anyparticular window.We report our results for the runup (as measured by CAR) of tech targets for each of the windows by quartilesin Table 4. Panel A of Table 4 shows results when using the market model. For the window (−50, −1), the lowest quartile has amean of −26.346%, which suggests that the stock price runup is negative for at least one fourth of all target firms. In fact, for all

Table 2Summary statistics of sample firms (N = 833).

Variable Mean Median StdDev

Deal Value ($ million) 1311.80 259.732 5284.47Acquirer Market Cap. ($ million) 20,229.32 2,012.60 49,564.67Target Market Cap. ($ million) 895.584 160.263 4085.65% target shares acquired 86.106 100 33.982Target debt ratio 0.397 0.348 0.289Target cash flow ratio −0.895 0.087 12.62Target ROA −0.088 0.017 0.326Target idiosyncratic risk 0.051 0.029 0.083Target trading volume/shares outstanding 0.011 0.008 0.01Target capital expenditure/sale 0.134 0.042 0.866Target R&D expense/sale ratio 0.601 0.11 5.123Target intangible asset ratio 0.124 0.02 0.181

This table reports the summary statistics of some target characteristics

Table 3Summary statistics of runup.

Windows Method to estimate abnormal returns Mean t-stat Median M-sign

(−50, −1) Market-model 10.841% 9.70*** 9.303% 116.50***Fama–French 3-factor model 10.748% 9.65*** 8.190% 124.50***

(−20, −1) Market-model 7.548% 10.66*** 5.924% 142.50***Fama–French 3-factor model 7.514% 10.58*** 6.085% 145.50***

(−10, −1) Market-model 5.625% 10.08*** 3.849% 130.50***Fama–French 3-factor model 5.580% 9.97*** 3.360% 125.50***

This table reports the results the summary statistics of the target's runup. Runup is measured by the target's CARs in the (−50, −1), (−20, −1) and (−10, −1)days windows estimated from the market model and from the Fama–French 3-factor model alternatively. The estimation period is (−270,−65) days prior to theannouncement date of the merger. T-statistics and M-sign rank test statistics are reported. *, ** and *** indicate the significance levels of 10%, 5% and 1%,respectively.

43J. Madura, T. Ngo / Journal of High Technology Management Research 25 (2014) 36–53

windows assessed, the lowest quartile runup is negative. This is an important finding, because it demonstrates that runup ofmany targets does not fit the typical generalization in past studies when focusing only on the mean. Furthermore, the highestquartile for the windown (−50, −1) has a mean runup of 51.777%, which suggests that the level of private information prior tosome merger announcements is much larger than the typical generalization when focusing only on the mean.

Panel B of Table 4 shows results when using the Fama–French model. The lowest quartile for the window (−50, −1) has amean runup of−26.164%, which is similar to results from using the market model in Panel A. Again, at least one-fourth of all techtarget firms experience a negative runup prior to the merger announcement, regardless of the window assessed. The mean runupfor the highest quartile is 51.523%.

5.2. Negative runup sample

We segment our sample into tech targets that experience a negative versus positive runup based on the (−50, −1) windowprior to the merger announcement and provide these results in Table 5. When applying the market model, our negative runupsample (columns 2 and 3 in Panel A of Table 5) consists of 300 tech targets, or 36.014% of the entire sample. The mean runup forthis subsample is−19.430%. When applying the Fama–French model, our Negative Runup Sample (columns 2 and 3 in Panel B ofTable 5) consists of 292 tech targets, or 35.052% of the entire sample. The mean runup for this subsample is −19.822%.

Table 5 shows the results for various windows leading up to the announcement. Notice that for the Negative Runup Sample inPanel A, the runup becomesmore negative as the window is extended backward in time.When using the market model, themean(median) runup for this subsample is −1.457% (−1.021%) over the 10-day window before the merger announcement, −4.944%(−3.886%) over the 20-day window before the announcement, −7.858% (−6.063%) over the 30-day window before theannouncement, −13.004% (−9.40%) over the 40-day window before the announcement, and −19.430% (−13.818%) over the50-day window before the announcement. Results are similar for the Fama–French model shown in Panel B.

It is difficult to measure or interpret the level of private information about tech targets for which there is a negative runup,because it is not possible for information leakage to be less than zero. Since the level of private information associated with animpending merger should be no less than zero for a target firm, one possible interpretation of a negative runup is that any level offavorable private information about an anticipated takeover of the firm is overwhelmed by unfavorable information about badprospects for the firm. In this case, perhaps investors are less willing to act on private information for some target firms that

Table 4Mean and median runup by quartiles of runup.

Window (−50, −1) Window (−20, −1) Window (−10, −1)

Quartile Mean Median Mean Median Mean Median

Panel A—CARs are estimated from market model1 −26.346% −21.277% −15.285% −11.319% −11.253% −8.022%2 0.885% 1.113% 1.474% 1.557% 0.632% 0.749%3 17.226% 16.776% 11.119% 11.125% 7.854% 7.760%4 51.777% 45.531% 32.994% 28.381% 25.349% 19.668%

Panel B—CARs are Estimated from Fama–French 3-Factor Model1 −26.164% −20.518% −15.488% −11.681% −11.214% −8.034%2 0.791% 1.323% 1.476% 1.596% 0.572% 0.760%3 17.018% 17.000% 11.152% 10.962% 7.735% 7.533%4 51.523% 43.811% 33.027% 28.014% 25.308% 19.565%

This table reports the summary statistics for each quartile of target's runup. Runup is measured by the target's CARs in the (−50, −1) day window estimatedfrom the market model (in Panel A) and from the Fama–French 3-factor model (in Panel B), respectively. The estimation period is (−270, −65) days prior to theannouncement date of the merger.

Table 5Positive runup versus negative runup.

Panel A—CARs are estimated from market model

Negative runup (N = 300) Positive runup (N = 533) Converted runup (N = 833)

Window Mean Median Mean Median Mean Median

(−50, −1) −19.430% −13.818% 29.954% 23.557% 17.839% 9.303%(−40, −1) −13.004% −9.400% 25.605% 20.816% 16.162% 7.485%(−30, −1) −7.858% −6.063% 21.107% 16.564% 14.262% 7.010%(−20, −1) −4.944% −3.886% 15.358% 12.009% 11.513% 5.924%(−10, −1) −1.457% −1.021% 10.142% 6.928% 8.574% 3.849%

Panel B—CARs are estimated from Fama–French 3-factor model

Negative runup (N = 292) Positive runup (N = 541) Converted runup (N = 833)

Window Mean Median Mean Median Mean Median

(−50, −1) −19.822% −15.456% 29.138% 22.632% 17.696% 8.190%(−40, −1) −13.040% −10.751% 25.051% 20.417% 16.125% 7.552%(−30, −1) −8.051% −6.995% 20.665% 15.191% 14.282% 7.459%(−20, −1) −4.821% −3.887% 15.080% 11.772% 11.539% 6.085%(−10, −1) −1.584% −1.104% 10.045% 6.751% 8.531% 3.360%

This table reports the summary statistics of the target's runup for the negative sample, positive runup sample and the converted runup sample. Runup ismeasured by the target's CARs in the (−50,−1) day window estimated from the market model (in Panel A) and from the Fama–French 3-factor model (in PanelB), respectively. The estimation period is (−270, −65) days prior to the announcement date of the merger. The negative sample includes only negative runupobservations. The positive sample includes only positive runup observations. The converted sample includes positive runup observations and negative runupobservations that are set to 0.

44 J. Madura, T. Ngo / Journal of High Technology Management Research 25 (2014) 36–53

experience abnormally weak stock price movements because they question whether an impending merger for this type of targetwill actually be completed.

Given the large number of tech targets that exhibit a negative and pronounced runup, the inclusion of these targets in a sampledilutes the estimated level of private information in the entire sample. For example, when calculating the mean level of privateinformation released before a merger announcement, a target with a −5% stock price runup offsets the leakage represented by a5% stock price runup, which creates the appearance of zero leakage on average. In addition, when assessing the variation inprivate information leakage among firms, a target with a −10% runup would be perceived as having a lower level of informationleakage than a target with−5% runup. Any target with a negative stock price runup would be interpreted as having a lower levelof information leakage than a target with a zero runup.

Of all the related studies that we cite that measure information leakages prior to mergers, only the study Jarrell and Poulsen(1989) acknowledges the potential estimation problems when a sample contains some targets that experience a negative runup.In their analysis of targets subjected to tender offers during the 1981–1985 period, they state that it is difficult to interpret thedegree of information leakage from a negative runup, because it indicates “less than no anticipation” (p. 240).

Since the leakage of private information should not be less than zero, we apply a transformation to tech targets that experiencea negative runup, similar to the adjustment applied by Jarrell and Poulsen (1989) in their study of tender offers. We exclude techtargets that experience a negative runup, which essentially allows our sample to include only tech targets that show evidence of aprivate information leakage. This screening creates what we refer to as the “Positive Runup Sample.”

We also consider an alternative approach used by Jarrell and Poulsen (1989), in which we retain the tech targets with anegative runup for a particular window prior to the merger announcement, but set the negative runup of these targets to zero.This conversion implies that tech targets with a negative runup had zero (not negative) information leakage. This conversioncreates what we refer to as the “Converted Runup Sample”, which includes all 833 tech targets.

5.3. Positive Runup Sample

The estimated runup for the Positive Runup Sample is shown in columns 4 and 5 of Table 5. When applying the market model(Panel A), there are 533 targets, or 63.99% of our initial sample that experience a positive runup within the (−50, −1) windowand therefore qualify for the Positive Runup Sample. The mean runup from applying the market model for this sample over the(−50, −1) window is 29.954%. Panel B of Table 5 shows that when applying the Fama–French model, there are 541 targets(64.945% of our initial sample) that qualify for the Positive Runup Sample, and the mean runup over the (−50, −1) window forthis sample is 29.138%.

When using the Positive Runup Sample, the level of private information increases when a longer window is used to measurerunup. Table 5 shows that the mean runup for this sample is about 10% over the 10-day window before the mergerannouncement, about 15% over the 20-day window before the announcement, about 21% over the 30-day window before theannouncement, about 25% over the 40-day window before the announcement, and about 30% over the 50-day window before the

45J. Madura, T. Ngo / Journal of High Technology Management Research 25 (2014) 36–53

announcement. The estimated runup is very similar (less than 1 percentage point difference) for the market model (Panel A) andFama–French model (Panel B) for all windows when focusing on the Positive Runup Sample.

The mean runup for the Positive Runup Sample is about double that of our initial sample for windows of 30 days or less prior to themerger announcement date. For the 50-daywindow, themean runup for the Positive Runup Sample is about 2.76 times that of our initialsample (29.954% as compared to 10.841%). Thus, the average magnitude of information leakage when considering only targets thatshow evidence of a leakage is much larger than the leakage when considering all targets, because it avoids the dilution effect describedearlier.

5.4. Converted Runup Sample

The estimated runup for the Converted Runup Sample (in which a target's negative runup for a particular window is convertedto zero) is shown in columns 6 and 7 of Table 5. The sample size here is the same as for our initial sample in Table 3 because alltargets are retained in the sample, even if their runup is negative. However, the estimated mean runup is naturally higher for theConverted Runup Sample than for the initial sample because of converting the negative runup of any target in our sample to zeroin the manner of Jarrell and Poulsen (1989).

When measured by the market model (Panel A of Table 5), the mean runup for the Converted Runup Sample is 14.262% overthe 30-day (−30, −1) window, 16.162% over the 40-day window, and 17.839% over the 50-day window. Results are similarwhen using the Fama–French model (Panel B of Table 5).

Notice that for any window, the mean runup is lower for the Converted Runup Sample than it is for the Positive Runup Sample.This is because the targets with negative runup are retained in the sample but assigned a runup of zero in the Converted RunupSample, while they are removed from the Positive Runup Sample. Now that we have created our two samples, we can applymultivariate analysis to explain the variation in the runups for each sample separately.

6. How information leakage varies among tech targets

Next, we consider how private information leakage (as measured by runup) prior to merger announcements varies amongtech target firms. Since our goal is to capture the complete stock price runup per tech target, we use a 50-day window for much ofour analysis. As a basis of comparison, Schwert (1996) used a 42-day window for much of his analysis on a sample of target firms.We replicate our analyses based on an estimated runup over a 40-day window (results not shown to conserve space), and ourresults are robust.

6.1. Variation in information leakage caused by industry differences

Table 6 displays the mean runup in the (−50,−1) day window for the targets categorized by tech industry (per panel) withinthe tech sector. Results show much variation in mean runup among tech industries, which suggests that the level of privateinformation released prior to mergers varies by types of tech targets. When applying the market model to the Positive RunupSample, the mean runup is highest for the tech targets in the computers industry (35.313%) and relatively low for tech targets inthe drug industry (28.501%) and the category of unclassified tech targets (27.664%). Similar results are derived from applying theFama–French model to derive the runup.

When applying the market model to the Converted Runup Sample, the mean runup is highest for the computers industry(14.211%) and is relatively low for the drugs industry (7.222%) and the category of unclassified tech targets (9.934%). Similarresults are found when applying the Fama–French model to derive the runup.

Table 6Runup by tech groups.

Panel A—Positive sample Panel B—Converted sample

Tech groups Market-model Fama–French 3-factor model Tech groups Market-model Fama–French 3-factor model

Business service 31.309% 29.772% Business Service 10.465% 10.254%(N = 135) (25.017%) (22.237%) (N = 245) (10.056%) (8.130%)Chips 27.912% 25.877% Chips 12.539% 11.730%(N = 82) (21.219%) (20.523%) (N = 135) (9.372%) (9.046%)Computers 35.313% 35.641% Computers 14.211% 14.706%(N = 68) (28.460%) (30.583%) (N = 116) (11.426%) (10.557%)Drugs 28.501% 28.927% Drugs 7.222% 7.002%(N = 44) (24.924%) (22.498%) (N = 82) (9.596%) (8.657%)Other 27.664% 27.317% Others 9.934% 10.106%(N = 204) (20.734%) (21.888%) (N = 255) (6.358%) (6.459%)

This table reports the mean and median (in parentheses) of the target's runup by each tech group. Runup is measured by the target's CARs in the (−50, −1) daywindow estimated from the market model and from the Fama–French 3-factor model alternatively. The estimation period is (−270, −65) days prior to theannouncement date of the merger. The positive sample includes only positive runup observations. The converted sample includes positive runup observationsand negative runup observations that are set to 0.

46 J. Madura, T. Ngo / Journal of High Technology Management Research 25 (2014) 36–53

6.2. Variation in leakage associated with tech characteristics

To offer more insight on the underlying characteristics of tech targets that can influence the level of private information, weestimate the runup in the (−50, −1) day window for quartiles of observable firm-specific characteristics that are measured bycontinuous variables. Results for runup estimated frommarket model and from Fama–French 3-factor model are reported in Table 7,Panel A and Panel B, respectively. For the Positive Runup Sample, the runup is inversely related to target size, and intangible assetratio, and is positively related to the target's risk andM&A activity (at the .10 level). The results show no relationship between runupand the other characteristics. For the Converted Runup Sample, the runup is inversely related to the target's size (at the .10 level),positively related to the target's risk, and not significantly related to the other variables.

6.3. Variation in leakage associated with government initiatives

Wemeasure themean level of private information (asmeasured by runup) for particular timeperiods to determine if the governmentinitiatives could have influenced the level of private information leaked about tech firms prior tomerger announcements. Results for eachof the three government initiatives are displayed in Table 8, separately for the Positive Runup Sample and the Converted Runup Sample.We report the t-statistics and the non-parametric Wilcoxon statistics from the test for the differences in mean and median runup,respectively.

Table 8 shows the results for runup in the (−50,−1) day window estimated from themarket model (in Panel A) and from theFama–French 3-factor model (in Panel B). The mean runup of tech firms prior to merger announcements is higher in the period

Table 7Runup by quartiles of target characteristics.

Sample Runup by quartiles of SIZE RD INTANG CAPX RISK DEBT VOLUME M&A ACTIVITY

Panel A—Runup estimated from market modelPositive Q1 37.050% 29.277% 35.454% 31.489% 18.081% 29.968% 30.933% 28.240%

Q 2 32.001% 27.558% 26.824% 26.442% 23.376% 26.870% 29.071% 25.124%Q 3 26.330% 27.916% 27.520% 26.496% 29.913% 29.242% 27.194% 32.407%Q4 24.037% 35.058% 24.753% 36.301% 48.344% 33.735% 32.625% 34.419%Q4–Q1 −13.013% 5.781% −10.701% 4.812% 30.263% 3.767% 1.692% 6.179%t-stat −3.86*** 1.64 −3.89*** 1.26 9.08*** 1.03 0.48 1.65*Wilcoxon −2.64*** 0.81 −3.08*** 1.28 7.92*** 1.52 0.7 1.74*

Converted Q1 13.435% 10.501% 12.188% 10.788% 7.843% 11.247% 7.448% 10.185%Q 2 12.067% 11.655% 9.218% 9.811% 8.483% 9.201% 16.080% 8.569%Q 3 10.386% 10.463% 10.310% 13.190% 10.221% 10.514% 9.543% 12.724%Q4 7.746% 10.745% 9.665% 11.183% 16.820% 12.403% 10.300% 12.827%Q4–Q1 −5.689% 0.244% −2.523% 0.395% 8.977% 1.156% 2.852% 2.642%t-stat −1.7* 0.07 −0.95 0.11 2.57** 0.35 0.84 0.78Wilcoxon −1.12 −0.12 −0.48 0.07 2.02** 0.21 1.08 0.92

Panel B—Runup estimated from Fama–French 3-factor modelPositive Q1 36.401% 29.463% 34.400% 31.126% 17.093% 30.638% 30.494% 27.684%

Q 2 31.320% 26.359% 26.014% 25.434% 23.432% 24.969% 27.671% 25.442%Q 3 25.207% 26.727% 26.796% 25.519% 29.556% 27.539% 27.051% 31.145%Q4 23.256% 34.004% 24.224% 35.005% 46.366% 33.417% 31.346% 32.862%Q4–Q1 −13.145% 4.541% −10.176% 3.879% 29.273% 2.779% 0.852% 5.178%t-stat −3.79*** 1.25 −3.6*** 1.01 8.45*** 0.76 0.24 1.35Wilcoxon −3.02*** 0.20 −3.02*** 1.06 7.35*** 1.11 0.47 1.28

Converted Q1 13.173% 10.907% 11.941% 11.105% 7.229% 12.371% 7.515% 9.843%Q 2 12.388% 11.230% 8.912% 8.495% 9.362% 8.255% 15.227% 9.117%Q 3 10.041% 9.851% 10.444% 13.052% 10.223% 10.022% 9.455% 12.414%Q4 7.967% 11.001% 9.673% 11.774% 16.179% 12.347% 10.801% 12.602%Q4–Q1 −5.206% 0.094% −2.268% 0.669% 8.950% −0.024% 3.286% 2.759%t-stat −1.56 0.03 −0.86 0.19 2.57** −0.01 0.99 0.81Wilcoxon −1.23 −0.16 −0.48 0.04 1.89* 0.09 1.08 0.65

This table reports the mean target's runup for each quartile of target's characteristics. Runup is measured by the target's CARs in the (−50, −1) day windowestimated from the market model (in Panel A) and from the Fama–French 3-factor model (in Panel B), respectively. The estimation period is (−270, −65) daysprior to the announcement date of the merger. The positive sample includes only positive runup observations. The converted sample includes positive runupobservations and negative runup observations that are set to 0. SIZE is the natural logarithm of the target's market capitalization at the end of the fiscal yearpreceding the announcement date. RD is the target's ratio of R&D expense to sale at the end of the fiscal year preceding the announcement date. INTANG is thetarget's ratio of intangible assets to total assets at the end of the fiscal year preceding the announcement date. CAPX is the target's ratio of capital expenditure tosales at the end of the fiscal year preceding the announcement date. RISK is the idiosyncratic risk of the target in the estimation period from (−270, −65) daywindow. CAPX is the target's ratio of capital expenditure to sales at the end of the fiscal year preceding the announcement date. DEBT is the target's ratio of totaldebt to total assets at the end of the fiscal year preceding the announcement date. VOLUME is the average ratio of the trading volume to number of outstandingshares in the estimation period. M&A ACTIVITY is measured by the natural logarithm of the deal values of all mergers in the same industry in the preceding year.T-statistics and non-parametric Wilcoxon statistics for the difference between the highest quartile and the lowest quartile are reported. *, ** and *** indicate thesignificance levels of 10%, 5% and 1%, respectively.

Table 8Impact of government initiatives on runup.

Government initiatives Positive sample Converted sample

Before After Difference Before After Difference

Mean Mean Mean Diff t-stat Wilcoxon Mean Mean Mean Diff t-stat Wilcoxon

Panel A—Runup is estimated from market modelRegulation FD 31.045% 28.895% −2.150% −0.91 −1.67* 12.918% 9.065% −3.853% −1.72* −2.24**

(N = 237) (N = 230) (N = 384) (N = 443)SOX 33.525% 24.793% −8.733% −3.8*** −4.02*** 13.454% 7.357% −6.096% −2.78*** −2.73***

(N = 276) (N = 191) (N = 476) (N = 537)Galleon 30.750% 20.122% −10.628% −3.3*** −2.52** 11.315% 5.331% −5.984% −1.98* −1.65*

(N = 432) (N = 35) (N = 767) (N = 66)

Panel B—Runup is estimated from Fama–French 3-factor modelRegulation FD 29.678% 28.613% −1.065% −0.44 −1.16 12.545% 9.211% −3.334% −1.50 −2.13**

(N = 237) (N = 230) (N = 384) (N = 443)SOX 32.349% 24.498% −7.851% −3.34*** −3.72*** 13.303% 5.199% −5.961% −2.71*** −2.79***

(N = 276) (N = 191) (N = 476) (N = 537)Galleon 29.924% 19.426% −10.499% −3.16*** −2.42** 11.298% 4.351% −6.947% −2.25** −1.89*

(N = 432) (N = 35) (N = 767) (N = 66)

This table reports the results from testing the impact of government initiatives on target's runup. Runup ismeasuredby the target's CARs in the (−50,−1) daywindowestimated from themarketmodel (in Panel A) and from the Fama–French 3-factormodel (in Panel B), respectively. The estimation period is (−270,−65) days prior tothe announcement date of themerger. The positive sample includes only positive runup observations. The converted sample includes positive runup observations andnegative runup observations that are set to 0. T-statistics and non-parametric Wilcoxon statistics for the significant difference in themean andmedian are reported. *,** and *** indicate the significance levels of 10%, 5% and 1%, respectively.

47J. Madura, T. Ngo / Journal of High Technology Management Research 25 (2014) 36–53

before Regulation FD than after Regulation FD. However, the difference in mean runup is not consistently significant acrossmodels.

Regarding the Sarbanes–Oxley (SOX) Act, the mean runup of tech firms prior to merger announcements in the Positive RunupSample is about 8% lower in the period after SOX, and the difference in means is significant, regardless of the model used tomeasure runup. For the Converted Runup Sample, the mean runup of tech firms prior to merger announcements is about 6% lowerin the period after SOX, and the difference is significant, regardless of the model used to measure runp.

Regarding the Galleon case, the mean runup of tech firms prior to merger announcements in the Positive Runup Sample isabout 10% lower in the period after the Galleon case, and the difference in means is significant, regardless of the model used tomeasure runup. For the Converted Runup Sample, the mean runup of tech firms prior to merger announcements is about 6% (7%)lower in the period after the Galleon case when applying the market (Fama–French) model to measure runup, and the differenceis significant.

6.4. Multivariate analysis of information leakage of tech targets

In order to assess how the level of private information is related to each tech characteristic and government initiative whilecontrolling for other characteristics, we apply a multivariate analysis. The dependent variable RUNUP is the stock price runup overthe (−50, −1) day window estimated from the market model (and alternatively from the Fama–French 3-factor model). Todetermine whether the level of private information leaked prior to the merger announcement is related to tech firm and timeseries characteristics, we apply the following multivariate model:

RUNUPi ¼ α þ β1SIZE þ β2RDþ β3INTANGþ β4CAPX þ β5RISK þ β6DEBTþ β7CALIFORNIAþ β8FIN þ β9VOLUME þ β10OPTIONþ β11M&AACTIVITY þ β12FDþ β13SOX þ β14GALLEON þ εi

6.4.1. Results for the Positive Runup SampleResults from applying multivariate models to the Positive Runup Sample are disclosed in Table 9. Panel A provides results

based on using the market model to measure RUNUP, while Panel B provides results based on using the Fama–French model tomeasure RUNUP. For each panel, results from applying 3 different multivariate models are reported. Each model isolates theeffects of a specific government initiative while also accounting for tech and trading characteristics.

The models in Panel A explain between 17% and 18% of the variation in the runup within the sample, and the F-statistic issignificant for each model. As shown in Panel A, the coefficient of INTANG is negative and significant for two of the three models,which offers some evidence that information leakages are more pronounced for tech targets with a smaller level of intangibleassets. The coefficient of CAPX is positive and significant in one model.

The results for the remaining tech and trading characteristics are consistent across the three models. The coefficient of RISK ispositive and significant, consistent with the univariate analysis. This supports the hypothesis that the level of private information

Table 9Multivariate analyses of runup in (−50, −1) day window for the Positive RunUp Sample.

Model 1 Model 2 Model 3

Param. Est. t-stat Param. Est. t-stat Param. Est. t-stat

Panel A—Runup is estimated from the market modelIntercept 0.236 3.226*** 0.236 3.279*** 0.239 3.331 ***SIZE −0.016 −1.001 −0.014 −0.906 −0.016 −1.020RD −0.020 −1.534 −0.016 −1.398 −0.021 −1.698INTANG −0.132 −1.977 * −0.078 −1.397 −0.134 −2.526 **CAPX 0.077 1.554 0.062 1.427 0.081 1.730 *RISK 1.323 3.689*** 1.279 3.76*** 1.294 3.619 ***DEBT 0.087 2.432 ** 0.090 2.464 ** 0.088 2.426 **CALIFORNIA 0.009 0.331 0.014 0.575 0.008 0.318FIN −0.010 −0.323 −0.028 −0.752 −0.010 −0.310VOLUME 2.941 2.299 ** 3.023 2.438 ** 2.917 2.260 **OPTION 0.021 0.655 0.024 0.759 0.022 0.712M&A ACTIVITY 0.014 1.003 0.015 1.001 0.011 0.751FD −0.017 −0.533SOX −0.066 −2.217 **GALLEON −0.043 −2.896***F-stat 14.47*** 19.63*** 25.57***Adj. R-squared 0.174 0.184 0.175N 445 445 445

Panel B—Runup is estimated from the Fama–French 3-factor modelIntercept 0.254 3.76*** 0.256 3.858 *** 0.260 3.940 ***SIZE −0.020 −1.383 −0.019 −1.295 −0.021 −1.390RD −0.029 −2.494 ** −0.025 −2.460 ** −0.029 −2.614 **INTANG −0.140 −1.927 * −0.085 −1.410 −0.125 −2.082 **CAPX 0.102 2.331 ** 0.086 2.287 ** 0.101 2.453 **RISK 1.202 3.77 *** 1.177 3.828 *** 1.184 3.685 ***DEBT 0.085 2.324 ** 0.087 2.261 ** 0.085 2.248 **CALIFORNIA −0.003 −0.111 0.002 0.102 −0.003 −0.100FIN −0.003 −0.111 −0.021 −0.565 −0.008 −0.261VOLUME 3.801 2.82 *** 3.828 2.834 *** 3.724 2.734 **OPTION 0.022 0.670 0.026 0.795 0.025 0.794M&A ACTIVITY 0.009 0.665 0.011 0.749 0.008 0.556FD 0.002 0.074SOX −0.052 −1.642GALLEON −0.047 −2.793***F-stat 14.01*** 12.54*** 13.85***Adj. R-squared 0.161 0.167 0.163N 445 445 445

This table reports the results from the cross-sectional analyses of runup using the positive runup sample. The dependent variable is the target runup in the (−50,−1)daywindowestimated from themarketmodel (in Panel A) and from the Fama–French 3-factormodel (in Panel B), respectively. The estimation period is (−270,−65)days prior to the announcement date of the merger. The positive sample includes only positive runup observations. The converted sample includes positive runupobservations and negative runup observations that are set to 0. SIZE is the natural logarithm of the target's market capitalization at the end of the fiscal year precedingthe announcement date. RD is the target's ratio of R&D expense to sale at the end of the fiscal year preceding the announcement date. INTANG is the target's ratio ofintangible assets to total assets at the end of the fiscal year preceding the announcement date. CAPX is the target's ratio of capital expenditure to sales at the end of thefiscal year preceding the announcement date. RISK is the idiosyncratic risk of the target in the estimation period from (−270,−65) day window. CAPX is the target'sratio of capital expenditure to sales at the end of the fiscal year preceding the announcement date. DEBT is the target's ratio of total debt to total assets at the end of thefiscal year preceding the announcement date. CALIFORNIA is the dummy variable for targets incorporated in California. FIN is the dummy variable formergers inwhichthe acquirers use borrowed funds to finance the deal. OPTION is set equal to 1 for targets that have traded options, and zero otherwise. VOLUME is the average ratio ofthe trading volume to number of outstanding shares in the estimation period. M&A ACTIVITY is measured by the natural logarithm of the deal values of all mergers inthe same industry in the preceding year. FD is the dummy for period after the introduction of Regulation FD. SOX is the dummy for period after the introduction ofSarbanes–Oxley Act. GALLEON is thedummy for period after the inception of Galleon case. T-statistics are calculated basedupon standard errors corrected for clusteringeffects by year. *, ** and *** indicate the significance levels.

48 J. Madura, T. Ngo / Journal of High Technology Management Research 25 (2014) 36–53

prior to a merger announcement is more pronounced for targets that exhibit a higher level of risk. The coefficient of DEBT is positiveand significant, which supports our hypothesis that the level of private information leaked is larger formore highly levered tech firms.The coefficient of VOLUME is positive, which implies that the level of private information leaked is positively related to the tradingvolume of tech targets. Informed traders can more easily avoid signaling with their trades based on information leakages when thetrading volume is greater. The remaining tech firm or trading variables in the multivariate model are not significant.

Model 1 shows the results from testing the effect of Regulation FD, while accounting for other characteristics. The coefficientfor FD is negative but not significant, offering no support for our hypothesis that the level of private information changes inresponse to Regulation FD.

Model 2 shows the results from testing the effect of the Sarbanes–Oxley Act, while accounting for other characteristics. Thecoefficient for SOX is negative and significant, which is consistent with the univariate results, and supports our hypothesis thatthe level of private information prior to tech merger announcements reduces since SOX.

49J. Madura, T. Ngo / Journal of High Technology Management Research 25 (2014) 36–53

Model 3 shows the results from testing the effect of the Galleon case, while accounting for other characteristics. The coefficientfor Galleon is negative and significant, which supports our hypothesis that the level of private information reduces since theGalleon case.

Panel B shows that when using the Fama–French model to measure the dependent variable RUNUP, the multivariate modelsexplain about 16% of the variation, and all models are significant based on the F-statistics. The coefficient of RD is consistentlynegative and significant across models in Panel B, which suggests that tech targets with a smaller investment in research anddevelopment experience a larger runup. The coefficient of CAPX is consistently positive and significant in all models in Panel B,which suggests that tech targets that spendmore on capital expenditures experience a larger runup. The RISK, VOLUME, and DEBTvariables remain significant in all models in Panel B, consistent with the results shown in Panel A.

Table 10Multivariate analyses of runup in (−50, −1) day window for the Converted RunUp Sample.

Independent variables Converted sample

Model 1 Model 2 Model 3

Param. Est. t-stat Param. Est. t-stat Param. Est. t-stat

Panel A—Runup is estimated from the market modelIntercept 0.149 2.429 ** 0.148 2.291 ** 0.146 2.329 **SIZE −0.015 −1.363 −0.014 −1.312 −0.016 −1.472RD −0.003 −0.596 −0.002 −0.511 −0.003 −0.725INTANG −0.020 −0.248 0.007 0.096 −0.037 −0.469CAPX 0.020 1.038 0.017 0.986 0.024 1.192RISK 0.061 0.223 0.041 0.167 0.030 0.114DEBT 0.073 1.186 0.074 1.176 0.078 1.288CALIFORNIA 0.070 2.906 *** 0.073 3.284 *** 0.068 3.036 ***FIN 0.036 1.203 0.028 0.869 0.039 1.191VOLUME −2.474 −1.115 −2.410 −1.074 −2.272 −1.022OPTION 0.037 1.330 0.037 1.228 0.035 1.155M&A ACTIVITY 0.010 0.519 0.008 0.465 0.005 0.300FD −0.041 −1.377SOX −0.062 −2.610 **GALLEON −0.059 −2.570 **F-stat 3.51*** 3.46*** 3.91***Adj. R-squared 0.030 0.034 0.029N 786 786 786

Panel B—Runup is estimated from the Fama–French 3-factor modelIntercept 0.141 2.465 ** 0.140 2.359 ** 0.140 2.429 **SIZE −0.015 −1.323 −0.014 −1.273 −0.016 −1.450RD −0.003 −0.633 −0.002 −0.543 −0.004 −0.735INTANG −0.008 −0.096 0.024 0.322 −0.018 −0.223CAPX 0.017 0.810 0.013 0.717 0.020 0.939RISK 0.039 0.152 0.021 0.090 0.007 0.027DEBT 0.069 1.073 0.069 1.058 0.074 1.176CALIFORNIA 0.066 2.670 ** 0.069 3.083 *** 0.064 2.773 ***FIN 0.033 1.151 0.024 0.768 0.033 1.093VOLUME −1.171 −0.564 −1.127 −0.541 −0.978 −0.477OPTION 0.023 0.741 0.024 0.718 0.022 0.680M&A ACTIVITY 0.009 0.549 0.009 0.501 0.005 0.317FD −0.037 −1.488SOX −0.064 −2.630 **GALLEON −0.074 −3.224 ***F-stat 3.05*** 3.57*** 3.31***Adj. R-squared 0.024 0.029 0.025N 786 786 786

This table reports the results from the cross-sectional analyses of runup using the converted runup sample. The dependent variable is the target runup in the (−50,−1)daywindowestimated from themarketmodel (in Panel A) and from the Fama–French 3-factormodel (in Panel B), respectively. The estimation period is (−270,−65)days prior to the announcement date of the merger. The positive sample includes only positive runup observations. The converted sample includes positive runupobservations and negative runup observations that are set to 0. SIZE is the natural logarithm of the target's market capitalization at the end of the fiscal year precedingthe announcement date. RD is the target's ratio of R&D expense to sale at the end of the fiscal year preceding the announcement date. INTANG is the target's ratio ofintangible assets to total assets at the end of the fiscal year preceding the announcement date. CAPX is the target's ratio of capital expenditure to sales at the end of thefiscal year preceding the announcement date. RISK is the idiosyncratic risk of the target in the estimation period from (−270,−65) day window. CAPX is the target'sratio of capital expenditure to sales at the end of the fiscal year preceding the announcement date. DEBT is the target's ratio of total debt to total assets at the end of thefiscal year preceding the announcement date. CALIFORNIA is the dummy variable for targets incorporated in California. FIN is the dummy variable formergers inwhichthe acquirers use borrowed funds to finance the deal. VOLUME is the average ratio of the trading volume to number of outstanding shares in the estimation period.OPTION is set equal to 1 for targets that have tradedoptions, and zero otherwise.M&AACTIVITY ismeasured by the natural logarithmof the deal values of all mergers inthe same industry in the preceding year. FD is the dummy for period after the introduction of Regulation FD. SOX is the dummy for period after the introduction ofSarbanes–Oxley Act. GALLEON is thedummy for period after the inception of Galleon case. T-statistics are calculated basedupon standard errors corrected for clusteringeffects by year. *, ** and *** indicate the significance levels.

50 J. Madura, T. Ngo / Journal of High Technology Management Research 25 (2014) 36–53

Regarding the government initiatives, Model 1 shows that the coefficient for FD is positive but not significant, offering nosupport for the hypothesis that the level of private information changes in response to Regulation FD. These results differ from theunivariate analysis that shows a lower runup after Reg FD, but are consistent with the multivariate results in Panel A.

Model 2 shows that the coefficient for SOX is negative but falls short of being significant, although it is significant in Panel A.Model 3 shows that the coefficient for GALLEON is negative and significant, which is consistent with the univariate results and theresults in Panel A, and supports our hypothesis that the level of private information declines following the Galleon case.

6.4.2. Results for the Converted Runup SampleResults from applying multivariate models to the Converted Runup Sample are disclosed in Table 10. Panel A provides results

based on using the market model to measure RUNUP, while Panel B provides results based on using the Fama–French model tomeasure RUNUP. The models in Table 10 explain about 3% of the variation in RUNUP among the Converted Runup Sample, versus16% to 18% of the variation in RUNUP among the Positive Runup Sample in Table 9, but are still significant based on the F-statistics.

Some of themultivariate results for the Converted Runup Sample in Table 10 are distinctly different from the results for the PositiveRunup Sample in Table 9. As shown in Panel A of Table 10, the coefficient of CALIFORNIA is consistently positive and significant in allmodels in Panel A, which suggests that the level of private information prior to a merger announcement is more pronounced for techtargets that are located in California. The other variables that represent the tech firmor trading characteristics are not significant in any ofthe models.

Regarding the variables representing government initiatives, the coefficient for FD (in Model 1) is negative but not significant.When applying Model 2, the coefficient for SOX is negative and significant, which supports our hypothesis that the level of privateinformation changes in response to SOX. When applying Model 3, the coefficient for GALLEON is negative and significant, whichsupports our hypothesis that the level of private information changes in response to Galleon.

In Panel B of Table 10 where dependent variable RUNUP is estimated from the Fama–French model, the results are very similarto those in Panel A, with the same variables significant and in the same direction in every model. Most importantly, Model 2shows that the coefficient of the SOX variable is negative and significant, which reinforces the previous findings of a lower RUNUPsince the Sarbanes–Oxley Act. In addition, Model 3 shows that the coefficient of the GALLEON variable is negative and significant,which reinforces the previous finding of a lower RUNUP since the inception of the Galleon case. While the results for many controlvariables are distinctly different in the Converted Runup Sample (in Table 10) than in the Positive Runup Sample (in Table 9), thereduction in information leakage prior to merger announcements since the Sarbanes–Oxley Act and Galleon case is documentedin both samples.

7. How informed trading returns vary among targets

We also assess how informed trading returns vary among tech targets. For this purpose, we first apply a simulated tradingstrategy to estimate the return that could have been earned by informed traders who have private information about an impendingmerger. Our simulated trading strategy is based on the assumption that informed traders purchase shares at the beginning of therunup window (day −50) and sell their shares one day after the announcement (day +1). We use this window in an attempt toensure that we fully capture the information leakage prior to the publicized bid. The measurement reflects the cumulativeabnormal return CAR (−50, +1).

Since we are simulating a strategy in which traders capitalize on private information that is leaked, we do not want to includeany tech targets for which there is no information leakage. Therefore, we focus only on the Positive Runup Sample, because thatsample only includes tech targets for which there is evidence of an information leakage.

We apply our multivariate model to explain the variation in informed trading returns, as measured over the (−50, +1)window.Since the share price response to the merger announcement is very large, the behavior of informed trading returns (which capturesthis announcement effect) could be substantially different from the behavior of the runup. Thus, the results derived when applying amultivariate model to the informed trading returns could deviate substantially from results derived when applying a multivariatemodel to explain variation in the information leakage prior to the merger announcement (as measured by the runup).

Threemultivariatemodels are applied,with eachmodel isolating a particular government initiative. Results are disclosed in Table 11.Themodels explain about 50% of the variation in informed trading returns among tech targets. Regarding the technology characteristics,the coefficient of RISK is consistently positive and significant across the models in Panel A. In addition, the coefficient of VOLUME isconsistently positive and significant across themodels in Panel A. Recall that these variables are also positive and significant in explainingthe variation in information leakage in the Positive Runup Sample (as shown in Table 9).

Regarding the government initiatives, Model 1 shows that the coefficient for FD is not significant, implying no evidence oflower informed trading returns on tech target firms after Regulation FD. Model 2 shows that the SOX variable is not significant,implying no evidence of lower informed trading returns on tech target firms after the Sarbanes–Oxley Act. Model 3 shows that theGALLEON variable is positive and significant, which implies a higher informed trading returns on tech target firms after theGalleon case. Thus, while private information leakages of tech targets have declined since the Sarbanes–Oxley Act and Galleoncase, informed trading returns have not declined.

We also applied multivariate models to explain the variation in informed trading returns in the Positive Runup Sample asmeasured by the Fama–Frenchmodel in place of the market model. Results are disclosed in Panel B. Themodels explain about 48%of the variation in informed trading returns. The results are very similar to those in Panel A, except that the coefficient of the FDvariable is positive and significant at the .10 level in Model 1. These results reinforce our general conclusions that while private

Table 11Multivariate analyses of trading profit.

Panel A—Trading profit is estimated from the market model Panel B—Trading profit is estimated from the Fama–French 3-factor model

Model 1 Model 2 Model 3 Model 1 Model 2 Model 3

Param. Est. t-stat Param. Est. t-stat Param. Est. t-stat Param. Est. t-stat Param. Est. t-stat Param. Est. t-stat

Intercept 0.318 5.023 *** 0.321 5.097 *** 0.314 4.873 *** 0.339 5.027 *** 0.343 5.113 *** 0.337 4.913 ***SIZE −0.006 −0.452 −0.006 −0.459 −0.005 −0.369 −0.011 −0.940 −0.011 −0.946 −0.010 −0.830RD −0.005 −0.534 −0.004 −0.517 −0.002 −0.302 −0.013 −1.636 −0.013 −1.617 −0.009 −1.278INTANG −0.073 −0.953 −0.069 −0.868 −0.055 −0.891 −0.083 −0.976 −0.077 −0.892 −0.046 −0.673CAPX 0.018 0.559 0.017 0.533 0.009 0.322 0.041 1.405 0.040 1.371 0.027 1.017RISK 4.144 10.514 *** 4.188 10.925 *** 4.200 10.732 *** 4.002 9.260 *** 4.066 9.877 *** 4.068 9.593 ***DEBT 0.088 1.374 0.085 1.284 0.086 1.287 0.088 1.393 0.082 1.267 0.084 1.266CALIFORNIA 0.014 0.461 0.014 0.428 0.016 0.519 0.003 0.099 0.002 0.059 0.006 0.203FIN 0.045 1.049 0.046 0.981 0.042 1.030 0.049 1.206 0.051 1.102 0.041 1.052VOLUME 3.981 2.290 ** 3.854 2.172 ** 3.975 2.244 ** 4.933 2.698 ** 4.747 2.539 ** 4.872 2.610 **OPTION −0.019 −0.390 −0.016 −0.335 −0.018 −0.366 −0.018 −0.398 −0.014 −0.318 −0.015 −0.316M&A ACTIVITY 0.007 0.450 0.011 0.664 0.013 0.817 0.002 0.104 0.007 0.424 0.010 0.618FD 0.045 1.186 0.065 1.845 *SOX 0.032 0.905 0.048 1.388GALLEON 0.066 3.766 *** 0.062 3.883 ***F-stat 27.21*** 27.91*** 27.31*** 29.73*** 23.23*** 24.24***Adj. R-squared 0.502 0.501 0.501 0.481 0.478 0.477N 445 445 445 445 445 445

This table reports the results from the cross-sectional analyses of runup using the positive runup sample. The dependent variable is the trading profit in the (−50, +1) day window estimated from themarket model (in Panel A)and from the Fama–French 3-factor model (in Panel B), respectively. The estimation period is (−270,−65) days prior to the announcement date of the merger. SIZE is the natural logarithm of the target's market capitalization atthe end of the fiscal year preceding the announcement date. RD is the target's ratio of R&D expense to sale at the end of the fiscal year preceding the announcement date. INTANG is the target's ratio of intangible assets to total assetsat the end of the fiscal year preceding the announcement date. CAPX is the target's ratio of capital expenditure to sales at the end of the fiscal year preceding the announcement date. RISK is the idiosyncratic risk of the target in theestimation period from (−270,−65) day window. CAPX is the target's ratio of capital expenditure to sales at the end of the fiscal year preceding the announcement date. DEBT is the target's ratio of total debt to total assets at theend of the fiscal year preceding the announcement date. CALIFORNIA is the dummy variable for targets incorporated in California. FIN is the dummy variable for mergers in which the acquirers use borrowed funds to financethe deal. VOLUME is the average ratio of the trading volume to number of outstanding shares in the estimation period. OPTION is set equal to 1 for targets that have traded options, and zero otherwise.M&AACTIVITY ismeasured bythe natural logarithm of the deal values of all mergers in the same industry in the preceding year. FD is the dummy for period after the introduction of Regulation FD. SOX is the dummy for period after the introduction of Sarbanes–Oxley Act. GALLEON is the dummy for period after the inception of Galleon case. T-statistics are calculated based upon standard errors corrected for clustering effects by year. *, ** and *** indicate the significance levels.

51J.M

adura,T.Ngo

/JournalofH

ighTechnology

Managem

entResearch

25(2014)

36–53

52 J. Madura, T. Ngo / Journal of High Technology Management Research 25 (2014) 36–53

information leakages of tech targets have declined since the Sarbanes–Oxley Act and Galleon case, informed trading returns havenot declined.

We reconcile our multivariate analysis of informed trading returns with our multivariate analysis of the information leakage.The increased share price response of tech targets to merger announcements at least offsets the reduction in information leakagesince the Sarbanes–Oxley Act, and overwhelms the reduction in information leakage since the Galleon case. Therefore, thereduction in information leakages of tech targets since the Sarbanes–Oxley Act and the Galleon case does not cause a reduction inpotential informed trading profits.

8. Summary and recommendations for future research

Tech firms are commonly subject to private information leakages, as a result of expert networks or other means by whichinsiders spread information. We investigate the level of private information surrounding the tech target firms that reachesinvestors prior to public merger announcements. The stock price runup prior to public merger announcements can reveal thelevel of private information (if any) that is released.

Our study is unique because it focuses on firms in the tech sector, and therefore gives attention to how technology or othercharacteristics can attract attention from informed traders. Our study is also distinguished from other studies that assess insidertrading because of the manner in which we measure the leakage of private information. Specifically, we follow Jarrell and Poulsen(1989) and apply two alternative methods to screen our sample. First, we compile a Positive Runup Sample by excluding techtargets that experience a negative runup, because a negative measurement of leakage implies less than zero private informationleaked, and could be misleading. Second, we compile an alternative sample (Converted Runup Sample) in which we retain all techtargets even if they experience a negative runup before their corresponding merger announcements, but convert the negativerunup to zero. We also assess how trading returns from capitalizing on information leakages vary among tech targets.

We find much variation in the level of private information among tech targets. In fact, more than one-third of the tech targetsexperience a negative runup. When excluding these firms for which there is no evidence of an information leakage, the meanlevel of private information for those targets in which there is evidence of a leakage prior to merger announcements is muchlarger than has been estimated for targets on average by other studies.

When assessing the Positive Runup Sample, we find that tech targets are subject to greater leakages of private informationwhen they have low investment in research and development, high capital expenditures, high risk, high debt, and high tradingvolume. These firms may be subject to more asymmetric information and more mispricing, which allows for larger stock pricegains from using expert networks or other means to obtain private information about impending mergers. When assessing theConverted Runup Sample, we find that tech targets are subject to greater leakages of private information when they have highrisk or are based in California.

Our multivariate analysis applied to the Positive Runup Sample and the Converted Runup Sample confirms the reduction in privateinformation leakages prior to merger announcements following the Sarbanes–Oxley Act, and following the Galleon case, whilecontrolling for the other characteristics. These results support our hypotheses that these government initiatives can discourage therelease of private information to informed traders, or the use of that private information by traders. However, we find no reduction ininformed trading returns from capitalizing on private information leakages prior to merger announcements following the Sarbanes–OxleyAct orGalleon case.We reconcile these results by considering that the informed trading returns as defined in our simulated tradingstrategy are composed of the target's information leakage (as measured by runup) and its share price response to the mergerannouncement. Since the reduction in the information leakage prior to merger announcements following the Sarbanes–Oxley Act andGalleon case is offset by the increased share price response to the merger announcement, informed trading returns have not declined.

While the government has effectively reduced the private information leakage about tech targets, it has not reduced thepotential reward to those informed traders who still illegally use private information. However, its efforts to enforce regulationsthat restrict the use of private information have accentuated the risk (of criminal or civil penalties) faced by any informed traderswho still illegally pursue the potential reward.

The battle between the government and traders with illegal insider trading has not ended. Opportunities for illegal insidertrading still exist if participants within expert networks are willing to take the risk of disclosing inside information. And the stakesremain high because of the intense competition among hedge funds to achieve higher returns and attract more capital. InNovember 2013, SAC Capital Advisors (a large hedge fund) pleaded guilty to criminal fraud charges because it did allowed or didnot prevent insider trading activity by its managers. In this landmark case, the fund will have to pay a penalty of $1.8 billion, andwill have to close its fund to outside investors. [See http://articles.latimes.com/2013/nov/08/business/la-fi-sac-plea-20131108].Future research can assess whether this case discouraged insider trading to a greater degree, and reduced information leakages.

Much of the illegal insider trading in recent years was detected by wiretaps and by guilty parties implicating others.Participants of expert networks who remain willing to disclose inside information in the future may use other methods tocommunicate their inside information so that they are not detected by wiretaps. This could present a major challenge to thegovernment, and could cause information leakages to increase over time. Future research is warranted to measure intertemporalshifts in information leakages as a means of determining whether the government maintains control over illegal insider trading.

While our study focused on insider trading prior to tech mergers, another common form of illegal activity in the technologysector is insider trading prior to technological news. Many tech firms rely heavily on research and development, where there canbe information leakages from expert networks prior to the publicized results of successful (or failed) medical research trials. Theappeal of these events to hedge funds is the substantial change in valuation in biotech firms that may occur in response to the

53J. Madura, T. Ngo / Journal of High Technology Management Research 25 (2014) 36–53

result of medical trials. Future research should be focused on these events as well to assess whether the publicized insider tradingcases have reduced the information leakages associated with these events over time.

References

Acharya, V. V., & Johnson, T. C. (2010). More insiders, more insider trading: Evidence from private equity buyouts. Journal of Financial Economics, 98, 500–523.Admati, A. R., & Pfleiderer, P. (1988). A theory of intraday trading patterns: Volume and price variability. Review of Financial Studies, 1, 3–40.Ahuja, G., & Katila, R. (2001). Technological acquisitions and the innovation performance of acquiring firms: A longitudinal study. Strategic Management Journal,

22, 197–220.Amihud, Y., & Mendelson, H. (1986). Asset pricing and the bid-ask spread. Journal of Financial Economics, 17, 223–249.Anderson, C. W., & Garcia-Feijoo, L. (2006). Empirical evidence on capital investment, growth options, and security returns. Journal of Finance, 61(1), 171–194.Andres-Alonso, P., Azofra-Palenzuela, V., & Fuente-Herrero, G. (2006). The real options component of firmmarket value: The case of the technological corporation.

Journal of Business Finance & Accounting, 33(1), 203–219.Anilowski, C., Feng, M., & Skinner, D. (2007). Does earnings guidance affect market returns? The nature and information content of aggregate earnings guidance.

Journal of Accounting and Economics, 44(1–2), 36–63.Barclay, M., & Warner, J. (1993). Stealth trading and volatility: Which trades move prices? Journal of Financial Economics, 34, 281–305.Barron, O. E., Byard, D., Kile, C., & Riedl, E. J. (2002). High-tech intangibles and analyst forecasts. Journal of Accounting Research, 40, 289–312.Bhattacharya, S., & Ritter, J. R. (1983). Innovation and communication: Signaling with partial disclosure. Review of Economic Studies, 50, 331–346.Brigida, M., & Madura, J. (2012). Sources of target stock price run-up prior to acquisitions. Journal of Economics and Business, 64, 185–198.Cassiman, B., Colombo, M., Garrone, P., & Veugelers, R. (2005). The Impact of M&A on the R&D process: An empirical analysis of the role of technological and

market relatedness. Research Policy, 34, 195–220.Chan, L. K. C., Lakonishok, J., & Sougiannis, T. (2001). The stock market valuation of research and development expenditures. Journal of Finance, 61, 2431–2456.Chira, I., & Madura, J. (forthcoming). Impact of the Galleon case on insider trading prior to mergers. Journal of Financial Research (forthcoming).Coad, A., & Rao, R. (2008). Innovation and firm growth in high-tech sectors: A quantile regression approach. Research Policy, 37, 633–648.Desyllas, P., & Hughes, A. (2008). Sourcing technological knowledge through corporate acquisition: Evidence from an international sample of high technology

firms. Journal of High Technology Management Research, 18, 157–172.Dittmar, A. K., & Dittmar, R. F. (2008). The timing of financing decisions: An examination of the correlation of financing waves. Journal of Financial Economics, 90,

59–83.Easley, D., & O'Hara, M. (2004). Information and the cost of capital. Journal of Finance, 59, 1553–1583.Eberhart, Allan C., Maxwell, W. F., & Siddique, A. R. (2004). An examination of long-term abnormal stock returns and operating performance following R&D

increases. Journal of Finance, 59, 623–650.Grimpe, C., & Hussinger, K. (2008). Pre-empting technology competition through firm acquisitions. Economics Letters, 100, 189–191.Grossman, S. J., & Hart, O. (1981). The allocational role of takeover bids in situations of asymmetric information. Journal of Finance, 36, 253–270.Gupta, A., & Misra, L. (1988). Illegal insider trading: Is it rampant before corporate takeovers? Financial Review, 23, 453–464.Harford, J. (2005). What drives merger waves? Journal of Financial Economics, 77, 529–560.Helwege, J., & Liang, N. (1996). Is there a pecking order? Evidence from a panel of IPO firms. Journal of Financial Economics, 40, 429–458.Higgins, M. J., & Rodriguez, D. (2006). The outsourcing of R&D through acquisitions in the pharmaceutical industry. Journal of Financial Economics, 80, 351–383.Ho, Y. K., Tjahjapranata, M., & Yap, C. M. (2006). Size, leverage, concentration, and R&D investment in generating growth opportunities. Journal of Business, 79(2),

851–876.Jabbour, A., Jalilvand, A., & Switzer, J. (2000). Pre-bid price runups and insider trading activity: Evidence from Canadian acquisitions. International Review of

Financial Analysis, 9, 21–43.Jarrell, G., & Poulsen, A. B. (1989). Stock trading before the announcements of tender offers: Insider trading or market anticipation? Journal of Law, Economics, and

Organization, 5, 225–248.Jayaraman, N., Frye, M. B., & Sabherwal, S. (2001). Informed trading around merger announcements: An empirical test using transaction volume and open interest

in options markets. Financial Review, 36, 45–74.Keown, A. J., & Pinkerton, J. M. (1981). Merger announcements and insider trading activity: An empirical investigation. Journal of Finance, 36, 855–869.King, M. (2009). Prebid runups ahead of Canadian takeovers: How big is the problem? Financial Management, 39, 699–726.Kohers, N., & Kohers, T. (2000). The Value Creation Potential of High-tech Mergers. May/June: Financial Analysts Journal 40–50.Kohers, N., & Kohers, T. (2001). Takeovers of technology firms: Expectations vs. reality. Financial Management, 30, 35–54.Leland, H. E., & Pyle, D. H. (1977). Informational asymmetries, financial structure, and financial intermediation. Journal of Finance, 32, 371–387.Libby, R., & Tan, H. (1999). Analysts' reactions to warning of negative earnings surprises. Journal of Accounting Research, 37, 415–435.Makri, M., Hitt, M., & Lane, P. (2010). Complement technologies, knowledge relatedness, and invention outcomes in high technology mergers and acquisitions.

Strategic Management Journal, 31, 602–628.Meulbroek, L. (1992). An empirical analysis of illegal insider trading. Journal of Finance, 47, 1661–1699.Oriani, R., & Sobrero, M. (2008). Uncertainty and the market valuation of R&D within a real options logic. Strategic Management Journal, 29, 343–361.Ornaghi, C. (2009). Mergers and innovation in big pharma. International Journal of Industrial Organization, 27, 70–79.Pindado, J., Queiroz, V., & Torre, C. (2010). How do firm characteristics influence the relationship between R&D and firm value. Financial Management, 39(2),

757–782.Powell, R. G., & Yawson, A. (2005). Industry aspects of takeovers and divestitures: Evidence from the UK. Journal of Banking & Finance, 29, 3015–3040.Schwert, G. W. (1996). Markup pricing in mergers and acquisitions. Journal of Financial Economics, 41, 153–192.Sevilir, M., & Tian, X. (2011). Acquiring innovation. Working Paper.Sinha, P., & Gadarowski, C. (2010). The efficacy of regulation fair disclosure. The Financial Review, 45(2), 331–354.Stahl, J. (2010). Mergers and sequential innovation: Evidence from patent citations. Working Paper.Trigeorgis, L. (1993). Real options and interactions with financial flexibility. Financial Management, 22(3), 202–224.Tsai, K. H., & Wang, J. C. (2005). Does R&D performance decline with firm size? A re-examination of terms of elasticity. Research Policy, 34, 966–976.Zuckerman, G., & Pulliam, S. (2010). How an SEC crackdown led to rise of ‘expert networks’. The Wall Street Journal (December 17, C1, P1).