Wolves at the Door: A Closer Look at Hedge Fund Activism

Abstract

Most investor coordination remains undisclosed. I provide empirical evidence on the extent and

consequences of investor coordination in the context of hedge fund activism, in which potential benefits

and costs from coordination are especially pronounced. In particular, I examine whether hedge fund

activists orchestrate “wolf packs,” i.e. groups of investors willing to acquire shares in the target firm before

the activist’s campaign is publicly disclosed via a 13D filing, as a way to support the campaign and

strengthen the activist’s bargaining position. Using a novel hand-collected dataset, I develop a method to

identify the formation of wolf packs before the 13D filing. I investigate two competing hypotheses: the

Coordinated Effort Hypothesis (wolf packs are orchestrated by lead activists to circumvent securities

regulations about “groups” of investors) and the Spontaneous Formation Hypothesis (wolf packs

spontaneously arise because investors independently monitor and target the same firms at about the same

time). A number of tests rule out the Spontaneous Formation Hypothesis and provide support for the

Coordinated Effort Hypothesis. Finally, the presence of a wolf pack is associated with various measures of

the campaign’s success.

1

1. Introduction

Coordination among investors is a critical aspect of external governance because it allows investors to

overcome the classic free-rider problem that limits their incentives to monitor and engage in activism (Shleifer and

Vishny 1986). At the same time, investor coordination has historically faced a number of legal impediments (Black

1990). Indeed, recent survey evidence indicates that while 59% of institutional investors acknowledge coordinating

their actions with other investors, most of those that do not coordinate blame it primarily on legal concerns related

to acting in concert (McCahery, Sautner and Starks 2016). Despite the importance of investor coordination,

academic research on its occurrence and effects is fairly limited because initiatives aimed at promoting investor

coordination are rarely publicly disclosed, partly because of those legal concerns (for exceptions, see Song and

Szewczyk 2003 and Dimson, Karakaz and Li 2015). Thus, researchers try to infer the occurrence of investor

coordination from observing shareholders’ behavior in settings in which coordination is likely to take place (Crane,

Koch and Michenaud 2017).

In this study, I identify hedge fund activism as one of such settings. Hedge fund activists typically hold a

relatively small stake in target firms (on average about 6%, see Brav, Jiang and Kim 2009). Thus, obtaining support

from other investors is critical for the activists’ ability to pressure the target firms to acquiesce to their requests.

One way to obtain such support is to coordinate a “wolf pack,” that is, to recruit other investors who are willing to

accumulate shares and support the campaign before it is publicly disclosed by the activist via a 13D filing (Briggs

2007, Coffee and Palia 2016). By joining the activist’s campaign before it is publicly disclosed, those investors may

benefit from the run-up in stock price typically accompanying such disclosures (Klein and Zur 2009, Brav et al.

2009). Also, continued coordination with the activist throughout the campaign could allow those investors to

profitably time their trades. In other words, the incentives to coordinate are particularly pronounced for both parties,

making hedge fund activism a powerful setting to examine whether some form of coordination takes place and its

consequences. On the other hand, this type of coordination, if detected, could results in the activists and the wolf

pack members being considered a “group” for securities regulation purposes, with a series of negative legal

ramifications that potentially could dampen the benefits of coordination (see footnote 1). Furthermore, even if one

found that some investors acquire shares in the target firm before the public disclosure of the campaign, such a

2

finding would not necessarily be evidence of coordination since it is possible that those investors simply share the

activists’ investment preferences. Thus, it is an empirical question whether wolfs packs are formed around activists’

campaigns and, if so, whether they are a symptom of investor coordination.

To address these questions, I first examine the trading pattern prior to 13D filings to identify the formation of

wolf packs, and then explore two competing hypotheses in relation to their formation. The first hypothesis, which

I label Spontaneous Formation Hypothesis, is that wolf packs arise spontaneously because investors independently

monitor and target the same firms at about the same time. Under this hypothesis, if an event (such as news that a

particular company is underperforming) is significant enough to trigger the reaction of one activist, it is likely to

also trigger the simultaneous reaction of other like-minded activists, resulting in the formation of a (non-

coordinated) wolf pack. This hypothesis is consistent with economic models such as Brav, Dasgupta and Matthews

(2018), and is often mentioned by activists to explain why under-performing companies may be targeted by several

activists (Hoffman, Viswanatha and Benoit 2015). An opposing hypothesis, which I label Coordinated Effort

Hypothesis, is that activist hedge funds orchestrate the formation of wolf packs (e.g., Briggs 2007, Coffee and Palia

2016). Under this hypothesis, the lead activist (the 13D filer) recruits other investors to join the campaign before

the 13D filing becomes public, effectively offering the jump in stock price that typically follows the 13D filing in

exchange for those investors’ support. This arrangement allows the lead activist to accumulate a larger percentage

of de facto ownership (and thus increase the chances of a successful campaign) without the need to purchase more

equity and without triggering adverse regulations based on ownership thresholds (Coffee and Palia 2016). In

particular, when multiple investors work together, securities regulations require these investors to report their

collective ownership as a “group.” An informally coordinated wolf pack allows the lead activist to bypass this

“group” requirement, and thus avoid crossing ownership thresholds that might trigger takeover defenses and insider

trading regulations.1

1 An example of takeover defenses triggered by the ownership threshold is the shareholder rights plan, commonly known as a poison pill, which typically gives shareholders (other than the activist) rights to buy more shares at a discount when an activist buys more than a certain percentage of the company’s shares. An

example of an insider trading regulation based on ownership level is the “short-swing profit” rule. An activist will be subject to this rule (Section 16(b) of the

Securities Exchange Act of 1934) when she acquires more than 10% of shares outstanding. It entitles shareholders to recover short-swing profits that are based on a purchase and sale or a sale and purchase, within six months, of the stock of a “reporting company.” While the average holding period of the activist is

usually longer than six months (see Brav et al. 2009), the activist may not want to lose the option to turn over the position quickly.

3

To investigate these hypotheses, I examine a sample of 1,922 activist hedge funds’ campaigns—all the

campaigns in the SharkRepellent database from 1998 through 2014 in which an activist filed Schedule 13D. The

first step is to identify the formation of wolf packs. To do so, I examine trading volume in the 60-day period prior

to the 13D filing. An especially important day during that window is the “trigger date”, that is, the date when the

activist takes an action (e.g., crossing the 5% ownership threshold) that triggers the requirement to file a 13D form

within the subsequent 10 days. Similar to other studies, I find that on the trigger date share turnover is about 325%

of normal trading volume. Since the trigger date is not publicly observable until the 13D is filed, the abnormal

turnover cannot be a reaction to public news of the activist’s campaign (see Figure 1). At the same time, it is not

necessarily evidence of wolf pack formation. As noted by Bebchuk, Brav, Jackson and Jiang (2013), it may simply

indicate that the lead activist accumulates most of its holdings on the trigger date. To shed light on the reasons for

the abnormal trading volume, I exploit the fact that activists must report on a Schedule 13D any purchase or sale of

the target firm’s equity daily for at least 60 days before the filing date, therefore including the trigger date. Using

this hand-collected information, I split the share turnover on the trigger date into two separate components: trades

by the 13D filer and trades by other investors. I find that, even after removing trades by the 13D filers, the remaining

average share turnover is about 250% of normal trading volume. Hence, the bulk of abnormal trading volume on

the trigger date reflects trades by other investors, consistent with the presence of a wolf pack.

Having developed a method to detect the presence of a wolf pack, I then investigate the potential mechanisms

behind its formation. In particular, I first examine the Spontaneous Formation Hypothesis, i.e. the notion that wolf

packs arise spontaneously because investors independently monitor and target the same firms at about the same

time. The fact that abnormal trading by other investors is concentrated on the trigger date casts doubts on this

hypothesis. Generally speaking, it seems unlikely that many investors would independently and spontaneously

decide to accumulate shares in the target firms on the same day—and even less likely that they would do so exactly

on the day the 13D filer crosses the 5% threshold (which is not publicly known).2 However, I consider three

scenarios consistent with this possibility. The first is that both the activist and the wolf pack members are responding

2 See online appendix. Although both the 13D filer and other investors start accumulating their position about 40 days before the trigger date, there is a sharp

increase in trading on the trigger date. Furthermore, this activity levels off immediately afterwards.

4

to the arrival of a news item that dramatically improves the expected profitability of investing in a firm, which

causes them to independently acquire shares of that firm at the same time. However, I find a similar level of

abnormal share turnover on the trigger date in a subset of 751 campaigns in which there is no Factiva news regarding

the target firm during the 11-day period prior to, and including, the trigger date. The second scenario deals with

order flow information, i.e. the possibility that the 13D filer’s trades on the trigger date cause other smaller investors

to trade simultaneously. To examine this scenario, I exploit the fact that some 13D filers do not trade on the trigger

date. This is because 13D filing is triggered not only by a change in ownership (crossing the 5% threshold), but also

by a change in intent, from “passive” to “active” (both triggers must be satisfied before an investor must file a 13D).

There is a subset of 351 campaigns in which the filers already owned more than 5% of the target firm before the

trigger date, but decided to switch their objective from passive to active on the trigger date, without any trading

activity.3 Even for this subsample, I continue to find a high level of share turnover by other investors on the trigger

date. Finally, prior literature suggests that hedge funds can create value by using their private information to create

new strategic plans or actions, which they force on management (e.g. Klein and Li 2015). If the abnormal trading

reflected private information common to multiple investors at the same time, then it should be more pronounced

for smaller targets and/or targets with large amounts of intangible assets, for which private information is likely to

be more extensive. However, I find a similarly high level of abnormal turnover by other investors on the trigger

date even in a subsample of large targets with no intangible assets.

Next, I investigate the Coordinated Effort Hypothesis, according to which the lead activist (the 13D filer)

would recruit other investors to join the campaign before the 13D filing becomes public, effectively offering the

jump in stock price that typically follows the 13D filing in exchange for those investors’ support. Short of observing

explicit communication between lead activists and other investors, this hypothesis is not directly testable.

Nonetheless, I offer five pieces of evidence that, collectively, may be interpreted as consistent with it. First, by

using a proprietary dataset containing institutional investors’ trades, I find that institutions that buy shares on the

trigger date are more likely to be those with a prior relationship with the 13D filer. In particular, an institution is

3 Investors without an active intent must file a Form 13G once they have acquired more than a 5% holding. From the moment that these investors switch their

intent from passive to active, they have up to 10 days to change their filings from 13G to 13D (see Rule 13d-1(e), Exchange Act).

5

about 8% more likely to accumulate shares on the trigger date if it also has done so in a prior campaign led by the

same activist.4 While it is possible that the same activist-institution pairs repeat across different campaigns because

they employ similar investment strategies, there is no reason to expect the same pairs to repeat specifically on the

trigger date. Second, if wolf pack formation is coordinated by the lead activist, then presumably she would inform

members of the upcoming campaign only after she has finished accumulating her position. Consistent with this

prediction, I find that the abnormal trading volume on the trigger date is ~36% higher for campaigns in which the

13D filers have already finished their position (i.e. campaigns in which the 13D filer did not acquire any stock

between trigger date and 13D filing date).5 Third, under the Coordinated Effort Hypothesis, the activist’s need to

form a wolf pack is higher when the target firm is more “resistant.” Indeed, I find that the abnormal trading volume

on the trigger date is higher when the target firm has stronger takeover defenses. Fourth, if the wolf pack is

coordinated, one would expect the wolf pack members to support the lead activist throughout the campaign.

Consistent with this prediction, I find that in the three years after the campaign filing date, wolf pack members

(defined as those investors who purchased shares on the trigger date) are 11% more likely than non-wolf pack

members to vote against management. Fifth, I find that in the two years after the 13D filing date wolf pack members

earn 2.57% higher abnormal returns than non-wolf pack members. More informed trading is consistent with wolf

pack members’ continued involvement in the campaign.

In the last part of the study, I examine whether the presence of wolf packs is associated with the outcome of

the campaign. For this purpose, I define as “campaigns with wolf packs” those campaigns in which trading by other

investors is in the top quartile of the sample distribution. After controlling for other determinants identified in prior

studies, I find that campaigns with wolf packs are associated with i) greater likelihood of the activist obtaining board

seats and other stated objectives of the campaign, and ii) 8.4% higher buy-and-hold abnormal returns over the

duration of the campaign. Similar to other studies examining the outcomes of hedge fund activism, endogeneity

concerns caution against causality inferences. For example, it is possible that pack members choose to join

4 Among all trades that take place on the trigger date, the probability that any underlying trade is a “buy” is ~45%, i.e. the conditional probability that an

investor buys on the trigger date is ~45%. 5 However, even for campaigns in which the lead activists have not finished their position, I continue to find the occurrence of abnormal trading. This is

consistent with the idea that in addition to the intentional coordination, there may be some unintentional leakage of information.

6

campaigns that are expected to be successful, thereby explaining the ex-post higher success rate of these campaigns.

If this reverse causality explanation is true, then I would expect to observe more wolf packs in campaigns with the

highest expected benefits. Using stock returns at the time of 13D filings as a proxy for expected benefits (Klein and

Zur 2009), I fail to find evidence consistent with such a selection scenario.

This study contributes to the literature on hedge fund activism. While a number of studies focus on the causes

and consequences of hedge fund activism (e.g. Klein and Zur 2009, Klein and Zur 2011, Brav et al. 2009), there is

little evidence on how activists successfully pressure target firms to acquiesce to their requests, in spite of their

relatively limited holdings (about 6% on average (Brav et al. 2009)). Anecdotal evidence suggests that part of the

success may be due to wolf packs forming around hedge fund targets and providing the activist with the support

needed to achieve its objectives (e.g. Coffee and Palia 2016). To the best of my knowledge, this is the first large-

sample study to empirically examine this phenomenon.6 In particular, I devise a way to detect the formation of wolf

packs using publicly available data, document their frequency, examine the mechanism behind their formation

(coordinated effort versus spontaneous formation), and investigate whether the presence of wolf packs is associated

with the outcome of the activist’s campaign.7

This study also contributes to the policy debate surrounding investor coordination and, especially, hedge fund

activism. The SEC has recently expressed concern about whether investors are using tactics, such as an informally

coordinated wolf pack, to circumvent the “group” definition for the purpose of reporting ownership levels, and thus

avoid crossing ownership thresholds that may trigger takeover defenses and other regulations. By providing greater

support for the Coordinated Effort Hypothesis, my findings suggest that these concerns are warranted. Importantly,

my study also outlines an empirical approach that government agencies can implement to detect wolf pack activism.

Finally, this study contributes to a relatively limited literature on investor coordination. Most of these studies

focus on cases of explicit (publicly disclosed) coordination. (e.g. Song and Szewczyk 2003; Dimson, Karakaz and

6 Becht, Franks, Grant and Wagner (2017) examine a special case of disclosed wolf packs, that is, campaigns in which multiple-schedule 13Ds are sequentially

filed for the same company, and find that campaigns with multiple 13D filers are more successful than campaigns with a single 13D filer. My study focuses

instead on undisclosed wolf packs in campaigns with a single 13D filer. The two types of campaigns are qualitatively different. Those with multiple, sequential 13D filings tend to last longer (747 days vs. 404 days for campaigns with a single 13D filer) and likely capture cases in which a second 13D filer (another

activist) joins a struggling campaign to increase the probability of success. By contrast, my study examines the role of undisclosed wolf packs at the time of

the first activist event, which has been the subject of debate among commentators and policy makers but has not been examined empirically. 7 Also, by showing that only about 25% of the trading volume on the trigger date is driven by the activist’s trades, with the rest reflecting trades by other

investors, my study sheds light on the reasons for the high share turnover on the trigger date documented in prior studies (see Bebchuk et al. 2013).

7

Li 2015). Empirical evidence on implicit coordination is relatively scarce. A notable exception is Crane, Koch and

Michenaud (2017) (CKM), who focus on how existing shareholders coordinate their votes on proxy items. This

study adds to CKM by examining implicit coordination in a different setting, where coordination i) requires

investors (wolf pack members) to buy shares, a more “costly” commitment than voting; ii) requires investors to

agree on the merits and goals of the campaign, which involves more complex, firm-specific strategic issues than

the standard governance topics covered by proxy votes, and iii) may have more significant legal consequences for

the parties involved. In other words, the stakes are higher.

2. Sample selection and descriptive statistics

I use data from SEC Schedule 13Ds and SharkRepellent.net to construct a comprehensive sample of activist

campaigns between 1998 and 2014. As shown in Table 1, I start with 3,744 unique activism events. Since I focus

on trading by other investors before public disclosure of the campaigns, I remove 304 campaigns in which the

trigger date and the 13D filing date are the same. For each remaining events, I manually download all 13D filings

from SEC.gov and collect the following information: the filing and trigger dates; the identity and Central Index Key

(CIK) of the hedge fund; the name, CIK, CUSIP, and SIC code of the target firm, and the percentage of shares

owned by the activist at the time of 13D filing. For each stock traded in the dataset, I collect returns, share price,

trading volume, and shares outstanding from CRSP, and book value of equity from Compustat. I remove 151

campaigns in which a 13D cannot be located, 201 campaigns with missing variables from CRSP/Compustat, and

528 campaigns in which the trigger date is not reported. After excluding Real Estate Investment Trusts (SIC 6798),

blank check entities (SIC 6770), trusts (SIC 6792), and American Depositary Receipts (ADRs), I am left with 2,293

distinct campaigns. I also exclude 366 activism events in which multiple 13Ds are filed during the same campaign,

that is, another 13D is filed (by another activist) after the first 13D filing date and prior to the end date of the related

campaign.8 These 366 events are the type of campaigns that Becht et al. (2017) classify as “disclosed” wolf pack

events. Finally, I exclude five campaigns for which daily trades by the 13D filer are not available.

8 These 366 events include 170 initial 13Ds and 196 subsequent 13Ds. For the 170 initial campaigns in which at least one 13D is filed subsequently, the average time between the initial and subsequent 13D is 501.3 days. The length of these initial campaigns is 746.7 days, significantly longer than the rest of the

sample (403.5 days).

8

The remaining 1,922 campaigns are initial campaigns without any subsequent 13Ds and constitute my final

sample. These campaigns comprise 340 individual activists and 1,753 unique firms, with the 20 most prominent

activists representing about 50% of all campaigns.

The target companies in my sample are comparable to those in other studies of activism (Brav, Jaing, Partnoy

and Thomas 2008, Klein and Zur 2009). At the time of the 13D filing on average target firms have a market value

of $933.6 million, institutional holding of 44%, and 3.3 analysts following the firm (see Table 2). Also, at the time

of the 13D filing, on average 13D filers own 8.8% of the shares outstanding, with about 60% of this amount (5.4%)

being purchased in the 60 days before the filing date. Notably, at the time of 13D filing, more than 85% of the

activists hold less than 10% of shares outstanding (untabulated). This is consistent with the argument that poison

pills and the short-swing profit rule constrain the amount of shares that can be accumulated by a lead activist (see

Section 4.2). Finally, Table 2 shows that most filers take advantage of the 10-day filing delay allowed under

schedule 13D, with the average delay being 7.6 days and over 50% of the sample filing more than nine days after

the trigger date9 (I will describe the other variables in Table 2 when they are used later in the study).

[Insert Table 2]

3. How common are wolf packs? Evidence on the accumulation of shares by other investors

Similar to Coffee and Palia (2016) and Brav et al. (2018), I refer to a wolf pack as a loose network of investors

who accumulate shares in the target firm before the 13D filing. To identify the accumulation of shares by other

investors, I first examine trading volume around the trigger date. In Figure 2, I plot the trading volume for the full

sample of 1,922 campaigns for the 60-day period centered on the trigger date. The variable on the y-axis,

𝑇𝑢𝑟𝑛𝑜𝑣𝑒𝑟_𝑎𝑙𝑙 =𝑉𝑜𝑙𝑖,𝑡

𝐴𝑣𝑔(𝑉𝑜𝑙𝑖,𝑡−120…..𝑉𝑜𝑙𝑖,𝑡−60) , which is calculated for each day of the campaign and is presented as a

percentage of normal trading volume. Hence, if 𝑇𝑢𝑟𝑛𝑜𝑣𝑒𝑟_𝑎𝑙𝑙 =1, it means there is no abnormal turnover on that

date.10 Normal trading volume is estimated as the average trading volume between day −120 and −60 relative to

9 These figures are higher than in previous studies because I remove all campaigns where the filing date and the trigger date are the same (see Table 1). 10 Many studies (e.g., Beaver 1968, DeFond, Hung and Trezevant 2007) use abnormal trading volume to measure the information content of a given event (e.g. an earnings announcement). In this setting, though, information content is less relevant, since the activist’s campaign is supposed to be a privately known event

prior to public disclosure.

9

each date. Figure 2 shows a significant spike in turnover on the trigger date: the average turnover is about 325% of

normal volume.

[Insert Figure 2]

Since the trigger date is not publicly observable until the 13D is filed, the high level of turnover cannot be a

public reaction to the activist’s campaign. Two prior studies have documented significant abnormal share turnover

on the trigger date before a 13D filing, but they differ in their interpretations. Coffee and Palia (2016) see this high

level of turnover as evidence of wolf pack formation. In contrast, Bebchuk et al. (2013) interpret it as evidence of

activist hedge funds accumulating most of their holdings on the trigger date.

To distinguish between these explanations, I subtract from the total volume the trades by the 13D filer. If other

investors drive at least part of the abnormal share turnover, then I expect to see a significant level of abnormal

turnover even after removing the 13D filer’s trades.

I identify trades by 13D filers by hand collecting the relevant information from Schedule 13Ds, which include

trading information for at least 60 days before the filing date.11 Figure 3 presents the daily trading volume by other

investors (total volume less trading by 13D filer) for the period between day -30 and day +10 relative to the trigger

date (unlike Figure 2, I stop at day 10, the filing date, because I can only observe trades by the lead activist up to

the filing date). The variable on the y-axis, 𝑇𝑢𝑟𝑛𝑜𝑣𝑒𝑟_𝑂𝑡ℎ𝑒𝑟𝑡, is the turnover driven by other investors, presented

as a percentage of normal trading volume (defined as in Figure 2). Consistent with the existence of wolf packs,

Figure 3 shows that, even after removing trades by the 13D filers, the average trading volume on the trigger date is

still about 240% of normal trading volume, which is 140% higher than the average 𝑇𝑢𝑟𝑛𝑜𝑣𝑒𝑟_𝑂𝑡ℎ𝑒𝑟𝑡 (~100%) for

all “non-trigger date” days examined in this figure. This finding also implies that only about 25% of the total trading

volume on the trigger date is driven by the 13D filer’s trades.

[Insert Figure 3]

4. Mechanism of wolf pack formation

11 In the vast majority of cases, transaction data are reported on a daily basis. When transaction data are reported at higher-than-daily frequencies, I aggregate

to the daily level. In particular, I collect the following data: date of each transaction, transaction type (purchase or sell), transaction size, transaction price, class

of the transaction (e.g., common stock, options, warranty), whether the transaction happened in an open market or a private transaction, and the entity making the trade. See online appendix for an example of a typical 13D filing trading schedule.

10

4.1 Spontaneous formation hypothesis

Even absent any coordination, economic theories provide an explanation for why wolf packs might emerge:

different investors might independently target a similar set of firms at about the same time. The most applicable

model is provided by Brav et al. (2018). In their model, there are two types of players—a large activist and many

small activists—and the large activist’s campaign will succeed if the number of shares owned by all activists is

larger than the shares held or controlled by management. A pack can then form around the lead activist without any

explicit communication or intentional coordination by the lead activist. Small and large activists often monitor the

same companies and determine their targets using similar criteria. Therefore, if small activists are aware of an event

that may trigger a lead activist’s engagement, the small activists will act even though the small activists by

themselves may not have sufficient shares to overcome management. Similarly, a large activist may start a campaign

even though he may not be able to acquire sufficient shares to overcome management by himself, knowing that

there will be support from other smaller activists.

There are two scenarios under which this type of spontaneous formation can occur. The first is that, after

independently performing their research, multiple activists come to the same conclusions and end up targeting the

same firms at about the same time. However, the evidence of substantial trading by other investors on the trigger

date (Figure 2 and 3) casts doubts on this possibility. While it is plausible that other investors independently decide

to accumulate shares in the same firms targeted by the 13D filers, it seems unlikely that many investors would

decide to do so at the same time and to do it exactly on the trigger date, which is not a publicly observable event.

The second, more plausible scenario, is that the simultaneous actions may be driven by some sudden change

in market conditions that are observable to both lead activist and other investors. In the following subsections, I

examine this possibility in detail.

4.1.1 Spontaneous formation: Reaction to the arrival of news?

A potential reason for a sudden change in market conditions is the arrival of public news regarding the target

firm on or immediately before the trigger date. In this case, both the 13D filer and the independent investors

accumulate shares in the target company simultaneously on the trigger date because they have the same reaction to

news related to the target firm.

11

To examine this explanation, I identify a subset of 751 campaigns with no Factiva news regarding the target

firm during the 11-day window from day −10 to and including the trigger date. If the documented share turnover is

mostly due to public news arrival, there should be little or no abnormal trading on the trigger date for this subset of

campaigns. However, as shown in Figure 4, trading by other investors for this “no news” subsample is about 230%

of normal trading volume, which is 130% higher than the average 𝑇𝑢𝑟𝑛𝑜𝑣𝑒𝑟_𝑂𝑡ℎ𝑒𝑟𝑡 (~100%) for all “non-trigger

date” days examined in this figure and similar in magnitude to the full sample (Figure 3). Hence, it does not appear

that the abnormal turnover by other investors is driven solely by the arrival of public news about the target firm.

[Insert Figure 4]

4.1.2 Spontaneous formation: Reaction by Kyle-type traders?

Another type of news which may trigger simultaneous action by small activists is order flow information. On

the trigger date, the 13D filer’s trades account for 25% of the total trading volume on average. A Kyle-type small

activist (see Kyle 1985) may interpret the large order flow by the 13D filer as suggesting an upcoming campaign

and thus buy shares in the target firm. To examine this possibility, I exploit the fact that not all 13D filers trade on

the trigger date. This is because the mandatory 13D filing is triggered, not only by a change in ownership (crossing

the 5% threshold), but also by a change in intent, from “passive” to “active.” Both triggers must be satisfied before

an investor must file a 13D. That is, there is a subset of 351 campaigns with 13D filers (“13G switchers”) who

already own more than 5% of the target firm before the trigger date but decide to switch their investment objective

from passive to active on the trigger date.12 Thus, it is the change in objective that triggers the 13D filing, not a

change in holdings.13 If the documented abnormal turnover is entirely or mostly due to Kyle-type traders, there

should be little or no abnormal turnover on the trigger date for this subset of campaigns. However, as shown in

Figure 5, trading by other investors for this subsample (the 13G switchers) is about 240% of normal trading volume,

which is ~130% higher than the average 𝑇𝑢𝑟𝑛𝑜𝑣𝑒𝑟_𝑂𝑡ℎ𝑒𝑟𝑡 (~110%) for all “non-trigger date” days examined in

12 These investors have filed a 13G in the past. Investors without an active intent must file a 13G once they have acquired more than a 5% holding. From the moment that these investors switch their intent, they have up to 10 days to change their filings from 13G to 13D (see Rule 13d-1(e), Exchange Act). 13 The average announcement return for this subset of campaigns is about 1.4% (3 days of abnormal return centered on the filing date).

12

this figure and similar in magnitude to the full sample (Figure 3). Hence, it is unlikely that the abnormal turnover

by other investors is driven only by Kyle-type traders.

[Insert Figure 5]

4.1.3 Spontaneous formation: Asymmetric information?

Supporters of hedge fund activism argue that activists can create value by converting their private

information into new strategic plans and forcing company management to adopt them. Therefore, it is possible that

the simultaneous reaction may be driven by the arrival of information that is available only to the activists’

community. According to this explanation, I would expect abnormal trading to be concentrated in firms for which

activists have the highest level of private information.

To examine this possibility, I limit my subsample to big firms (higher median level of total assets) with no

intangible assets. Prior literature finds that these firms have a low level of information asymmetry and hence there

is a low probability that activists will be privately informed (for e.g. Chari, Jagannathan and Ofer 1988, Aboody

and Lev 2000). As shown in Figure 6, even within this subsample of firms with low information asymmetry,

abnormal turnover by other investors remains high, at about 230% of normal trading volume, which is ~120%

higher than the average 𝑇𝑢𝑟𝑛𝑜𝑣𝑒𝑟_𝑂𝑡ℎ𝑒𝑟𝑡 (~110%) for all “non-trigger date” days examined in this figure and is

similar in magnitude to the full sample (Figure 3). Hence, again there is no support for the Spontaneous Formation

Hypothesis.

[Insert Figure 6]

4.1.4 Multivariate analysis of daily trading volume

In the previous analyses, the proxy for normal (and thus abnormal) trading volume is based only on past

trading volume. In this section, I perform a multivariate analysis to examine whether my inferences hold after

controlling for various determinants of trading volume. In particular, I estimate the following pooled ‘campaign-

day’ regression, in which each observation represents a trading date within the 60 days before the 13D filing,

resulting in 115,320 observations (= 1,922 activist events x 60 days), with standard errors clustered by activist and

firm:

13

𝑇𝑢𝑟𝑛𝑜𝑣𝑒𝑟_𝑂𝑡ℎ𝑒𝑟𝑡 = 𝐼𝑛𝑡𝑒𝑟𝑐𝑒𝑝𝑡 + 𝑇𝑟𝑖𝑔𝑔𝑒𝑟_𝐷𝑎𝑡𝑒 + 𝐶𝑜𝑛𝑡𝑟𝑜𝑙𝑠 + 𝐹𝐸. (1)

The dependent variable, 𝑇𝑢𝑟𝑛𝑜𝑣𝑒𝑟_𝑂𝑡ℎ𝑒𝑟𝑡, measures the turnover driven by other investors (other than the

13D filer), scaled by normal trading volume (as in Figure 3). Hence, if no abnormal turnover is detected,

𝑇𝑢𝑟𝑛𝑜𝑣𝑒𝑟_𝑂𝑡ℎ𝑒𝑟𝑡 will be 1.

My main independent variable of interest is Trigger_Date, an indicator variable equal to 1 if that particular

trading date is the trigger date, and 0 otherwise. This variable captures the difference in abnormal turnover between

the trigger date and every other date in the 60-day window prior to a 13D filing after controlling for other

determinants of trading volume. I divide these determinants into the following categories: 1) liquidity, 2) arrival of

news and momentum, and 3) other firm-specific characteristics (see Appendix A for more details and data sources).

Collin-Dufresne and Fos (2015) document a high level of liquidity on the days that the 13D filer trades. If such

a spike in liquidity simultaneously raises the probability of a firm’s shares being purchased by other investors, then

share turnover may be higher on the trigger date. Therefore, I include the following proxies to control for liquidity:

Amihudt, the ratio of stock return to trading volume on day t (Amihud 2002); Log (MV), the natural logarithm of the

firm market value at the beginning of the calendar year, and Institutional Salest, the percentage of shares outstanding

sold by institutional investors on day t (Gantchev and Jotikasthir (2017) show that the increase in liquidity is driven

by sales from other institutional owners).

As mentioned in Section 4.1.1, the arrival of news on or immediately before the trigger date may induce both

the 13D filer and independent investors to accumulate shares in the target company simultaneously. Thus I include

the following news proxies as controls: 10Kt, 8Kt, and 10Qt are indicator variables for Forms 10-K, 8-K, and 10-Q

that are filed on day t; I/B/E/S forecastt is an indicator variable for the issuance of I/B/E/S analysts’ forecasts on day

t; Management Guidancet is an indicator variable for the issuance of management guidance on day t, and Newst(t-1,t-

2) is the number of news items about the target firm on date t (t-1, t-2). In addition, order flow and stock returns may

convey information about the future stock price of the underlying firms and therefore may explain trading by other

investors. I include the following momentum proxies as controls: Volt−1, the percentage of shares outstanding traded

on day t−1, and Abn_rett(t-1), the excess return from a four-factor model on date t (t−1).

14

I also include a number of other firm characteristics that may be correlated with share turnover: Bid-Ask

Spreadt, the absolute difference between the bid (low) and ask (high) on day t; 13D Filer Holdings, the percentage

of shares outstanding held by the 13D filer on the filing date;14 Institutional Holding, the percentage of shares

outstanding held by institutional investors in the most recent quarter (source: Thomson Reuters 13F Filings), and

Analyst Following, the number of analysts following the firm. Finally, I include year fixed effects to control for

time trends, industry fixed effects (Fama-French 12 industries) to control for time-invariant industry characteristics,

and weekday fixed effects to control for changes in trading across weekdays.

The results indicate that the abnormal turnover by other investors on the trigger date persists even after

controlling for all the above determinants of trading volume. As shown in Table 3, Panel A, Column (1) the

coefficient of Trigger_Date is approximately 1.23, implying that on the trigger date share turnover by other

investors is 123% higher than on other days in the 60-day window, on average. This confirms, in a multivariate

setting, the evidence from Figure 3.

Next, in Columns (2)-(4) I divide the campaigns into various subsamples, essentially re-examining in a

multivariate setting the three scenarios discussed in Sections 4.1.1 to 4.1.3. In particular, in Column 2, I divide my

sample into campaigns with and without news (based on Factiva search) in the 11 days up to and including the

trigger date. The indicator variable Trigger date & no news takes the value of 1 if the campaign has no news and

that particular date is a trigger date, and 0 otherwise. In Column (3), I divide my sample into 13G switchers and

non-switchers. The indicator variable Trigger date & 13G switcher (non-switcher) takes the value of 1 if the filer

is a 13G switcher (non-switcher) and that particular date is a trigger date, and 0 otherwise. Finally, in Column (4),

I divide my sample into low- and high-information asymmetry campaigns. The indicator variable Trigger date &

Low Infor. Asymmetry takes the value of 1 if the target firm has a low level of information asymmetry (higher than

sample median level of total assets and no intangible assets) and that particular date is a trigger date, and 0 otherwise.

The coefficients on, respectively, Trigger date & no news, Trigger date & 13G switcher and Trigger date &

Low Infor. Asymmetry are significant at 1.19, 1.21 and 1.23, implying that, even for these subsets of campaigns, on

14 The 13D Filer Holdings is included as a proxy for the accumulation of shares by the 13D filer during the event window.

15

the trigger date share turnover is 119%, 121%, and 123% higher than on other days in the 60-day window, on

average. Finally, F-tests reported in Panel B also indicate no differences in the coefficient on Trigger date among

the subsamples within each partition, except that trading by other investors is higher on the trigger date in the

presence of news about the target firm (but, as noted above, it is economically and statistically significant even

when there is no news). Overall, the multivariate analysis supports my inferences from Figure 3-6, and is not

consistent with the Spontaneous Formation Hypothesis.

[Insert Table 3]

4.2 Coordinated effort hypothesis

Market observers often allege that lead activists muster wolf packs. In such a scenario, the activist recruits

several other investors to join the campaign before public disclosure of the campaign via the 13D filing, effectively

offering the jump in stock price that typically follows the 13D filing in exchange for their support. These pre-

disclosure arrangements can be done either explicitly as alleged by media accounts (see, for example, Pulliam,

Chung, Benoit and Barry 2014, Hoffman et al. 2015), or implicitly via various forms of indirect communication

and signaling. However, it is important for the arrangement to take place informally, to avoid forming a “group”

under Section 13(d)(3) of the Securities Exchange Act of 1934.

At first sight, it may appear that constituting a wolf pack would not be in the best interest of the lead activist

since she bears all the costs of engagement but only reaps a small percentage of benefits, a typical free-rider problem

(e.g., Admati, Pfleiderer and Zechner 1994). However, there are two reasons why this form of informal coordination

is attractive to the lead activist. First, the lead activist may be financially constrained, and thus unable to acquire

sufficient shares to implement changes in the target company. Second, even if not financially constrained, regulatory

barriers such as the “short-swing profit rule”15 and takeover defenses, such as “poison pill”16, make it difficult for

activists to acquire ownership that exceeds a certain holding threshold. For example, once an activist acquires more

15 Section 16(b) of the Securities Exchange Act of 1934 entitles shareholders to recover short-swing profits that are based on a purchase and sale or a sale and

purchase, within six months, of the stock of a “reporting company.” The average holding period of the activist is usually longer than six months (see Brav et

al. 2009). Yet the activist may not want to lose the option to turn over the position quickly. The definition of “group” is the same under Section 13(d) and Section 16(b). Group activity in both cases is governed by Section 13(d)(3) of the Exchange Act. 16 A shareholder rights plan, commonly known as a poison pill, is a tool used by boards of directors to deter activists. Typically, such a plan gives shareholders

(other than the activist) rights to buy more shares at a discount if an activist buys a certain percentage or more of the company’s shares. Third Point LLC vs. Ruprecht 2014 held that the lowest statutory limit for a poison pill is 10% of shares outstanding. If every other shareholder can buy more shares at a discount,

this dilutes the activist’s interest.

16

than 10% of a target’s shares, she is subjected to the “short-swing profit rule,” which may force the activist to

surrender any short-swing profits to the target company. By arranging a wolf pack, the lead activist can increase the

percentage of voting shares under its effective control without incurring these problems (Coffee and Palia 2016).

As for the pack members, learning about the impending 13D filing without being treated as a formal 13D

“group” member creates an opportunity for profitable trading. As mentioned earlier, the market usually reacts

positively to a 13D filing. Furthermore, being an informal member allows investors to trade profitably without

incurring the risk of future lawsuits because the target company will usually not know of their existence (Coffee

and Palia 2016).

Lacking access to private communications between investors, providing direct evidence on coordination is

extremely difficult. In this section, I devise five different tests that examine patterns of behavior that would be

consistent with the coordinated effort mechanism. The first three refer to the period prior to the start of the activist

campaign, while the last two focus on the subsequent period. In particular, under the hypothesis that the wolf pack

is coordinated, I will argue that: (1) investors are more likely to be ‘recruited’ as pack members if they have been

pack members in a past campaign led by the same lead activist; (2) the lead activist is more likely to coordinate a

pack if he has already accumulated his position as of the trigger date (else, the price pressure from pack members’

trading would make it more costly for the lead activist to acquire the rest of her target stake); (3) the lead activist is

more likely to form a wolf pack when its expected benefits are higher, that is, when targeting firms with stronger

takeover defenses; (4) after the 13D filing date, wolf pack members (as proxied by investors purchasing shares on

the trigger date) are more likely to “support” the campaign by voting against management recommendations at

annual meetings, and (5) after the 13D filing date (i.e. during the campaign), wolf pack members continue to earn

abnormal returns relative to other investors, a symptom of their continuing involvement in the campaign.

Below I explain in greater detail why these predictions follow from the Coordinated Effort Hypothesis and

then I empirically examine each of the five predictions.

4.2.1a Pre-campaign evidence of coordination: The role of prior relationships between lead activist and pack

members

17

Under the Coordinated Effort Hypothesis, the wolf pack is orchestrated by the lead activist and presumably

she will prefer to limit the potential free-riding problem17 by recruiting members with a pre-existing relationship.18

Accordingly, in this subsection, I examine whether past relationships between the investor buying shares (the

presumed pack member) and the lead activist increase the likelihood that such investor will be part of the wolf pack.

To perform this type of analysis, I need (1) to obtain the identity of the investor executing each trade, and (2)

define a ‘past relationship’ with the lead activist.

With respect to (1), the identity of the investors trading on a given date is not publicly available. To overcome

this data limitation, I obtain from a consulting firm a proprietary dataset that includes transaction-level trading data

for more than 500 institutional investors between January 1, 1998 and December 31, 2010 (Gantchev and Jotikasthir

(2017), Green (2006) and Klein and Li (2015) use a similar proprietary dataset). For stocks covered by both this

dataset and CRSP, this dataset accounts for roughly 10-15% of the total CRSP trading volume. The dataset includes

the complete transaction history for its institutional clients (for the period in which the institutional client remains

a subscriber). There are two ways an institution can be listed in the database. It can invest on behalf of a plan sponsor

that subscribes, or it can subscribe directly. Each observation corresponds to an executed trade. For each trade, the

database reports the date of the trade, the execution price of the trade, the stock traded, the number of shares traded,

whether the trade was a buy or a sell, and a unique client identity code for the institution making the trade. The

client identifier corresponds to the plan sponsor or money manager who is a client of the consulting firm. The client

identifier is a permanent numeric code, which allows me to track a given client both in the cross section and through

time. Over the sample period, the dataset covers a total of 535 unique managers and their trades in the stock of

5,231 unique firms. On average, 374 managers and 3,475 firms are listed per year. Each day, about 4.4 different

institutions trade a particular stock, with an average trade size of $265,188 (untabulated descriptive statistics).

With respect to (2), using this dataset I create an empirical proxy for a past relationship as follows: for each

potential activist-institution pair (among the institutions covered by the dataset), I define as Past Relationship the

17 Activists’ campaigns are especially prone to the free-riding problem because the lead activist incurs significant up-front cost (such as acquiring adequate

shares and preparing for the campaign), but part of the return expected from the campaigns can be realized by the members immediately on the 13D filing date

without contributing to the actual campaign (e.g. by voting to support the lead activist’s agenda). 18 The pre-existing relationship limits the free-riding problem by aligning the interests between the pack leader and the members. For example, the possibility

of being excluded from future campaigns creates a disincentive for members to engage in opportunist behaviors.

18

number of times that a particular institution has “participated” in a prior campaign led by that particular activist

within the last year. I classify an institution as a “participant” if that institution purchased shares on the trigger date

of the previous campaign.

Next, I estimate the following regression:

𝐵𝑢𝑦𝑖𝑛𝑔 = 𝑃𝑎𝑠𝑡 𝑅𝑒𝑙𝑎𝑡𝑖𝑜𝑛𝑠ℎ𝑖𝑝 + 𝐶𝑜𝑛𝑡𝑟𝑜𝑙𝑠, (2)

where Buying is an indicator variable which equals 1 if the trading institution accumulated a positive number

of shares in the target on the trigger date, and 0 otherwise; and Past relationship is defined as described above. In

essence, I examine whether institutions that trade on the trigger date of a campaign by an activist are more likely to

buy shares (rather than sell them) if they bought shares on the trigger date of a prior campaign by the same activist.

Note that the above regression is limited to those 1,233 campaigns in which at least one of the institutions covered

by the proprietary dataset (other than the 13D filer) traded on the trigger date (also, since estimating Past

Relationship requires at least one lag year of data and the dataset starts in 1998, I restrict my sample to 1999-2010).

This results in a total of 3,553 observations, implying that on average in each campaign, about 3 institutions covered

by the dataset trade shares of the target firm on the trigger date.

Table 4, Panel A, Column (1) estimates Eq. (2) as a probit regression and finds an average marginal effect

of about 8%. This implies that a one standard deviation increase in Past Relationship leads to an 8% increase in the

probability that the institution will buy shares of the firm targeted by the same activist if trading on the trigger date

(the average probability that an institution will buy a target stock conditioned upon trading on the trigger date is

~45%). The results are similar using a standard OLS regression (Column 2).

While indirect, this pattern is consistent with the notion that lead activists are more likely to tip off investors

with whom they have a prior relationship. While it is possible that the same activist-institution pairs repeat across

different campaigns because they employ similar investment strategies (such as acquiring shares in firms with a

high level of liquidity), there is no reason to expect the same pair to repeat specifically on the trigger date.

Nonetheless, in the next subsection, I try to devise a test in which the incentive for wolf pack formation is unrelated

to factors such as liquidity.

4.2.1b Pre-campaign evidence of coordination: Position accumulated by lead activist and wolf pack formation

19

Under the Coordinated Effort Hypothesis, the lead activist has discretion on the timing of pack formation.

Presumably, a lead activist would inform potential pack members about the upcoming campaign only after she has

acquired her position. Doing so before the position is accumulated would push the stock price upward and make it

more costly to achieve her target stake. If so, I would expect to see a positive relationship between the degree to

which the 13D filer has accumulated her position prior to the trigger date and the extent of abnormal trading by

other investors on the trigger date. To examine this prediction, I employ the following linear regression:

𝑇𝑢𝑟𝑛𝑜𝑣𝑒𝑟_𝑂𝑡ℎ𝑒𝑟𝑇𝑟𝑖𝑔𝑔𝑒𝑟𝐷𝑎𝑡𝑒 = 13𝐷 𝐹𝑖𝑙𝑒𝑟 𝐷𝑜𝑛𝑒 + 𝐶𝑜𝑛𝑡𝑟𝑜𝑙𝑠 (3)

13D Filer Done is an indicator variable which takes the value of 1 if the 13D filer does not acquire any shares

between the day subsequent to the trigger date and the filing date, and 0 otherwise. As shown in Panel B, Column

(1), the abnormal share turnover is ~36% higher for campaigns in which the 13D filer has acquired her position

prior to the trigger date. As an alternative measure for wolf pack formation, I replace 𝑇𝑢𝑟𝑛𝑜𝑣𝑒𝑟_𝑂𝑡ℎ𝑒𝑟𝑇𝑟𝑖𝑔𝑔𝑒𝑟𝐷𝑎𝑡𝑒

with the indicator variable Spike which takes the value of 1 if 𝑇𝑢𝑟𝑛𝑜𝑣𝑒𝑟_𝑂𝑡ℎ𝑒𝑟𝑇𝑟𝑖𝑔𝑔𝑒𝑟𝐷𝑎𝑡𝑒 is greater than 1, and 0

otherwise. As shown in Panel B, Column (2), the probability that we observe a Spike on the trigger date is about

5% higher when the 13D filer has finished his position. These findings are consistent with the pattern we would

expect to observe if the wolf pack is coordinated by the lead activist.

4.2.1c Pre-campaign evidence of coordination: Takeover defenses and wolf pack formation

As mentioned previously, the key benefit of a coordinated wolf pack is that it allows the lead activist to

“control” a larger number of shares outstanding. The additional “control” increases the bargaining power that the

activists have over management and is especially useful when targeting more “resistant” firms. As a result, under

the Coordinated Effort Hypothesis, evidence of wolf pack formation (as captured by abnormal turnover by other

investors on the trigger date) should be greater when the target firm is more resistant (as captured by the existence

of a poison pill or other takeover defenses).19

To examine this prediction, I employ the following linear regression:

19 A premise of this test is that, although the existence of a wolf pack can improve the lead activist’s probability of success, lead activists will use the wolf

pack tactic (or use it to a greater degree) only when necessary. This is because engaging in wolf pack activities may be costly for lead activists. First, other

members may start accumulating shares, pushing prices upward and making it costlier for the lead activists to achieve their target stake; second, by arranging a wolf pack the lead activists may incur additional litigation risk. (See Godnick and Gussman 2006 for examples of the potential litigation risks associated

with forming a wolf pack.)

20

𝑇𝑢𝑟𝑛𝑜𝑣𝑒𝑟_𝑂𝑡ℎ𝑒𝑟𝑇𝑟𝑖𝑔𝑔𝑒𝑟𝐷𝑎𝑡𝑒 = 𝑃𝑜𝑖𝑠𝑜𝑛 𝑃𝑖𝑙𝑙 + 𝐶𝑜𝑛𝑡𝑟𝑜𝑙𝑠 (4)

I use share turnover by other investors on the trigger date, 𝑇𝑢𝑟𝑛𝑜𝑣𝑒𝑟_𝑂𝑡ℎ𝑒𝑟𝑇𝑟𝑖𝑔𝑔𝑒𝑟𝐷𝑎𝑡𝑒, as an empirical

proxy for wolf pack formation. Poison Pill is an indicator equal to 1 if the target firm responds to the campaign by

adopting a poison pill or if it already has a poison pill in place. The regression results for Eq. (4) are presented in

Panel C, Column 1. The coefficient on Poison Pill is positive and significant, suggesting that lead activists are more

likely to employ the wolf pack tactic with companies that are more likely to adopt a poison pill. As an alternative

measure for corporate defense, in Column 2 I replace Poison Pill with Bullet Proof Rating, a proprietary index from

FactSet which takes into account various proactive takeover defenses undertaken by the target.20 The rating scale is

from 0 to 10, with a 10 representing the most formidable defenses. Again, the coefficient on Bullet Proof Rating is

positive and significant at the 10% level.

Overall, Panel C is consistent with lead activists being more likely to form wolf packs when dealing with

better-defended companies, i.e. when a wolf pack may be more beneficial.

[Insert Table 4]

4.2.2a Post-campaign evidence of coordination: Voting behavior of wolf pack members

The previous three tests refer to the initial formation of the wolf pack and suggest that its circumstances are

consistent with the Coordinated Effort Hypothesis. However, under this hypothesis, one would also expect to

observe that pack members do indeed support the lead activist’s campaign after the 13D filing. To examine whether

this is the case, lacking a more direct measure of support, I analyze the post-campaign voting behavior of wolf pack

members at the annual meetings. I focus on mutual funds’ voting records because mutual funds are the only

institutional investors who must disclose their votes at the annual meetings. The voting data are obtained from ISS

and are available between 2003 and 2010. In my sample, 730 campaigns take place between 2003 and 2010. Of

those 420 have identifiable trading data and identifiable traders from my proprietary dataset. Within these 420

campaigns, I find a total of 52 campaigns with ISS voting records. These 52 campaigns translate into a total of 185

20 The rating does not take into consideration ownership and voting rights, the takeover laws which govern the state in which a company is incorporated, or

whether a company has opted out of coverage of applicable state takeover laws.

21

campaign-manager observations. The low overlaps between the databases are driven by two factors: (1) ISS only

covers the S&P 1500 firms and (2) ISS only contains votes from mutual funds.

In Table 5 Panel A, I compare the frequency of votes in favor of management recommendations on all items

up for a vote at the annual meetings taking place within three years after the beginning of the campaign (13D filing

date). I divide mutual funds into two groups: those that purchased shares on the trigger date (my proxy for wolf

pack members) and those that did not. It is well known that mutual funds generally vote in favor of management.

However, the results show that the wolf pack members are significantly less likely than non-wolf pack members to

vote in favor of management recommendations. Admittedly, this is a small sample and is only indirect evidence of

support for the campaign, but it is consistent with the notion that pack members are at least vocally opposing

management throughout the campaign.

4.2.2b Post-campaign evidence of coordination: Abnormal returns by wolf pack members

Finally, I devise another test that may speak to the involvement that the wolf pack members have in the

campaign after the initial formation. After the positive market reaction from the 13D filing, there are several reasons

for the pack members to continue their support for the lead activist. First, there is an implicit understanding between

the lead activist and the pack members that they will provide support throughout the campaign. If this support is

not provided, it is unlikely that the pack members will be recruited again in a future campaign (loosely consistent

with this argument, the test in section 4.2.1a shows that lead activists are more likely to recruit members with whom

they have a past relationship). Second, and more importantly, being involved in the campaign creates additional

opportunity for further trading gains. Thus, under the Coordinated Effort Hypothesis, I expect that pack members

continue to earn higher abnormal returns than non-pack members after the initial 13D filing.

To test this prediction, I proceed as follows. First, using the proprietary dataset described earlier, for each

of the 1,233 campaigns I identify all the investors who have made at least one trade within the two-year period

starting one year prior to the 13D filing date and ending one year after (i.e. from day -365 to day +365 relative to

the filing date), including, of course, investors trading on the trigger date. This results in 17,022 campaign-investor

observations. Within this subset of observations, I divide institutional investors into two groups: those that purchase

22

shares on the trigger date (my proxy for wolf pack members) and those that do not. For each group of investors, I

calculate the CAR earned between filing date +1 to filing date + 365.21

As shown in Panel B, over the course of the campaign wolf pack members earn significantly higher abnormal

returns than non-wolf pack members, consistent with wolf pack members being more informed and thus better able

to time their trades during the campaign relative to other investors, as one would expect if they were actively

involved with the campaign and thus in line with the Coordinated Effort Hypothesis. Since wolf pack members are

more likely to be those who have a prior relationship with the lead activist (Section 4.2.1a), in Panel C I further

divide the wolf pack members into those who have a prior relationship with the lead activists and those who do not.

Consistent with a past relationship leading to more active participation, I find that wolf pack members with prior

relationships earn significantly better returns than wolf pack members without a prior relationship.

Overall, while admittedly indirect and suggestive, the five tests presented above indicates patterns of behavior

consistent with the Coordinated Effort Hypothesis.

[Insert Table 5]

5. Additional analyses: Wolf packs and activist campaigns’ outcome

Hedge fund activists typically hold a relatively small stake in target firms (about 6%, see Brav et al. 2009).

Yet, prior studies show that they have been highly successful in pressuring target firms to acquiesce to their requests.

For example, Klein and Zur (2009) and Brav et al. (2009) show that hedge fund activists achieve at least one stated

objective in 60-70% of cases. Some have suggested that part of this success may be due to the formation of a wolf

pack (Briggs 2007; Coffee and Palia 2016). In this section, I examine the validity of this conjecture.

5.1 Empirical analysis

To estimate the relationship between the existence of a wolf pack and the activist campaign’s outcome I

employ the following linear regression:

21The groups’ return is calculated as follows: First, I estimate the daily position of each institutional investor by aggregating all his past trades, starting from day -365 relative to the filing date. Second, on each day, I calculate each group’s position by summing the daily position across all investors in that group.

Third, I estimate the group’s return on each day. If the aggregate position for the group is positive, then the group’s return on that day is the abnormal return

of the target firm. If the aggregate position is zero, then the group’s return on that day is zero. If the aggregate position is negative, then the group’s return on that day is the abnormal return of the target firm multiplied by negative one. Finally, I estimate each group’s campaign return by cumulating the group’s daily

return over the campaign period, starting from one day after the filing date until 365 days after (filing date +1 day to filing date + 365 days).

23

𝑂𝑢𝑡𝑐𝑜𝑚𝑒 = 𝑊𝑜𝑙𝑓𝑃𝑎𝑐𝑘 + 𝐶𝑜𝑛𝑡𝑟𝑜𝑙𝑠 (5)

I use three measures of the outcome of the activist campaign. The first is based on the description of the

campaign’s objective and outcome provided in the “Comment” section of the SharkRepellent database. In

particular, I manually classify each campaign into one of the following seven categories: success, partial success,

settled, withdrawn, failed, ongoing, and not enough information (see Appendix B for examples). Then, after

removing campaigns coded as ongoing and not enough information (reducing the sample to 1,484 campaigns), I

create the indicator variable 𝑆𝑢𝑐𝑐𝑒𝑠𝑠 which is set to 1 if the campaign is coded as a success, partial success, settled,

or withdrawn (all indicating cases where the activist has likely achieved at least part of its objectives), and 0 if the

campaign is coded as failed.22

The second and third measures rely on the evidence that the most frequently sought-after objective by the

hedge fund activists is board representation (Klein and Zur 2009).23 In particular, the second is the indicator variable

𝑊𝑜𝑛_𝑆𝑒𝑎𝑡, which is set to 1 if the activists gain at least one board seat, and 0 otherwise (I define the variable only

for a subset of 716 campaigns in which the activist requested at least one board seat, as reported in SharkRepellent).

The third is # 𝑆𝑒𝑎𝑡𝑠 𝑊𝑜𝑛, which is the number of seats gained by the lead activist, as reported in SharkRepellent

(again, within the subset of 716 campaigns in which the activist requested at least one board seat). As shown in

Table 2, 74.4% of the campaigns are coded as 𝑆𝑢𝑐𝑐𝑒𝑠𝑠, and 𝑊𝑜𝑛_𝑆𝑒𝑎𝑡 is equal to 1 in 65.5% of the campaigns in

which the activists request board representation, with an average # 𝑆𝑒𝑎𝑡𝑠 𝑊𝑜𝑛 of 1.455.

As a proxy for the likely presence of a wolf pack, I use the indicator variable Wolf Pack, set equal to 1 if

𝑇𝑢𝑟𝑛𝑜𝑣𝑒𝑟_𝑂𝑡ℎ𝑒𝑟 on the trigger date is in the top quartile, and 0 otherwise. As shown in Table 2, the median value

for 𝑇𝑢𝑟𝑛𝑜𝑣𝑒𝑟_𝑂𝑡ℎ𝑒𝑟 is 1.08, implying that over 50% of the campaigns have some form of abnormal turnover on

the trigger date; the top quartile is selected to ensure that the indicator variable Wolf Pack is capturing campaigns

with a high level of abnormal turnover (in untabulated tests, I use two alternative definition to capture Wolf Pack:

(1) I redefine Wolf Pack campaigns as those that have higher than median level of 𝑇𝑢𝑟𝑛𝑜𝑣𝑒𝑟_𝑂𝑡ℎ𝑒𝑟, and (2) I

22 Results are similar if I exclude campaigns coded as settled or withdrawn and thus code as 𝑆𝑢𝑐𝑐𝑒𝑠𝑠 only campaigns coded as success or partial success. 23 In Klein and Zur (2009), changing board composition is the most sought-after objective by hedge fund activists and it makes up 27% of their sample. In my

sample, 37% of the campaigns requested board representation.

24

replace Wolf Pack with the continuous variable 𝑇𝑢𝑟𝑛𝑜𝑣𝑒𝑟_𝑂𝑡ℎ𝑒𝑟 . Both of these alternative measures yield

qualitatively similar results).

Since prior literature finds that proxy advisors influence shareholder votes (e.g., Ertimur, Ferri and Oesch

2013, Alexander, Chen, Seppi and Spatt 2010) and thus can affect activists’ campaigns, I control for the

recommendations by Institutional Shareholder Services (ISS) and Glass Lewis & Co. (GL), as reported in

SharkRepellent, for the subset of campaigns with proxy contests and available data on proxy advisors’

recommendations. The indicator variable ISS Recommendation (Glass Lewis Recommendation) takes on the value

of 1 when ISS (GL) makes a recommendation for the lead filer, and 0 otherwise.24 I also include an indicator variable

With ISS or GL Recommendations which has the value of 1 when ISS or GL issued a recommendation, to control

for the potential effect that the existence of an ISS and GL recommendation may have on the campaign outcome,

and 0 otherwise. I also control for a series of campaign-specific characteristics which have been shown to affect a

campaign’s outcome (see Shivdasani 1993, Ahmed and Duellman 2007, Anderson, Mansi and Reeb 2004). In

particular, Hostile Offer, Lawsuit, and Unsolicited Offer are indicator variables that take the value of 1 if the lead

activist makes a hostile offer, files a lawsuit, or makes an unsolicited offer, respectively, and 0 otherwise; Classified

Board is an indicator variable that takes the value of 1 if the target firm has a classified board, and 0 otherwise, and

Poison Pill is an indicator variable that takes the value of 1 if a poison pill was either in effect or adopted in response

to the 13D filer’s campaign, and 0 otherwise. The source for these variables is SharkRepellent (see Appendix A for

details).

Finally, my main variable of interest, Wolf Pack, is based on the turnover on the trigger date. Since part of

this turnover may be driven by changes in market conditions unrelated to wolf pack formation but somehow

correlated with campaign outcomes, I also include the same control variables from Eq. (1) so as to control for

correlated omitted variable biases. As discussed in Section 4.1.4, these variables relate to the arrival of news,

changes in liquidity, momentum, and other firm-specific characteristics.

24 SharkRepellent obtains the ISS and GL recommendation from press releases or proxy filings (usually the recipient of a positive recommendation will disclose

it; see Appendix A for more details). In my sample of 1,922 campaigns, there are 378 campaigns with ISS or GL recommendations.

25

Table 6 presents the results. As shown in Column (1), using a probit regression, the average marginal effect

of the Wolf Pack on 𝑆𝑢𝑐𝑐𝑒𝑠𝑠 is positive and significant at 7.5% (as noted earlier and in Table 2 the average success

rate in the sample is about 74%). Column (2) (probit regression) shows that the average marginal effect of Wolf

Pack on the likelihood of obtaining a board seat (𝑊𝑜𝑛_𝑆𝑒𝑎𝑡) is positive and significant at 8.5% (as noted earlier

and in Table 2, the average probability of winning a board seat is about 66%).

Column (3) shows a positive and significant association between Wolf Pack and # 𝑆𝑒𝑎𝑡𝑠 𝑊𝑜𝑛, with the

coefficient implying that campaigns accompanied by a wolf pack gain an average of 0.26 more board seats (as noted

earlier and in Table 2, the average number of seats won by an activist when requesting a board seat is 1.46 ).

One of the ultimate objectives for hedge fund activists is to earn a positive return on their stock holdings.

Therefore, in Column (4) I introduce an alternative measure of success: the buy-and-hold abnormal stock return

(BHAR) for the duration of the campaign. I use the 13D filing date as the first day of the activist campaign and the

end date reported by SharkRepellent as the last day of the campaign.25 The results indicate that the average BHAR

for a wolf pack campaign is 8.4% higher than a non-wolf pack campaign, as indicated by the positive and

statistically significant coefficient on Wolf Pack. In untabulated tests, I find similar results when using market-

adjusted returns (5.5% higher for wolf pack campaigns) or raw returns (6.9% higher).

[Insert Table 6]

5.2 Reverse causality?

Similar to other studies on hedge fund activism, this part of my analysis is subject to endogeneity concerns.

Prior literature finds that firms targeted by activists tend to perform better after the activist campaigns. However, it

remains possible that “improved performance” is because of the activist’s ability to select firms that will turn around

in the future, rather than a consequence of the activist’s intervention. In my study I find that, within targeted firms,

25 The end date as reported by SharkRepellent is the date that signals the end of a campaign. For campaigns that are associated with a proxy fight, this date is usually the date that the proxy fight went to a vote or otherwise ended. For nonproxy fight campaigns, this is the most logical date that signals the end of the

campaign. For example, if the activist requested the company to seek a buyer, the end date would be the date that the company agreed to be acquired. For

campaigns that are missing an end date in SharkRepellent, I use the filing date of the last schedule 13D/A as the end date of the campaign. The SEC requires 13D filers to report any material changes in holdings (1% or more) under the schedule 13D/A, and the last schedule 13D/A is usually filed because the activist

is unwinding his/her position.

26

those with a wolf pack perform better than those without it. The endogeneity concern in my setting is that pack

members are merely joining campaigns that are more likely to be successful, rather than causing such success.

However, if this were true, then I would expect to observe a greater frequency of wolf packs in campaigns

with a high level of expected payoffs. Klein and Zur (2009) correlate stock returns around the campaign’s

announcement date (the 13D filing date) with subsequent outcomes and conclude that the market can differentiate

ex-ante between successful and unsuccessful campaigns. Thus, I use the announcement date return as an ex-ante

measure for the campaign’s expected payoff, and I examine whether such a measure predicts the presence of a wolf

pack, as assumed by the selection story described above. In particular, I employ the following linear regression:

𝑊𝑜𝑙𝑓𝑃𝑎𝑐𝑘 = 13𝐷𝐴𝑛𝑛𝑜𝑢𝑛𝑐𝑒𝑚𝑒𝑛𝑡 𝑅𝑒𝑡𝑢𝑟𝑛 + 𝐶𝑜𝑛𝑡𝑟𝑜𝑙𝑠 (6)

As before, I proxy for the likely presence of a wolf pack using the indicator variable Wolf Pack, set equal to 1

if 𝑇𝑢𝑟𝑛𝑜𝑣𝑒𝑟_𝑂𝑡ℎ𝑒𝑟 on the trigger date is in the top quartile, and 0 otherwise. I measure the 13D Announcement

Return as the cumulated four-factor abnormal return starting from one trading day before the 13D filing date and

ending one trading day after. As shown in Table 7, the coefficient of 13D Announcement Return is not statistically

significant at conventional levels in either Column 1 (probit) or Column 2 (O.L.S.). Therefore, it does not appear

that pack members are selecting campaigns that are ex-ante more likely to be successful (although ex-post these

wolf pack campaigns are more successful), inconsistent with a reverse causality story. Thus, subject to the caveat

that my research design does not allow causal inferences, my evidence of an association between the presence of

wolf packs and subsequent campaign outcomes in Table 6 is at least consistent with the possibility that wolf packs

positively affect campaign outcomes.

[Insert Table 7]

6. Conclusion

I find evidence consistent with the existence of wolf packs by documenting share accumulation by other

investors before public disclosure of 13D filings and especially around the trigger date. I then examine two

competing hypotheses regarding the mechanism of wolf pack formation: spontaneous formation versus coordinated

effort. I find little support for the former: a large portion of the documented abnormal turnover remains unexplained

even after controlling for the arrival of news, momentum trading, liquidity, and other firm-specific factors. In

27

contrast, five pieces of evidence are consistent with the Coordinated Effort Hypothesis. First, investors who

accumulate shares prior to the 13D filing (i.e., the public announcement) are more likely to be those who have a

prior relationship with the lead activist. Second, wolf packs tend to concentrate in targets with poison pills and

better takeover defenses, i.e. cases in which the lead activist is more likely to benefit from the wolf pack’s support.

Third, wolf packs are more likely to form on the trigger date when the lead activist is done accumulating its position,

as one would expect if the pack was coordinated by the lead activist. Fourth, wolf pack members continue to support

and be involved with the campaign after the initial filing, as evidenced by their voting behavior and the profitability

of their trades during the campaign. Finally, the presence of wolf packs is associated with the success of the

campaign (in terms of whether the activist achieves his stated objectives and in terms of post-campaign stock

returns), suggesting that the use of wolf packs improves the activist’s probability of success. However, similar to

other hedge fund activism studies, I cannot fully rule out the possibility that wolf pack activists are merely selecting

better campaigns. My findings contribute to the literature on investor coordination and on hedge fund activism and

may be relevant to policymakers concerned with the implications of this form of activism.

28

References

Aboody D, Lev B (2000) Information asymmetry, R&D, and insider gains. The Journal of Finance 55: 2747-2766.

Admati AR, Pfleiderer P, Zechner J (1994) Large shareholder activism, risk sharing, and financial market

equilibrium. Journal of Political Economy 102: 1097-1130.

Ahmed AS, Duellman S (2007) Accounting conservatism and board of director characteristics: An empirical

analysis. Journal of Accounting and Economics 43: 411-437.

Alexander CR, Chen MA, Seppi DJ, Spatt CS (2010) Interim news and the role of proxy voting advice. Review of

Financial Studies 23: 4419-4454.

Amihud Y (2002) Illiquidity and stock returns: cross-section and time-series effects. Journal of Financial Markets

5: 31-56.

Anderson RC, Mansi SA, Reeb DM (2004) Board characteristics, accounting report integrity, and the cost of debt.

Journal of Accounting and Economics 37: 315-342.

Beaver WH (1968) The information content of annual earnings announcements. Journal of Accounting Research

6: 67-92.

Bebchuk L, Jackson R (2012) The law and economics of blockholder disclosure. Harvard Business Law Review 2:

39-60.

Bebchuk LA, Brav A, Jackson RJ Jr, Jiang W (2013) Pre-disclosure accumulations by activist investors: Evidence

and policy. Journal of Corporation Law 39: 1-34.

Becht M, Franks J, Grant J, Wagner H (2017) The returns to hedge fund activism: An international study. Review

of Financial Studies 30: 2933–2971.

Black B (1990) Shareholder passivity reexamined. Michigan Law Review 89: 520-608.

Brav A, Dasgupta A, Mathews R (2018) Wolf pack activism. Working Paper.

Brav A, Jiang W, Kim H (2009) Hedge fund activism: A review. Foundations and Trends® in Finance 4: 185-246.

Brav A, Jiang W, Partnoy F, Thomas R (2008) Hedge fund activism, corporate governance, and firm performance.

The Journal of Finance 63: 1729-1775.

Briggs TW (2007) Corporate governance and the new hedge fund activism: An empirical analysis. Journal of

Corporation Law 32: 681-723,725-738.

Bushee BJ (1998) The influence of institutional investors on myopic R&D investment behavior. The Accounting

Review 73: 305-333.

Celarier M (2013) Now, it gets ugly: Ackman to sic SEC on Soros ‘trade.’ New York Post. (August 1).

Chari V, Jagannathan R, Ofer A (1988) Seasonalities in security returns: The case of earnings announcements.

Journal of Financial Economics 21: 101-121.

29

Coffee J, Palia D (2016) The wolf at the door: The impact of hedge fund activism on corporate governance. Journal

of Corporation Law 41: 545-607.

Collin-Dufresne P, Fos V (2015) Do prices reveal the presence of informed trading? The Journal of Finance 70:

1555-1582.

Crane AD, Koch A, Michenaud S (2017) Institutional investor cliques and governance. Journal of Financial

Economics. Forthcoming.

DeFond M, Hung M, Trezevant R (2007) Investor protection and the information content of annual earnings

announcements: International evidence. Journal of Accounting and Economics 43, 37-67.

Dimson E, Karakas O, Li X (2015) Active ownership. The Review of Financial Studies 28: 3225-3268.

Ertimur Y, Ferri F, Oesch D (2013) Shareholder votes and proxy advisors: Evidence from say on pay. Journal of

Accounting Research 51: 951-996.

Gantchev N, Jotikasthir C (2017) Institutional trading and hedge fund activism. Management Science 64: 2473-

2972.

Godnick HG, Gussman W (2006) Beware the counterattack against activist investors: The group trap. Schulte Roth

& Zabel LLP.

Green TC (2006) The value of client access to analyst recommendations. Journal of Financial and Quantitative

Analysis 41: 1-24.

Hoffman L, Viswanatha A and Benoit D (2015) SEC probes activist funds over whether they secretly acted in

concert. The Wall Street Journal. (June 4).

Klein A, Zur E (2009) Entrepreneurial shareholder activism: Hedge funds and other private investors. The Journal

of Finance 64: 187-229.

Klein A, Zur E (2011) The impact of hedge fund activism on the target firm's existing bondholders. The Review of

Financial Studies 24: 1735-1771.

Klein AL, Li T (2015). Acquiring and trading on complex information: How hedge funds use the Freedom of

Information Act. Working Paper.

Kyle AS (1985) Continuous auctions and insider trading. Econometrica 53: 1315-1335.

McCahery J, Sautner Z, Starks L (2016) Behind the scenes: The corporate governance preferences of institutional

investors. The Journal of Finance 71: 2905-2932.

Orol R (2017) Bill Ackman is watching for these six Washington policies. The Street, Inc.

Pulliam S, Chung J, Benoit D, Barry R (2014) Activist investors often leak their plans to a favored few: Strategically

placed tips help build alliances for campaigns at target companies. The Wall Street Journal. (March 26).

Securities Exchange Act of 1934. United States of America.

Shivdasani A (1993) Board composition, ownership structure, and hostile takeovers. Journal of Accounting and

Economics 16: 167-198.

30

Shleifer A, Vishny R (1986) Large shareholders and corporate control. Journal of Political Economy 94: 461-488.

Song W, Szewczyk S (2003). Does coordinated institutional investor activism reverse the fortunes of

underperforming firms? The Journal of Financial and Quantitative Analysis 38: 317-336.

31

Appendix A: Variables definitions

Firm characteristics

Log(MV) Natural log of market value of the target firm.

Institution Holding This is the % of shares outstanding owned by all 13F institutions in the most recent

quarter.

Analyst Following The number of I/B/E/S analyst that issued an earnings forecast for the target firm

during the past quarter.

Institutional Salest The percentage of shares outstanding sold by institutional investors on day t, see

Gantchev and Jotikasthira (2017).

Amihudt This is the Amihud illiquidity ratio, estimated at date t.

10Kt Indicator variable which takes the value of 1 if a Form 10-K was filed on day t.

8Kt Indicator variable which takes the value of 1 if a Form 8-K was filed on day t.

10Qt Indicator variable which takes the value of 1 if a Form 10-Q was filed on day t.

Earnings

Announcementt

Indicator variable which takes the value of 1 if day t is the date of the quarterly

earnings announcement.

I/B/E/S forecastt Indicator variable which takes the value of 1 if an analyst in I/B/E/S issues an

earnings forecast on day t.

Management

Guidancet

Indicator variable which takes the value of 1 if management issues guidance on day

t.

# Newst The number of news items reported on the date t about the target firm in the 13D

filing. Obtained from Factiva searches.

# Newst-1 The number of news items reported on the date t-1 about the target firm in the 13D

filing. Obtained from Factiva searches.

# Newst-2 The number of news items reported on the date t-2 about the target firm in the 13D

filing. Obtained from Factiva searches.

Abn_Rett The abnormal return estimated using the four-factor model (SML, HML, Market,

Momentum) on the trigger date.

Abn_Rett-1 The abnormal return estimated using the four-factor model (SML, HML, Market,

Momentum) one trading date prior to the trigger date.

Volt-1 Normal trading volume estimated as the average of volume between trading day -

120 to day -60.

Bid-Ask Spreadt The absolute difference between the bid and ask on date t.

13D filer-related variables

13D Filer Holdings The % of shares outstanding held by the 13D filer at filing date.

13D Filer Shares -

60 to filing date

This is the % of common shares accumulated by the 13D filer from 60 days prior to

filing date to the filing date.

Filing Delay The number of days between trigger date and filing date.

32

Proxies for coordination between lead activists and wolf pack members

Past Relationship The number of times that a particular institution has participated in a prior campaign

led by that particular activist within the last year; an institution is treated as a

participant if that institution purchased shares on the trigger date of the previous

campaign.

Poison Pill Indicator variable obtained from SharkRepellent: takes the value of 1 if a poison pill

was either in effect or adopted in response to the 13D filer's campaign, 0 otherwise.

Bullet Proof Rating A proprietary rating by FactSet that measures how well a company is defended; a

high bullet proof index suggests that the company is well defended.

13D Filer Done Indicator variable which takes the value of 1 if the 13D filer has finished his position

on the trigger date (the 13D filer trades no shares between trigger date+1 and filing

date).

Proxies for outcome of the activist campaign

Success Indicator variable: takes the value of 1 if activist achieved at least a part of their

demand, settled the campaign, or withdrawn the campaign, 0 if the campaign failed.

Won Seat Indicator variable: takes the value of 1 if activist won at least one seat, 0 otherwise.

# Board Seat Won This is the number of board seats won by the lead activist as reported in

SharkRepellent.

Proxies for the presence of Wolf Packs

Turnover_Othertrigger

date

(Total Trading Volumetriggerdate-13D Filer Tradetriggerdate)/(Normal Trading

Volumetriggerdate).

Wolf Pack Indicator variable: takes the value of 1 if " Turnover_Othertrigger date " is in the upper

quartile, 0 otherwise.

Determinants of campaign outcome

Glass Lewis

Recommendation

Indicator variable obtained from SharkRepellent: takes the value of 1 if Glass Lewis

make a recommendation for the 13D filer, 0 otherwise. Usually the winning side of

the recommendation will disclose the ISS recommendation in a press release or

proxy filing, but this is not always the case. When the Glass Lewis recommendation

is not available I code this as 0. For a board seat proxy fight, if Glass Lewis

recommends half or more of the dissident nominees, SharkRepellent will consider

Glass Lewis as supporting the activist. In cases in which the proxy fight or other

activist campaign include both proposals and director nominations, SharkRepellent

assigns more weight to Glass Lewis’ recommendation regarding director elections

in determining which part Glass Lewis’ supports.

ISS

Recommendation

Indicator variable obtained from SharkRepellent: takes the value of 1 if ISS make a

recommendation for the 13D filer, 0 otherwise. Usually the winning side of the

recommendation will disclose the ISS recommendation in a press release or proxy

filing, but this is not always the case. When the ISS recommendation is not available

I code this as 0. For a board seat proxy fight, if ISS recommends half or more of the

dissident nominees, SharkRepellent will consider ISS as supporting the activist. In

cases where the proxy fight or other activist campaign include both proposals and

director nominations, SharkRepellent assigns more weight to ISS’ recommendation

regarding director elections in determining which part ISS supports.

33

With ISS and GL

Recommendation

An indicator variable which takes the value of 1 if ISS or Glass Lewis has a

recommendation available.

Classified Board Indicator variable obtained from SharkRepellent: takes the value of 1 if board is

classified, 0 otherwise.

Unsolicited Offer Indicator variable obtained from SharkRepellent: takes the value of 1 if an

unsolicited offer is made, 0 otherwise.

Hostile Offer Indicator variable obtained from SharkRepellent: takes the value of 1 if a hostile

offer is made, 0 otherwise.

Lawsuit Indicator variable obtained from SharkRepellent: takes the value of 1 if a lawsuit is

filed, 0 otherwise.

Letter to

Shareholder

Indicator variable obtained from SharkRepellent: takes the value of 1 if a letter to

shareholder was made, 0 otherwise.

34

Appendix B: Examples from SharkRepellent database

Classifications Comments

Successful Example 1: Campaign to maximize shareholder value included a letter to

management proposing that company engage in a share repurchase program. Shortly

after campaign initiation, the board announced cost reduction measures and a

500,000 share repurchase program.

Example 2: Dissident campaign included a notice that it may speak to management

regarding Board representation and business plans. The dissident group later

suggested one dissident representative to replace Board member who resigned, and

this representative was elected.

Example 3: Dissident conducted discussions with company regarding operational

and structural changes, including a spin-off. Company announced plans to spin off

its Financial Services business and later expanded its Board to appoint five Dissident

recommended individuals.

Partly

successful

Example 1: Dissident won three of four seats up for election to 13-person Board;

ISP never launched tender offer after the company refused to exempt the tender offer

from company's "poison pill" and Delaware freeze-out provision.

Example 2: Annual meeting proxy fight to replace four of 10 directors settled for

three seats on the 11-seat board. Previously, Starboard had started a written consent

solicitation, and company adopted poison pill with 15% trigger after Starboard

accumulated 14.8% stake.

Example 3: Roumell has nominated two candidates to the six-seat board for 2014

annual meeting. Company agreed to add one nominee (previously, Roumell urged

company to repurchase shares. After third party made an unsolicited offer, Roumell

urged for sale process.).

Settled Example 1: Proxy fight to elect two dissident nominees settled. As part of the

settlement agreement, the company agreed to repurchase 111,000 common shares

from the dissident for $20.25 per share (a premium of 11.75% over the preceding

30-day average market price).

Example 2: Proxy fight settled. As part of the settlement agreement, Board size

increased from 10 to 13, and three dissidents elected to Board.

Example 3: Longview, 9% holder, urged PETM to review strategic alternatives and

explore the sale of the company. Company reviewed strategic alternatives and then

agreed to be acquired by a private group led by BC Partners.

Example 4: Maguire, 5.3% holder, requested one board seat and urged company to

replace CEO, implement cost restructuring plan, and review strategic alternatives,

including a sale. Parties entered into a settlement agreement providing for the

mutually agreeable director.

Withdrawn Example 1: Proxy fight for three seats on eight-seat board was voluntarily

withdrawn after Progress announced a plan to divest assets and buy back shares, as

Starboard had requested.

35

Example 2: Proxy fight for two seats on the seven-person board at the 2009 annual

meeting was withdrawn; dissident decided not to nominate its candidates at the

annual meeting.

Example 3: Dissident campaign urged the company to seek a sale of its subsidiary

banks, otherwise threatened to seek board representation. Dissident withdrew

campaign after the company announced it agreed to be acquired by First Financial

Corporation.

Fail Example 1: Lenox's one director nominee was defeated at the 2010 annual meeting.

Although its proposal requesting board declassification received more votes cast for

its approval than against it, the proposal was defeated after counting abstentions as

votes against.

Example 2: Campaign urged Board to enhance shareholder value and specifically

liquidate company's investment in Ready Mix, Inc. Despite dissident's opposition,

management nominees were elected at the annual meeting and the shareholder

proposal defeated.

Example 3: Dissident two-person slate not elected, as no nominee received required

majority of votes present at the meeting; incumbent directors thus continued to serve.

Western's nonbinding declassification proposal passed.

Example 4: Proxy fight for one board seat was unsuccessful. Company had adopted

10% trigger poison pill in response to Biglari's 9.3% stake.

Not Enough

Information

Example 1: 13D filer - No publicly disclosed activism

Ongoing Example 1: Basswood Capital, 9.03% holder, disclosed it may engage in

discussions with the company's board, management, other shareholders, industry

observers, and potential acquirers regarding the company's future plans to increase

shareholder value.

Example 2: Dissident campaign included a letter to the board advising it to seek a

director candidate from its larger shareholders. Dissident also advised the company

to have at least one conference call each year in which all shareholders can

participate.

Example 3: Atlantic, 5.1% holder, disclosed that it engaged and would continue to

engage in discussions with Oil States' management and board regarding the

company's business, corporate governance, and board composition for the purpose

of increasing shareholder value.

36

Figure 1 - Timeline for 13D filing

Figure 1 is a timeline for a typical 13D filing. The filing date is the date on which the 13D is submitted to the SEC

and made publicly available. The trigger date is the date on which the 13D filer triggered the filing requirement.

There are two triggers for 13D filings: (1) the investor acquires more than 5% of any class of security of a publicly

traded company, and (2) the investor has an interest in influencing the management of the company. Once both

triggering events are satisfied, the investor has up to 10 days to file Form 13D with the SEC.

Trigger Date

37

Figure 2 - Total turnover around the trigger date

Figure 2 shows the average daily share turnover (𝑇𝑢𝑟𝑛𝑜𝑣𝑒𝑟_𝐴𝑙𝑙) over the 60-day period at the time of the trigger

date (the day on which the activist triggers the 13D filing requirement). 𝑇𝑢𝑟𝑛𝑜𝑣𝑒𝑟_𝐴𝑙𝑙 =𝑉𝑜𝑙𝑖,𝑡

𝐴𝑣𝑔(𝑉𝑜𝑙𝑖,𝑡−120…..𝑉𝑜𝑙𝑖,𝑡−60) is

the total daily trading volume deflated by the normal trading volume for the firm. Normal trading volume

𝐴𝑣𝑔(𝑉𝑜𝑙𝑖,𝑡−120 … . . 𝑉𝑜𝑙𝑖,𝑡−60) is calculated as the rolling average of -120 to -60 days trading volume.

𝑇𝑢𝑟𝑛𝑜𝑣𝑒𝑟_𝐴𝑙𝑙 =1 implies that there is no abnormal trading on that particular day. The figure includes 1,922 activist

events between 1998 and 2014 (see Table 1). All variables are winsorized at the 1% and 99% level.

0

0.5

1

1.5

2

2.5

3

3.5

-30 -28 -26 -24 -22 -20 -18 -16 -14 -12 -10 -8 -6 -4 -2 0 2 4 6 8 10 12 14 16 18 20 22 24 26 28 30

𝑇𝑢𝑟𝑛𝑜𝑣𝑒𝑟

_𝐴𝑙𝑙

Trading Days Relative to Trigger Date

38

Figure 3 - Turnover by other investors around the trigger date: full sample

Figure 3 shows the average daily share turnover by other investors (𝑇𝑢𝑟𝑛𝑜𝑣𝑒𝑟_𝑂𝑡ℎ𝑒𝑟𝑡) at the time of the trigger

date (the day on which the activist triggers the 13D filing requirement). 𝑇𝑢𝑟𝑛𝑜𝑣𝑒𝑟_𝑂𝑡ℎ𝑒𝑟𝑡 =𝑂𝑡ℎ𝑒𝑟_𝑉𝑜𝑙𝑖,𝑡

𝐴𝑣𝑔(𝑉𝑜𝑙𝑖,𝑡−120…..𝑉𝑜𝑙𝑖,𝑡−60) is the daily trading volume by other investors (total volume net of the volume traded by the

13D filer) deflated by the normal trading volume for the firm. In particular, Other_Vol is calculated as the total

trading daily volume less the daily volume traded by the activist (manually collected from the schedule 13D).

Normal trading volume 𝐴𝑣𝑔(𝑉𝑜𝑙𝑖,𝑡−120 … . . 𝑉𝑜𝑙𝑖,𝑡−60) is calculated as the rolling average of -120 to -60 days

trading volume. 𝑇𝑢𝑟𝑛𝑜𝑣𝑒𝑟_𝑂𝑡ℎ𝑒𝑟 =1 implies that there is no unexplained abnormal trading on that particular day.

The figure includes 1,922 activist events between 1998 and 2014 (see Table 1). All variables are winsorized at the

1% and 99% level. Note: because 13D filers are only required to disclose their trading up until the filing date,

whenever a filer files prior to the 10-day cutoff, for the purpose of this figure I assume that the 13D filer makes no

trade between the filing date and day +10 from the trigger date.

0

0.5

1

1.5

2

2.5

3

-30 -28 -26 -24 -22 -20 -18 -16 -14 -12 -10 -8 -6 -4 -2 0 2 4 6 8 10

𝑇𝑢𝑟𝑛𝑜𝑣𝑒𝑟

_𝑂𝑡ℎ𝑒𝑟

Trading Days Relative to Trgger Date

39

Figure 4 - Turnover by other investors around the trigger date: subset of campaigns with no news

Figure 4 shows the average daily share turnover by other investors (𝑇𝑢𝑟𝑛𝑜𝑣𝑒𝑟_𝑂𝑡ℎ𝑒𝑟𝑡, defined in Figure 3) at the

time of the trigger date (the day on which the activist triggers the 13D filing requirement) for the subset of 751

campaigns with no public news in the 11 days up to and including the trigger date, based on a Factiva news search.

Note: because 13D filers are only required to disclose their trading up until the filing date, whenever a filer files

prior to the 10 days cutoff, for the purpose of this figure I assume that the 13D filer makes no trade between the

filing date and day +10 from the trigger date.

0

0.5

1

1.5

2

2.5

-30 -28 -26 -24 -22 -20 -18 -16 -14 -12 -10 -8 -6 -4 -2 0 2 4 6 8 10

Turn

ove

r_O

ther

Trading Days Relative to Trigger Date

40

Figure 5 - Turnover by other investors around the trigger date: subset of campaigns with no

trades by the 13D filer on the trigger date (13G switchers)

Figure 5 shows the average daily share turnover by other investors (𝑇𝑢𝑟𝑛𝑜𝑣𝑒𝑟_𝑂𝑡ℎ𝑒𝑟𝑡, defined in Figure 3) at the

time of the trigger date (the day on which the activist triggers the 13D filing requirement) for the subset of 351

campaigns without any trading by the 13D filer on the trigger date (these are cases in which the activist already

owns 5% of the shares but triggers the filing requirement because it changes its intent from ‘passive’ to ‘active’).

Note: because 13D filers are only required to disclose their trading up until the filing date, whenever a filer files

prior to the 10 days cutoff, for the purpose of this figure I assume that the 13D filer makes no trade between the

filing date and day +10 from the trigger date.

0

0.5

1

1.5

2

2.5

3

-30 -28 -26 -24 -22 -20 -18 -16 -14 -12 -10 -8 -6 -4 -2 0 2 4 6 8 10

𝑇𝑢𝑟𝑛𝑜𝑣𝑒𝑟

_𝑂𝑡ℎ𝑒𝑟

Trading Days Relative to Trigger Date

41

Figure 6 - Turnover by other investors around the trigger date: subset of campaigns with low

information asymmetry

Figure 6 shows the average daily share turnover by other investors (𝑇𝑢𝑟𝑛𝑜𝑣𝑒𝑟_𝑂𝑡ℎ𝑒𝑟𝑡, defined in Figure 3) at the

time of the trigger date (the day on which the activist triggers the 13D filing requirement) for the subset of 450

campaigns in which the level of information asymmetry is low. Note: because 13D filers are only required to

disclose their trading up until the filing date, whenever a filer files prior to the 10 days cutoff, for the purpose of

this figure I assume that the 13D filer makes no trade between the filing date and day +10 from the trigger date.

0

0.5

1

1.5

2

2.5

-30 -28 -26 -24 -22 -20 -18 -16 -14 -12 -10 -8 -6 -4 -2 0 2 4 6 8 10

Turn

ove

r_O

ther

Trading Days Relative to Trigger Date

42

Table 1 - Sample selection

Selection procedure Campaigns removed Total

SharkRepellent.net Campaigns between 1998 to 2014 3744

Campaigns where trigger date = filing date 304 3440

Campaigns where no 13D is found 151 3289

Missing variables from CRSP/Compustat 201 3088

Missing trigger date 528 2560

Remove SIC 6798, 6770, 6792, ADRs 267 2293

Subsequent campaigns 196 2097

Initial Campaigns with subsequent campaigns 170 1927

13D transaction table is at lower-than-daily frequency 5 1922

43

Table 2 - Descriptive statistics

N Mean Median Std.

Firm characteristics

MV (millions) 1,922 933.556 209.348 2651.816

Institution Holding 1,922 0.440 0.444 0.361

Analyst Following 1,922 3.271 2.000 5.090

13D filer-related variables

13D Filer Holdings filing date 1,922 0.088 0.063 0.067

13D Filer Shares -60 to filing date 1,922 0.054 0.040 0.068

Filing Delay 1,922 7.612 9.000 3.740

Proxies for outcome of the activist campaign

Success 1,484 0.744 1.000 0.437

Won Seat 716 0.655 1.000 0.476

# Seats Won 716 1.455 1.000 1.348

Proxies for the presence of Wolf Packs

Turnover Other trigger date 1,922 2.450 1.080 5.204

Wolf Pack 1,922 0.250 0.000 0.433

Determinants of campaign outcome

Glass Lewis Recommendation 1,922 0.020 0.000 0.098

ISS Recommendation 1,922 0.042 0.000 0.154

Classified Board 1,922 0.469 0.000 0.499

Unsolicited Offer 1,922 0.041 0.000 0.199

Hostile Offer 1,922 0.014 0.000 0.116

Lawsuit 1,922 0.039 0.000 0.194

Letter to Shareholder 1,922 0.077 0.000 0.266

44

Table 3 - Mechanism of wolf pack formation: changes in market conditions

Panel A

(1) All (2) NoNews (3) 13G

Switchers

(4) Low

Information

Asymmetry

Intercept 1.046 *** 1.038 *** 1.044 *** 1.048 *** [13.532] [13.463] [13.589] [12.751]

Main Variable Trigger date 1.231 ***

[7.932] Trigger date & (No News)

(13GSwitcher) (Low Infor.

Asymmetry)

1.194 *** 1.209 *** 1.225 ***

[5.416] [3.484] [4.359]

Trigger date & (With News) (Non-

Switcher) (High Infor. Asymmetry)

1.361 *** 1.284 *** 1.231 *** [8.282] [5.961] [4.46]

Liquidity Institutional Salest 0.783 *** 0.793 *** 0.785 *** 0.769 *** [5.197] [5.236] [5.207] [5.231]

Log (MV) -0.064 *** -0.064 *** -0.064 *** -0.064 *** [15.459] [14.124] [15.421] [14.093]

Amihudt 0.004 0.003 0.004 0.003 [0.544] [0.389] [0.603] [0.445]

Arrival of News and Momentum 10Kt 0.212 ** 0.197 ** 0.214 ** 0.198 ** [2.094] [2.007] [2.105] [2.021]

8Kt 0.601 *** 0.539 *** 0.605 *** 0.538 *** [5.236] [4.681] [5.23] [4.652]

10Qt 0.171 *** 0.163 *** 0.172 *** 0.165 *** [3.732] [3.448] [3.744] [3.439]

Earnings Announcementt 0.275 ** 0.038 ** 0.293 ** 0.039 *** [2.079] [0.355] [2.123] [0.364]

I/B/E/S forecastt 0.418 *** 0.408 *** 0.414 *** 0.409 *** [4.1] [3.354] [4.046] [3.377]

Management Guidancet 0.326 ** 0.326 ** 0.323 ** 0.324 ** [2.408] [2.576] [2.393] [2.559]

# Newst 0.021 ** 0.019 ** 0.021 ** 0.019 ** [2.01] [2.124] [2.009] [2.123]

# Newst-1 -0.014 *** -0.012 *** -0.014 *** -0.012 *** [3.003] [2.817] [3.008] [2.817]

# Newst-2 -0.004 -0.003 -0.004 -0.003 [1.112] [1.182] [1.093] [1.17]

Abn_Rett 5.070 *** 4.949 *** 5.087 *** 4.947 *** [3.074] [3.052] [3.088] [3.056]

Abn_Rett-1 -0.763 *** -0.681 *** -0.760 *** -0.682 *** [3.036] [2.657] [3.033] [2.663]

Volt-1 0.004 *** 0.004 *** 0.004 *** 0.004 *** [6.614] [6.496] [6.586] [6.504]

45

Other firm characteristics Bid Ask Spreadt -0.163 *** -0.173 *** -0.164 *** -0.174 *** [2.679] [2.991] [2.696] [3.022] 13D Filer Holdings 0.002 * 0.002 * 0.002 0.002 * [1.671] [1.833] [1.476] [1.849]

Institution Holding -0.115 *** -0.115 *** -0.115 *** -0.113 *** [3.447] [3.378] [3.431] [3.33]

Analyst Following -0.005 ** -0.005 ** -0.006 *** -0.005 ** [2.526] [2.280]

[2.568] [2.26]

Adj. R^2 0.114 0.110

0.111

0.101

0.110

0.108

0.107

0.105 Adj. R^2 without FE

No. (13GSwitcher)(No

News)(BigFirms)(LowIntangible) N/A 751 351 450

No. (Non-Switcher)(With

News)(SmallFirms)(HighIntangible) N/A 1,171 1,571 1,472

No. Total Campaigns 1,922 No. Observations 115,320 Fixed Effects Year, Industry, Weekday Std. Error Cluster Activist Firm

Panel B

All No vs. Have News 13G vs. Non-

Switchers

Low vs. High Infor.

Asymmetry

Difference between Coefficients [F Value] N/A 0.167 *** 0.075 -0.006

[5.160] [0.010] [-0.020]

Panel A presents a pooled campaign-day regression with 115,320 observations; each observation represents one

trading date within the 60 days prior to the 13D filing, and a total of 1,922 unique campaigns are represented. The

campaigns are obtained from SharkRepellent database; I removed campaigns with the same trigger date and filing

date, and campaigns with multiple 13Ds. The estimation period is from January 1998 to December 2014. Columns

(1) – (4) are estimated using standard O.L.S. regression standard errors are clustered by activist and firm. The

dependent variable 𝑇𝑢𝑟𝑛𝑜𝑣𝑒𝑟_𝑂𝑡ℎ𝑒𝑟𝑡 (𝑇𝑜𝑡𝑎𝑙 𝑇𝑟𝑎𝑑𝑖𝑛𝑔 𝑉𝑜𝑙𝑢𝑚𝑒−13𝐷 𝐹𝑖𝑙𝑒𝑟 𝑇𝑟𝑎𝑑𝑒

𝑁𝑜𝑟𝑚𝑎𝑙 𝑇𝑟𝑎𝑑𝑖𝑛𝑔 𝑉𝑜𝑙𝑢𝑚𝑒) is calculated on the date t, if there is

no abnormal turnover this variable would equal 1. In Column (1), the main variable of interest Trigger_date is an

indicator variable equal to 1 if that date is a trigger date (the date which trigger the filing obligation for 13D filers),

and 0 otherwise. In Columns (2) – (4), the main variable of interest Trigger_date & NoNews (13G switcher)( Low

Infor. Asymmetry) is an indicator variable equal to 1 if the 13D campaign is either a without news (13G switcher)

(low information asymmetry) campaign and that date is a trigger date, and 0 otherwise. Panel B presents the

corresponding F Test between the coefficients Trigger_date & NoNews (13GSwitcher)(Low Infor. Asymmetry) and

Trigger_date & With News (Non-Switcher)(High Infor. Asymmetry). Note: because 13D filers are only required to

disclose their trading for 60 calendar days prior to the filing date, therefore not all filers provide their trading

information between “day – 60 from trigger date” and “day – 60 from filing date,” for the purpose of the regressions

in Table 3, I assume that the 13D filers make no trade within this period. All variables are winsorized at the extreme

at the 1% and 99% level. For a description of the control variables, please refer to Appendix A.

46

Table 4 – Pre-campaign evidence of coordination

Panel A: The role of prior relationships between lead activist and pack members

Dependent Variable Buying Buying

(1) Probit [dydx] (2) O.L.S.

Past Relationship 0.079 * 0.081 ***

[2.38] [2.58]

Controls All the controls variables in Table 3 are also included

Adj. [Pseudo] R^2 [0.030] 0.021

Fixed Effects FF 12 Industries FF 12 Industries

Std. Cluster None Activist, Firm

No. Campaigns 1,233 1,233

No. Observations 3,553 3,553

Panel B: Position accumulated by lead activist and wolf pack formation

Dependent Variable 𝑇𝑢𝑟𝑛𝑜𝑣𝑒𝑟_𝑂𝑡ℎ𝑒𝑟𝑇𝑟𝑖𝑔𝑔𝑒𝑟𝐷𝑎𝑡𝑒 Spike

(1) O.L.S. (2) Probit [dydx]

13D Filer Done 0.359 * 0.054 **

[1.734] [2.5]

Controls All the controls variables in Table 3 are also included here

Adj. [Pseudo] R^2 0.115 [0.071]

Fixed Effects FF 12 Industries FF 12 Industries

Std. Cluster Activist, Firm None

No. Campaigns 1,922 1,922

No. Observations 1,922 1,922

Panel C: Takeover defenses and wolf pack formation

Dependent Variable 𝑇𝑢𝑟𝑛𝑜𝑣𝑒𝑟_𝑂𝑡ℎ𝑒𝑟𝑇𝑟𝑖𝑔𝑔𝑒𝑟𝐷𝑎𝑡𝑒 𝑇𝑢𝑟𝑛𝑜𝑣𝑒𝑟_𝑂𝑡ℎ𝑒𝑟𝑇𝑟𝑖𝑔𝑔𝑒𝑟𝐷𝑎𝑡𝑒

(1) O.L.S. (2) O.L.S.

Poison Pill

0.502 *

[1.803]

Bullet Proof Ratings 0.026 * [1.488]

Controls All the controls variables in Table 3 are also included here

Adj. [Pseudo] R^2 0.115 0.109

Fixed Effects FF 12 Industries FF 12 Industries

Std. Cluster Activist, Firm Activist, Firm

No. Campaigns 1,922 863

No. Observations 1,922 863

Panel A is based on 1,233 campaigns between 1999 and 2010, inclusive. Trading data and the identity of the

institutional investors executing each trade are obtained from a consulting firm. The main variable Past Relationship

is the number of times that a particular institution has participated in a prior campaign led by that particular activist

within the last year; an institution is treated as a participant if that institution purchased shares on the trigger date

of the previous campaign. The dependent variable Buying is an indicator variable which takes the value of 1 if the

institution purchased shares in the target firm on the trigger date, and 0 otherwise. Panel B is based on 1,922

campaigns as described in Figure 2. In Column (1) the dependent variable 𝑇𝑢𝑟𝑛𝑜𝑣𝑒𝑟_𝑂𝑡ℎ𝑒𝑟𝑇𝑟𝑖𝑔𝑔𝑒𝑟𝐷𝑎𝑡𝑒 is

(𝑇𝑜𝑡𝑎𝑙 𝑇𝑟𝑎𝑑𝑖𝑛𝑔 𝑉𝑜𝑙𝑢𝑚𝑒−13𝐷 𝐹𝑖𝑙𝑒𝑟 𝑇𝑟𝑎𝑑𝑒

𝑁𝑜𝑟𝑚𝑎𝑙 𝑇𝑟𝑎𝑑𝑖𝑛𝑔 𝑉𝑜𝑙𝑢𝑚𝑒) calculated on the trigger date; if there is no abnormal turnover this variable

equals 1. The main variable is 13D Filer Done, an indicator variable which takes the value of 1 if the 13D filer has

finished his position on the trigger date (the 13D filer trades no shares between trigger date +1 and filing date). In

47

Column (2) the dependent variable is Spike, an indicator variable which takes the value of 1 if 𝑇𝑢𝑟𝑛𝑜𝑣𝑒𝑟_0𝑡ℎ𝑒𝑟 is

above 1 on the trigger date. Panel C is based on 1,922 campaigns as described in Figure 2. In Column (1) the main

variable of interest is Poison Pill, an indicator variable which takes the value of 1 if either the target adopted a

poison pill in response to the campaign or a poison pill was already in place. Column (2) includes 863 campaigns

with bullet proof index provided by FactSet. The main variable of interest is Bullet Proof Rating, a proprietary

rating by FactSet that measures how well a company is defended; a high bullet proof index suggests that the

company is well defended. In Columns (1) and (2) the dependent variable is 𝑇𝑢𝑟𝑛𝑜𝑣𝑒𝑟_𝑂𝑡ℎ𝑒𝑟𝑇𝑟𝑖𝑔𝑔𝑒𝑟𝐷𝑎𝑡𝑒. All

variables are winsorized at the extreme 1% level. For a description of the control variables, please refer to Appendix

A.

48

Table 5 – Post-campaign evidence of coordination

Panel A: Voting behavior of wolf pack members

Wolf Pack Non-Wolf Pack Differences t-stat/F-stat

Campaign-Mutual Fund

Observations 116 69

% vote for management

Mean 77% 88% -11%** -1.67

Median 100% 100% 0%* -1.42

Panel B: Post-campaign abnormal returns by wolf pack members

Wolf Pack Non-Wolf Pack Differences t-stat/F-stat

Campaign-Investor

Observations 1,598 15,424

Post-campaign abnormal return

Mean 8.18% 5.61% 2.57%** 1.86

Median 3.51% 0.28% 3.23%* 1.56

Panel C: Post-campaign abnormal returns by wolf pack members with prior relationship

Wolf Pack +

Prior Relationship

Wolf Pack +

No Prior Relationship Differences t-stat/F-stat

Campaign-Investor

Observations 852 1,280

Post-campaign abnormal return

Mean 9.17% 6.97% 2.19%* 1.73

Median 4.52% 1.75% 2.77%* 1.31

Panel A above is based on 185 campaign-mutual fund observations (52 unique campaigns) between 2003 and 2010,

inclusive. It is the interaction between the trading data obtained from a consulting firm and ISS Voting Analytics.

The unit of analysis is the percentage of votes by each mutual fund for each firm each year. The sample is divided

into two groups, those that purchased shares on the trigger date (wolf pack) and those that did not (non-wolf pack).

The main variable % of vote for management is the number of items that the mutual fund voted in favor of

management, divided by the total number of items up for votes that year. Only votes which are cast within three

years of the campaigns’ filing date are included. Panel B is based on 17,022 campaign-investor observations. I

include only investors who have made at least one trade within the two-year period starting from one year prior to

the 13D trigger date and ending one year after (trigger date - 365 days to trigger date + 365 days). Panel C is based

on 1,598 campaign-investor observations; only investors who purchased shares on the trigger date are included

(wolf pack). The sample is divided into two groups: those that have a prior relationship with the lead activist and

those that do not. In Panels B and C, the main variable Post-campaign abnormal return is the cumulated abnormal

return earned by each group between filing date + 1 to filing date + 365.26

26The groups’ return is calculated as follows: First, I estimate the daily position of each institutional investor by aggregating all his past trades. Second, on each day, I calculate each group’s position by summing the daily position across all investors in that group. Third, I estimate the group’s return on each day.

If the aggregate position for the group is positive, then the group’s return on that day is the abnormal return of the target firm. If the aggregate position is zero,

then the group’s return on that day is zero. If the aggregate position is negative, then the group’s return on that day is the abnormal return of the target firm multiplied by -1. Finally, I estimate each group’s campaign return by cumulating the group’s daily return over the campaign period, starting from one day after

the filing date until 365 days after (filing date +1 day to filing date + 365 days).

49

Table 6 – Presence of wolf packs and campaign outcomes

Dependent variable Success Won seats # Seats B.H.A.R.

(1) Probit [dydx] (2) Probit [dydx] (3) O.L.S. (4) O.L.S.

Wolf Pack 0.075 ** 0.085 *** 0.258 *** 0.084 ***

[2.219] [3.021] [2.615] [2.632]

Proxy advisors

ISS Recommendation 0.069 0.182 *** 0.540 *** -0.097 **

[1.502] [4.901] [3.434] [-1.116]

Glass Lewis Recommendation 0.116 ** -0.053 0.289 0.092 **

[2.264] [1.233] [1.246] [1.028]

With ISS or GL Recommendation -0.179 *** 0.134 *** 0.030 0.020

[4.38] [3.346] [0.122] [0.539]

Activist tactics

Hostile Offer -0.030 -0.108 0.377 0.093

[0.32] [1.248] [0.559] [0.662]

Lawsuit 0.047 0.067 -0.160 -0.010

[1.012] [1.322] [0.689] [0.215]

Classified Board -0.015 -0.017 -0.331 *** -0.027

[0.718] [1.032] [3.274] [1.029]

Poison Pill -0.019 0.045 ** 0.204 0.053

[0.819] [2.445] [1.196] [1.639]

Unsolicited Offer -0.256 *** 0.024 -0.389 * 0.056

[3.779] [0.483] [1.727] [0.651]

Liquidity

Institutional Sales trigger date 1.827 *** 2.011 ** 5.806 0.419

[2.725] [2.402] [2.259] [1.527]

Log (MV) 0.010 -0.022 *** 0.025 0.505

[1.067] [2.77] [0.518] [1.409]

Amihud trigger date 0.000 0.001 -0.006 -0.001

[0.241] [0.451] [0.694] [1.497]

Arrival of news

10K trigger date 0.174 *** 0.083 -0.084 -2.727 ***

[2.628] [0.562] [0.249] [3.892]

8K trigger date -0.279 -0.234 1.097 -0.009

[1.391] [1.129] [1.82] [0.66]

10Q trigger date -0.043 0.164 * -0.116 0.017

[0.405] [1.91] [0.4] [0.892]

Earnings Announcement trigger date 0.087 0.254 -0.693 ** 0.221

[0.822] [1.293] [1.979] [1.115]

I/B/E/S forecast trigger date 0.042 -0.078 -0.321 0.127

[0.625] [1.071] [1.035] [0.839]

Management Guidance trigger date 0.155 *** 0.114 1.148 0.228

[2.59] [0.641] [0.87] [0.848]

# News trigger date -0.003 -0.001 -0.011 -0.089

[1.229] [0.495] [1.458] [0.823]

# News trigger date-1 0.004 -0.001 0.015 -0.011

[1.403] [0.544] [0.709] [0.135]

# News trigger date-2 -0.004 0.001 0.048 -0.095

[0.834] [0.759] [1.335] [0.602]

50

Abn_Ret trigger date -0.036 -0.106 -0.992 -0.004 *

[0.244] [1.296] [0.975] [1.677]

Abn_Ret trigger date -1 0.189 0.047 0.828 0.000

[0.828] [0.239] [0.631] [0.062]

Vol trigger date -1 0.000 0.000 0.005 * 0.020

[0.541] [1.322] [1.719] [1.599]

Other firm characteristics

Bid Ask Spread trigger date 0.043 -0.012 0.113 0.249 **

[0.486] [0.198] [0.186] [2.126]

13D Filer Holdings 0.001 0.006 *** 0.021 * -0.002

[0.584] [3.282] [1.73] [0.671]

Institution Holding 0.061 * -0.008 0.165 0.122 **

[1.849] [0.293] [0.731] [2.178]

Analyst Following 0.001 0.013 *** -0.001 0.005

[0.439] [4.662] [0.043] [1.148]

Adj. [Pseudo] R^2 [0.089] [0.119] 0.124 0.056

Fixed Effect None None FF 12 Industries FF 12 Industries

No. Observations 1,484 716 716 1,484

Std. Error Cluster None None Activist, Firm Activist, Firm

In the table above, the estimation period is from January 1998 to December 2014. The main variable of interest is

Wolf Pack, an indicator variable which takes the value of 1 when 𝑇𝑢𝑟𝑛𝑜𝑣𝑒𝑟_0𝑡ℎ𝑒𝑟𝑡𝑟𝑖𝑔𝑔𝑒𝑟_𝑑𝑎𝑡𝑒 is in the top quartile,

and 0 otherwise. Columns (1) and (4) include 1,484 campaigns with sufficient information in SharkRepellent

database to determine the campaign outcome. The dependent variable Success is an indicator variable which equals

1 if the activist achieved at least part of what was requested, and 0 otherwise. Columns (2) and (3) include 716

campaigns in which the lead activist requested at least one board seat. In Columns (2) the dependent variable won

seats is an indicator variable which equals 1 if the activist won at least one board seat, and 0 otherwise. In Column

(3) the dependent variable #seats is the raw number of seats won by the lead activist. All variables are winsorized

at the extreme 1% level. For a description of the control variables, please refer to Appendix A. In Column (4) the

dependent variable B.H.A.R. is the four-factor (momentum, size, book to market and market factors) abnormal

return for the underlying target, starting from the 13D filing date to the last date of the campaign. The last date of

the campaign is the campaign end date as reported in SharkRepellent. The abnormal returns are estimated using the

standard two-step method; the estimation window is 255 days, ending 46 days before the return date. All variables

are winsorized at the extreme 1% level. For a description of the control variables, please refer to Appendix A.

51

Table 7 – Assessing reverse causality

Dependent variable Wolf Pack Wolf Pack

(1) Probit [dydx] (2) O.L.S.

13D Announcement Return 0.089 0.075 [0.6] [0.348] Liquidity

Institutional Sales trigger date 0.126 *** 0.368 ** [3.78] [2.275]

Log (MV) -0.015 -0.018 [1.22] [1.363]

Amihud trigger date -0.043 ** -0.036 *** [2.16] [3.025]

Arrival of news

10K trigger date 0.101 0.063 [0.49] [0.202]

8K trigger date -0.281 -0.343 * [1.49] [1.71]

10Q trigger date 0.229 *** 0.227 ** [2.64] [2.166]

Earnings Announcement trigger date -0.173 -0.174 [0.74] [0.62]

I/B/E/S forecast trigger date 0.016 0.084 [0.22] [0.743]

Management Guidance trigger date 0.521 *** 0.543 *** [3.454] [3.622]

# News trigger date 0.014 *** 0.013 *** [6.46] [3.689]

# News trigger date-1 -0.003 -0.004 [0.94] [0.573]

# News trigger date-2 0.001 0.000 [0.11] [0.011]

Abn_Ret trigger date -0.024 -0.109 [0.11] [0.352]

Abn_Ret trigger date -1 -0.573 -0.451 [1.64] [1.147]

Vol trigger date -1 0.003 *** 0.002 *** [3.64] [4.055]

Other firm characteristics

Bid Ask Spread trigger date 0.134 0.189 [1.22] [1.561]

13D Filer Holdings -0.002 -0.001 [1.01] [0.757]

Institution Holding -0.046 -0.029 [0.75] [0.402]

Analyst Following -0.003 -0.003 [0.8] [0.721]

Adj. [Pseudo] R^2 [0.121] 0.092

Fixed Effects FF 12 Industries

52

No. Observations 1,484 1,484

Std. Error Cluster None Activist, Firm

Table 7 above includes the same sample of campaigns as in Table 6. The dependent variable Wolf Pack is an

indicator variable which takes the value of 1 when 𝑇𝑢𝑟𝑛𝑜𝑣𝑒𝑟_0𝑡ℎ𝑒𝑟𝑡𝑟𝑖𝑔𝑔𝑒𝑟_𝑑𝑎𝑡𝑒 is in the top quartile, and zero

otherwise. The main variable of interest 13D Announcement Return is the three-day, four-factor abnormal return

centered around the 13D filing date (filing day-1 to filing day+1). All variables are winsorized at the extreme 1%

level. For a description of the control variables, please refer to Appendix A.