Reputation and Investor Activism
Abstract:
1 Introduction
Investor activism is an important mechanism by which a company's shareholders can affect corporate decision making. To legally compel corporate managers to change policies, an activist investor must engage in a proxy fight during which they solicit support for their proposals from other shareholders. Proxy fights involve significant costs and therefore are infrequent: 91% of the activist campaigns in our sample do not involve proxy fights.1 Given the infrequency of proxy fights, a natural intuition is that target managers would rarely concede to activist demands. However, we show that activist campaigns without proxy fights are surprisingly effective. This leads to an important question: how do activist campaigns affect corporate policy despite the infrequency of observed proxy fights?
We argue campaigns by high reputation activists affect corporate policy by threatening, explicitly or implicitly, to initiate a proxy fight. Target managers "settle" with activists when the threat of a proxy fight is sufficiently credible because proxy fights are costly to the firm and its managers. Rather than incur the costs of fighting, managers prefer to settle by increasing dividends, repurchasing shares, changing board composition, engaging in merger or acquisition activity, or enacting other changes to the firm's operational, financial, or governance structure. This mechanism is summarized well by the following quote that describes the most hostile proxy contests during 2014:
"What is also noteworthy about the 30 or so contests ...is that with the exception of Trian, none of the biggest shareholder hedge fund activists ...are involved. For the most part this is because companies settled quickly in the face of attacks by these giants." - New York Times DealBook 4/21/2015
For the threat of a proxy fight to be effective, target managers must believe the threat is credible, which we argue is related to the activist's reputation in addition to simple observable characteristics like their size. We model reputation arising due to an information asymmetry in which managers do not know the activist's cost associated with proxy fights (their "type") and must infer it from observed behavior. We formally define reputation as the posterior belief about an activist's type conditional on previous interactions, as first suggested in Kreps & Wilson (1982) and Milgrom & Roberts (1982). The activist's type encompasses the monetary costs, non-monetary costs such as effort, and any offsetting benefits such as enjoyment associated with proxy fights. Consistent with our model, prominent activists often make public statements advertising their low subjective cost of proxy fights:
"I enjoy the hunt much more than the `good life' after the victory." - Carl Icahn
"We will pay any price, bear any burden, meet any hardship." - William Ackman, quoting John F. Kennedy when discussing a proxy fight with Target Corp.
In our model, an activist sequentially engages two management teams. In each engagement, the activist buys a portion of the target firm's shares, and proposes a positive NPV project that would involve a private cost to the manager. The manager then decides whether to settle and incur the private cost to pursue the project, or refuse and risk provoking a proxy fight with the activist. Managers settle with high reputation activists because, conditional on having to do the project, the manager would rather not go through a proxy fight. Managers refuse low reputation activists because if the activist decides not to fight, the manager avoids the costs associated with both the project and the proxy fight.
In addition to showing how reputation affects target managers' decision to settle, our model illustrates how activists build reputation. Managers in the second activist campaign estimate the activist's type based on the outcome of the first campaign. Therefore, activists without a strong prior reputation can improve their reputation by initiating a proxy fight. Activists with a sufficiently strong prior reputation, by contrast, are settled with by both managers. We therefore predict that activists with a recent track record of initiating proxy fights, or extracting settlements from targets, are more successful in future campaigns.
Activists in our model can also signal their type through the size of the stake they purchase in the target, which is costly because their demand for shares moves prices upwards. However, stake size alone is not a sufficient signal of reputation because if it were, managers would settle upon observing large positions, making the activist's type irrelevant and attracting campaigns from weak as well as strong activists. We therefore predict that the fraction of the target firm the activist purchases is not related to campaign success.
We find strong empirical support for our model's predictions using a sample of 2,199 activist campaigns, each initiated by an activist filing form 13-D with the SEC. Our first main result is that campaigns by high reputation activists are more successful even when they do not feature a proxy fight. To measure campaign success, we analyze corporate 8-K filings using S&P's Capital IQ data and examine the propensity of activist targets to increase payouts, change management, change board composition, engage in a merger or acquisition, or otherwise reorganize. We construct two empirical proxies for activist reputation motivated by our model: a dummy variable indicating the activist recently engaged a different target firm in a proxy fight (Recent Fight), and the average success of recent campaigns by the activist that did not feature a proxy fight (Settle Rate).
We find campaigns are at least 28% more successful when the activist has a Recent Fight and 7% more successful when the activist has a one-standard deviation higher Settle Rate. These effects hold regardless of whether the campaign involves a proxy fight, supporting our model's prediction that managers are more likely to settle with high reputation activists.
Our results show that reputation helps explain variation in activist campaign success across different categories of institutional investors. Consistent with the evidence in Gillan & Starks (2007) and Klein & Zur (2009), we find that hedge funds are more prolific and successful activists than those with different institutional structures. Kahan & Rock (2007), Yermack et al. (2010), Gantchev (2013), and others argue that hedge funds are more effective activists because they have fewer conflicts of interest, and because lighter regulation allows for lock-up provisions and greater portfolio concentration.2 We argue these characteristics make hedge funds more successful activists by allowing them the flexibility to build reputation and the time horizon to benefit from it. We provide empirical evidence supporting this channel by showing hedge funds have higher average Recent Fight and Settle Rate than other activists.
We also document the surprising result that campaigns in which the activist purchases a larger fraction of the target firm's shares outstanding are no more successful than those in which activists purchase smaller stakes. Theories on large shareholders, for example in Shleifer & Vishny (1986) and Admati et al. (1994), predict larger positions make shareholders better able to influence manager behavior. In our model, by contrast, because weak activists imitate the equilibrium quantity of shares purchased by strong activists, we predict larger positions have no influence on manager behavior. Instead, consistent with our empirical evidence, the only way activists successfully improve their reputation and induce managers to settle is by engaging in proxy fights.
Finally, we argue that even the observed 9% frequency of proxy fights in activist campaigns is puzzling in a static setting because more than half occur in campaigns where the activist's position in the target is smaller than $50 million, meaning they would need the target firm's stock price to rise 21% just to offset the $10.7 million cost that Gantchev (2013) estimates activists incur for proxy contests. Average target stock returns around proxy fights are far below 21%, with estimates ranging from 4% to 8% (see Brav et al. (2008), Brav et al. (2009), and Gantchev (2013)), suggesting many activists in our sample incur losses when engaging in proxy fights. While puzzling in a static setting, we show that in a dynamic setting activists sometimes initiate proxy fights when expecting a net loss as an investment in reputation that allows them to profit from settling with future targets. Our empirical evidence that campaigns by activists with a recent fight are more successful indicates this investment in reputation pays off.
In our empirical results, we address a natural alternative hypothesis: activists with a track record of proxy fighting, or being settled with, are better at picking stocks that will subsequently institute the changes we measure as campaign success regardless of whether the activist intervenes. Under this hypothesis, high reputation activist investors do not cause their targets to act through activism, but instead predict and profit from these actions as non-activist speculators. We address this possibility directly by estimating and controlling for the expected number of actions the target would take in the absence of the activist. Another related hypothesis is that there are persistent differences in activist skill unrelated to reputation. This hypothesis would explain the correlation between the success rate of past non-proxy campaigns, Settle Rate, and future campaign success. However, this hypothesis would not explain why the propensity to engage in proxy fights, Recent Fight, predicts campaign success after controlling for past campaign success and other measures of activist skill. We further address this alternative, and the stock-picking alternative, by showing that within-activist variation in reputation is positively related to within-activist variation in campaign success.
2 Literature Review
Our model draws from seminal theories defining reputation as the posterior belief about an agent's type conditional on previous interactions, as first suggested in Kreps & Wilson (1982) and Milgrom & Roberts (1982). The key feature of these models that differentiates them from the costly signaling framework originating in Spence (1973) is that the weaker types successfully imitate stronger types in a dynamic setting by taking costly reputation-building actions. This reputation concept has been applied to many settings (e.g. debt issuance in Diamond (1989) and investment banking in Chemmanur & Fulghieri (1994)), but to our knowledge we are the first to apply it to investor activism.
Our model adds a unique perspective to the theoretical literature on investor activism, which primarily focuses on activism and monitoring for a single firm by a large shareholder (examples include Burkart et al. (1997), Maug (1998), Aghion et al. (2004), and Admati & Pfleiderer (2009)). In these papers, large shareholders are effective monitors and activists because they reduce the free-rider problem associated with activists capturing only a fraction of the value they create. Like in our model, activists in Admati & Pfleiderer (2009) use a credible threat to convince managers to act without a shareholder vote, namely the threat to sell their shares and cause a price decline (the "Wall Street Walk"). Our model differs because we study the threat of a proxy fight and use a repeated game with information asymmetry to endogenize the credibility of the threat.
More-recent work argues large blockholders may be less effective activists because they are non-competitive (Edmans & Manso (2011)) and are often money managers whose concern with their own investors makes the threat of exit non-credible (Dasgupta & Piacentino (2015)). Levit (2014) extends this literature by examining communication and exit as alternate channels to avoid the costs associated with proxy fights, and Corum & Levit (2016) studies the role of activists in facilitating takeovers.
A key ingredient in our model and the discussion of our empirical results is the cost of a proxy fight. Gantchev (2013) estimates the net cost to activists of each stage of a campaign using a system of nested logistic regressions based on a sequential decision model. Because the goal of the Gantchev (2013) model is to estimate these costs, while the goal of our model is to describe the role of reputation in the strategic interaction between managers and activists, the two models are quite different. The Gantchev (2013) model features a single target firm, a more granular negotiation process, and estimates rather than models the decision functions of both parties. In contrast, we model the dynamics of multiple campaigns, collapse the negotiation to one decision by each party, and most importantly endogenize the activist and target firm decisions in a strategic equilibrium framework.
Our empirical analysis supports and extends a rapidly growing literature on investor activism, which is largely focused on hedge funds. Our results suggest activist campaigns, particularly hedge fund campaigns, catalyze significant actions at target firms, consistent with the findings in Brav et al. (2008) and many other papers surveyed in Brav et al. (2009). Two contemporaneous papers, Krishnan et al. (2016) and Boyson et al. (2016), examine activist hedge fund reputation and experience empirically. Krishnan et al. (2016) measures activist reputation by the dollar value of recent activist positions. They find that short-term stock returns and long-term firm performance are both stronger following interventions by high-reputation funds. Boyson et al. (2016) measures experience using the number of prior campaigns in their sample initiated by the hedge fund, and shows activists with more experience produce larger announcement CARs in the future as well as better long-term firm performance.
Our paper differs from the contemporaneous empirical papers on the impact of reputation and experience on activism's success in at least four important ways. First, we provide a theoretical framework to define and understand activist reputation. Second, we use unique empirical measures of reputation motivated by our model. Third, we include non-hedge fund activists in our sample and compare them with hedge funds in terms of reputation and campaign success. Fourth, we show our reputation measures remain significant in a horse-race regression for predicting campaign success with alternative measures of reputation, skill, and experience.
3.1 Description
There are three agents, an activist A and two managers M1 and M2, all risk-neutral. Each manager controls a company whose shares outstanding are normalized to one and whose existing asset values are normalized to zero. The managers have access to a project with net present value (NPV) $$ V>0$$ that they do not take without intervention by A because it entails private cost $$ B>V$$.3 The activist A approaches the managers sequentially, playing the stage game in Model Figure 1 with M1 and then with M2.
Model Figure 1: Stage game tree
This figure details the stage-game that underlies the two activist-manager interactions that comprise our model. The payoffs for each of the three possible outcomes are described, in detail, in the Model section.
In each stage game, A moves first and buys Q shares of M1's company in the secondary market for a per-share price $$ \lambda Q$$, where $$ \lambda>0$$ is the stock's illiquidity. Any $$ Q>0$$ constitutes an implicit threat to engage in a proxy fight if the manager forgoes the project. The manager then decides whether or not to undertake the project (settle or refuse).
If the manager settles, they undertake the project and the game ends, making the payoffs:
| $$\displaystyle \left[ \Pi_{A,\textit{settle}}, \Pi_{M,\textit{settle}} \right]$$ | $$\displaystyle = \left[ Q\left(V-\lambda Q \right), V - B \right].$$ | (1) |
We assume for simplicity that the managers internalize the full benefit of the project V, perhaps due to performance-based bonuses, making their net payoff when the project occurs V-B. Our equilibrium results are identical up to scaling constants under the alternative assumption that the manager only internalizes a fraction $$ 0 < \phi < 1$$ of the benefits, making their net payoff from the project $$ \phi V - B$$.
If the manager refuses, A decides whether to initiate a proxy fight (fight or fold). For simplicity, we assume proxy fights are always successful and therefore result in firm value increasing by V.4 However, proxy fights also have private costs for both the activist ($$ F_A>0$$) and the manager ($$ F_M>0$$). Therefore, if A decides to fight, the payoffs are:
| $$\displaystyle \left[ \Pi_{A,\textit{fight}}, \Pi_{M,\textit{fight}} \right]$$ | $$\displaystyle = \left[ Q\left(V-\lambda Q \right) - F_A , V - B - F_M \right].$$ | (2) |
If the manager refuses and the activist does not fight (fold), the manager ignores the project and the payoffs are:
| $$\displaystyle \left[ \Pi_{A,\textit{fold}}, \Pi_{M,\textit{fold}} \right]$$ | $$\displaystyle = \left[ Q \left( - \lambda Q \right), 0 \right].$$ | (3) |
All parameters are common knowledge except A's cost of fighting FA, which we refer to as their "type" and takes one of two values: $$ F_L < F_H$$. Activists with cost FL are more likely to fight and we therefore refer to them as aggressive and activists with cost FH as cautious. Aggressive activists may have lower costs associated with proxy fights because they have the knowledge and experience necessary to more-easily initiate a successful fight. Alternatively, they can be interpreted as "commitment types" (as in Fudenberg & Levine (1992) or Schmidt (1993)) that initiate proxy fights because they intrinsically enjoy the conflict and attention that comes with them.
Our main construct of interest is reputation, which we define as the probability A is aggressive conditional on the outcome of the first game:
| $$\displaystyle r_1 \equiv \mathbb{P}\left( F_A = F_L \middle \vert \text{ outcome of stage game with } M_1 \right).$$ | (4) |
To compute r1, the manager also relies on the unconditional probability A is aggressive, $$ r_0 \equiv \mathbb{P}\left( F_A = F_L \right)$$, a parameter we refer to as initial reputation.
3.2 Second Stage Equilibria
We solve the model using backward induction, beginning with the second stage game and then turning our attention to the first stage game. There are two second-stage equilibria, the "low reputation equilibrium" that prevails for low r1 and the "high reputation equilibrium" for high r1. Theorem 1 describes the low reputation equilibrium.
- Aggressive A choose $$ Q^* \equiv \frac{V}{2\lambda} + \frac{F_L}{2V}$$ and fight if M2 refuses.
- Cautious A choose $$ Q^*$$ with probability $$ q_{A,2} \equiv \frac{F_M}{B-V} \frac{r_1}{1-r_1}$$, $$ Q = 0$$ with probability $$ 1-q_{A,2}$$, and fold if M2 refuses.
- M2 refuses when $$ Q < Q^*$$ or $$ Q^* < Q < \frac{F_H}{V}$$, and settles when $$ Q\geq \frac{F_H}{V}$$. When $$ Q = Q^*$$, M2 settles with probability $$ y_2^* = \frac{1}{2} + F_L \frac{\lambda}{2V^2}$$.
Aggressive A play a simple pure strategy, fighting if M2 refuses and choosing $$ Q^*$$, the quantity for which the manager settles with non-zero probability. Cautious A play a mixed strategy, "bluffing" by choosing $$ Q^*$$ with probability $$ q_{A,2}$$ and sitting out by choosing Q=0 with probability $$ 1-q_{A,2}$$. Cautious A are indifferent between these strategies because M2 also plays a mixed strategy, settling with probability $$ \frac{1}{2} + F_L \frac{\lambda}{2V^2}$$ in response to $$ Q = Q^*$$. If M2 refuses, the cost FH is too high and so the cautious types fold. Given these strategies, Q=0 reveals that A is the cautious type and M2 optimally refuses, while $$ Q = Q^*$$ indicates an aggressive type with probability $$ \frac{B-V}{B-V+F_M}$$, making M2 indifferent between settling and refusing. Proofs of Theorem 1 and our other equilibrium results are in Appendix A, along with a discussion of M2's off-equilibrium beliefs.
In the high reputation equilibrium, the probability A is the aggressive type is high enough that M2 prefers settling if aggressive and cautious types pool at a single quantity of shares. As a result, the high reputation equilibrium is as characterized in Theorem 2.
- Aggressive A choose $$ Q^*(\Pi_H)$$ and fight if M2 refuses.
- Cautious A choose $$ Q^*(\Pi_H)$$ and fold if M2 refuses.
- M2 refuses when $$ Q < Q^*(\Pi_H)$$ or $$ Q^*(\Pi_H) < Q < \frac{F_H}{V}$$, and settles when $$ Q\geq \frac{F_H}{V}$$. When $$ Q = Q^*(\Pi_H)$$, M2 settles with probability $$ y_2^*(\Pi_H) \geq \frac{1}{2} + F_L \frac{\lambda}{2V^2}$$.5 When $$ r_1 = \frac{B-V}{B-V+F_M}$$, any $$ \Pi_H \in \left[0,\frac{V^2}{4 \lambda}\right]$$ and corresponding $$ Q^*(\Pi_H)$$ and $$ y_2^*(\Pi_H)$$ are an equilibrium. When $$ r_1 > \frac{B-V}{B-V+F_M}$$, $$ \Pi_H = \frac{V^2}{4 \lambda}$$, $$ Q^* = \frac{V}{2 \lambda}$$, and $$ y_2^* = 1$$.
There are two key differences between the high and low reputation equilibria. The first is the cautious type bluffs with probability one in the high reputation equilibrium. The second key difference is, because A is very likely to be aggressive, M2 weakly prefers settling and does so with a higher probability than in the low reputation equilibrium. If r1 is strictly greater than the cutoff $$ \frac{B-V}{B-V+F_M}$$, M2 settles with probability one. If r1 equals $$ \frac{B-V}{B-V+F_M}$$, we have a set of admissible equilibrium where M2 settles with probability $$ y_2^*(\Pi_H) \geq \frac{1}{2} + F_L \frac{\lambda}{2V^2}$$ and cautious A receive any expected payoff $$ \Pi_H$$ between 0 and $$ \frac{V^2}{4\lambda}$$. This flexibility allows us to solve the dynamic model.
3.3 First Stage Equilibria
We now characterize the equilibria in the first stage game between A and M1. The equilibria are different than the second stage game because A internalizes the impact of their actions on r1. As detailed in Appendix A, r1 only affects A's expected payoffs in the second stage to the extent it influences whether the high reputation equilibrium prevails. Therefore, in the first stage game, A considers the extent to which their decisions affect whether r1 is below, at, or above the high reputation cutoff, $$ \frac{B-V}{B-V+F_M}$$.
Like the second stage game, the first stage game has two equilibria, one for low r0 and one for high r0. Theorem 3 describes the low initial reputation equilibrium.
- Aggressive A choose $$ Q^* \equiv \frac{V}{2\lambda} + \frac{F_L}{2V}$$ and fight if M1 refuses.
- Cautious A choose $$ Q^*$$ with probability $$ q_{A,1} \equiv \frac{r_0}{1-r_0} \frac{F_M}{B-V} \frac{2(B-V) + F_M}{B-V}$$ and $$ Q = 0$$ with probability $$ 1-q_{A,1}$$, and fold if M1 refuses when Q=0. They fight with probability $$ q_R \equiv \frac{B-V}{2(B-V)+F_M}$$ if M1 refuses when $$ Q = Q^*$$.
- M1 refuses when $$ Q < Q^*$$ or $$ Q^* < Q < \frac{F_H}{V}$$, and settles when $$ Q\geq \frac{F_H}{V}$$. When $$ Q = Q^*$$, M1 settles with probability $$ y_1^* = \frac{1}{2} + F_L \frac{\lambda}{2V^2}$$.
- If M1 settles, $$ r_0 < r_1 <\frac{B-V}{B-V+F_M}$$ and the low reputation equilibrium prevails with M2. If M1 refuses and A folds, $$ r_1 = 0$$ and the low reputation equilibrium prevails with M2. If M1 refuses and A chooses to fight, $$ r_1 = \frac{B-V}{B-V+F_M}$$ and the high reputation equilibrium prevails with M2 where $$ \Pi_H = F_H - \frac{v^2}{2 \lambda} - \frac{F_L}{2}$$.
The first stage game between A and M1 is different from the second stage game in one critical way: cautious A sometimes choose to fight. By fighting, cautious A increase r1 and benefit from playing the high reputation equilibrium in the second stage game. In equilibrium, cautious A are indifferent between fighting and not in the first stage because the increase in their expected second stage payout and their share of the project's NPV exactly offsets their private cost FH. Therefore, they play a mixed strategy after choosing $$ Q = Q^*$$ and being refused by M1, fighting for reputation with probability qR. As a result of these strategies, the low reputation equilibrium prevails in the second stage unless M1 refuses and A chooses to fight.
The high initial reputation equilibrium occurs when r0 is large enough M1 prefers to settle when both types always purchase the same quantity, resulting in the pooling equilibrium described in Theorem 4.
- Aggressive A choose $$ Q^* \equiv \frac{V}{2\lambda}$$ and fight if M1 refuses.
- Cautious A choose $$ Q^*$$ and fight with probability $$ q_R \equiv \frac{B-V}{2(B-V)+F_M}$$ if M1 refuses.
- M1 refuses when $$ Q < Q^*$$ and settles when $$ Q \geq Q^*$$.
- Reputation $$ r_1 = r_0$$, and the corresponding equilibrium prevails in the second stage.
The key feature of the high reputation equilibrium is that both types of activist always choose $$ Q^*$$, and M1 always settles. This equilibrium can be thought of as describing the interaction between managers and activists with pre-existing strong reputations.
3.4 Empirical Predictions
The primary contribution of the model is to formalize our notion of activist reputation, provide a mechanism for its development over time, and show how it makes activism an effective governance tool even in campaigns that do not result in a proxy fight. In our model, the only ways A can achieve a high enough reputation that the manager settles with high probability in the second stage game is by choosing to fight in the first stage game, or by having previously built a strong enough reputation to play the high reputation equilibrium in both games. This leads to the following empirical predictions:
An important aspect of Empirical Prediction 1 is that high reputation activists are more successful in future campaigns through two channels. The first is a mechanical channel, that high reputation activists are more likely to initiate proxy fights when refused. The second is an equilibrium channel, that managers are more likely to settle with high reputation activists. We distinguish between these channels in our empirical analysis by controlling for whether a campaign features a proxy fight.
Our model also illustrates why aggressive activists cannot credibly signal through the size of the stake they buy in the target. The reason is the cost of the signal, the liquidity premium $$ \lambda Q$$ paid when buying shares, is identical for both types of activists.6 As a result, the cautious activist always imitates the aggressive activist and deviates from potential separating equilibria in which there is a quantity only the aggressive activist buys and the manager settles upon observing. Our model therefore predicts that, controlling for reputation, position size has no effect on managers' likelihood of settling.
A final contribution of the model is to provide an explanation for the puzzling mismatch between activist hedge fund spending on proxy fights, which Gantchev (2013) estimates is $10.7m, and their median position sizes in these cases, around $50m. Because equilibrium payoffs are large for high reputation activists, it can be optimal to fight in the first stage game even if the cost FA is too high for it to be profitable in a static environment.
4.1 Sample of Activist Campaigns
The sample for our main empirical tests consists of 2,199 activist campaigns, each comprised of an activist-target pair, which we assemble using activist 13-D and 13-F filings.7 We initially identify 133,799 13-D filings between 2000 and 2014 using the Wharton Research Data Services (WRDS) SEC Analytics tool. Of these, we hand-match 5,442 13-D filer Central Index Key (CIK) identifiers with an institutional asset manager from the Thomson Reuters S34 13-F holdings file, and match 18,109 registrant (target) CIK identifiers with a security from the Center for Research in Security Prices (CRSP) monthly file.8 The intersection of our activist and target matches yields a preliminary sample of 4,647 13-D filings.
In cases where multiple activists file 13-Ds for the same target on the same day, we collapse the multi-activist consortium to a single observation for the purposes of our tests. In some campaigns, other activists initiate follow-on campaigns for the same target within one year. Following Brav et al. (2016) and the popular press, we refer to these as `wolf packs' and define the binary variable Wolf Packj,i,t+1 to indicate a multi-activist campaign. Multi-activist consortia and wolf packs comprise roughly 11% of the activist campaigns in our sample, and excluding these campaigns does not alter our results. We exclude 963 13-D filings in which the CRSP classification for the target security is not common equity (CRSP share codes 10 or 11) and exclude 16 13-D filings involving special acquisition companies (SPACs, SIC codes 6770 and 6726). We remove these campaigns because the outcome variables we study (defined below) are not relevant for closed-end mutual funds and SPACs. See Bradley et al. (2010) analysis of activist campaigns of this type. We also exclude 13-D filings in which target companies do not have covariate data in Compustat during the quarter of the activist's 13-D filing. These filters and refinements result in a sample of 2,199 13-D filings involving 367 unique activists and 1,805 unique target firms, summarized in Table 1, which is among the most comprehensive assembled to date.9
4.2 Sample of Activist Proxy Fights
We identify activist campaigns that involve proxy fights by constructing a sample of activist proxy filings using similar procedures to those outlined above for our 13-D sample. Using the WRDS SEC analytics tool we identify 188,160 proxy filings between 2000 and 2015.10 We apply similar filters as outlined above for our 13-D sample, resulting in a sample of 1,991 proxy filings. We use this sample to compute the binary variable Proxyj,i,t+1, which indicates the activist fund j that filed a 13-D with respect to target company i at date t also filed a proxy document with the SEC with respect to company i within the following year.
We find that these 1,991 proxy filings correspond to only 329 unique proxy contests, which implies that activists file over six proxies during the average proxy fight. Proxy contests generally involve multiple proxy filings as the activist and the target respond to one another's analysis and recommendations. We find that 198 (9% ) of the 2,199 campaigns in our sample involve proxy contests, which is very close to the fraction in Gantchev (2013).
4.3 Target Variables: Measures of Activist Campaign Success
For each target i in our sample we construct Actioni, t+1 to measure activist campaign success using target company 8-K filings via the S&P Capital IQ database. Capital IQ classifies 8-K filings into `key development types' according to the information they contain. We use these classifications to calculate five dummy variables indicating events related to activist campaign success: payout increases (Payouti, t+1, key development types 36, 94, 214, and 232), CEO turnover (CEOi, t+1, type 101), business reorganizations (Reorgi, t+1, type 21), mergers and acquisitions (M&Ai, t+1, type 80), and board turnover (Boardi, t+1, type 172). Each variable indicates target i files at least one event-classified 8-K in the year following date t. We calculate Actioni, t+1 by summing the five event indicator variables.
We define activism success without regard for the activist's initial demands because negotiations often result in different actions by target firms than those initially demanded by the activist, and ex-ante demands are often not well defined.11 Table 1 shows summary statistics for Actioni, t+1, as well as the five sub-category indicator variables outlined above.
We also use Actioni, t+1 to construct $$ \widehat{Action}_{i,t+1}$$ for each target firm to measure the propensity of each target firm to take the actions we interpret as campaign success without activist intervention. We construct $$ \widehat{Action}_{i,t+1}$$ by estimating the predictive regressions outlined in Equation (5) using the entire Compustat cross-section in each quarter t:
| $$\displaystyle Action_{i,t+1}$$ | $$\displaystyle = \alpha + \beta\,X_{i,t} + \epsilon_{i,t},$$ | (5) |
where $$ X_{i,t}$$ is a vector of the target firm characteristics described in Appendix B. We define $$ \widehat{Action}_{i,t+1}$$ for each of the activist targets in our sample as the predicted value for $$ Action_{i,t+1}$$ based on estimates of Equation (5) for quarter t using all firms in Compustat. Table 1 shows summary statistics for $$ \widehat{Action}_{i,t+1}$$, as well as the characterstics in $$ X_{i,t}$$.
4.4 Activist Variables: Measures of Reputation and Signaling
We use our model to inform the construction of three variables characterizing the activist's reputation and stake size at the beginning of each campaign. Our first reputation measure, Recent Fightj,t, captures reputation building in our model by measuring the recent propensity of activist j to engage in a proxy fight with target firms. Our model suggests the ideal measure would be the number of recent proxy fights scaled by the number of times that the activist made demands and was refused. However, such a denominator is unobservable. We therefore rely on a simpler measure, defining Recent Fightj,t as a dummy variable indicting that activist fund j initiated a proxy fight in the year prior to t. In addition to capturing reputation building in our model, this measure reflects the extensive SEC filings described above which tend to be widely covered in the financial press and are therefore likely to be salient to target management teams. Our second activist reputation measure, Settle Ratej,t, captures pre-existing reputation in our model by measuring the activist's success rate in recent campaigns not involving a proxy fight. We define Settle Ratej,t as the one-year rolling total Actioni, t+1 from activist j's non-proxy campaigns divided by the average number of campaigns initiated by activist j in the three years prior to date t.12 Because we only include actions in campaigns for which activist j has not initiated a proxy fight, we interpret Settle Ratej,t as a measure of the frequency with which previous targets of the activist settled without a proxy context.
For campaigns in which an activist is filing a 13-D for the first time in our sample period, we assign the unconditional sample average for both Settle Ratej,t and Recent Fightj,t. This convention is consistent with our model, in which a target firm interacting with an activist in the first period assigns the activist an average reputation (r0).
An alternative mechanism by which activists can signal their level of seriousness to target managers in our model is the size of the stake in the target firm they purchase. We measure this directly, defining Target Sharej,i,t as the maximum fraction of the target firm i's shares outstanding activist j owns over the first year of the campaign. Table 1 shows the median Target Sharej,i,t is 6.8%, which comfortably exceeds the 5% ownership threshold required for activists to file a 13-D. We exclude campaigns in which the activist holds more than 50% of the shares of the target. Our results are robust to dropping the 9% of our sample in which the activist owns more than 20% of the target.
We also classify each activist by their institutional structure: hedge fund, mutual fund, private equity fund, broker-dealer, or "other" by hand-matching institutional asset managers from the Thomson Reuters holdings data with asset managers from the Factset LionsShares holdings data. Our sample includes only six campaigns from pension fund activists, including one campaign involving CalPERS, which we classify as "other."13 We present summary statistics for all activist variables in Table 1.
5 Results
Table 1 shows that the average value of Proxyj,i,t+1 is 9%, indicating that 91% of the campaigns we study do not feature a proxy fight. Nevertheless, activists are successful in affecting firm behavior far more than 9% of the time, as documented by prior research including Brav et al. (2008) and other papers surveyed in Brav et al. (2009). In our sample, Table 1 shows that campaigns produce an average Actioni, t+1 of 0.707, which is significantly higher than the average $$ \widehat{Action}_{i,t+1}$$ in the full Compustat universe of 0.552. We hypothesize the effectiveness of non-proxy activist campaigns is attributable to targets settling in the face of credible threats by high reputation activists. We test this hypothesis and our model's related empirical predictions using the analysis described in this section.
5.1 Distinguishing Between Treatment Effects and Selection Effects
Our empirical analyses examine the relation between activist reputation and target manager actions during activist campaigns. Our model predicts activists with a track record of proxy fighting, or being settled with, will enjoy greater activist campaign success as a result of treatment effects: the activist causes the target to institute the changes we measure as campaign success either by initiating a proxy fight or extracting a settlement. However, a natural alternative hypothesis involves a selection effect: high reputation activists are especially skilled at picking, and then passively investing in, companies that subsequently take actions that we measure as campaign success.
We examine the extent to which our measures of activist reputation relate to target selection skill using the regressions presented in Table 2 and outlined in Equation (6). These regressions relate $$ \widehat{Action_{i,t+1}}$$, our measure of the propensity of each target firm to take the actions we interpret as campaign success without interference from activists, to each of our activist reputation measures:
| $$\displaystyle \widehat{Action_{i,t+1}} = \beta_0 + \beta_{1}\,{Recent Fight}_{j,t}+\beta_{2}\,{Settle Rate}_{j,t}+\beta_{3}\,{Target Share}_{j,i,t}+\epsilon_{j,i,t}.$$ | (6) |
Columns (1) and (2) of Table 2 show that both $$ {Recent Fight}_{j,t}$$ and $$ {Settle Rate}_{j,t}$$ are positively related to $$ \widehat{Action_{i,t+1}}$$, indicating higher reputation activists select target firms with a greater propensity for subsequent actions. Somewhat surprisingly, Column (3) of Table 2 shows there is a negative relation between Target Sharej,i,t and $$ \widehat{Action_{i,t+1}}$$, indicating activists take smaller positions in firms likely to make substantial changes without their intervention. Column (4) shows that the relations between $$ \widehat{Action_{i,t+1}}$$ and $$ {Recent Fight}_{j,t}$$, $$ {Settle Rate}_{j,t}$$, and Target Sharej,i,t all hold incremental to each other, and Column (5) shows they hold when controlling for whether the campaign results in a proxy fight.
The strong positive relations between $$ \widehat{Action_{i,t+1}}$$ and both $$ {Recent Fight}_{j,t}$$ and $$ {Settle Rate}_{j,t}$$ across all specifications in Table 2 illustrate that the selection alternative could drive a spurious relation between our reputation measures and Actioni, t+1. We address this concern in our main empirical tests by controlling for $$ \widehat{Action_{i,t+1}}$$. The identifying assumption we make is that $$ \widehat{Action_{i,t+1}}$$ accurately measures the average Actioni, t+1 that would prevail without activist intervention conditional on observable information at time t.
5.2 Reputation and Activist Campaign Outcomes
Empirical Predictions 1, 2, and 3 state that Recent Fight and Settle Rate, but not Target Share, are positively related to activist campaign success. We empirically examine each of these predictions using variants of the pooled OLS regression outlined in Equation (7):
| $$\displaystyle {Action}_{i,t+1} = \beta_0 + \beta_{1}\,{Recent Fight}_{j,t}+\beta_{2}\,{Settle Rate}_{j,t}+\beta_{3}\,{Target Share}_{j,i,t}+\epsilon_{j,i,t}.$$ | (7) |
The results in Columns (1) and (2) of Table 3 show that campaigns featuring activists with stronger reputations, as measured by Recent Fightj,t and Settle Ratej,t, are more successful.14 As an illustration of economic significance, the coefficient estimate of 50.06 on Recent Fightj,t in Column (1) suggests that a Recent Fight correlates with an increase of 0.5006 in Actioni, t+1, 71% of the unconditional average (0.707, see Table 1). Similarly, the coefficient estimate of 12.43 on Settle Ratej,t in Column (2) suggests that a one standard deviation increase in Settle Ratej,t correlates with an increase of 0.1243 in Actioni, t+1, 18% of the unconditional average.
Column (3) of Table 3, by contrast, shows that campaigns where activists purchase larger fractions of their targets are no more successful, consistent with Empirical Prediction 3. This result is surprising because models of activism involving threat of exit, such as Admati & Pfleiderer (2009), predict that campaigns by larger shareholders are more likely to succeed, and theories of large shareholders, such as Shleifer & Vishny (1986), predict larger shareholders will have a greater impact on corporate governance. Our results suggest that, in the context of shareholder activism, the signaling value of stake size is minimal.
There are two potential channels through which high reputation activists are more successful, both occurring in our model. The first is a direct effect of persistence in proxy fighting behavior. If activists with a track record of proxy fights or settlements are more likely to initiate proxy fights in the future, a pattern we confirm in later tests, they will mechanically have more-successful campaigns because they initiate more proxy contests. The second channel for the success of campaigns by high reputation activists is the equilibrium effect driven by target managers settling with them more frequently. We separate these channels empirically by controlling for whether the campaign results in a proxy contest.
Column (5) of Table 3 shows that campaigns involving proxy fights are substantially more successful, resulting in an average of 0.85 additional actions by target firms. More importantly for our thesis, controlling for this direct effect slightly reduces but does not eliminate the relation between reputation and campaign success. Together, these patterns imply that high reputation activists are more successful primarily because they are more likely to extract settlements but also because they are more likely to fight if refused.
Finally, in Column (6) of Table 3, we address the selection effects discussed above by including $$ \widehat{Action}_{i,t}$$ as a control variable.15 The coefficients for $$ {Recent Fight}_{j,t}$$ and $$ {Settle Rate}_{j,t}$$ are smaller in Column (6) relative to Column (5), reflecting the fact that activists with a $$ {Recent Fight}_{j,t}$$ or high $$ {Settle Rate}_{j,t}$$ select targets with a higher $$ \widehat{Action}_{i,t}$$, as documented in Table 2. However, the coefficients for both $$ {Recent Fight}_{j,t}$$ and $$ {Settle Rate}_{j,t}$$ remain statistically and economically significant when controlling for activist skill at selecting targets and for the direct effect of proxy fights, indicating high reputation activists are better able to extract settlements from targets, as predicted by our model.
We further rule out the alternative stock selection hypothesis by showing that within-activist variations in reputation predict within-activist variation in campaign success. To do so, Table 4 repeats the analysis in Table 3 but with activist fixed effects.16 Across all columns of Table 4, we find a positive coefficient on Recent Fightj,t with significance fluctuating due to each activist in our sample pursuing an average of only six campaigns, limiting our statistical power. These results suggest that, assuming activist stock selection skill is constant through time, reputation effects are separate and distinct from activist's ability to pick stocks that will have high subsequent Actioni,t even without activist intervention.
5.3 Reputation Across and Within Institutional Investor Structures
Table 5 illustrates how our measures of reputation and campaign success vary across different activist institutional structures. Hedge fund activists are the most prevalent in our sample, comprising nearly two thirds of observed campaigns, and are different in many ways from other categories of activists. The first differences are in our measures for reputation, Recent Fightj,t and Settle Ratej,t, which are strikingly higher for hedge funds relative to other institutional structures. Target Sharej,i,t, by contrast, is similar among hedge fund campaigns and the full sample of campaigns. Despite purchasing similar fractions of their targets as other types of activists, hedge fund campaigns are far more effective, as measured by average Actioni, t+1. Together, these differences across institutional structure illustrate that hedge funds' ability to build and utilize reputation despite managing significantly less capital than other activists makes them especially effective, and prolific, activists.
Prior research, for example Yermack et al. (2010), argues hedge funds are more effective activists because lighter regulation allows hedge funds to hold more concentrated portfolios and employ longer-term investment strategies. These regulatory advantages also make hedge funds particularly well suited to initiate costly proxy fights as long-term investments in reputation and take concentrated portfolio positions to profit from reputation. The evidence in Table 5 that hedge fund activists have stronger reputation and allocate larger fractions of their portfolios to activism is consistent with reputation being the channel through which hedge funds are more successful activists than other institutional structures.
The differences across institutional structure in reputation and campaign success raise the question of whether our pooled evidence in Tables 3 and 4 is driven by variation across institutional structure alone. We show that variation within institutional structure is also predictive using the regressions presented in Table 6, which include fixed effects for each category of activist. The large and highly significant coefficient for Hedge Fundi,t in Table 6 indicates that hedge funds are significantly more effective activists relative to the other categories, reflecting the across-structure variation shown in Table 5. By contrast, the negative coefficient for Mutual Fundi,t indicates that mutual funds are less effective activists. Importantly, Recent Fighti,t is significant across all columns of Table 6, indicating that within-structure variation in activist reputation has a significant bearing on activist campaign success. Settle Ratei,t also positively predicts campaign success across all specifications, though the statistical significance is more marginal.
5.4 Other Measures of Activist Reputation, Skill, and Experience
In Table 7, we compare our measures of activist reputation to other activist characteristics and alternative measures of activist reputation, skill, and experience. Column (1) of Table 7 shows the results of a regression similar to Equation (7) but with Active Ratej,t, the number of activist j's recent activist campaigns scaled by the activist's number of positions as the independent variable. Consistent with the evidence in Boyson et al. (2016) that experienced activists are more successful, we find Active Ratej,t positively predicts campaign success. However, this effect vanishes in Column (6) after incorporating additional controls including our reputation measures.
Column (2) of Table 7 examines another potential measure of activist reputation, Return Avgj,t, the average market reaction to prior campaign initiations.17 We find that past market reactions have a positive but statistically insignificant bearing on campaign success. This indicates Return Avgj,t is a noisier measures of reputation than Recent Fightj,t and Settle Ratej,t, perhaps because announcement returns reflect the specifics of the target's situation and activist demands in addition to reputation.18 The results in Column (3) of Table 7 show that the activist's recent investment performance, as measured by Perfj,t, is not significantly related to campaign success. This result suggests that skills relating to investment performance are distinct from skills relating to activist effectiveness.
We find an additional notable result from Table 7: campaigns that attract follow-on activists are extremely effective. Specifically, Column (4) of Table 7 shows a strong positive relation between Wolf Packj,i,t+1, which indicates follow-on activists join the campaign, and campaign success. This relation supports the theory outlined in Brav et al. (2016), which argues that the presence of additional follow-on activists has a significant impact on activist campaign success.
The results in Column (5) suggest that Portfolio Sizej,t has no bearing on campaign success. While on the surface this is somewhat surprising, these results support both Table 5 and Table 6 which show that hedge funds are much more effective activists than activists with other institutional structures despite managing much smaller portfolios. On the other hand, Column (5) also shows that portfolio position count (Positionsj,t) and churn rate (Churn Ratej,t) are negatively and positively associated with greater campaign success, respectively. These patterns are largely driven by hedge funds having more-concentrated and higher-turnover portfolios relative to other activists.
Most importantly, Column (6) of Table 7 shows that our measures of reputation, Recent Fightj,t and Settle Ratej,t, predict campaign success incrementally to all of the above alternative measures of activist reputation, skill, and experience.
5.5 Other Measures of Campaign Outcomes
In Table 8 we examine the impact of our measures of activist reputation on the different responses by target firms we include in our Actioni, t+1 measure. Specifically, Table 8 presents regressions similar to those in Equation (7) but with dependent variables for each of the five sub-categories of Action: Payouti, t+1, CEOi, t+1, Reorgi, t+1, M&Ai, t+1, and Boardi, t+1. Columns (2) through (6) of Table 8 shows that both Recent Fightj,t and Settle Ratej,t are positively related to each type of target action, with statistical significance that fluctuates due to each individual outcome being less frequent.
5.6 Stake Size and Reputation
Next, we provide further evidence to support Empirical Prediction 3, that both aggressive and cautious activists purchase similar stake sizes, and as a result that campaign success is not sensitive to stake size. We assess the determinants of activists' choice of stake size using the regressions of the form:
| $$\displaystyle {Target Share}_{i,t+1} = \beta_0 + \beta_{1}\,{Recent Fight}_{j,t}+\beta_{2}\,{Settle Rate}_{j,t}+\beta_3 {Proxy}_{j,i,t+1} + \epsilon_{j,i,t}.$$ | (8) |
To the extent that activists who are more likely to fight purchase larger stake sizes, we would expect Target Sharej,i,t to be increasing in Recent Ratej,t, Settle Ratej,t, or Proxyj,i,t+1. Table 9 shows that this is not the case. Across all specifications, our reputation measures and future proxy behavior have either a negative an insignificant relation with Target Sharej,i,t, which is consistent with our model's prediction that cautious activists imitate the position sizes of aggressive activists and as a result `jam' the signal represented by stake size.
A related question from target management's perspective is which variables are useful in predicting whether an activist campaign results in a proxy fight. This question is an important aspect of managers' decision whether or not to settle with the activist. Our model predicts that activists with strong reputations are more likely to fight if refused. However, as discussed above, Empirical Prediction 3 states that propensity to fight is unrelated to stake size. We test these two predictions using regressions of the form:
| $$\displaystyle {Proxy}_{j,i,t+1} = \beta_0 + \beta_{1}\,{Recent Fight}_{j,t}+\beta_{2}\,{Settle Rate}_{j,t}+\beta_3 {Target Share}_{j,i,t} + \epsilon_{j,i,t}.$$ | (9) |
Table 10 shows that Recent Fightj,t is a significantly positive predictor of future proxy fights, consistent with propensity to fight being a persistent characteristic. In univariate analysis, we find a significant relation between Settle Ratej,t and future proxy fights, however this relation becomes insignificant when controlling for $$ \widehat{Action}_{i,t+1}$$. This is the result of two offsetting effects. On one hand, activists with a strong enough reputation that they are frequently settled with are more likely to fight if refused, predicting a positive relation between Settle Ratej,t and Proxyj,i,t+1. On the other hand, activists with a strong enough reputation that they are frequently settled with often get what they want without a proxy fight, predicting a negative relation between Settle Ratej,t and Proxyj,i,t+1.
Finally, Table 10 provides additional support for Empirical Prediction 3, as we find no relation between Target Sharej,i,t and future proxy fighting behavior.
5.7 Activist Position Sizes
Figure 1 shows a histogram of the dollar value of the activist's position size in our sample of 198 campaigns featuring a proxy fight, and indicates that half of the proxy fights in our sample occur in campaigns where the activist position is smaller than $50 million. As a point of comparison, Gantchev (2013) estimates the average cost of a proxy fight for an activist is $10.7 million. Survey evidence from the 1980s and recent anecdotal evidence from activist hedge funds estimate this cost is at least $5 million in today's dollars.19 Based on the Gantchev (2013) estimate, an activist initiating a proxy fight with median position size would need the target firm's stock to appreciate 21% just to offset the cost of the proxy campaign. Therefore, the fact that we observe proxy fights in this subsample is consistent with Empirical Prediction 4 from our model, which states that activists will sometimes engage in proxy fights at a short-term loss as an investment in their reputation.
6 Conclusion
We present a theory of an activist hedge fund's reputation, how it is developed, and how it impacts activist campaigns. In our model, reputation is managers' belief about the willingness of the activist to initiate proxy fights, which evolves as the activist engages two targets sequentially. In equilibrium, some activists incur the cost of proxy fights despite minimal short-term benefits as an investment in their reputation. A strong reputation is valuable to these activists because, unlike their position size in the target, a reputation as a proxy fighter allows them to convince future managers to settle without a proxy fight.
We find extensive empirical support for our model's predictions. Campaigns by high reputation activists result in abnormal payout increases, CEO and board changes, reorganizations, and M&A activity. The fraction of targets' shares purchased by activists is independent of their reputation, does not predict the likelihood of a proxy fights, and has no bearing on the success of the campaign. Combined, our theory and evidence indicate reputation plays a central role in the strategic dynamics and effectiveness of activist campaigns.
Activist Variables
| Recent Fight$$\displaystyle _{j,t}$$ | $$\displaystyle =$$ Dummy variable indicating activist j initiated a proxy fight within the year prior to t. | |
| Settle Rate$$\displaystyle _{j,t}$$ | $$\displaystyle =$$ The total Actioni, t+1 from activist j's non-proxy campaigns in the prior year divided by the average number of campaigns initiated by j in the prior three years. | |
| Target Share$$\displaystyle _{j,i,t}$$ | $$\displaystyle =$$ The max percent of target i's shares outstanding that activist j acquires during the activist campaign. | |
| Portfolio Size$$\displaystyle _{j,t}$$ | $$\displaystyle =$$ Value, in $ millions, of activist j's equity holdings positions from form 13-F. | |
| Number of Positions$$\displaystyle _{j,t}$$ | $$\displaystyle =$$ The number of activist j's equity holdings positions from form 13-F. | |
| Churn Rate$$\displaystyle _{j,t}$$ | $$\displaystyle =$$ Quarterly turnover of activist j's portfolio as defined in Gaspar, Massa, and Matos (2005). | |
| Proxy$$\displaystyle _{j,i,t+1}$$ | $$\displaystyle =$$ Dummy variable indicating activist j initiatiates a proxy fight with target i in the year following the activist j's campaign initiation. | |
| Position Size$$\displaystyle _{j,i,t}$$ | $$\displaystyle =$$ The size, in $ millions, of the position taken by activist j in company i's stock. | |
| Perf$$\displaystyle _{j,t}$$ | $$\displaystyle =$$ Activist j's prior 3-year average portfolio performance as implied by quarterly equity holdings filings. | |
| Active Rate$$\displaystyle _{j,t}$$ | $$\displaystyle =$$ The number of campaigns initiated by activst j in the year prior to t divided by Number of Positionsj,t. | |
| Return Avg$$\displaystyle _{j,t}$$ | $$\displaystyle =$$ The average abnormal three-day return surrounding the announcement of the initiation of activist j's campaigns over the year prior to t. | |
| Wolf Pack$$\displaystyle _{j,i,t+1}$$ | $$\displaystyle =$$ Dummy variable indicating at least one activist other than j files a 13-D for target i in year following t. |
Target Variables
| Size$$\displaystyle _{i,t}$$ | $$\displaystyle =$$ The natural log of i's CRSP market capitalization. | |
| Book-to-Market$$\displaystyle _{i,t}$$ | $$\displaystyle =$$ The equity book-to-market ratio: book equity from Compustat divided by CRSP market capitalization. | |
| ROA$$\displaystyle _{i,t}$$ | $$\displaystyle =$$ Operating income (prior year) divided by total assets from Compustat. | |
| Capex/Assets$$\displaystyle _{i,t}$$ | $$\displaystyle =$$ Additions to PP&E from the prior year's cash flow statement divided by total assets from Compustat. | |
| Net Leverage$$\displaystyle _{i,t}$$ | $$\displaystyle =$$ Debt minus cash divided by total assets from Compustat. | |
| Illiquidity$$\displaystyle _{i,t}$$ | $$\displaystyle =$$ The quarterly illiquidity measure from Amihud (2002) using daily data. | |
| Inst Ownership$$\displaystyle _{i,t}$$ | $$\displaystyle =$$ The percent of shares held by all 13-F filing institutions. | |
| $$\displaystyle \widehat{\text{Action}_{i,t+1}}$$ | $$\displaystyle =$$ The predicted value for Actioni,t+1, as defined in Section 4.3. |
Campaign Outcome Variables
| Action$$\displaystyle _{i,t+1}$$ | $$\displaystyle =$$ The sum of Payouti,t+1, CEO Changei,t+1, Board Changei,t+1, Reoganizationi,t+1, and M&Ai,t+1. | |
| Payout$$\displaystyle _{i,t+1}$$ | $$\displaystyle =$$ Dummy variable indicating a dividend initiation, dividend increase, special dividend, stock repurchase program initiation, or stock repurchase program increase within the year following t, measured using 8-K filings with Capital IQ key development type codes 36, 94, 214, and 232. | |
| CEO Change$$\displaystyle _{i,t+1}$$ | $$\displaystyle =$$ A dummy variable indicating CEO turnover within the year following t, code 101. | |
| Board Change$$\displaystyle _{i,t+1}$$ | $$\displaystyle =$$ A dummy variable indicating board of directors turnover, or a new director appointment, within the year following t, code 172. | |
| Reoganization$$\displaystyle _{i,t+1}$$ | $$\displaystyle =$$ The announcement of a reorganization within the year following t, code 21. | |
| M&A$$\displaystyle _{i,t+1}$$ | $$\displaystyle =$$ The announcement of a merger or acquisition within the year following t, code 80. |
Bibliography
The Journal of Finance 54(3):833-874.
Review of Financial Studies 22(7):2445-2485.
Journal of Political Economy pp. 1097-1130.
Review of Finance 8(3):327-353.
Available at SSRN 2693145 .
Econometrica: Journal of the Econometric Society pp. 647-661.
Available at SSRN 2727371 .
Journal of Financial Economics 95(1):1-19.
Robert H. Smith School Research Paper No. RHS 2529230.
Foundations and Trends in Finance 4(3):185-246.
The Journal of Finance 63(4):1729-1775.
The Quarterly Journal of Economics pp. 693-728.
Journal of Finance pp. 57-79.
Available at SSRN 2600340 .
The Journal of Finance 70(6):2853-2896.
Journal of Corporate Finance .
The Journal of Political Economy pp. 828-862.
Review of Financial Studies 24(7):2395-2428.
The Review of Economic Studies 59(3):561-579.
Journal of Financial Economics 107(3):610-631.
Journal of Applied Corporate Finance 19(1):55-73.
Journal of Financial Economics 3(4):305-360.
University of Pennsylvania Law Review pp. 1021-1093.
The Journal of Finance 64(1):187-229.
Journal of Economic Theory 27(2):253-279.
Journal of Corporate Finance 40:296-314.
Available at SSRN 1969475 .
The Journal of Finance 53(1):65-98.
Journal of Economic Theory 27(2):280-312.
Econometrica pp. 325-351.
The Journal of Political Economy pp. 461-488.
The Quarterly Journal of Economics pp. 355-374.
Journal of Financial and Quantitative Analysis 22(01):17-32.
The Journal of Economic Perspectives 21(2):175-194.
Vanderbilt Law Review En Banc 61:1.
Annual Review of Financial Economics 2(1):103-125.
Figure 1: Size of Positions Resulting In Proxy Fights
This figure presents a histogram of the dollar value of the activist's position in the target for campaigns resulting in a proxy fight. We define an activist campaign resulting in a proxy fight based on proxy filings with the SEC by both activist and target, as described in the Data Section. Of the activist campaigns in our sample, 96 result in proxy fights. The data include activist campaigns initiated between 2000-2014.
Table 1: Activist Campaign Summary Statistics
This table shows summary statistics for our sample of activist campaigns. We summarize activist j characteristics, target company i characteristics, and campaign outcomes. All variables are described in Appendix B. The data include 2,199 activist campaigns initiated during 2000-2014.
| Mean | Median | Std Dev | Min | Max | |
|---|---|---|---|---|---|
| Activists | |||||
| Recent Fightj,t | 0.12 | 0.00 | 0.31 | 0.0 | 1.0 |
| Settle Ratej,t | 54.0% | 39.0% | 65.5% | 0.0% | 500.0% |
| Target Sharej,i,t | 8.4% | 6.8% | 6.3% | 0.0% | 49.5% |
| Proxyj,i,t+1 | 0.09 | 0.00 | 0.29 | 0.0 | 1.0 |
| Portfolio Sizej,t ($ mln) | $8,507 | $1,689 | $23,898 | $3 | $596,961 |
| Number of Positionsj,t | 425 | 75 | 746 | 1 | 4,895 |
| Churn Ratej,t | 52.8% | 36.3% | 44.9% | 0.0% | 282.4% |
| Position Sizej,i,t ($ mln) | $88 | $19 | $265 | $0 | $5,307 |
| Perfj,t | 0.7% | 0.0% | 5.2% | -13.9% | 18.9% |
| Active Ratej,t | 5.5% | 2.4% | 8.4% | 0.0% | 100.0% |
| Return Avgj,t | 1.9% | 1.4% | 3.3% | -6.9% | 15.9% |
| Wolf Packj,t | 0.11 | 0.00 | 0.31 | 0.0 | 1.0 |
| Targets | |||||
| Sizei,t ($ mln) | $1,567 | $258 | $10,810 | $1 | $428,700 |
| Book-to-Marketi,t | 1.4 | 0.6 | 6.0 | 0 | 100 |
| ROAi,t | -0.8% | 5.0% | 23.7% | -193% | 80% |
| Capex/Assetsi,t | 4.6% | 2.7% | 6.5% | 0% | 75% |
| Net Leveragei,t | 1.4% | 5.3% | 38.6% | -100% | 200% |
| Illiquidityi,t | 1.1 | 0.0 | 6.5 | 0.0 | 118.4 |
| Inst Ownershipi,t | 61% | 66% | 28% | 0% | 100% |
| $$ \widehat{\text{Action}_{i,t+1}}$$ | 55.2 | 53.9 | 23.8 | 0.0 | 167.9 |
| Campaign Outcomes | |||||
| Actioni, t+1 | 70.7 | 0.0 | 86.4 | 0.0 | 400.0 |
| Payouti, t+1 | 16.5 | 0.0 | 37.1 | 0.0 | 100.0 |
| CEO changei, t+1 | 19.6 | 0.0 | 39.7 | 0.0 | 100.0 |
| Board changei, t+1 | 7.1 | 0.0 | 25.8 | 0.0 | 100.0 |
| Reorganizationi, t+1 | 15.9 | 0.0 | 36.6 | 0.0 | 100.0 |
| M&Ai, t+1 | 11.5 | 0.0 | 31.9 | 0.0 | 100.0 |
Table 2: Activist Reputation and Target Selection
Table presents pooled OLS regressions assessing the relation between activist reputation and the ex-ante propensity of targets to take actions we interpret as campaign success without activist intervention. The dependent variable is $$ \widehat{Action_{i,t+1}}$$, the fitted value in a cross-sectional regression among all Compustat firms in quarter t of $$ Action_{i,t+1}$$ on a vector of firm characteristics $$ X_{i,t}$$ described in Appendix B, multiplied by 100. The independent variables are: $$ \textit{Recent Fight}_{j,t}$$, a dummy variable indicating activist j initiated a proxy fight in the prior year; $$ \textit{Settle Rate}_{j,t}$$, the total Actioni, t+1 from activist j's non-proxy campaigns divided in the prior year divided by the average number of campaigns initiated by j in the prior three years; Target Sharej,i,t, the peak percent of target i's shares outstanding that activist j acquires during the activist campaign; and $$ \textit{Proxy}_{j,t,t+1}$$, an indicator for whether the campaign results in activist j initiating a proxy fight. All continuous independent variables are de-meaned and scaled by their standard deviation. The data include 2,199 activist campaigns initiated during 2000-2014. T-statistics are presented in parenthesis, *** indicates significance at 1% level, ** indicates 5%, and * indicates 10%.
| (1) $$ \widehat{Action_{i,t+1}}$$ |
(2) $$ \widehat{Action_{i,t+1}}$$ |
(3) $$ \widehat{Action_{i,t+1}}$$ |
(4) $$ \widehat{Action_{i,t+1}}$$ |
(5) $$ \widehat{Action_{i,t+1}}$$ |
|
|---|---|---|---|---|---|
|
Recent Fight$$ _{j,t} $$ |
8.92*** (3.70) |
6.36*** (2.82) |
3.69* (1.80) |
||
|
Settle Rate$$ _{j,t} $$ |
4.43*** (4.45) |
4.16*** (4.42) |
4.03*** (4.34) |
||
|
Target Share$$ _{j,i,t} $$ |
-2.54*** (-4.44) |
-2.47*** (-4.40) |
-2.50*** (-4.49) |
||
|
Proxy$$ _{j,i,t+1} $$ |
10.53*** (5.34) |
||||
| Intercept |
54.14*** (28.21) |
55.25*** (31.20) |
55.14*** (29.59) |
54.47*** (30.80) |
53.83*** (29.71) |
| Cluster SE | Quarter | Quarter | Quarter | Quarter | Quarter |
| Obs | 2,199 | 2,199 | 2,199 | 2,199 | 2,199 |
| F-Statistic | 13.66 | 19.80 | 19.70 | 12.74 | 15.30 |
| R2 | 0.013 | 0.034 | 0.011 | 0.051 | 0.066 |
Table 3: Activist Reputation and Campaign Outcomes
This table presents pooled OLS regressions assessing how measures of activist reputation relate to activist campaign effectiveness. The dependent variable is Actioni, t+1, the number of significant actions (defined in Section 4.3) that occur in the year following the start of the activist campaign, multiplied by 100. The primary independent variables of interest are: $$ \textit{Recent Fight}_{j,t}$$, a dummy variable indicating activist j initiated a proxy fight in the prior year; $$ \textit{Settle Rate}_{j,t}$$, the total Actioni, t+1 from activist j's non-proxy campaigns in the prior year divided by the average number of campaigns initiated by j in the prior three years; and Target Sharej,i,t, the peak percent of target i's shares outstanding that activist j acquires during the activist campaign. Other variables are defined in Appendix B. All continuous independent variables are de-meaned and scaled by their standard deviation. The data include 2,199 activist campaigns initiated during 2000-2014. T-statistics are presented in parenthesis, *** indicates significance at 1% level, ** indicates 5%, and * indicates 10%.
| (1) Action$$ _{i,t+1} $$ |
(2) Action$$ _{i,t+1} $$ |
(3) Action$$ _{i,t+1} $$ |
(4) Action$$ _{i,t+1} $$ |
(5) Action$$ _{i,t+1} $$ |
(6) Action$$ _{i,t+1} $$ |
|
|---|---|---|---|---|---|---|
|
Recent Fight$$ _{j,t} $$ |
50.06*** (6.44) |
45.01*** (5.52) |
23.63*** (3.61) |
19.85*** (3.44) |
||
|
Settle Rate$$ _{j,t} $$ |
12.43*** (5.78) |
10.29*** (4.60) |
9.18*** (4.44) |
5.06*** (2.75) |
||
|
Target Share$$ _{j,i,t} $$ |
-0.80 (-0.46) |
-0.01 (-0.01) |
-0.22 (-0.14) |
2.34 (1.57) |
||
|
Proxy$$ _{j,i,t+1} $$ |
84.57*** (9.11) |
73.76*** (8.14) |
||||
|
$$ \widehat{Action_{i,t+1}}$$ |
1.03*** (13.89) |
|||||
| Intercept |
64.89*** (22.66) |
70.91*** (24.12) |
70.70*** (22.50) |
65.64*** (24.60) |
60.49*** (23.19) |
5.29 (1.30) |
| Cluster SE | Quarter | Quarter | Quarter | Quarter | Quarter | Quarter |
| Obs | 2,199 | 2,199 | 2,199 | 2,199 | 2,199 | 2,199 |
| F-Statistic | 41.53 | 33.40 | 0.211 | 21.59 | 34.98 | 118.3 |
| R2 | 0.032 | 0.020 | 0.000 | 0.045 | 0.117 | 0.192 |
Table 4: Activist Reputation and Campaign Outcomes Within Activist
This table presents OLS regressions assessing how measures of activist reputation relate to activist campaign effectiveness when including activist fixed effects. The dependent variable is Actioni, t+1, the number of significant actions (defined in Section 4.3) that occur in the year following the start of the activist campaign, multiplied by 100. The primary independent variables of interest are: $$ \textit{Recent Fight}_{j,t}$$, a dummy variable indicating activist j initiated a proxy fight in the prior year; $$ \textit{Settle Rate}_{j,t}$$, the total Actioni, t+1 from activist j's non-proxy campaigns divided in the prior year divided by the average number of campaigns initiated by j in the prior three years, and Target Sharej,i,t, the peak percent of target i's shares outstanding that activist j acquires during the activist campaign. All continuous independent variables are de-meaned and scaled by their standard deviation. The data include the subset of the 2,199 activist campaigns initiated between 2000-2014 which involve activists that have initiated more than two campaigns. T-statistics are presented in parenthesis, *** indicates significance at 1% level, ** indicates 5%, and * indicates 10%.
| (1) Action$$ _{i,t+1} $$ |
(2) Action$$ _{i,t+1} $$ |
(3) Action$$ _{i,t+1} $$ |
(4) Action$$ _{i,t+1} $$ |
(5) Action$$ _{i,t+1} $$ |
(6) Action$$ _{i,t+1} $$ |
|
|---|---|---|---|---|---|---|
|
Fight Rate$$ _{j,t} $$ |
15.68* (1.86) |
16.12* (1.92) |
12.60 (1.63) |
9.21 (1.33) |
||
|
Settle Rate$$ _{j,t} $$ |
-1.48 (-0.58) |
-1.80 (-0.71) |
-2.73 (-1.17) |
-3.66 (-1.39) |
||
|
Target Share$$ _{j,i,t} $$ |
-3.35 (-1.31) |
-3.22 (-1.24) |
-3.80 (-1.62) |
-0.90 (-0.40) |
||
|
Proxy$$ _{j,i,t+1} $$ |
73.20*** (6.68) |
63.54*** (6.18) |
||||
|
$$ \widehat{Action_{i,t+1}}$$ |
0.99*** (15.45) |
|||||
| Intercept |
68.28*** (25.78) |
70.34*** (28.77) |
70.19*** (28.95) |
68.06*** (26.06) |
62.29*** (23.74) |
8.74** (2.30) |
| Fixed Effects | Activistj | Activistj | Activistj | Activistj | Activistj | Activistj |
| Cluster SE | Quarter | Quarter | Quarter | Quarter | Quarter | Quarter |
| Obs | 1,889 | 1,889 | 1,889 | 1,889 | 1,889 | 1,889 |
| F-Statistic | 3.471 | 0.340 | 1.728 | 2.140 | 16.23 | 75.41 |
| R2 | 0.238 | 0.236 | 0.237 | 0.239 | 0.278 | 0.328 |
Table 5: Campaign Summary Statistics By Activist Institutional Structure
This table shows averages of the relevant variables summarizing our sample of activist campaigns grouped according to the activist's institutional structure. We summarize activist j characteristics, target company i characteristics, and campaign outcomes. All variables are described in Appendix B. The data include 2,199 activist campaigns initiated during 2000-2014.
| Hedge Fund |
Mutual Fund |
Private Equity |
Broker Dealer |
Other |
All |
|
|---|---|---|---|---|---|---|
| # Campaigns | 1,379 | 485 | 70 | 69 | 196 | 2,199 |
| # Unique Activists | 266 | 35 | 12 | 19 | 35 | 367 |
| Activists | ||||||
| Recent Fightj,t | 0.16 | 0.06 | 0.06 | 0.00 | 0.03 | 0.12 |
| Settle Ratej,t | 64.3% | 35.0% | 25.8% | 35.7% | 45.5% | 54.0% |
| Target Sharej,i,t | 8.6% | 6.2% | 11.4% | 10.8% | 10.6% | 8.4% |
| Proxyj,i,t+1 | 0.12 | 0.02 | 0.06 | 0.03 | 0.05 | 0.09 |
| Portfolio Sizej,t ($ mln) | $2,598 | $22,292 | $1,687 | $50,444 | $3,647 | $8,507 |
| Number of Positionsj,t | 200 | 925 | 41 | 2,503 | 172 | 425 |
| Churn Ratej,t | 65.6% | 24.4% | 23.4% | 61.1% | 40.3% | 52.8% |
| Position Sizej,i,t ($ mln) | $94 | $82 | $44 | $55 | $86 | $88 |
| Perfj,t | 1.0% | 0.5% | -1.5% | 0.2% | 0.0% | 0.7% |
| Active Ratej,t | 7.3% | 2.5% | 7.1% | 0.3% | 2.0% | 5.5% |
| Return Avgj,t | 2.4% | 1.1% | 0.9% | 0.8% | 1.3% | 229.2% |
| Wolf Packj,t | 0.12 | 0.09 | 0.04 | 0.07 | 0.08 | 0.11 |
| Targets | ||||||
| Sizei,t ($ mln) | $1,725 | $1,130 | $539 | $3,274 | $1,306 | $1,567 |
| Book-to-Marketi,t | 1.3 | 1.5 | 2.2 | 1.1 | 1.7 | 1.4 |
| ROAi,t | -0.9% | 4.8% | -24.6% | -10.0% | -1.9% | -0.8% |
| Capex/Assetsi,t | 4.4% | 4.9% | 4.3% | 7.4% | 4.4% | 4.6% |
| Net Leveragei,t | -0.2% | 6.8% | -20.3% | 2.1% | 6.5% | 1.4% |
| Illiquidityi,t | 1.0 | 1.0 | 2.1 | 0.6 | 2.2 | 1.1 |
| Inst Ownershipi,t | 62% | 65% | 44% | 49% | 57% | 61% |
| $$ \widehat{\text{Action}_{i,t+1}}$$ | 57.4 | 52.2 | 45.7 | 50.6 | 51.7 | 55.2 |
| Campaign Outcomes | ||||||
| Actioni, t+1 | 85.4 | 40.4 | 44.3 | 53.6 | 57.7 | 70.7 |
| Payouti, t+1 | 20.5 | 9.3 | 5.7 | 7.2 | 13.3 | 16.5 |
| CEO changei, t+1 | 22.6 | 13.2 | 17.1 | 18.8 | 16.3 | 19.6 |
| Board changei, t+1 | 9.6 | 2.3 | 4.3 | 4.3 | 3.6 | 7.1 |
| Reorganizationi, t+1 | 18.6 | 8.9 | 11.4 | 17.4 | 15.3 | 15.9 |
| M&Ai, t+1 | 14.1 | 6.8 | 5.7 | 5.8 | 9.2 | 11.5 |
Table 6: Activist Reputation and Campaign Outcomes by Activist Category
This table presents OLS regressions assessing how measures of activist reputation relate to activist campaign effectiveness within activist institutional structure. The dependent variable is Actioni, t+1, the number of significant actions (defined in Section 4.3) that occur in the year following the start of the activist campaign, multiplied by 100. Hedge Fundj,t, Mutual Fundj,t, Private Equityj,t, and Broker Dealerj,t are indicator variables for activist j's institutional structure. Other variables are described in Appendix B. All continuous independent variables are de-meaned and scaled by their standard deviation. The data include 2,199 activist campaigns initiated during 2000-2014. T-statistics are presented in parenthesis, *** indicates significance at 1% level, ** indicates 5%, and * indicates 10%.
| (1) Action$$ _{i,t+1} $$ |
(2) Action$$ _{i,t+1} $$ |
(3) Action$$ _{i,t+1} $$ |
(4) Action$$ _{i,t+1} $$ |
(5) Action$$ _{i,t+1} $$ |
(6) Action$$ _{i,t+1} $$ |
|
|---|---|---|---|---|---|---|
|
Fight Rate$$ _{j,t} $$ |
41.01*** (6.01) |
37.76*** (5.22) |
18.84*** (3.17) |
15.79*** (2.91) |
||
|
Settle Rate$$ _{j,t} $$ |
8.70*** (4.12) |
7.26*** (3.33) |
6.66*** (3.27) |
2.99 (1.61) |
||
|
Target Share$$ _{j,i,t} $$ |
-2.64 (-1.49) |
-1.81 (-1.04) |
-1.74 (-1.08) |
0.76 (0.49) |
||
|
Proxy$$ _{j,i,t+1} $$ |
79.31*** (8.61) |
69.44*** (7.67) |
||||
|
$$ \widehat{Action_{i,t+1}}$$ |
0.99*** (13.19) |
|||||
|
Hedge Fund$$ _{j,t} $$ |
22.69*** (4.19) |
25.64*** (4.58) |
27.29*** (4.76) |
20.52*** (3.80) |
17.01*** (3.13) |
14.23*** (2.75) |
|
Mutual Fund$$ _{j,t} $$ |
-18.17*** (-3.17) |
-15.39*** (-2.78) |
-18.59*** (-3.04) |
-18.13*** (-3.13) |
-15.86*** (-2.89) |
-15.47*** (-3.16) |
|
Private Equity$$ _{j,t} $$ |
-14.38 (-1.50) |
-10.45 (-1.10) |
-12.74 (-1.32) |
-11.90 (-1.26) |
-12.36 (-1.27) |
-7.64 (-0.90) |
|
Broker Dealer$$ _{j,t} $$ |
-2.51 (-0.25) |
-2.19 (-0.22) |
-3.35 (-0.33) |
-1.44 (-0.14) |
-0.65 (-0.07) |
-0.53 (-0.06) |
| Intercept |
56.27*** (9.59) |
58.57*** (10.33) |
58.19*** (9.64) |
57.97*** (10.10) |
54.71*** (9.94) |
2.73 (0.47) |
| Cluster SE | Quarter | Quarter | Quarter | Quarter | Quarter | Quarter |
| Obs | 2,199 | 2,199 | 2,199 | 2,199 | 2,199 | 2,199 |
| F-Statistic | 31.97 | 29.32 | 27.33 | 23.85 | 27.80 | 73.29 |
| R2 | 0.072 | 0.061 | 0.052 | 0.079 | 0.142 | 0.211 |
Table 7: Other Reputation Measures and Campaign Outcomes
This table presents OLS regressions assessing how alternative measures of activist reputation relate to activist campaign effectiveness. Actioni, t+1 and the independent variables are defined in Appendix B. All continuous independent variables are de-meaned and scaled by their standard deviation. The data include 2,199 activist campaigns initiated during 2000-2014. T-statistics are presented in parenthesis, *** indicates significance at 1% level, ** indicates 5%, and * indicates 10%.
| (1) Action$$ _{i,t+1} $$ |
(2) Action$$ _{i,t+1} $$ |
(3) Action$$ _{i,t+1} $$ |
(4) Action$$ _{i,t+1} $$ |
(5) Action$$ _{i,t+1} $$ |
(6) Action$$ _{i,t+1} $$ |
|
|---|---|---|---|---|---|---|
|
Active Rate$$ _{j,t} $$ |
4.77** (2.81) |
1.77 (0.94) |
||||
|
Return Avg$$ _{j,t} $$ |
2.94 (1.07) |
-0.16 (-0.06) |
||||
|
Perf$$ _{j,t} $$ |
1.05 (0.57) |
0.85 (0.47) |
||||
|
Wolf Pack$$ _{j,i,t+1} $$ |
20.11** (2.84) |
18.19** (2.69) |
||||
|
Port Size$$ _{j,t} $$ |
1.26 (1.02) |
0.29 (0.24) |
||||
|
Positions$$ _{j,t} $$ |
-10.22*** (-5.19) |
-7.82*** (-3.41) |
||||
|
Churn Rate$$ _{j,t} $$ |
6.62*** (3.34) |
6.02*** (3.25) |
||||
|
Recent Fight$$ _{j,t} $$ |
41.26*** (4.56) |
|||||
|
Settle Rate$$ _{j,t} $$ |
3.26* (2.00) |
|||||
|
Target Share$$ _{j,i,t} $$ |
1.16 (1.25) |
|||||
|
Proxy$$ _{j,i,t+1} $$ |
79.41*** (9.98) |
79.59*** (9.86) |
80.46*** (9.71) |
76.65*** (8.78) |
76.04*** (9.27) |
66.15*** (8.18) |
|
$$ \widehat{Action_{i,t+1}}$$ |
1.05*** (13.25) |
1.06*** (14.17) |
1.08*** (13.87) |
1.06*** (14.11) |
1.03*** (15.32) |
0.98*** (13.09) |
| Intercept |
5.75 (1.29) |
5.05 (1.22) |
4.06 (0.91) |
3.05 (0.70) |
6.88* (1.85) |
7.07 (1.55) |
| Cluster SE | Quarter | Quarter | Quarter | Quarter | Quarter | Quarter |
| Obs | 2,199 | 2,199 | 2,199 | 2,199 | 2,199 | 2,199 |
| F-Statistic | 329.6 | 394.6 | 338.9 | 301.8 | 311.1 | 229.8 |
| R2 | 0.186 | 0.184 | 0.183 | 0.188 | 0.200 | 0.215 |
Table 8: Different Campaign Outcomes
This table presents OLS regressions assessing how activist characteristics relate to different activist campaign outcomes. The dependent variables are Actioni, t+1, the number of significant actions (defined in Section 4.3) that occur in the year following the start of the activist campaign, as well as each different type comprising Actioni, t+1: Payouti, t+1, indicating the target announces a significant increase in payout; CEOi, t+1, indicating the target changes CEO; Reorgi, t+1, indicating the target announces a significant reorganization; M&Ai, t+1, indicating the target is acquired; and Boardi, t+1 indicating the target announces of a new member of the board of directors. Each dependent variable is multiplied by 100. Independent variables are defined in Appendix B. All continuous independent variables are de-meaned and scaled by their standard deviation. The data include 2,199 activist campaigns initiated during 2000-2014. T-statistics are presented in parenthesis, *** indicates significance at 1% level, ** indicates 5%, and * indicates 10%.
| (1) Action$$ _{i,t+1} $$ |
(2) Payout$$ _{i,t+1} $$ |
(3) CEO$$ _{i,t+1} $$ |
(4) Reorg$$ _{i,t+1} $$ |
(5) M&A$$ _{i,t+1} $$ |
(6) Board$$ _{i,t+1} $$ |
|
|---|---|---|---|---|---|---|
|
Recent Rate$$ _{j,t} $$ |
19.85*** (3.44) |
1.11 (0.41) |
1.98 (0.68) |
7.52** (2.39) |
1.30 (0.51) |
7.61*** (3.67) |
|
Settle Rate$$ _{j,t} $$ |
5.06*** (2.75) |
1.57* (1.86) |
0.90 (0.97) |
1.35 (1.60) |
1.29* (1.97) |
-0.20 (-0.33) |
|
Target Share$$ _{j,i,t} $$ |
2.34 (1.57) |
0.35 (0.57) |
0.88 (1.06) |
0.63 (0.82) |
-0.07 (-0.11) |
-0.14 (-0.36) |
|
Proxy$$ _{j,i,t+1} $$ |
73.76*** (8.14) |
7.20** (2.37) |
13.12*** (3.52) |
6.73* (1.90) |
10.68*** (3.03) |
35.12*** (9.40) |
|
$$ \widehat{Action_{i,t+1}}$$ |
1.03*** (13.89) |
0.91*** (12.89) |
1.13*** (12.30) |
1.02*** (14.13) |
1.23*** (6.22) |
2.60*** (5.22) |
| Intercept |
5.29 (1.30) |
-0.38 (-0.31) |
1.31 (1.01) |
-0.35 (-0.39) |
1.45 (1.21) |
-0.53 (-0.92) |
| Cluster SE | Quarter | Quarter | Quarter | Quarter | Quarter | Quarter |
| Obs | 2,199 | 2,199 | 2,199 | 2,199 | 2,199 | 2,199 |
| F-Statistic | 118.3 | 44.14 | 38.57 | 61.15 | 14.24 | 30.30 |
| R2 | 0.192 | 0.100 | 0.058 | 0.144 | 0.039 | 0.254 |
Table 9: Size of Stakes Purchased by Activists in Target Companies
This table presents OLS regressions relating the proportion of the target purchased by an activist hedge fund to our measures of activist reputation. The dependent variable is Target Sharej,i,t, the peak percent of target i's shares outstanding that activist j acquires during the activist campaign. The primary independent variables of interest are: $$ \textit{Recent Fight}_{j,t}$$, a dummy variable indicating activist j initiated a proxy fight in the prior year; and $$ \textit{Settle Rate}_{j,t}$$, the total Actioni, t+1 from activist j's non-proxy campaigns divided in the prior year divided by the average number of campaigns initiated by j in the prior three years. Other controls variables are defined in Appendix B. All continuous independent variables are de-meaned and scaled by their standard deviation. The data include 2,199 activist campaigns initiated during 2000-2014. T-statistics are presented in parenthesis, *** indicates significance at 1% level, ** indicates 5%, and * indicates 10%.
| (1) Tgt Shr$$ _{j,i,t} $$ |
(2) Tgt Shr$$ _{j,i,t} $$ |
(3) Tgt Shr$$ _{j,i,t} $$ |
(4) Tgt Shr$$ _{j,i,t} $$ |
(5) Tgt Shr$$ _{j,i,t} $$ |
|
|---|---|---|---|---|---|
|
Recent Fight$$ _{j,t} $$ |
-1.39** (-2.88) |
-1.48*** (-3.04) |
-1.53** (-2.70) |
-1.64** (-2.85) |
|
|
Settle Rate$$ _{j,t} $$ |
0.10 (0.89) |
0.17 (1.44) |
0.17 (1.40) |
0.17 (1.60) |
|
|
Proxy$$ _{j,i,t+1} $$ |
0.20 (0.35) |
0.26 (0.46) |
|||
|
$$ \widehat{Action_{i,t+1}}$$ |
-0.10 (-1.01) |
||||
|
Port Size$$ _{j,t} $$ |
-0.03*** (-4.76) |
||||
|
Positions$$ _{j,t} $$ |
0.25 (1.64) |
||||
|
Churn Rate$$ _{j,t} $$ |
-1.04*** (-4.44) |
||||
| Intercept |
8.56*** (35.79) |
8.40*** (37.73) |
8.57*** (35.83) |
8.56*** (38.91) |
10.21*** (26.59) |
| Cluster SE | Quarter | Quarter | Quarter | Quarter | Quarter |
| Obs | 2,199 | 2,199 | 2,199 | 2,199 | 2,199 |
| F-Statistic | 8.286 | 0.788 | 5.222 | 3.768 | 14.33 |
| R2 | 0.005 | 0.000 | 0.005 | 0.005 | 0.038 |
Table 10: Proxy Fights by Activists
This table presents OLS regressions relating the frequency of proxy fights to our measures of activist reputation. The dependent variable is $$ \textit{Proxy}_{j,t,t+1}$$, an indicator for whether the campaign resulted in activist j initiating a proxy fight. The primary independent variables of interest are: $$ \textit{Recent Fight}_{j,t}$$, a dummy variable indicating activist j initiated a proxy fight in the prior year; $$ \textit{Settle Rate}_{j,t}$$, the total Actioni, t+1 from activist j's non-proxy campaigns divided in the prior year divided by the average number of campaigns initiated by j in the prior three years; and Target Sharej,i,t, the peak percent of target i's shares outstanding that activist j acquires during the activist campaign. Other controls variables are defined in Appendix B. All continuous independent variables are de-meaned and scaled by their standard deviation. The data include 2,199 activist campaigns initiated during 2000-2014. T-statistics are presented in parenthesis, *** indicates significance at 1% level, ** indicates 5%, and * indicates 10%.
| (1) Proxy$$ _{j,t+1} $$ |
(2) Proxy$$ _{j,t+1} $$ |
(3) Proxy$$ _{j,t+1} $$ |
(4) Proxy$$ _{j,t+1} $$ |
(5) Proxy$$ _{j,t+1} $$ |
|
|---|---|---|---|---|---|
|
Recent Fight$$ _{j,t} $$ |
25.87*** (6.77) |
25.28*** (6.63) |
23.61*** (6.58) |
||
|
Settle Rate$$ _{j,t} $$ |
2.51*** (3.41) |
1.31* (1.90) |
0.31 (0.43) |
||
|
Target Share$$ _{j,i,t} $$ |
-0.26 (-0.42) |
0.25 (0.40) |
0.33 (0.51) |
||
|
$$ \widehat{Action_{i,t+1}}$$ |
1.27** (2.08) |
||||
|
Port Size$$ _{j,t} $$ |
0.14*** (5.40) |
||||
|
Positions$$ _{j,t} $$ |
1.00* (1.68) |
||||
|
Churn Rate$$ _{j,t} $$ |
-2.58*** (-5.37) |
||||
| Intercept |
5.99*** (9.92) |
9.04*** (12.06) |
9.00*** (11.67) |
6.09*** (10.06) |
-1.44 (-1.06) |
| Cluster SE | Quarter | Quarter | Quarter | Quarter | Quarter |
| Obs | 2,199 | 2,199 | 2,199 | 2,199 | 2,199 |
| F-Statistic | 45.79 | 11.62 | 0.176 | 15.90 | 18.69 |
| R2 | 0.077 | 0.008 | 0.000 | 0.079 | 0.100 |