Private Politics and Public Interest: NGOs, Corporate Campaigns, and Social Welfare 1

Transcription

1 Private Politis and Publi Interest: NGOs, Corporate Campaigns, and Soial Welfare 1 Jose Miguel Abito 2 David Besanko 3 Daniel Diermeier 4 November 22, 214 Abstrat We present a theory of orporate ampaigns and private regulation, with a fous on welfare impliations. An ativist indues private regulation through ampaigns aimed at harming the rm s reputation. In equilibrium, ampaigns have both stati and dynami e ets, whih have fairly di erent but omplementary e ets on soial welfare. The anonial ase in whih an ativist ampaign inreases welfare involves a rm that ares intensely about proteting itself against reputation loss, an ativist that is not exessively passionate, and a ampaign aimed at addressing a high-stakes negative externality. Finally, ampaigns resulting in private regulation that is essentially redistributive neessarily redue welfare. JEL L31, D6 1 We are grateful to Davin Raiha for his helpful omments, and we thank partiipants at the 14th Annual Strategy and the Environment Conferene, Kellogg Shool of Management, Northwestern University, April 214 for their useful suggestions and questions. 2 University of Pennsylvania, Wharton Shool, Department of Business Eonomis & Publi Poliy, Philadelphia, PA 1914, abito@wharton.upenn.edu. 3 Northwestern University, Kellogg Shool of Management, Department of Management & Strategy, 21 Sheridan Rd., Evanston, IL. 628, d-besanko@kellogg.northwestern.edu (orresponding author). 4 Northwestern University, Department of Managerial Eonomis and Deision Sienes (MEDS), Kellogg Shool of Management, Northwestern University, 21 Sheridan Rd., Evanston, IL. 628, d-diermeier@kellogg.northwestern.edu and Canadian Institute for Advaned Researh, CIFAR Program on Institutions, Organizations, and Growth, CIFAR, Toronto, ON M5G 1Z8. 1

2 1 Introdution The regulation of eonomi ativity is one of the main arenas of politial ompetition. The impetus for hanges to regulatory regimes frequently originates with onerned itizens, often motivated by soial or ethial onerns. Traditionally, onerned itizens have used publi institutions suh as legislatures, exeutive agenies, and ourts to advane their agenda. But in reent deades private politis has emerged as a new regulatory mehanism (e.g. Baron 21, Baron 23, Baron 212, Baron and Diermeier 27, Feddersen and Gilligan 21, Ingram, Yue, and Rao, 21, King and MDonnell 212, King and Peare 21). 5 Private politis does not operate through publi institutions. Instead, it is haraterized by the interation of private entities suh as rms, ativists and NGOs. One of the key outomes of private politis is private regulation, the voluntary adoption of rules that onstrain ertain ompany ondut without the involvement of publi agents. 6 Private regulation plays a growing role in the regulation of global ommere. For example, fourteen perent of the world s temperate forests and seven perent of global sheries are governed by private erti ation systems (Vandenbergh 213). Issues that have given rise to various forms and levels of private regulation inlude environmental protetion, human rights, disrimination, working onditions, data privay, produt safety, endangered speies, and animal welfare. Private regulation is partiularly widely used in ases where publi institutions are missing or governane proesses are underdeveloped or orrupt. Those inlude issues that transend a single government suh as regulation of labor praties within a multinational supply hain or limate hange that require intergovernmental oordination among governments but easily undermined by the free-rider problem, the lak of an adequate framework for international law governing multinational orporations, and the inability of existing multilateral organizations to impose santions. Other examples an be found in ases of intra-state on it or weak regulatory or state apaity. One suh example is the attempt to redue the availability of on it diamonds, whih are used to fund ivil wars in West Afria. In these settings private regulation often takes the form of expliit standards and rules-making proesses that resemble publi regulation but without any referene to governmental ators. Forestry Initiative. Examples are the Equator Priniples or the Sustainable But private regulation also ours in ountries with well-developed regulatory apaity. In the U.S., for example, more money is spent on private environmental inspetions than federal enforement e orts (Vandenbergh 213). Ativists and observers have argued that this shift towards private solutions may re et the inreasing di ulty to adopt new regulations. Vandenbergh (213) points out that in the U.S., no major environmental statutes has been enated sine the Clean Air At Amendments in 199. Mihael Brune, long-time exeutive diretor of the Rainforest Ation Network (RAN) and urrently exeutive diretor of the Sierra Club, ommented that Companies were more responsive to publi opinion than ertain legislatures were. We felt we ould reate more demoray in the marketplae than in the government. (Baron and Yurday 24). Private regulation an our at the level of individual rms or entire industries. An example of the former 5 Maxwell, Lyon, and Hakett (21) all this self regulation. Vogel (21) presents the losely related idea of ivil regulation. Vandenberg (213) uses the term private governane. 6 In some ases ompanies and industries adopt self-regulation to forestall publi regulattion (Lyon and Maxwell 22). Our fous will be on self-regulation to prevent harm from the ations of private agents (ativists, NGOs) not publi agents (regulators, legislators, ourts). For a reent model that studies the interation between private politis and publi regulators see Egorov and Harstad (212). 2

3 is Apple s agreement to improve labor onditions at the Chinese fatories of its main supplier Foxonn in the wake of a global media outrage that involved allegations of illegal overtime, inadequate safety onditions and poor workers housing, as well as a string of reported suiides. Another example is the variety of hanges to business praties undertaken by MDonald s over the years in areas ranging from safety, environmental stewardship, labor onditions at suppliers, animal welfare, the use of antibiotis in food animals, as well as obesity and healthy living in response to ativist ampaigns and in antiipation of both publi and private politis (Baron 26). Given the sale of global supply-hains, private regulation involving a single rm an have a signi ant impat. Walmart, for example, has over 1, Chinese suppliers and would be China s eighth largest trading partner if it were a ountry (Vandenbergh 213). Walmart uses supply-hain ontrating on a large sale from labor onditions to energy and emission requirements, pakaging, sustainable sheries, or on it-free diamonds, to name just a few. Vandenbergh (27) studied the supply hain praties of 74 large (and mainly U.S.) rms in eight setors (e.g., retailing and automobile manufaturing) and found that more than half the ompanies imposed environmental requirements on their suppliers that exeeded regulatory requirements. A reent example of industry-wide private regulation was triggered by the ollapse of the Rana Plaza fatory building in Dhaka, Bangladesh on April 24, 213. More than 1, garment workers, mostly women, died. The Rana Plaza fatory was a supplier to many international retailers and brands inluding Walmart, Disney, The Gap, and H&M. Global media overage quikly pointed to lax building onstrution and safety standards, and poor working onditions. After onsiderable media outrage and ativist pressure, ompanies either withdrew from Bangladesh or agreed to improved safety standards. Private regulation is based on agreements between private parties, and suh agreements sometimes onstitute legally binding ontrats. For example, in the wake of the Rana Plaza ollapse, more than 7 ompanies, mostly leading European retail brands, established the Aord on Fire and Building Safety in Bangladesh, a legally binding agreement to improve working onditions and safety standards among overseas suppliers. In other ases, the agreements are not legally binding. U.S. retailers and brands, for example, largely did not join the Aord, but joined in a separate agreement, the Alliane of Bangladesh Worker Safety, a voluntary agreement without legal fore. To be e etive, agreements that are not legally binding must be self-enforing, e.g. through the redible ativist threat to restart a ampaign against a ompany that fails to omply. Whether agreements are designed to be legally binding or not depends on various fators inluding di erent liability exposures and legal traditions (e.g. Kaeb 28). But whether enshrined in formal ontrats or voluntary agreements, private regulatory mehanisms in pratie play the standard-setting, implementation, monitoring, enforement, and adjudiation roles traditionally played by publi regulatory regimes (Vandenbergh 213; p. 15). Private regulation an have substantial onsequenes. The 211 protests over the prie of ottage heese in Israel led to a 2 perent immediate prie redution that was sustained into 213 (Hendel, Lah, and Spiegel 213). Similarly, an analysis of Indonesian suppliers in the textile, footwear, apparel setors showed a 3 perent wage inrease for suppliers to major brands suh as Nike, Adidas, and Reebok ompared to domesti manufaturers in the same setor (Harrison and Sorse 21). 7 Major brands, espeially Nike, had been aused sine the 199s for sweat-shop working onditions at their o -shore suppliers. After a sustained ampaign, Nike and other retailers ommitted to odes of ondut for their suppliers and established monitoring programs. Harrison and Sorse (21) also show that the rise in wages did not lead to a detetable loss of employment, but was e etively redistributive in nature, analogous to fored pro t-sharing. 7 Indonesian manufaturers in the textile, footwear, apparel setors also have higher wage rates, about 1-2 perent, ompared to other manufaturing setors (Harrison and Sorse 21). 3

4 More generally, as the sope of private regulation is growing, sholars have raised onerns whether it is an adequate response to the market and governane failures that an arise in a global eonomy (e.g. Hau er 21, Bhagwati and Narlikar 213). For example, in the ontext of labor onditions, some ritis argue that private regulation will undermine the ompetitiveness of low-ost suppliers and reverse labor fore partiipation for women in traditional soieties (e.g. Bhagwati and Narlikar 213). Others have argued that private regulation may underut the pressure for publi regulation, espeially in the ontext of global environmental onerns suh as limate hange. Suh laims are di ult to assess in the absene of a model. For example, any welfare analysis of private regulation must be lear about its omparison ase. First-best omparisons are often unrealisti sine the alternative to private politis is often the omplete absene of regulation due to lak of governane apaity or a dysfuntional or orrupt politial proess that is inapable of implementing rst-best regulatory solutions. An additional dimension of the welfare impat of private regulation is the proess by whih private regulation often omes about. Rather than engaging through traditional publi hannels, ativists have started to target ompanies diretly via orporate ampaigns. Suh ampaigns usually onsist of a spei demand, e.g. inrease wage rates, aompanied by an impliit or expliit threat to impose harm on the ompany if management does not omply with the demand (Baron and Diermeier 27). 8 Though ampaigns sometimes involve disruption of operations or onsumer boyotts, the most ommon way that ativists seek to harm rms is by a eting their reputations. To be suessful, orporate ampaigns need to attrat media attention, either through overage in the news media or by leveraging soial media suh as Faebook and Twitter. Campaigns a et private regulation (and thus soial welfare) in potentially two di erent ways. Antiipating the threat posed by a ampaign, a rm might inrease its urrent level of private regulation to defuse the threat (e.g., by reating a bu er of goodwill or by shaping a narrative that ounters an ativist s laims). In addition, beause reputation has persistene, if the ampaign atually damages the rm s reputation, the rm s future path of private regulation may di er from what it would have been had the ampaign not ourred. This paper presents a theory of orporate ampaigns and private regulation, with a fous on their welfare impliations. To the extent that rms are highly motivated to protet their reputations, the orporate ampaign is a potentially powerful inentive mehanism for induing rms to take soially bene ial ations. If publi regulation is weak or non-existent, ampaigns ould serve the soial interest. However, the harm that a ampaign an do to a ompany s reputation may be ostly, e.g., by making it more di ult for the rm whose reputation is impaired to reruit talent or aess apital. Moreover, the objetives of ativists are not neessarily the objetives of soiety as a whole; ativists may be muh more passionate about abating a soial harm than a welfare-maximizing planner would be. Thus, a ampaign might indue rms to overdiret resoures toward addressing problems that ativists are passionately about. For these reasons, it is an open question to what extent orporate ampaigns would be expeted to inrease soial welfare. Indeed, it seems oneivable that they ould even derease welfare, in whih ase soiety would be better o without them, even in a ontext in whih there is no viable publi regulation. Beause ampaigns an a et reputation and private regulation in both the present and the future, we analyze a dynami model. Spei ally, we study a three-period game with two players: a pro t-maximizing rm and an ativist whose objetive is to ahieve abatement of an unregulated soial harm. Eah period the rm an inur osts to voluntarily abate the soial harm (i.e., engage in private regulation) as a way to stohastially improve its reputation. 9 At the same time, the ativist an initiate a ampaign in eah 8 Most orporate ampaigns tend to rely exlusively on threatened harm rather than promised rewards (e.g. endorsements). See Baron and Diermeier (27) for details. 9 Beause the game terminates in the third period, the interesting ation takes plae in the rst and seond periods. However, 4

5 period that ould undermine the rm s reputation. The rm s single-period pro ts are assumed to inrease in reputation, but at a diminishing rate, what we all diminishing stati returns to reputation. Beause reputation enhanement is valuable, the rm would engage in some degree of private regulation even without ativist pressure, though its amount may be less than the soially e ient level beause it does not internalize the soial bene t of abatement. In equilibrium, the ativist will be shown to initiate a ampaign only in period 1. The rst-period ampaign arises beause seond-period private regulation dereases in reputation an impliation of diminishing stati returns to reputation and if the rm s reputation is tarnished, it will engage in more abatement ativity, whih is the ativist s objetive. If there were onstant or inreasing stati returns to reputation, the interests of the ativist and the rm would be aligned, and there would be no ampaign. 1 With a ampaign by the ativist in the rst period, the rm not only faes headwinds in its e orts to improve its reputation, it also faes the risk that its reputation ould deline between periods 1 and 2. We show that diminishing stati returns to reputation asade bak to make the rm s (endogenous) value funtion at the beginning of the seond period inreasing and onave in reputation, what we all diminishing dynami returns to reputation. With this endogenously indued risk aversion in reputation, the rm will hoose a higher level of private regulation than it would have, had there been no ativist. Private regulation in our model thus stems from two oneptually distint soures: a baseline level aimed at burnishing reputation, and an additional level (arising when the ativist launhes a ampaign) that serves to mitigate the risk of reputation loss reated by the ativist ampaign. In this sense, private regulation in our model re ets two of the drivers identi ed by Hau er (21): reputation enhanement and risk management. 11 In our model, there are two distint ways an ativist ampaign ould inrease the expeted present value of soial welfare relative to the benhmark in whih there is no private regulation (the plausible benhmark in a global setting). First, if private regulation in the initial period is less than the stati abatement optimum, the inrease in rst-period private regulation ontributes to higher rst-period soial welfare. Seond, despite the inrease in rst-period private regulation, the presene of the ativist an be shown to stohastially redue the rm s reputation in the seond period. This unfavorable shift in the distribution of the rm s seond-period reputation has two o setting e ets. On the one hand, it means that the rm is more likely to undertake higher levels of abatement in period 2 than it would have in the absene of the ativist. On the other hand, it redues the disounted present value of the rm s pro t and possibly, too, the welfare of other onstituenies, suh as onsumers or employees, who derive more value when the rm s reputation is higher rather than lower. The balane of these fores is summarized by a term we all the soial return to reputation, the extent to whih seond-period soial welfare goes up or down when the rm s reputation inreases. When the soial return to reputation is negative, the stohasti redution in the rm s reputation ontributes to higher soial welfare. Depending on parameter values, the presene of the ativist ould inrease or derease the expeted present value of soial welfare relative to the no-ativist benhmark. The typial ase in whih the ativist plays a onstrutive role is one in whih the soial marginal bene t of abatement is high, the marginal ost urve of abatement is at, diminishing stati returns to reputation are pronouned, and the ativist is not exessively passionate about the ause of abating the soial harm. Indeed, we establish that a su iently high soial marginal bene t of abatement is a neessary ondition for the ativist s presene to be soially bene ial, and we show that when private regulation is primarily redistributive, as found in the third period matters beause the prospet of improving pro tabillity in the third period motivates the rm to engage in private regulation in period 2. 1 In fat, with inreasing stati returns to reputation, the ativist would want to assist the rm in building its reputation beause that would indue the rm to undertake more abatement ativity. 11 The third driver identi ed by Hau er (21) is the more rapid spread of best praties for how to voluntarily regulate. 5

6 Harrison and Sorse (21), it must neessarily derease soial welfare. 12 The organization of the remainder of the paper is as follows. Setion 2 desribes the model of ompetition between the rm and the ativist. Setion 3 analyzes the equilibrium of the three-period game. Setion 4 explores the question of whether the ativist s ampaign bene ts soiety. Setion 5 presents a number of robustness heks and potential extensions of the model. Setion 6 summarizes and onludes. Proofs of all propositions are in the Appendix The model We onsider a model with two ators a rm and an ativist and two key features. First, the rm s operations ontribute to a visible, but unregulated soial harm, suh as a negative environmental externality or unsafe onditions in an upstream supplier s plant. Seond, the extent to whih the rm abates the harm denoted by x and the intensity of the ativist s ampaign against the rm denoted by y a ets the evolution of the rm s reputation over time. Our notion of reputation is onsistent with that employed in the soiology and management literatures: it is a soial onstrut re eting the publi s subjetive beliefs and attitudes about a ompany (Bermiss, Zaja, and King 213; Clardy 212). It is not a posterior belief about hidden information (suh as produt quality) as in a model with informed and uninformed players. 14 Throughout the paper, we use the terms abatement ativity and private regulation interhangeably. Moreover, we use those terms broadly: in our model, private regulation an enompass not only voluntary hanges in prodution operations or produt harateristis that redue environmental externalities or safety risks, but also odes of ondut whose primary impat is to transfer surplus from the rm to workers or onsumers in the form of higher wages or lower pries. 2.1 Reputational dynamis Beause reputation is fundamentally a dynami phenomenon, our model is dynami. Spei ally, the interation between the ativist and the rm is modeled as a dynami stohasti game that plays out over a nite horizon t = 1; : : : ; T, and for simpliity we assume T = 3, whih (as will be seen below) is the smallest number of periods needed to generate interesting behavior by the ativist. The strength of the rm s reputation is r, where r 2 I 2 f 1; : : : ; ; 1; : : : ; 1g is integer-valued and is the state variable in the model. We summarize the impat of reputation on the rm s pro tability by a redued-form pro t funtion (r), where () is stritly inreasing and stritly onave in r. We thus assume a stronger reputation is an asset to the rm, either beause it translates into higher demand for the rm s produts, gives the rm an edge in reruiting exeutive talent, allows it to eonomize on other brand investment ativity suh as advertising, or makes it easier to nd partners for business deals that rely on some degree of trust (Fombrun 1996). In what follows, we let r = (r), and r r+1 r. Thus, r 1 < r < r+1, while r 1 > r > r+1 :We refer to this latter ondition as diminishing stati returns to reputation (DSRR), stati beause it pertains to the properties of the single-period pro t funtion. When DSRR prevails the single-period 12 As we disuss below, we use a standard Kaldor-Hiks welfare funtion of the type typially employed in welfare analyses in the regulation and industrial organization literatures. (onsumer plus produer surplus plus/minus external bene ts/osts). If instead we used a more general utilitarian welfare funtion with dereasing marginal utilities, even if private regulation is purely redistributive, the ativist ampaign ould potentially inrease soial welfare. 13 Additional derivations and alulations as noted below are in the On-line Appendix to this paper. 14 See Board and Meyer-Ter-Vehn (213) for an elegant model in whih a rm s reputation is the market s belief about the rm s produt quality. A hallenge in adapting this type of model to the setting we study is how to develop a ompelling mehanism by whih an ativist ould distort the Bayesian learning proess. 6

7 pro t hit from reputational loss is more signi ant than the single-period pro t bump from reputational improvement. Put another way, to assume DSRR is to assume that the rm is risk averse with respet to its reputation. This strikes us as a plausible assumption. That said, we illustrate below the impliations of onstant and inreasing stati returns to reputation. The pro t inrement to reputation r plays an important role in the analysis. In what follows, it will be useful to parameterize r as follows: r = r 1 ; (1) where 2 (; 1) indiates the extent of DSRR: the lower is, the more signi ant is DSRR, and as! 1, DSRR disappears in the limit. Reputational dynamis are determined by the following proess: r t = r t 1 + f t a t, where f t 2 f; 1g is a positive shok to the rm s reputation, and a t 2 f; 1g is a negative shok to the rm s reputation. Let p = Pr(f t = 1) = P (x t ) and q = Pr(a t = 1) = Q(y t );where P () =, P (x) >, Q() =, and Q (y) >. Thus, more private regulation x in a given period inreases the probability of a positive shok in that period, and greater ampaign intensity y in that period inreases the probability of a negative shok. 15 If a positive and negative shok our in the same period, they o set eah other, and the rm s reputation remains unhanged. Beause there is a one-to-one relationship between x and p and y and q, respetively, it will be analytially onvenient to model the rm and ativist as hoosing p and q diretly. Strething the terminology somewhat, hereafter we refer to p and q as private regulation and ampaign intensity, respetively. Letting h u (p; q); h s (p; q), and h d (p; q) denote the probabilities that the rm s reputation inreases, stays the same, and dereases from one period to the next, we have h u (p; q) = p(1 q); h d (p; q) = (1 p)q, and h s (p; q) = 1 h u (p; q) h d (p; q): 2.2 The rm and soiety The soial bene t w(p) = w(p 1 (p)) of private regulation is given by w(p) =!p, where! > is the soial marginal bene t of abatement e ort. This soial marginal bene t re ets both the real external bene ts of private regulation (e.g., the soial bene ts of redued arbon emissions), as well as the net balane of any transfers of surplus among the rm and its stakeholders (e.g., the net hange in worker plus produer surplus due to a ode of ondut requiring the rm s upstream suppliers to pay higher wages). The total ost to the rm of private regulation the sum of implementation, ompliane, and administrative osts plus the osts of publiizing the rm s e orts so as to boost reputation is given by (p) = (P 1 (p)). 16 For simpliity we assume it is quadrati: (p) = 2 p2 ; where >. Let p S = min! ; 1 optimum, the level of private regulation that maximizes the (stati) net bene t of abatement,!p 1 denote the stati abatement 15 An alternative formulation would allow for the possibility that urrent reputational shoks ould depend on the entire history of private regulation and ampaign intensity, i.e., Pr(f t = 1) = P (x t; x t 1 ; : : : ; x 1 ) and Pr(a t = 1) = Q(y t; y t 1 ; ; : : : ; y 1 ). This formulation, whih inludes our setup as a speial ase, adds notational omplexity but relatively little additional insight. For this reason, we fous on the ase in whih hanges in reputation from the urrent to the next period depend only on urrent period private regulation and ampaign intensity. 16 The ost (p) does not inlude transfers of surplus (e.g., lower produer surplus due higher wages) from the rm to stakeholders. These are inluded, along with their o setting bene ts to reipients, in w(p). 2 p2. 7

8 2.3 The ativist An ativist ampaign tries to draw publi attention to the fat that the rm s business operations reate a soial harm. By hoosing a higher level of ampaign intensity (e.g., by devoting more resoures to organize volunteers or working harder to draw attention to the ampaign so that it is more likely to be overed in the media), the ativist inreases the likelihood that the rm s reputation will su er a negative shok. The total ost to the ativist of a ampaign is given by (q) = (Q 1 (q)) = q2 2, where. We assume that the ativist s ampaign tatis have no diret ost to the rm. These tatis are used only to draw attention to the harm the rm reates in the traditional or soial media in the hope of hurting the rm s reputation. We further assume that the ativist does not derive a diret bene t from hurting the rm s reputation. That is, it is no happier when the rm has a lower reputation than it is when the rm has a higher reputation. Instead, we assume that the ativist is pragmati, i.e., its only soure of bene t is the level of abatement ativity the rm provides eah period. We assume the ativist maximizes the di erene between w(p) and the ost of mounting a ampaign (q), where 1 is the ativist s passion for the abatement ause. A ampaign bene ts a pragmati ativist only if, by ausing the rm s reputation to derease, the rm subsequently inreases its private regulation. 2.4 Equilibrium onditions The interation between the ativist and the rm is modeled as a dynami stohasti game. An equilibrium is desribed by f(p rt ; q rt ; u rt ; v rt )j (r; t) 2 I f1; 2; 3gg, where u rt and v rt are the rm s and ativist s values in state r, period t. Letting F 2 (; 1) and A 2 (; 1) denote the disount fators of the rm and ativist, respetively, the values u rt and v rt are given by the Bellman equations: u rt = max p rt2[;1] U rt(p rt ; q rt ) r p 2 rt 2 + F u r;t+1 + F fu r;t+1 h u (p rt ; q rt ) u r 1;t+1 h d (p rt ; q rt )g ; (2) v rt = max V rt(p rt ; q rt ) = w(p rt ) q rt2[;1] q 2 rt 2 + Av r;t+1 + A fv r;t+1 h u (p rt ; q rt ) v r 1;t+1 h d (p rt ; q rt )g ; where u rt u r+1;t u rt and v rt v r+1;t v rt, and it is understood that u r4 = v r4 =. U rt (p rt ; q rt ) and V rt (p rt ; q rt ) are stritly onave in p rt and q rt, respetively, so the Kuhn-Tuker rst-order onditions are neessary and su ient for a unique global optimum. Using the expression for the derivatives of h u (p; q) and h d (p; q) those onditions in state (r; t) an be expressed as: p rt + F [(1 q)u r;t+1 + qu r 1;t+1 ] + & p rt q rt + A [fp ( v r;t+1 ) + (1 p)( v r 1;t+1 )g] + & q rt & p1 rt = ; 8 >< >: 8 >< rt = ; >: & q1 p rt 2 [; 1] & p rt p rt = & p1 rt (1 p rt ) = & p rt ; & p1 rt 9 >= q rt 2 [; 1] & q rt q rt = & q1 rt (1 q rt ) = & q rt ; & q1 rt where & p1 rt, & p1 rt, & q rt, and & q1 rt are Lagrange multipliers. Hereafter, we let denote equilibrium values. We maintain the following assumption throughout our analysis: Assumption 1 For all r 2 I, > F (1 + F ) r : (3) >; ; (4) 9 >= >; : (5) 8

9 This assumption says that the marginal ost of private regulation at p = 1, exeeds the disounted gain from improving reputation. The impat of the assumption is to eliminate equilibria (with or without the ativist) whih involve orner solutions for private regulation (p = 1) Equilibrium analysis We proeed in two steps. We rst desribe what the rm would optimally do in the absene of an ativist (the no-ativist benhmark). This forms the foundation for analyzing the equilibrium with the ativist, whih follows. 3.1 No ativist benhmark We denote the no-ativist benhmark by the supersript. In the Appendix, we derive the rm s private regulation and value funtion: 18 p r3 = ; (6) p r2 = F r p r1 = F u r2 2 (; 1); (7) 2 (; 1): (8) u r3 = r : (9) u r2 = (1 + F ) r + 2 F ( r ) 2 : (1) 2 u r1 = r + F u 2 r2 2 p r1 + F p r1u r2: (11) That is, there is no private regulation in the terminal period, but positive amounts in the penultimate and initial periods. Key properties of the no-ativist solution (established in the Appendix) are as follows. In period 2, the level of private regulation is dereasing in reputation, i.e., p r+1;2 < p r2: (12) The rm s value is stritly inreasing in reputation in eah period, i.e., u rt > ; r 2 I; t = 1; 2; 3: (13) The inrement to value is greater the more patient the rm, F > ; r 2 I: The rm s seond period value funtion exhibits diminishing marginal returns to reputation, i.e., u r2 < u r 1;2. As a onsequene of the onavity of the seond-period value funtion, the rm s rst period private regulation is stritly dereasing in its reputation level, i.e., p r+1;1 < p r1: 17 For results where Assumption 1 is important, we disuss the impliations of it not holding. 18 Unless expliitly noted, the results in this and subsequently propositions that are expressed in terms of an arbitrary state r should be thought of as applying to all possible states in I. 9

10 The onavity of the seond-period value funtion is important for our subsequent analysis. It shows that the assumed property of DSRR endogenously asades bakward to make the rm s value funtion onave in reputation in period 2. For this reason, we refer to the result that u r 1;2 > u r2 as diminishing dynami returns to reputation (DDRR). With this property, a strong reputation is a deterrent to private regulation. Beause DDRR plays an important role in our analysis with an ativist, it is useful to know how it varies with key parameters. From (1) and using (1) we have: u r2 = r 1 ((1 + F ) F r 1 2 ) ; (14) so u r 1;2 u r2 = r 1 ((1 )(1 + F ) 2 F r ) : (15) 2 u r 1;2 u r2 = r 1 F 1 2(1 2 ) F r 1 < ; where the inequality follows beause Assumption 1 implies F r 1 < 1 and sine < 1, 2(1 ) 1 2. Thus, the more pronouned is DSRR (lower ), the more pronouned is DDRR. It is also straightforward to show that lim!1 u r 1;2 u r2 =, so that DSRR is a neessary ondition for DDRR. Di erentiating with respet to the disount fator F, we ( u r 1;2 u r2 = (1 ) r 1 F F r 1 (1 + ) 1 2 ) : Given Assumption 1, a su ient ondition for this to be positive is (1 + ) 1 the rm is su iently patient or su iently risk averse < 1 + F, whih holds if Thus, for su iently high disount fators and/or high degrees of risk aversion, DDRR beomes more pronouned the more patient the rm is. 3.2 The equilibrium with an ativist We now turn to the haraterization of the equilibrium with an ativist Preliminary results Let us rst establish results that immediately follow from the nite-horizon struture of the game. (Proofs are in the Appendix.) First, p r3 = qr3 = qr2 =, i.e., in the terminal period the rm does not engage in private regulation (as in the no-ativist ase), and in both the terminal and penultimate periods, the ativist does not mount a ampaign. Seond, the rm s private regulation in period 2 equals the level in the no-ativist ase, i.e., p r2 = p r2. Third, the equilibrium values for the rm and the ativist in periods 2 and 3 are as follows: u r3 = u r3 = r ; u r2 = u r2 = (1 + F ) r + 2 F ( r ) 2 ; 2 19 In fat, beause max 2(;1) (1 + ) 1 2 = 32=27, F > 5=27 ' :19 is su ient for this inequality to hold for all 2 [; 1]. 1

11 vr3 = ; vr2 = w(p r2): The rm s value in periods 2 and 3 orresponds to the values in the no-ativist benhmark whih, as noted above, is stritly inreasing in reputation level. The ativist s value, by ontrast, is dereasing in the rm s reputation in period 2, i.e., vr2 <. Beause the ativist bene ts only from a hange in the rm s behavior in the subsequent period, and beause the rm s behavior (trivially) does not vary in period 3, the ativist gains no bene t from a ampaign in period 2, whih is why qr2 =. By ontrast, the rm may want to engage in private regulation in period 2 beause the potential improvement in reputation would result in higher pro ts in period 3. An important impliation of vr2 < is that the interests of the rm and the ativist in period 1 are opposed. Beause the ativist prefers the rm to be in a lower reputation state in period 2 to a higher one, it bene ts from ations in period 1 that make a derease in the rm s reputation more likely. The ativist bene ts by weakening the rm s reputation in period 2 beause it motivates the rm to undertake more private regulation in that period. A weaker reputation keeps the rm hungry to improve its image. This preferene by the ativist sets the stage for a orporate ampaign in period First-period equilibrium and impliations for private regulation dynamis Let w r2 w(p r2) w(p r+1;2) =! F w r 1;2 w(p r 1;2) w(p r2) =! F [ r r+1 ] =! F r 1 (1 ) > : (16) [ r 1 r ] =! F r 1 (1 ) > wr2 > : (17) From the haraterization of vr2 above, vi2 = w i2 < for i = r 1; r. Thus, v i2 and u i2 are positive for i = r 1; r, and the Kuhn-Tuker onditions in (4)-(5) imply that & p r1 = &q r1 = and p r1 > and q r1 > : 2 Moreover, beause Assumption 1 implies F u r2 < F u r 1;2 < 1, we have p r1 < 1. Thus, the only possibility for a orner solution is for q to equal 1. From the Kuhn-Tuker onditions, we an therefore derive reation funtions for the rm and ativist given by: p R r1(q) = F qr1(p) R A = min u r2 + u r 1;2 u r2 q : (18) w r 1;2 wr 1;2 wr2 p ; 1 : (19) Beause u r 1;2 > u r2, the rm s reation funtion is inreasing in ampaign intensity. Moreover, to the extent that u r 1;2 is large relative to u r2, ampaign intensity will have a stronger impat on private regulation on the margin. In other words, the ativist provides the strongest inentives for the rm to engage in private regulation when DDRR is most signi ant. As shown above, DDRR tends to be large when the rm is patient and when DSRR is pronouned. Thus, patient rms for whih the loss of reputation is far more onsequential than gains to reputation tend to be most responsive to an ativist ampaign. Beause w r 1;2 > w r2, q R r1(p) is non-inreasing in p, and it is stritly dereasing when q R r1(p) < 1. Thus, all things being equal, the smaller the rm s antiipated abatement ativity, the greater the intensity 2 To see why, suppose to the ontrary that p r1 =. The omplementary slakness onditions in (4) would then imply & p1 r1 =. The rst-order ondition in (4) would then redue to F [(1 q)u r;t+1 + qu r 1;t+1 ] + & p r1 = ; a ontradition sine u r2 >, u r 1;2 > and &p r1. A similar proof establishes q r1 >. 11

12 of the ativist s ampaign (unless the ativist s intensity is at the highest possible level). This is onsistent with the empirial evidene in Lenox and Eesley (29) that environmental ativists tend to target rms with higher levels of greenhouse gas emissions. The rst-period equilibrium p r1; q r1 solves the reation funtions simultaneously and is either a orner equilibrium given by p r1 = F u r 1;2, qr1 = 1 or an interior equilibrium, p r1 2 (; 1), qr1 2 (; 1), given by: 21 p r1 = q r1 = F A h F A h F A i u r2 + A w r 1;2 u r 1;2 2 (p u r 1;2 u r2 w r 1;2 wr2 r1; p r 1;1): (2) i F u r2 wr 1;2 + F u r2 wr2 (21) u r 1;2 u r2 w r 1;2 wr2: A w r 1;2 The orner equilibrium arises if and only if: A " w r 1;2 1 F u r 1;2! + wr2 F u r 1;2 # 1: (22) Condition (22) will hold only if the ativist is su iently patient, passionate, or ost e ient, or if the soial bene t! of private regulation is su iently large. Figure 1 shows the ase of an interior equilibrium. We now haraterize the impat of the ativist on private regulation in the rst period and on expeted private regulation in the seond period. Proposition 1 For any initial reputation level r, the presene of the ativist (a) inreases private regulation in the rst period, i.e., p r1 > p r1, and (b) inreases expeted private regulation in the seond period, i.e., Er (p 2 ) > Er (p 2 ). Though the ativist pushes private regulation in the same diretion in periods 1 and 2, the hannels through whih these impats operate are di erent. The ativist inreases private regulation state by state in period 1, a stati e et. In period 2, the ativist has no stati e et; reall, p r2 = p r2, i.e., seond-period private regulation is the same in any given state with or without the ativist. The inrease in expeted private regulation in period 2 (whih, as we shall disuss, ours in spite of the inrease in private regulation in period one) arises beause the ativist ampaign (and the assoiated hange in rst-period private regulation) hanges the evolution of reputation between periods 1 and 2. Let us rst disuss the intuition for part (a) of Proposition 1. By launhing a ampaign, the ativist makes it more di ult for the rm to improve its reputation between periods 1 and 2. In isolation, exogenous hanges in fators that make it harder for the rm to improve its reputation would derease inentives for private regulation. However, the ativist also puts the rm at risk of a reputational loss, and given DDRR the rm gains more by avoiding this downside than it gains by inreasing its reputation. Aordingly, the inrease in private regulation an be interpreted as a form of risk mitigation in the fae of possible reputation loss. To explain part (b) of Proposition 1, it is useful to examine the expression for Er (p 2 ) Er (p 2 ): Er (p 2 ) Er (p 2 ) = h dr p r 1;2 p r2 h ur h ur (p r2 p r+1;2); 21 The expressions in (2) and (21) were found by solving the reation funtions simultaneously assuming q < 1. The derivation of the neessary and su ent ondition for the orner solution in (22) is in the Appendix. 12

13 where h ur h u (p r1; q r1) = p r1(1 q r1) and h ur h u (p r1; ) = p r1 are the probabilities the rm s reputation goes up when there is an ativist and when there is not, and h dr h d(p r1; q r1) = (1 p r1)q r1 is the probability that the rm s reputation delines when there is an ativist. (Without an ativist, this probability is.) Beause p r 1;2 > p r2 and p r2 > p r+1;2, the sign of E r (p 2 ) E r (p 2 ) depends on how the ativist shifts the distribution over reputational states in period 2. (Reall that when the initial state is r, the period 2 distribution would be over states r 1, r, and r + 1.) The ampaign reates a positive probability of reputation loss, whih tends to worsen the distribution, and it ould potentially interfere with the rm s e orts to improve its reputation, whih would also worsen the distribution. However, the inrease in private regulation in period 1 might o set this latter e et to suh an extent that the probability that the rm s reputation goes up is atually higher when there is an ativist than when there is not, i.e., h ur > h ur, a rebound e et. If so the ativist s ampaign ould oneivably result in a mean-inreasing spread of distribution over reputational states, and with less private regulation assoiated with higher reputational states, this would imply E r (p 2 ) < E r (p 2 ). But in proving part (b) of Proposition 1, we show that the rebound e et does not arise, i.e., h ur < h ur. The intuition for this as is follows. An inrease in reputational risk (higher q) leads the rm to inrease its private regulation in period 1 (the risk management motive). The rebound e et will not arise if the rate dp R r1 (q) dq at whih the rm inreases this private regulation is not too large. (Essentially the proof of part (b) of Proposition 1 formalizes not too large. ) The risk management motive is weaker and thus the rate of inrease dpr r1 (q) dq smaller to the extent that indued seond-period risk aversion, measured by u r 1;2 u r2; the onavity of the seond-period value funtion, is smaller. (Note from (18) that dpr r1 (q) dq is proportional to u r 1;2 u r2.) For parameter onstellations satisfying Assumption 1, seond-period private regulation is bounded away from one, reduing the likelihood that reputation and thus pro t will inrease between periods two and three. This attens the seond-period value funtion by muting the e et of the return to reputation, r, whih ultimately limits the steepness of rm s rst-period reation funtion. 22 The risk aversion of the rm with respet to its reputation that is implied by DSRR plays a key role in Proposition 1. Indeed, DSRR is the pragmati ativist s lever. It is what leads to the opposition of interest between the rm and the ativist that is the fuel for an ativist ampaign, and it is what indues the rm to hoose a higher level of private regulation in period 1 than it would have in the absene of the ativist. We an make this point even more forefully by onsidering the impliations of onstant or inreasing returns to reputation, i.e., r r 1 for all r 2 I. From (7), it follows that p r2 is onstant in r with onstant returns to reputation, and p r2 is inreasing in r with inreasing returns to reputation. From (16) and (17), we would have w i2 for i = r 1; r, from whih it would follow from (5) that qr r1(p) =. Indeed, when returns to reputation are inreasing, the interests of the rm and the ativist are fully aligned. Not only is there no need for the ativist to launh a ampaign to harm the rm s reputation, it would be ounterprodutive. If the ativist ould somehow help the rm improve its reputation, it would do so. This disussion an be summarized by the following proposition: 22 This, of ourse, raises the question of what would happen if Assumption 1 did not hold. It should be noted that Assumption 1 is su ient to prove part (b) of Proposition 1, but not neessary. However, if the assumption was signi antly violated at adjaent reputation levels r 1 and r, suh that p r2 = p r 1;2 = 1, then the equilibrium between the rm and the ativist would be uninteresting if the initial reputation level was r, as the ativist would have nothing to gain by launhing a ampaign. If, by ontrast, 1 was strongly violated reputation level r 1 suh that p i r 1;2 = 1 but held at r, so that p r2 < 1; then we would have qr1 > and p r1 hp 2 r1 ; p r 1;1. The rebound e et might or might not arise in this ase, but we would not be able to use the sequene of logi in the proof of part (b) of Proposition 1 to rule it out. 13

14 Proposition 2 When an ativist is pragmati, a neessary and su ient ondition for the ativist to launh a ampaign in period 1 is if the rm is risk averse in its reputation, i.e., DSRR holds. If the rm was not risk averse in its reputation, a ampaign would be ounterprodutive for a pragmati ativist, and the ativist would even prefer to help the rm build its reputation if stati returns to reputation were inreasing. It is oneivable that the relationship between r and r is s-shaped, with an initial region of inreasing marginal returns followed by a region of diminishing marginal returns to reputation. A relationship of this type ould arise if a rm must attain a ertain ritial reputation threshold in order to begin to bene t from further enhanements of reputation, followed by an eventual diminishment of the returns to reputation as reputation ontinues to rise. Proposition 2 implies that a pragmati ativist would avoid targeting a rm that is in the stage of inreasing returns to reputation beause as the rm s reputation grows over this range, it would inrease its private regulation. To the extent that inreasing returns to reputation is orrelated with the maturity of the rm s business or where it is in the produt life yle, this insight ould lead to testable impliations for what types of rms are targeted by ativists Comparative statis As an be inferred from Figure 1, any parameter hange that shifts the ativist s reation funtion rightward (leftward), but does not shift the rm s reation funtion, results in an equilibrium with a higher (lower) level of private regulation and a more (less) intense ampaign by the ativist in period 1. From (19), it is straightforward to establish r1 r1 r1 (p) >, r1 <. Further, these parameters do not a et p R r1(q). It follows that the more passionate the ativist (higher ), the more patient the ativist (higher A ), the more ost e ient the ativist (lower ), and the higher the soial marginal bene t (higher!), the greater the level of equilibrium private regulation and ampaign intensity in period Figure 1 also implies that a parameter hange that simultaneously shifts the rm s and ativist s reation funtion rightward will inrease equilibrium private regulation in the rst period but have an ambiguous impat on equilibrium ampaign intensity. This logi implies three additional omparative statis results: the more patient the rm (higher F ), the atter the marginal ost urve of abatement (lower ), and the weaker the rm s initial reputation, the greater the level of equilibrium private regulation in period Do ativist ampaigns bene t soiety? Are ativist ampaigns bene ial for soiety? Beause we have assumed that publi regulation is either absent or ine etive, in the absene of ativist ampaign, the only fore pushing the rm to engage in abatement ativity is reputation enhanement. Our benhmark, then, is the no-ativist equilibrium, and we evaluate the extent to whih the presene of the ativist inreases or dereases welfare relative to this equilibrium. 4.1 The impat of the ativist on disounted soial welfare We evaluate the impat of the ativist using expeted disounted soial welfare for an arbitrary starting reputation state r in period 1. Though this is not the only possible metri, it is both omprehensive and natural. 23 These results an also be veri ed by straightforward (and tedious) di erentiation of (2) and (21). 24 The derivation of these results is presented in the Appendix. 14

15 Per-period soial welfare equals the soial bene t of abatement,!p, plus the disounted value of rm pro t, minus (if relevant) the ost of the ativist ampaign, q2 2, plus any additional per period surplus r of agents whose welfare varies with the rm s reputation and the rm does not internalize. For example, if onsumer demand depended on the rm s reputation, r would inorporate onsumer surplus (taking into aount the prie the rm is able to harge beause of its reputation). The inrement to this non-internalized reputational spillover, r r+1 r ould oneivably be positive or negative, but the more natural ase is r, i.e., some of the soial bene ts of the rm s higher reputation are shared with parties that transat with the rm. 25 spillovers. 26 We let r = r where measures the degree of non-internalized reputational Throughout the rest of the analysis, we refer to r as onsumer surplus. To simplify notation without signi ant loss of insight, we assume that the soial disount fator equals the rm s disount fator. To begin the analysis, we note that beause the rm does not internalize the soial bene t of abatement or the bene ts of reputation building on onsumer surplus, the no-ativist equilibrium does not maximize soial welfare. 27 Proposition 3 Let p F rt be the rst-best level of abatement hosen by a soial welfare-maximizing planner. For t = 2; 3, p F rt > p rt i.e., the planner hooses stritly more abatement than the rm hooses in the absene of the ativist. For t = 1, if non-internalized reputational spillovers are su iently small spei ally, 1 < 1 p F [p r2 p r+1;2] F r1 > p rt, i.e., the planner hooses stritly more abatement in period 1 than arises in the absene of the ativist. Otherwise, p F r1 p rt. At rst blush, Proposition 3 seems to re et nothing more than the standard textbook intuition that a rm will underinvest in abatement e ort if it does not internalize its soial bene ts. But the proposition re ets deeper trade-o s sine there are irumstanes under whih the rm s abatement in the absene of the ativist might be greater than the rst-best level. This is beause the dynamially optimal abatement levels for both the rm and the planner depend not only on the stati soial marginal bene t of abatement but also on the impat of abatement on the future path of reputation that is embedded in the rm s and planner s value funtions. If reputational growth is more valuable to the rm than it is soially, it is oneivable that the rm would invest more in abatement in the rst period (for reputation development purposes) than a 1 planner would. If > 1, the planner s value funtion is less elasti with respet to reputation F [p r2 p r+1;2] than the rm s, leading to overinvestment in reputation building by the rm. Proposition 3 opens the door to the possibility that the ativist s presene ould inrease soial welfare. However, this is not a sure thing. For one thing, while the no-ativist private regulation is less than the soially e ient level of abatement in periods 2 and 3 (in period 3, there is a lassi textbook negative externality problem, whih asades bak to period 2), the presene of the ativist does not hange private regulation in these periods. In period 1, the ativist will (exept when the non-internalized reputational spillover is extremely large) push the rm in the right diretion, but it does so by reating the risk that the rm s reputation will fall between period 1 and 2, something a benevolent planner would not do. 28 The upshot is that to disern the welfare impat of an ativist ampaign, we must ompare one distorted equilibrium to another distorted equilibrium. 25 For example, if r is onsumer surplus, then r implies that a pro t-maximizing rm does not apture e.g., via higher pries more than the entire bene t onsumers enjoy due to the rm s enhaned reputation. In the On-line Appendix, we disuss the onditions on market demand that ensure this is the ase for a single-produt monopolist. 26 In the On-line Appendix, we disuss how our results would hange if <. 27 We haraterize the rst-best soial welfare maximization problem in the On-line Appendix, and we prove Proposition 3 there. 28 A planner would diretly ontrol abatement to maximize welfare and would not (as we show in the analysis of the rst-best problem in the Appendix) hoose positive levels of q. 15

16 4.1.1 Disounted soial welfare The expeted disounted present value of soial welfare in period 1, given that the rm s reputation is r, is given by W r1 = X i2i 3X t=1 t 1 (p it ) F 2 (q it ) 2 i + i +!p it 2 2 itjr ; where itjr is the probability that the rm s reputation is i at time t given it was initially r, and r1jr = 1. The impat of the ativist on disounted welfare, W r = W r1 W r1 an be expressed as W r = X i2i 3X t=1 X i2i t=1 t 1 (p F it )2 i + i +!p it 2 3X t 1 (p F it )2 i + i +!p it itjr 2 (qit )2 (p it )2 i + i +!p it itjr 2 2 itjr ; (23) o o where n itjr ji 2 I; t 2 f1; 2; 3g and n itjr ji 2 I; t 2 f1; 2; 3g are the probability distributions over states indued by equilibrium play without and with the ativist. The impat of the ativist has both a stati e et on disounted soial welfare (the rst term in (23)) and a dynami e et (the seond term in (23)). The stati e et is how the ativist in uenes private regulation and per-period welfare, for a xed reputational. Beause p rt = p rt for all r 2 I, t 2 f2; 3g; the stati e et operates only in the rst period, and beause r1jr = 1, it is given by the di erene in the net bene t of abatement in period 1 minus ampaign osts, h i.e.,!p r1 2 (p r1) 2i h i!p 2 r1 2 p (q r1 )2 r1 2. The dynami e et is the impat of the ativist on how the rm s reputational trajetory unfolds over time. It is driven by how the ativist hanges the indued probability distributions over states. Beause r1jr = r1jr = 1 this e et operates on welfare in the seond and third periods. In a model with only three periods, the stati e et and the dynami e et operate in non-overlapping periods. However, in a T -period model, the stati e et operates in periods 1; :::; T 2, and the dynami e et operates in periods 2; :::; T. Thus, the two e ets are more than just a relabeling of the impat of the ativist in di erent periods. They re et two oneptually distint hannels through whih the presene of the ativist an hange disounted expeted soial welfare: by hanging welfare statewise (by hanging the level of private regulation in eah reputation state) and by hanging the evolution of states through time (by hanging the transition probabilities, whih depend on both private regulation and ampaign intensity). Utilizing the haraterization of the equilibrium with and without the ativist, we an rewrite W r as 29 W r = ( h!p r1 2 (p r1) 2i h!p r1 i 2 2 p r1 + h ur h ur F r2 h dr F r 1;2; (q r1 )2 2 ) ; (24) where and i2 u i2 + s i2 w i2; i = r; r 1; s r2 (1 + F ) r + p r2 F r (25) is the disounted present value of onsumer surplus in period The On-line Appendix shows how we derive (24) from (23). A key insight in the derivation is that onditional on reputation in period 2, the probability distribution over period 3 states is the same with and without and ativist. This is beause, as established earlier, the ativist does not launh a ampaign in period 2. 16

17 Having established in proving part (b) of Proposition 1 that h ur h ur <, the diretion of the dynami e et depends entirely on the terms r2 and r 1;2. They are the hanges in seond-period soial welfare when the rm s reputation hanges from r to r + 1 and r 1 to r, respetively, or what we will all the soial returns to reputation. The omponents of i2, i = r; r 1, are: u i2 > and s i2 > : the inreases in the expeted disounted present value (over periods 2 and 3) of the rm s pro ts and onsumer surplus, respetively, due to the rm having a stronger reputation in period 2. w i2 =! p i2 p i+1;2 > : the soial ost of an improvement in the rm s reputation due to the lower level of private regulation the rm undertakes in period 2 when its reputation inreases. Beause u i2 ; s i2 ; and w i2 are eah positive, r2 and r 1;2 an either be positive or negative depending on whether the diret bene ts to the rm and onsumers from a higher reputation exeed or fall short of the soial ost that a stronger reputation reates beause it indues the rm to ut bak its private regulation Welfare omparison Even though the presene of the ativist inreases private regulation in the rst period and expeted private regulation in the seond period, disounted soial welfare ould either go up or down, depending on parameter values. (This an be seen in the numerial examples we present below.) Given that, we will identify parameter onditions that tend to make Wr dynami e ets disussed above. positive or negative. To do this, we take a deeper look at the stati and The dynami e et: how the ativist s presene a ets expeted seond-period welfare At the ore of the dynami e et are the soial returns to reputation whih, using (31) 3, (16), and (25), an be expressed as: r2 = i (r 1) r 1 ((1 + F )(1 + ) ( )2 F i (r 1) r 1 1 2! F [1 ] ) ; i = r; r 1; It an be shown that if r 1;2 >, then r;2 > ; and it an also be shown that if r2 <, then r 1;2 <. This implies that one of three possibilities an obtain: r2 <, r 1;2 < ; r2 >, r 1;2 > ; and r2 >, r 1;2 <. Under the rst two onditions, the impat of the dynami e et on soial welfare is unambiguous, so we haraterize the parameter onditions under whih these ases obtain. Rearranging (26) for i = r, a su ient ondition for r2 < and r 1;2 < whih implies that the presene of the ativist inreases seond-period expeted welfare is that the soial marginal ost of abatement is su iently large, i.e.: (26)! > (1 + F )(1 + ) F (1 ) ( ) (1 + ) F r 1 (; F ; ; ; r 1 ): (27) This ondition makes sense: if the soial marginal bene t of abatement is su iently large, the soial ost w i2 of the redued abatement due an improvement in a rm s reputation will be large and will dominate the soial bene t, u i2 + s i2, making the soial returns to reputation negative. It is straightforward to 3 This equation is in the Appendix. 17

18 @ r 1 <, > :31 In light of (27), it follows that the presene of the ativist is more likely to inrease seond-period expeted welfare: (a) the atter the marginal abatement ost funtion (i.e., the lower is ); (b) the smaller the non-internalized reputational spillover (i.e., the smaller is ); () the more reputationally risk averse the rm (i.e., the smaller is ); (d) the larger the return to reputation, r 1, and (e) the more patient the rm (i.e., the higher is F ). Analogously, using (26) for i = r 1, a su ient ondition for r2 > and r 1;2 > whih implies that the presene of the ativist dereases seond-period expeted welfare is that the soial marginal bene t of abatement is su iently small:! < (1 + F )(1 + ) F (1 ) ( ) (1 + ) F r 1 (; F ; ; ; r 1 ): (28) It an be @ > @ r 1 <, >. In light of (28), the presene of the ativist is more likely to derease seond-period expeted welfare: (a) the steeper the marginal abatement ost funtion (i.e., the larger is ); (b) the larger the non-internalized reputational spillover (i.e., the larger is ); () the less reputationally risk averse the rm (i.e., the larger is ); (d) the smaller the return to reputation, r 1, and (e) the less patient the rm (i.e., the lower is F ). The stati e et: how the ativist a ets rst-period welfare If private regulation without an ativist exeeds the stati abatement optimum p S, the presene of the ativist redues the net bene t of abatement in period 1 by exaggerating the over-delivery of abatement e ort. Thus, p r1 >! is a su ient ondition for a ampaign to derease the net bene t of abatement in period 1 and rst-period soial welfare as well (beause the ativist s ampaign is ostly). Using (8) and (31), we an rewrite this su ient ondition as! < F r 1 ((1 + F ) F r 1 2 ) (; F ; ; r 1 ): (29) Thus, if the soial marginal bene t of abatement! is small enough, the ativist s presene dereases the net bene t of abatement in period 1. It is straightforward to r 1 >. Given F (29), the presene of the ativist is more likely to redue rst-period welfare: (a) the steeper the marginal abatement ost funtion; (b) the greater the return to reputation, and () the higher the disount fator. The impat of on is ambiguous. Analogously, beause p r1 < p r1 < p r 1;1, if p r 1;1 is less than the stati abatement optimum p S, the presene of the ativist inreases the net bene t of abatement in period 1 by stimulating under-supplied abatement e ort. If q r1 is su iently small (whih ours if A or is small), the ost of the ampaign will be small, and the ativist s presene would deease rst-period welfare. We an rewrite the ondition p r 1;1 < p S as:! > F r 1 ((1 + F ) F r 1 2 ) (; F ; ; r 1 ); (3) whih implies that if! is large enough, the ativist s presene inreases the net bene t of abatement in period 1. It is straightforward to @r 1 >, >. Thus, the presene of the ativist is more likely to inrease rst-period welfare: (a) the atter the marginal abatement ost funtion; (b) the smaller the return to reputation; and () the smaller the disount fator. 31 The @ are not immediately obvious and are derived in the On-line Appendix to this paper. 18

19 Overall welfare omparison Figure 2 atalogs the possible welfare senarios, depending on the impat of the ativist on rst-period welfare and expeted seond-period welfare, the latter depending on the nature of the soial return to reputation. Using the gure and the results just presented, Table 1 summarizes how hanges in parameters a et the impat of the ativist on soial welfare. in eah period and overall. A + ( ) indiates that the parameter hange makes it more likely (less likely) that the ativist s presene inreases welfare: Parameter First-period Seond-period Overall hange welfare* expeted welfare welfare Larger soial marginal bene t of abatement (!) Steeper marginal abatement ost urve () Smaller non-internalized reputational spillover () no e et + + More reputationally risk aversion (lower ) ambiguous + ambiguous Larger return to reputation ( r 1 ) + ambiguous More patient rm ( F ) + ambiguous *Provided A or is su iently small. Table 1 Table 1 gives us three su ient onditions for the presene of the ativist to inrease (derease) disounted expeted welfare: A su iently large (small) soial marginal bene t of abatement; A su iently at (steep) marginal ost of abatement urve; A su iently small (large) reputational spillover to onsumers. The intuition for these rst two onditions will be learer after we disuss Proposition 4. To understand the impat of ;we note from (29) and (3) that hanges in reputational spillover have no impat on how the ativist a ets rst-period welfare. However, a smaller reputational spillover lowers the soial returns to reputation, making it more likely that they will be negative, thus making it more likely that the ativist s presene inreases seond-period soial welfare. The disussion so far has foused on su ient onditions for the ativist to inrease or derease soial welfare. We now o er a neessary ondition. Proposition 4 The lowest level of the soial marginal bene t of abatement! onsistent with Wr > is bounded away from. Thus, for the ativist s presene to inrease expeted disounted soial welfare, the soial marginal bene t of abatement must be su iently large. Expressed equivalently, for! > but extremely small, the presene of the ativist redues welfare. This result is not obvious. When! =, there is no ativist ampaign, and there is no apparent intuition that would suggest that an in nitessimal inrease in the soial marginal bene t of abatement! would redue welfare. Indeed, one might expet that a slight inrease in! from zero ould inrease soial welfare: equilibrium ampaign intensity q r1 would beome positive and that would be assoiated with an inrease in private regulation that would inrease the gross soial bene t of abatement,!p. But this intuition is inomplete. First, if rst-period private regulation without an ativist is already distorted above the stati abatement optimum p S whih must neessarily be the ase if! is positive but 19

20 su iently small 32 the stati e et of the ativist (i.e., the inrease in private regulation in the rst period) will have a rst-order negative impat on rst-period welfare. Seond, when! is positive but su iently small, the soial return to regulation must be positive (intuitively beause the soial ost of the drop in abatement due to an improvement in the rm s reputation is low). 33 Thus, the dynami e et of the ativist s presene is to redue seond-period expeted welfare by stohastially shifting the distribution of seond-period reputation in an unfavorable way. 34 As a onluding point, we note that this intuition is robust to more general funtion forms: if the soial marginal marginal bene t of abatement is su iently low, the rm is likely to over-supply private regulation relative to the stati abatement optimum, and the soial returns to reputation in the seond period are likely to be positive, irrespetive of funtional forms. 35 A possible reation to Proposition 4 is to question whether ases in whih an ativist ampaign inreases soial welfare are likely to be relatively rare? As a way to gain partial insight into this question, we onsider a ase in whih it might seem unlikely that the ativist ould inrease soial welfare: when the ativist-spei parameters and A are large enough so that ondition (22) holds, and q r1 = 1. We refer to this as the ase of a strong ativist. A strong ativist will either redue the rm s reputation from r to r 1 or leave it unhanged, and the diret ost of a ampaign is as high as it an be. One might expet, then, that a strong ativist is potentially unfavorable to soial welfare. However, as the next proposition indiates, even a strong ativist an inrease soial welfare under ertain irumstanes. Proposition 5 Suppose the ativist is strong (i.e., and A are su iently large so that q r1 1), and ondition (27) holds so there are negative soial returns to reputation. If is positive but su iently small, the presene of the ativist inreases soial welfare. A possible senario under whih the onditions in Proposition 5 might prevail would be where the issue on whih the rm s reputation depends has already been well reported in the media, so the additional ost to the ativist of drawing publi attention to it is low, and the marginal net bene t of even a little more abatement ativity by the rm is very high, possibly beause very little is urrently being done to mitigate the soial harm. A ampaign launhed against a high-pro le rm in the wake of a well-publiized aident in whih the rm or its suppliers are impliated as having done too little to prevent the aident might t this set of irumstanes. By ontrast, if the ativist is strong but there are positive soial returns to reputation, then the impat of the ativist on soial welfare is ambiguous: the strong ativist would inrease soial welfare in the rst period but derease it in the seond period. For the strong ativist to derease rst-period welfare overall disounted welfare in the fae of positive soial returns to reputation, the rm would need to do signi antly more private regulation in the rst period than the stati abatement optimum. A possible example when this set of irumstanes might arise is when the ativist is extremely passionate and private regulation takes the form of greenwashing ostly ativities suh relabeling produts or PR ampaigns aimed at improving the rm s reputation for being environmentally responsible without signi antly abating an underlying harm. 32 In the proof of Proposition 4, this is established by showing that (; F ; ; ; r 1 )j =F (1+ F ) r 1 >. 33 In the proof of Proposition 4, this is established by showing that (; F ; ; ; r 1 )j =F (1+ F ) r 1 >. 34 A tehnial point is in order here. Proposition 4 relies Assumption 1. If Assumption 1 were strongly violated and did not hold for states r and r 1, we would have p r 1;1 = p r1 = 1; in whih ase the equilibrium would involve no ampaign, and the ativist trivially has no welfare e et. arise. On the other hand, if Assumption 1 was violated for state r 1 but not state r (so that p r1 < 1), then the result in the proposition would ontinue to hold. 35 However, a ompliating fator is that the rebound e et annot be ruled out when analyzing the model with more general funtion forms. If so, the ativist s impat on seond- period expeted welfare does not depend singularly on the sign of the soial return to reputation, but rather on the sign of the entire hange in seond-period welfare, h ur h ur F r2 h dr F r 1;2. In this ase, the result in Proposition 4 might be weaker in that it ould be onditional on whatever parameter onditions in the more general ase prelude the rebound e et from arising or at least make it su iently weak. 2

21 Under these onditions, the soial marginal bene t of the rm s e ort will be small (so p S is low), but it is oneivable that the rm will inrease the odds of a reputation boost from its e orts. Finally, onsider the distributional e ets of the ativist s presene. We an rewrite (24) as the sum of the hange in the expeted present value of the gross bene t of abatement, plus the hange u r1 u r1 in the present value of the target rm s pro ts, plus the hange s r1 s r1 in the present value of onsumer surplus, minus the ost of the ampaign: Wr =! p r1 p r1 + F E r (p 2 ) Er (p 2 ) + u r1 u r1 + s r1 s (qr1) 2 r1 : 2 From Proposition 1, the hange in the expeted present value of the gross bene t of abatement is positive. Further, it is straightforward to show that u r1 u r1 < and s r1 s r1 <. Thus, the ativist s presene helps onstituenies that bene t from inreased abatement at the expense of the target rm and possibly its onsumers. When Wr <, the ativist s presene is akin to an imperfet orretive tax whose exess burden exeeds its externality reduing bene t. Summing up, this setion has identi ed su ient onditions for the ativist s presene to have a positive effet on soial welfare soial marginal bene t of abatement su iently large, the marginal ost of abatement su iently at, and non-internalized reputational spillovers small. The other parameters F ; ; ; r 1 have an ambiguous e et on how the ativist impats soial welfare. This setion also showed that a neessary ondition for the ativist to have a bene ial impat on soial welfare is that! is bounded away from. Thus, the abatement ativity the ativist wants to stimulate must have a su iently large bene t-ost ratio for the ativist s ampaign to be welfare-improving. Finally, the ativist s presene redistributes surplus away from the target rm and possibly its onsumers, toward onstituenies that bene t from the rm s abatement ativity. Numerial analysis To further illustrate how the ativist s presene impats soial welfare, we illustrate numerial alulations over portions of parameter spae. Figures 3-6 illustrate ombinations of! and for whih the presene of the ativist inreases or dereases disounted soial welfare relative to the no-ativist equilibrium. Eah gure orresponds to a partiular ombination of r 1 and. 36 The gures suggest the following regularities: An ativist ampaign inreases disounted soial welfare only if the soial marginal bene t of abatement! is su iently large. Conditional on! being large enough, the ativist ampaign is more likely to inrease disounted soial welfare the more risk averse the rm. A passionate ativist shrinks the range of (!; ) for whih a ampaign is soially bene ial. Higher returns to reputation r 1 expand the range of (!; ) for whih a ampaign is soially bene ial for a highly passionate ativist but not for a less passionate ativist. The numerial analysis is onsistent with the results above, but it also adds additional insight. ativist is more likely to inrease soial welfare the more risk averse is the target rm, provided that the marginal ost of abatement is su iently large. This suggests that soiety should welome ampaigns over high-stakes issues against high-pro le ompanies that are a lot about proteting their reputations. A key 36 The other parameter values whose values remained xed in these alulations are: F = :95, A = :95, = 8, = 175, r 1 =, = :25, and r = 1. An 21

22 aveat, however, is that a highly passionate ativist may undertake suh an intense ampaign that it ends up being ounterprodutive even when the soial marginal bene t of abatement is reasonably high. In suh ases, the soial bene t of the additional private regulation engendered by the ativist is o set by the osts of reputation loss to the rm and onsumers, and possibly, too, by the diret osts of the ampaign. To onserve spae, we do not show the results of alulations that vary F. However, a representative plot in (!; F ) spae is presented in the On-line Appendix, and it shows that a higher disount fator for the rm modestly expands the range over whih Wr expand that range. >. And as in the gures above, inreases in! also Private regulation that is redistributive or has negative net bene ts In the analysis so far we have assumed that the soial marginal bene t! is stritly positive. However, it is possible to imagine irumstanes under whih!. This would our, for example, if private regulation results in redistributive pro t sharing between rms and workers (! = ), as Harrison and Sorse (21) suggest was the ase in Indonesia with the odes of ondut adopted by major footwear brands suh as Nike and Reebok in the wake of anti-sweatshop ampaigns in the 199s. It would also be the ase if private regulation gives rise to a deadweight loss (! < ), as ould be the ase if a ode of ondut for upstream suppliers establishing a living wage aused signi ant job loss relative to any inrease in welfare from the redistribution of surplus from rms to employed workers. Our assumption that! > rules out both of these possibilities, and thus stak the dek against nding that ativist ampaigns ould redue soial welfare. As we showed above, ativist ampaigns an redue soial welfare even when! >, indiating that there are additional fores for why an ativist ampaign might redue soial welfare beyond the diret possibility that private regulation itself may soially undesirable. Still, the ase of! is interesting when viewed through the lens of our model, so it is useful to disuss it here. To inorporate this ase, while at the same time giving a pragmati ativist an ationable motive (whih it would not have given our urrent spei ation), we assume that there is a wedge " > between the true soial marginal bene t of abatement! and the marginal bene t of abatement b! =! + " pereived by the ativist. Thus, if the wedge is large enough, we an have b! >, while!. The wedge ould represent the fat that the ativist has expliitly redistributive motives that would not be aptured in the soial welfare funtion or that it is unaware of the invisible osts of private regulation. The key role of the wedge is to provide a rationale for a pragmati ativist to mount a ampaign. (With! and no other hanges to our model, we would have q r1 =.) Spei ally, under this formulation, the ativist s reation funtion beomes where bq R r1(p) = min A bw r 1;2 bw r 1;2 bw r2 p ; 1 bw r2 = wr2 + " F r 1 (1 ) bw r 1;2 = wr 1;2 + " F r 1 (1 ): If!, wr2 and wr 1;2, but if " is large enough, bq r1(p) R >, for some p, and we will have an equilibrium in whih qr1 >. Under these irumstanes, all our results pertaining to the equilibrium hold. Moreover, the welfare omparison is exatly as it was before sine " does not enter the soial welfare 22

23 funtion. Beause! must be su iently large for the ativist s presene to enhane welfare, we onlude that ampaigns that result in greater amounts of private regulation that are essentially redistributive (! = ) or involves diret soial osts that outweigh soial bene ts (! < ) will neessarily derease soial welfare. Proposition 6 If the ativist pereives a marginal bene t of abatement b! =! +" that exeeds the true soial marginal bene t of abatement! by a su iently large amount so that qr1 >, then if!, the ativist s ampaign dereases soial welfare. This result provides a welfare benhmark that ould be used to think about Harrison and Sorse s (21) ndings of large wage inreases for workers of Indonesian suppliers that followed the adoption of odes of ondut by leading U.S. and European footwear brands. It seems plausible that the ations needed to implement the odes of ondut involved at least some inremental administrative osts to the rms involved ( > ). But if the wage inreases did not lead to a detetable loss of employment, then as Harrison and Sorse suggest, they probably represented redistributive pro t sharing that transferred produer surplus from Indonesian suppliers to Indonesian workers (! = ). Our model highlights that while the ampaign that led to the odes of ondut arguably resulted in a more just outome, they probably redued soial welfare. This raises the question of whether, in a world of imperfet governane where publi regulation and redistribution are di ult to a et, there are better mehanisms than ampaigns to ahieve redistributions of surplus that are desired on grounds of justie and fairness. However, an important quali ation is in order. Our model assumes a welfare funtion that gives equal weight to a dollar of surplus for the rm, onsumers, and bene iaries of abatement. This is the standard Kaldor-Hiks welfare formulation that is widely used in the regulatory eonomis and industrial organization literatures and in ost-bene t analysis. It essentially uses potential Pareto improvements as the welfare metri. If we were to onsider more general welfare formulations e.g., a utilitarian soial welfare funtion in whih the marginal utilities of inome di ered aross parties then private regulation that a ets a pure transfer between the rm and a group of stakeholders would not be welfare neutral, i.e.,! 6=. Still, it is straightforward to reinterpret Proportion 6 to take into aount soial welfare funtions that embody di erent notions of distributive justie, suh as utilitarianism. For example, a surplus-neutral transfer from those with lower marginal utilities of inome (e.g., a rm s owners) to those with higher marginal utilities of inome ould inrease welfare (e.g., employees of an upstream supplier in a developing ountry), would inrease the utilitarian soial welfare funtion. If the transfer had no deadweight loss, then it would be as if! >, and if! were large enough, our model would imply that an ativist ampaign (e.g., for living wages) that indued a greater amount of purely redistributive private regulation would inrease soial welfare. On the other hand, if the redistribution was assoiated with a deadweight loss that was signi ant relative to the di erential in marginal utilities in inome, then the redistribution ould still redue soial welfare even with a utilitarian welfare formulation, and it would be as if! <. In suh a ase, our model implies that an ativist ampaign induing additional private regulation of this kind would redue soial welfare. 4.2 Is passion soially valuable? We have modeled the ativist s passion as the weight it gives to the gross bene ts of reduing a soial harm. As is evident from the analysis of the rst period equilibrium, this weight is isomorphi to the steepness of a pereived marginal ost of ampaign urve, with higher values of orresponding to a pereption by the ativist that the marginal ost of inreasing ampaign intensity is lower for any given level of intensity. As far as we know, the passion of atual ativist groups has not been measured, but within given movements, ativist groups do di er in the extent to whih they rely on onfrontational tatis to advane their aims, 23

24 and these di erenes ould arguably be mapped into di erenes in passion and/or the pereived opportunity ost of onfrontational ativity. For example, U.S. ompanies tend to view The Nature Conservany and the Environmental Defense Fund as ooperative partners with whom they an easily work on sustainability projets, while they view RAN, Dogwood Alliane, and Earth First as more onfrontational, with the latter two groups espeially prone to rely on name and shame tatis (Davies 214). 37 This raises the question of whether ativist passion is a valuable soial asset. Figures 3-6 seem to suggest that, broadly speaking, the answer to this question is no. As noted, a omparison of Figures 3 and 4 reveals that all other things equal, the ativist s presene is more likely to inrease soial welfare when it is less passionate ( = 1) than more ( = 5). But beause these gures only show ranges over whih the e et on welfare due to the presene of the ativist hanges, they do not allow us to asertain whether, on the margin, inreasing the passion of the ativist inreases the welfare gain from the ativist. To explore this issue, reall that inreasing the ativist s passion has the e et of shifting the ativist s reation funtion rightward and positioning the rst period equilibrium toward the northeast along the rm s reation funtion. Imagine, then, starting at an equilibrium (p r1(1); qr1(1)) orresponding to = 1, and then perturbing upward, resulting in a hange in the equilibrium dqr1 > and dp r1 = dpr r1 (q) dq dqr1 = p r 1;1 p r1 dq r1 >. The soial marginal bene t of inreasing the ativist s passion is given by dw r = [! p r1(1)] dp r1 q r1(1)dq r1 + dh ur F r2 dh dr F r 1;2; where dh ur and dh dr are the hanges in the probabilities that the rm s reputation inreases and dereases, respetively. The former is positive, while the latter is ambiguous. The sign of dwr is ambiguous, but there are irumstanes under whih it is positive. For example, if the soial returns to reputation are negative, rst period private regulation is less than the stati abatement optimum (i.e.,! p r1 > ), and inreased ativist passion does not derease the probability of redution in reputation (dh dr ), then dwr >. Figure 7 illustrates that these onditions are not vauous. It shows the relationship between the hange in welfare Wr and the ativist s passion for four di erent levels of the soial marginal bene t of abatement and for a very risk-averse target ( = :1). 38 If the soial marginal bene t of abatement is su iently high (! = 15 or 2), there is a range of passion parameters over whih the soial marginal bene t of passion is positive. In these ases, soiety ahieves a bigger welfare gain from a ampaign led by an ativist that overweights the welfare gains from abatement ( > 1) than by one whose objetive funtion inludes the true soial marginal bene t of abatement. On the other hand, if the soial marginal bene t of abatement is low enough, the ativist s presene redues soial welfare, no matter what the ativist s passion is, and furthermore, the soial marginal bene t of passion is everywhere negative. Our analysis thus suggests that ativist passion an be a valuable soial asset. A ampaign that targets a rm that bene ts signi antly from proteting its reputation may generate a larger welfare gain with a more passionate ativist, provided that the ativist is not too passionate. (Even with! = 2, beyond a ertain point, inreases in ativist passion redue the welfare gain.) 5 Robustness to hanges in assumptions and model spei ation In this setion, we disuss the robustness of the insights of the model by onsidering a number of natural extensions of the simple modeling struture we employ. 37 Other groups, suh as the Sierra Club and the National Audobon Soiety, fall somewhere in between. 38 The other parameter values are those used in the gures above: F = :95, A = :95, = 8, = 175, r 1 =, = :25, and r = 1. 24

25 5.1 More general funtional forms and in nite horizon dynamis Our model ould be analyzed with more general funtional forms than the linear-quadrati spei ation used here. We have done this analysis, and we obtain the same key result we obtain here: the presene of the ativist inreases rst period private regulation. However, we annot rule out the possibility of the rebound e et as we an with the linear-quadrati formulation. Still, the welfare trade-o s identi ed in the linear-quadrati spei ation hold more generally. In Abito, Besanko, and Diermeier (212), we analyze an in nite-horizon game between an ativist and a rm aimed at understanding how di erent ativist tatis ritiism and onfrontation a et reputational and private regulation dynamis. As in this paper, the ativist s presene has both a stati e et (hanges in private regulation statewise) and a dynami e et (hanges in transition probabilities and thus the path of reputation and private regulation over time). In the in nite-horizon model, as in our three-period model, the ativist tries to damage the rm s reputation in order to motivate the rm to rebuild its reputation by inreasing its private regulation. Critiism the analogue of ampaign intensity in our three-period model inreases private regulation on the margin, just as we nd here. However, onfrontation whih an ause a atastrophi ollapse in reputation that the rm is powerless to ounterat and has no analogue in our three-period model may work in the opposite diretion. This is beause a ollapse in reputation an redue the rm s expeted returns to private regulation by so muh that it more than o sets the risk management motive highlighted by our three-period analysis. The possibility of reputational ollapse ould be added to our three-period model, and we onjeture that it would operate the same way it operates in the in nite-horizon model. If so, it would inrease the set of irumstanes under whih the ativist s presene redues rst-period welfare and derease the set of onditions under whih the ativist inreases expeted seond-period welfare, with the overall impat on disounted soial welfare being ambiguous. Summing up, then, while the in nite-horizon model allows for riher reputational dynamis, we believe that as long as the ativist s opportunities for a eting reputation are modeled in a onsistent way, the welfare trade-o s in an in nite horizon model will parallel those in a three-period model. 5.2 Exogenous probability of reputation loss Firms fae reputational rises for reasons other than ativist ampaigns. This possibility ould be aptured by adding an exogenous probability of redution in reputation to the model. Under this spei ation, the rm has a risk management motive even in the absene of the ativist, and thus for any given reputation level, the rm will engage in more private regulation than in the no-ativist equilibrium without the exogenous probability of a redution in reputation. When the ativist is added to the mix under this spei ation, the exogenous probability of a reputation drop tends to make the ativist s ampaign intensity less than it would otherwise be beause the ativist an free ride on the exogenous probability of a redution in reputation. This would marginally improve the welfare properties of the equilibrium with the ativist beause it would, in e et, redue the ost of the ativist s ampaign. 5.3 Reputation enhanement based on solving a spei problem or eliminating a spei pratie Our model assumes that abatement ativity is valuable on an ongoing basis and the rm s reputation improves over time (stohastially) to the extent that it engages in a stream of private regulation over time. In pratie, however, some ativist ampaigns are aimed at induing a rm to solve a spei problem or end a spei pratie, and the rm s reputational improvement is not a funtion of how muh abatement 25

26 ativity it engages in but whether it solves the problem. It is not di ult to adapt our model to this ase. Let p be the probability that the rm atually solves the problem, onditional on not having solved it in the previous period. In the absene of the ativist, if the rm is pereived as having solved the problem, its reputation improves; otherwise, its reputation remains the same. The no-ativist equilibrium for this formulation is similar to the no-ativist equilibrium in our model, but the seond period ontinuation values would di er to re et the possibility that the game ould end after the rst period. Suppose that there is an ativist that endeavors to present evidene that the rm has made the problem worse, and suppose there is a probability q that this evidene beomes salient to the publi. This introdues three possible states: the rm solves the problem and the ativist s evidene does not beome salient, whih ours with probability p(1 q), resulting in an inrease in the rm s reputation from r to r + 1; the rm does not solve the problem and the ativist s evidene beomes salient, whih ours with probability (1 p)q, resulting in a derease in the rm s reputation from r to r 1, and the rm solves the problem and the ativist s evidene beomes salient, whih ours with probability 1 p(1 q) p(1 q). Though other assumptions are possible, a plausible assumption is that in the fae of this reputational noise the rm s reputation remains unhanged. Exept for the hanges in the ontinuation values just noted, this model has the same struture as the model we have analyzed. Changing the way in whih abatement ativity a ets reputation does not have a onsequential impat on the insights we derive from our model. 5.4 Multiple ativists and/or multiple rms Finally, our model ould be extended to inlude either multiple ativists and/or multiple rms. With multiple ativists, any one ativist group ould free ride on the e orts of other ativists. This suggests that overall ampaign intensity would be less than it would be with a single ativist. In irumstanes in whih a single ativist s presene enhanes welfare, the welfare gain under multiple ativists (if there is one) would be less, and in irumstanes in whih a single ativist s presene redues welfare, the welfare loss would be less with multiple ativists. The fragmentation of ativists under irumstanes in whih ampaigns are more likely to be soially ostly low soial marginal bene t of abatement, highly passionate ativists, and a target for whih building or repairing reputation does not matter that muh would be a good thing from soiety s perspetive. With multiple rms the issues are more subtle. If the ativist has a multi-tasking ost funtion, meaning that its total ost depends on sum of the ampaign intensities aross rms, in any period the ativist will alloate its intensity to a single rm. The rm that a pragmati ativist would be target would be the one for whih a drop in reputation would eliit the greatest inrease in private regulation. That would depend on rm-spei harateristis (e.g., the marginal ost of abatement and the degree of reputational risk aversion), but it would also depend on rms returns to reputation, whih ould depend on how hanges in reputation hange the strategi interations among rms. 6 Conlusions This paper presents a three-period model in whih private regulation is motivated by the rm s desire to improve its reputation. Reputational onerns are present without an ativist, but the ativist indues an additional motive for the rm to engage in private regulation: risk management to protet reputation. By launhing a ampaign in the rst period, the ativist reates a risk that the rm s reputation might deline between the rst and seond periods. This has two distint e ets. The stati e et derived from the risk management motive is that whatever the rm s rst period reputation is, it engages in more private 26

27 regulation in the rst period than it would have in the absene of an ativist. The dynami e et is that the ativist hanges the rm s reputational trajetory, making it more likely that the rm will have a lower reputation in the seond period, whih in turn leads it to undertake more private regulation in period 2 than it would have if its reputation had been higher. Comparing state by state, the rm s private regulation in period 2 when there is an ativist is idential to what it would be without the ativist. (Indeed, the ativist does not launh a ampaign in period 2 in any state.) The ativist s impat on seond period outomes lies in its ability to alter the ourse of history by putting the rm on a reputational path in whih it is hungrier to invest in reputation building. Overall, the ativist indues an inrease in the expeted value of private regulation in period 2. An ativist ampaign potentially reates soial value by induing more abatement ativity. We show that a rm that faes neither publi regulation nor pressure from ativists will hoose less than the rst-best level of abatement in periods 2 and 3, and will do so in period 1 if reputational spillovers are su iently small. This opens the door to the possibility that the ativist s presene, by inreasing abatement ativity in period 1 and (in expetation) period 2, an enhane soial welfare relative to the no-ativist equilibrium. However, the ativist ampaign is not a ostless vehile for inreasing private regulation. Not only is there a diret ost to an ativist ampaign, but the stati and dynami e ets of the ativist s ampaign themselves have soial osts. The stati e et an redue soial welfare if, in the absene of the ativist, the rm would have hosen a level of abatement exeeding the stati abatement optimum in whih ase the ativist worsens the rst-period distortion and redues the net bene t from abatement in the rst period or if the rm s abatement was less than the stati abatement optimum, but the ativist-indued inrease in private regulation is so large that the net bene t from abatement in the rst period goes down. The dynami e et has the bene t of making it more likely that the rm ends up in states in period 2 where it is highly motivated to engage in private regulation. But reating the inentive to keep the rm hungry has the ost of reduing the rm s pro ts and onsumer surplus. The net impat of the dynami e et is summarized by the soial returns to reputation. We show that these stati and dynami welfare e ets are omplementary. Both rst period and seond period welfare gains are more likely to arise when the soial marginal bene t of abatement is high and when the marginal ost urve of abatement is at. First period and seond period welfare losses from the ativist ampaign are more likely to our when the soial marginal bene t of abatement is low and the marginal ost urve of abatement is steep. Indeed, we show that the ativist s presene an inrease disounted soial welfare only if the soial marginal bene t of abatement is bounded away from. With a slight tweak to our model allowing the ativist to be motivated by a pereived soial marginal bene t of abatement that exeeds the atual soial marginal bene t, a wedge that might represent distributive onerns or a lak of awareness of ertain osts of private regulation we an show an ativist will launh a ampaign even though the soial marginal bene t of abatement is non-positive, and when it does, the ampaign must neessarily redue disounted soial welfare. This insight resonates with reent ritiisms of ativist ampaigns by some sholars, suh as Bhagwati and Narlikar (213). Our model suggests that when a ampaign leads a rm to adopt a ode of ondut that raise wages either in a distributionally neutral way (i.e., the wage inreases are essentially a pro t-sharing transfer between the rm and the upstream workers) or in a soially ostly way (i.e., the wage inreases reate a Harberger triangle by reduing employment opportunities for other workers), the ampaign must neessarily redue disounted soial welfare. This redution in welfare an be seen as the prie that soiety pays for the redistribution that the ativist ampaign aims to a et. However, our model goes beyond this point to suggest that even when ativists are not blind to the invis- 27

28 ible osts of private regulation and are about the true soial marginal osts abatements, ativist ampaigns may still redue soial welfare. In addition to settings in whih the soial marginal ost of abatement is low and the ativist passion is high, ampaigns may also be ounterprodutive when the rm is not espeially risk averse in its reputation and when the gain in pro t from an inrease in reputation is low. In these latter two ases, the additional motivation to invest in private regulation provided by reputation protetion is not espeially strong. The ativist reates a ost (a redution in the rm s reputation) with little orresponding bene t (inreases in private regulation in the rst period and little inremental gain in the seond period from plaing the rm in a position in whih it undertakes more private regulation). This suggests the anonial ase in whih ampaigns would serve the publi interest: high-stakes externalities for whih the soial marginal bene t of abatement is big; the rm is likely to be doing little voluntary abatement, but is highly risk averse and thus ares intensely about proteting its urrent reputation. An example might be a ampaign aimed at induing a high-pro le onsumer goods rm or retailer to ut bak on the use of an input that has been disovered to pose a serious but heretofore unregulated environmental, safety, or health risk. E orts on the part of ativist groups suh as the Safer Chemials, Healthy Family Coalition in the late 2s and early 21s to pressure retailers suh as Walmart to not sell baby bottles made with Bisphenol A (BPA) a plastis additive whose health risks hit the U.S. publi onsiousness in 28 is a possible andidate to t this pro le. The model also suggests the anonial ase in whih ampaigns would not serve the publi interest: settings in whih the rm s ativities reate low-stakes harm; the rm faes highly passionate ativists and is not partiularly risk averse about its reputation. An example that ts this pro le might be ampaigns aimed at preventing animal testing. 7 Appendix Derivation of the no-ativist benhmark: Period 3 is the terminal period, so there is no gain from undertaking private regulation, establishing (6) and (9). And sine r > by assumption, (13) holds for t = 3. Conditions (7) and (1) follow from solving the Kuhn-Tuker onditions in period 2 with q =. The fat that p r2 2 (; 1), is an impliation of Assumption 1: > F (1 + F ) r > F r. Condition (12) follows diretly from DSRR. To establish (13) for t = 2, note from (1) u r2 = u r+1;2 u r2 = (1 + F ) r F 2 ( r+1) 2 (1 + F ) r + 2 F 2 ( r) 2 h = (1 + F ) r + 2 F ( r+1 ) 2 ( r ) 2i : (31) 2 Noting that ( r+1 ) 2 ( r ) 2 = [ r+1 r ] [ r+1 + r ], with some slight algebrai rearrangement, (31) an be written as u r2 = (1 + F 2 F [ r+1 + r ] 2 ) r + 2 F [ r+1 + r ] r+1 > ; 2 where the inequality follows beause, F [ r+1+ r] 2 < 1 from Assumption 1 and thus 1+ 2 F [r+1+r] F. Conditions (8) and (11) also follow from solving the Kuhn-Tuker onditions in period 1 with q =. The 2 > 28

29 fat that p r1 2 (; 1) follows beause from (31) F u r2 = F (1 + F ) r < F (1 + F ) r + 3 F 2 2 h ( r+1 ) 2 < 1; ( r ) 2i where the rst inequality follows beause ( r+1 ) 2 < ( r ) 2 by DSRR and the seond follows from Assumption 1. To prove that u r1 >, we note that p r1 is feasible, though not neessarily optimal in reputation state r + 1. Thus, u p 2 r1 r+1;1 r F u r+1;2 + F u r+1;2p r1: Also Thus, u r1 = r p 2 r1 + 2 F u r2 + F u r2p r1: u r1 = u r+1;1 u r1 r + F u r2 1 p r1 + F u r+1;2p r1: Having just established u r2 > and u r+1;2 >, and sine r >, it follows that u r1 >, establishing (13) for t = 1. Using (31) and (7), we = r 1 p F r2 + F 2 ( r+1 ) 2 > : To prove that the rm s seond period value funtion exhibits diminishing marginal returns to reputation, we use (31) to get u r 1;2 u r2 = (1 + F ) [ r 1 r ] + 2 F nh( r ) 2 ( r 1 ) 2i h ( r+1 ) 2 ( r ) 2io 2 = (1 + F ) [ r 1 r ] + 2 F [r r 1 ] [ r + r 1 ] 2 [ r+1 r ] [ r+1 + r ] 8 r +r >< 1 9 F = F 1 [ r 1 r ] >= > ; (32) >: + 2 F 2 [ r r+1 ] [ r+1 + r ] >; where the last inequality follows beause Assumption 1 implies F r < 1 and F r 1 r +r 1 F 2 1 positive. 39 The result that p r+1;1 < p r1 follows immediately from (8). < 1, making Let s now establish that we have diminishing private regulation over time. Beause p r3 =, it su es to show that p r1 p r2. From (7) and (8) that inequality turns on the omparison between u r2 and r. From (31) u r2 = r + F r + F [ r+1 r ] F [ r+1 + r ] : 2 To prove the result, it su es to prove that F r + F [ r+1 r ] F [ r+1+ r] 2 >. This an be rewritten as 1 F [ r+1 + r ] F [ r+1 + r ] 2 F r + 2 F r+1 : Now, by Assumption 1, F r < 1 and F r+1 < 1, so F [ r+1+ r] 2 < 1. Thus 1 F [ r+1 + r ] F [ r+1 + r ] 2 F r + 2 F r+1 > F r+1 > : 39 Assumption 1 is not needed to prove this result. If we allow for orner solutions on p r2, we an still establish thatu r 1;2 u r2 >. 29

30 Hene u r2 > r : Finally, to establish oasting, we note that given u r2 > r, (8) and (7) imply p r2 < p r1. Moreover, as established above, p r+1;2 < p r2. Thus p r+1;2 < p r1. Proofs of preliminary results for the equilibrium with an ativist: The result that p r3 = qr3 follows immediately from the fat that the third period is the terminal period of the game. Thus, u r3 = r ; and (from 3) vr3 = : Now, using (3) again, we have v r2 = max qrt2[;1] w(p r2) q 2 r2 2, whih implies q r2 = and vr2 = w(p r2). The result that p r2 = p r2 follows immediately from qr2 =. The haraterization of equilibrium values follows immediately from bakward indution. The result that vr2 < arises beause vr2 = vr+1;2 vr2 = w(p r+1;2) w(p r2) and beause p r+1;2 < p r2 as noted earlier in our disussion of the no-ativist benhmark. Neessary and su ient ondition for a orner equilibrium in the rst period: If A " w r 1;2 1 F u r 1;2! + wr2 F u r 1;2 then qr1( R F u r 1;2 ) = 1. From (18), F u r 1;2 is the highest possible equilibrium value for p, and thus, p r1 F u r 1;2. Moreover, from (16) and (17), wr 1;2 > wr2, so the reation funtion qr1(p) R is non-inreasing in p. Thus, qr1 = qr1(p R r1) qr1( R F u r 1;2 ) = 1, but sine qr1(p) R annot be greater than 1, we must have qr1 = 1: Now, if qr1 = 1, then from (18), p r1 = F u r 1;2. Also, sine 1 = qr1 = qr1(p R r1), it must be the ase that n h min A i o wr 1;2 wr 1;2 wr2 F u r 1;2 ; 1 = 1; whih implies that A " w r 1;2 1 Proof of Proposition 1: Part (a) Condition (18) in onjuntion with (8) implies F u r 1;2! + wr2 F u r 1;2 p r1 = p r1 + q r1 p r 1;1 p r1 > p r1 ; beause from (19) qr1 > and as pointed out in the haraterization of the no-ativist equilibrium, p r 1;1 > p r1: Part (b): Note that the expression for Er (p 2 ) Er (p 2 ) is given by # # 1; 1: Er (p 2 ) Er (p 2 ) = h drp r 1;2 + (1 h ur h dr)p r;2 + h urp r+1;2 = h dr p r 1;2 p r2 h ur h ur (p r2 p r+1;2): (1 h ur )p r2 + h urp r+1;2 where h ur h u (p r1; qr1) = p r1(1 qr1) and h ur h u (p r1; ) = p r1 are the probabilities the rm s reputation goes up when there is an ativist and when there is not, and h dr h d(p r1; qr1) = (1 p r1)qr1 is the probability that the rm s reputation delines when there is an ativist. (Without an ativist, this probability is.) Sine p r2 > p r+1;2 and p r 1;2 > p r2, a su ient ondition for Er (p 2 ) Er (p 2 ) > is h ur h ur < : We will now prove that h ur h ur <. First, note that for a orner equilibrium, qr1 = 1, so h ur = :Thus, h ur h ur <, and Er (p 2 ) Er (p 2 ) > : Now, for an interior equilibrium, a su ient ondition for the ativist s presene to inrease expeted private regulation is the absene of the rebound e et, i.e., h ur h ur <. We rst establish that the rebound e et u will not arise if and only if r2 > 1 q u r1 2 q. To do so, de ne a funtion h r 1;2 r1 ur (q) p R r1(q)(1 q) p r1. This is the di erene in the transition probability from r to r + 1 when an ativist hooses an arbitrary q and the 3

31 rm reats optimally and when there is no ativist. Using (18) and (19), we note that: h ur () = : h ur (q r1) = h ur h ur: dh ur (q) dq dh ur () dq = dpr r1(q) (1 q) p R dq r1(q) = p r 1;1 p r1 (1 2q) p r1 : = p r 1;1 2p r1 : d 2 h ur (q) dq 2 = 2 p r 1;1 p r1 < : There are two possibilities for this funtion. If dhur() dq <, the funtion takes its maximum value for q <, and thus h ur (q) < for q 2 [; 1], and in partiular, h ur (qr1) = h ur h ur <. From above, this ase arises if p r 1;1 2p r1 < or equivalently, u r 1;2 2u r2 <., then h ur (q) takes on its maximum value at some q > and that maximum value will exeed. For this ase, the rebound e et does not arise only if the larger root q r + of h ur (q) is less than qr1 (beause h ur (q) is dereasing for q > q r +, and thus we would have = h ur (q r + ) > h ur (qr1) = h ur h ur). Using the expression for p R r1(q) in (18) we have If dhur() dq h ur (q) = F q u r 1;2 2u r2 The larger root is u r 1;2 2u r2. u r 1;2 u r2 Pulling these threads together, we have three ases: u r 1;2 2u r2 < ) h ur h ur < : u r 1;2 u r2 q : u r 1;2 2u r2 < u r 1;2 u r2 q r1 ) h ur h ur < : u r 1;2 2u r2 u r 1;2 u r2 q r1 ) h ur h ur : Thus, the neessary and su ient ondition for the rebound e et not to arise is u r 1;2 2u r2 < u r 1;2 u r2 q r1 ; or u r2 u r 1;2 > 1 q r1 2 qr1 : Note that the right-hand side of this expression is stritly dereasing in q r1 and thus 1 2 > 1 We now show that given Assumption 1, is. To establish this, note that (14) implies, u r2 u r 1;2 u r2 > 1 u 2 r 1;2 2 r 1 (1 + F ) > r 1 (1 + F ) > 1 2, whih then implies u r2 u r 1;2 if and only if (1 2 ) 2 F r 1 2 (1 2 ) 2 F r 1 ; 2 q r1 2 q r1 > 1 q r1 2 q r1. whatever q r1 whih is equivalent to > (22 1)(1 2 ) 2 F r 1 : 2(2 1)(1 + F ) But by Assumption 1, > (1+ F ) F r 1 > F r 1, whih in turn is greater than (22 1)(1 2 ) 2 F r 1 sine (22 1) (2 1) < 1, 1 2 < 1, and F 1+ F < 1: Thus u r2 u r 1;2 > q r1 2 q r1 (2 1)(1+ F ) whih, as just established, implies 31

32 h ur h ur <, so the rebound e et does not arise. It follows that Er (p 2 ) Er (p 2 ) > : Proof of omparative statis results: To prove that higher F results in greater rst-period private regulation, reall that in the disussion of the no-ativist > for i = r; r 1, whih from (18) r1 (q) >. From (16) and F > for i = r; r 1;whih r1 (p) >. Thus, a higher F F simultaneously shifts the rm s reation funtion rightward and the ativist s reation funtion rightward, implying that equilibrium private regulation in the rst period must inrease. To prove that a atter marginal abatement ost urve results in greater rst-period private regulation, with straightforward algebra it an be shown F u i2 < for i = r; r 1, and r1 <. 4 From (16) and (17), we see that < for i = r; r 1. Thus, a lower simultaneously shifts the rm s reation funtion rightward and the ativist s reation funtion rightward, implying that equilibrium private regulation in the rst period must inrease. Finally, to show that in the rst-period equilibrium, the weaker the rm s reputation, the greater its private regulation, i.e., p r 1;1 > p r1, we note that p R r1(q) = F (1 q)u r2 + qu r 1;2 and p R r 1;1 (q) = F (1 q)u r 1;2 + qu r 2;2. From DDRR, u r 2;2 > u r 1;2 > u r 1;2, so p R r 1;1(q) > p R r1(q) for any q 2 [; 1]. Thus, a derease in r shifts the rm s reation funtion rightward. We also note that qr1(p) R = n A o n min (1 p)w r 1;2 + pwr2 ; 1 and qr R 1;1(p) = min A (16) and (17), wr 1;2 > wr2, and it is straightforward to show that wr 2;2 > wr 1;2. Thus, (1 p)w r 2;2 + pw r 1;2 ; 1 o. From qr R 1;1(p) qr1(p) R for any p 2 [; 1]. Thus, a derease in r shifts the ativist s reation funtion rightward or not at all. As a result, p r 1;1 > p r1. Proof of Proposition 4: Beause! < (; F ; ; ; r 1 ) and! < (; F ; ; r 1 ) imply Wr <, it follows that! (; F ; ; ; r 1 ) or! (; F ; ; r 1 ) is a neessary ondition for Wr. Beause both (; F ; ; ; r 1 ) and (; F ; ; r 1 ) are inreasing in whose smallest possible value onsistent with Assumption 1 is F (1+ F ) r 1, it follows that the lowest possible value of! for the ativist to have a positive impat on soial welfare must be greater than minf(; F ; ; ; r 1 )j =F (1+ F ) r 1 ; (; F ; ; r 1 )j = F (1+ F ) r 1 ; g. We establish that this minimum is stritly positive by showing that (; F ; ; ; r 1 )j =F (1+ F ) r 1 and (; F ; ; r 1 )j =F (1+ F ) r 1 >. Substituting F (1 + F ) r 1 for in (28) and (29), and rearranging terms gives us, (; F ; ; ; r 1 )j =F (1+ F ) r 1 = (1 + ) r 1 (1 ) (; F ; ; r 1 )j =F (1+ F ) r 1 = F r F 1 (1 + F ) F (1 2 ) > : (1 + F ) 2 (1 ) > ; 2 sine 2 (; 1); F 2 (; 1); and 1 +. Proof of Proposition 5: When the ativist is strong, we have qr1 1, and from the disussion above, h ur h ur <. Moreover, h dr = (1 p r1) > beause from (18), p r1 2 (p r1; p r 1;1) < 1. Beause (27) holds, we have r2 < and r 1;2 <. The seond line of (24) is thus positive. Now, note that (27) implies! > (1 + F )(1 + ) F (1 ) ( ) (1 + ) F r 1 4 Using F u = F 2 r 3 F 3 ( r+1) 2 2 F 2 1 F r r < beause 1 F r < due to Assumption 1. 32

33 whih an be rewritten as! " (1 + ) 1 > + 1 (1 ) F ( ) 1 2 (1 + ) F r 1 # Sine F r 1 < 1, the term in brakets is greater than one and so! > (1+) (1 ) > 1. The impliation that! > implies that!p 2 p2 is stritly inreasing for all p 2 [; 1], and beause p r1 > p r1, it follows that!p r1 2 (p r1) 2 >!p 2, r1 2 p r1 whih establishes that the rst line of W r is positive. If is positive but su iently lose to zero, then Wr > : 33

34 8 Referenes 1. Abito, J.M., D. Besanko, and D. Diermeier (212) Corporate Reputational Dynamis, Private Regulation, and Ativist Pressure, Working Paper, Northwestern University. 2. Baron, D. P. (21) Private Politis, Corporate Soial Responsibility, and Integrated Strategy, Journal of Eonomis and Management Strategy, Vol. 1, No. 1, pp Baron, D. P. (23) Private Politis, Journal of Eonomis and Management Strategy,Vol. 12, No. 1, Baron, D. P. (29) A Positive Theory of Moral Management, Soial Pressure, and Corporate Soial Performane, Journal of Eonomis and Management Strategy;Vol. 18, No. 1, pp Baron, D. P. and E. Yurday (24) Strategi Ativism: The Rainforest Ation Network, Case P-44, Graduate Shool of Business, Stanford University, Stanford, CA. 6. Baron, D. P. and D. Diermeier (27) Strategi Ativism and Nonmarket Strategy, Journal of Eonomis and Management Strategy, Vol. 16, No. 3, pp Bhagwati, J., and A. Narlikar (213) Signing Up to Safety Laws Does Workers More Harm than Good, Finanial Times, July Bermiss, Y. S., E. J. Zaja, and B.G. King (213) Under Constrution: How Commensuration and Management Fashion A et Corporate Reputation Rankings, Organizational Siene (forthoming). 9. Board, S. and M. Meyer-Ter-Vehn (213) Reputation for Quality, Eonometria, Vol. 81, No. 6, pp Bond, J., and Kirshenbaum, R. (1998) Under the Radar: Talking to Today s Cynial Consumer, (New York: John Wiley & Sons). 11. Clardy, A. (212) Organizational Reputation: Issues in Coneptualization and Measurement, Corporate Reputation Review, Vol. 15, No. 4, pp Davies, J. (214) The 214 GreenBiz NGO Report: How Companies Rate Ativists, GreenBiz Group, Diermeier, D. (211) Reputation Rules: Strategies for Building Your Company s Most Valuable Asset, (New York: MGraw-Hill). 14. Eesley, C. and M. Lenox (25) Firm Responses to Seondary Stakeholder Ation. Best Paper Proeedings of the 25 Aademy of Management Conferene, (Honolulu, HI). 15. Eesley, C. and M. Lenox (26) Seondary Stakeholder Ations and the Seletion of Firm Targets, Best Paper Proeedings of the 26 Aademy of Management Conferene, (Atlanta, GA). 16. Feddersen, T. and T. Gilligan (21) Saints and Markets: Ativists and the Supply of Credene Goods, Journal of Eonomis and Management Strategy, Vol. 1, No. 1, pp Fombrun, C. J. (1996) Reputation: Realizing Value from the Corporate Image, (Boston: Harvard Business Shool Press). 34

35 18. Harrison, A. and Sorse, J. (21) Multinationals and Anti-sweatshop Ativism. Amerian Eonomi Review, Vol. 1, No. 1, pp Hau er, V. (21) A Publi Role for the Private Setor: Industry Self Regulation in a Global Eonomy (Washington, D.C.: Carnegie Endowment for International Peae). 2. Hendel, I., Lah, S. and Y. Spiegel (213). Soial Media and Buyers Power: The Cottage Cheese Boyott, Working Paper, Northwestern University. 21. Kaeb, C. (28) Emerging Issues of Human Rights Responsibility in the Extrative and Manufaturing Industries: Patterns and Liability Risks, Northwestern Journal of International Human Rights, Vol. 6, pp King, B. and M-H. MDonnell (212) Good Firms, Good Targets: The Relationship between Corporate Soial Responsibility, Reputation, and Ativist Targeting, Working Paper, Northwestern University. 23. King, B. and N. Peare (21) The Contentiousness of Markets: Politis, Soial Movements and Institutional Change in Markets. Annual Review of Soiology 36, pp King, B. and S. Soule (27) Soial Movements as Extra-Institutional Entrepreneurs: The E ets of Protests on Stok Prie Returns. Administrative Siene Quarterly Vol. 52, pp Lenox, M. and C. Eesley (29) Private Environmental Ativism and the Seletion and Response of Firm Targets, Journal of Eonomis & Management Strategy, Vol. 18, No. 1, pp Lyon, T.P. and J. W. Maxwell (22) Voluntary Approahes to Environmental Regulation: A Survey, in In M. Franzini & A. Niita (eds.), Eonomi Institutions and Environmental Poliy (Aldershot, Hampshire, UK: Ashgate Publishing Ltd.). 27. Maxwell, J. W., T. P. Lyon, and S. C. Hakett (2), Self-Regulation and Soial Welfare: The Politial Eonomy of Corporate Environmentalism, Journal of Law and Eonomis, Vol. 43, No. 2, pp Minor, D. and J. Morgan. (212) CSR as Reputation Insurane: Primum Non Noere, California Management Review, Vol. 53, No. 3, pp Vandenbergh, M.P. (27) The New Wal-Mart E et: The Role of Private Contrating in Global Governane, UCLA Law Review, Vol. 54, pp Vandenbergh, M.P. (213) Private Environmental Governane, Cornell Law Review;Vol. 99, pp Vogel, D. (21) The Private Regulation of Global Corporate Condut: Ahievements and Limitations, Business and Soiety, Vol. 49, No. 1 (Marh 21). 35

36 Figure 1: Firm and ativist reation funtions and rst-period equilibrium Figure 2: Possible senarios for rst-period, seond-period, and total welfare 36

37 Figure 3: Di erene in welfare in (!; ) spae for = 1; r 1 = 1: Figure 4: Di erene in welfare in (!; ) spae for = 5; r 1 = 1: 37