Good fences make good neighbors:
A longitudinal analysis of an industry self-regulatory institution
ABSTRACT
We extend theories of self-regulation of physical commons to analyze self-regulation of
intangible commons that arise in modern industry. We posit that when the action of one firm can
cause spillover harm to others in the industry, firms share a type of commons. We theorize that
the need to protect this commons can motivate self-regulation. Using data from the US chemical
industry, we find that spillover harm from industrial accidents increased after a major industrial
crisis and decreased following the formation of a new institution. Additionally, our findings
suggest that the institution lessened spillovers from participants to the broader industry.
Keywords: institutions; industry self-regulation; commons problems; cooperative strategy
Interest in self-regulatory institutions, whereby firms in an industry create and voluntarily
abide by a set of governing rules, has gone through a renaissance of late (for overviews, see
Gunningham & Rees, 1997; Prakash & Potoski, 2006; Wotruba, 1997). Scholars have
investigated self-regulatory institutions in industries as diverse as chemicals (King & Lenox,
2000; Lenox, 2006), hospitality and recreation (Rivera and de Leon, 2004), nuclear power (Rees,
1994), and maritime shipping (Furger, 1997). Drawing from the work of Elinor Ostrom (1990)
and Douglas North (1981), many of these scholars argue that self-regulatory institutions arise to
constrain individual actions that might harm the industry as a whole. Ostrom’s (1990) work on
community self-regulation has been particularly influential in framing recent research. She
demonstrates that those who share in common pool resources like fisheries or forests can unite to
create an institution that helps them avert the “tragedy of the commons” (Hardin, 1968).
However, the theoretical and empirical foundation of this growing stream of the
management literature remains uncertain and contradictory. First, the common pool resource
dilemmas that Ostrom considers are not apparent in many industries where self-regulation has
arisen. The literature has yet to establish whether the same logic that Ostrom applied to the
governance of shared physical resources can be extended to modern industries. Second,
empirical studies of self-regulation have often fallen victim to what Granovetter (1985)
dismissed as “bad functionalism” in that they infer the function that an institution was created to
fulfill not by observing the conditions that preceded its creation but rather by observing
conditions after its implementation. Third, research on some frequently studied institutions
seems to provide contradictory evidence with respect to their function. For several important
self-regulatory institutions, scholars have failed to find any evidence that the institution limits the
1
harmful practices of member firms, yet studies of these same institutions show that they provide
a benefit to firms in the industry (King & Lenox, 2000; Rivera & de Leon, 2004, Lenox, 2006).
In this article, we resolve these shortcomings and so strengthen the theoretical and
empirical foundation of the literature on self-regulatory institutions. First, we draw attention to
the existence of a novel type of “commons problem” that exists in many industries. We argue
that the action of one firm may affect other firms in its industry because stakeholders lack the
ability to impose precisely targeted sanctions, and this causes all firms to share a pooled risk.
Second, we avoid bad functionalism by developing a mechanism for measuring this shared risk
over a time period that spans both the emergence of our hypothesized commons problem and the
formation of the self-regulatory institution. Finally, by more precisely identifying how the
institution provides benefits to the industry, we provide new insight on how inconsistencies in
the existing literature may be resolved.
THEORY AND HYPOTHESES
Institutions are the “humanly devised constraints that structure political, economic and
social interaction” (North, 1990: 97). North (1990) separates institutions into those that operate
through formal constraints (e.g. rules, laws, and constitutions) and those that operate through
informal constraints (e.g. norms of behavior, conventions, and self-imposed codes of conduct).
Ingram and Clay (2000) refine North’s typology by suggesting that institutions should be
classified as (1) public or private, and (2) centralized or decentralized. Public institutions are
usually compulsory and are often run by the state. Private institutions – those run by
organizations or individuals – are voluntary in nature, because actors can opt in or out. In the
centralized form of these institutions, a central authority sets rules, incentives, and sanctions for
non-compliance. For example, many private institutions (e.g. for-profit firms) have a principal
2
that is ultimately in charge of internal procedures. Decentralized forms of these institutions lack
a powerful central authority and thus rely on the action of numerous independent actors to
encourage compliance with institutional rules. In many industrial settings, anti-trust laws forbid
centralized industry bodies from controlling and sanctioning member firm behaviors and so
industry self-regulation tends to take the form of private and decentralized institutions.
Until recently, many scholars dismissed the viability of self-regulatory institutions. The
influential analyses of Olson (1965) and Hardin (1968) suggested that since participation is
voluntary and free from enforcement by a central authority, each actor has incentive to defect
from agreements and so such institutions should generally fall victim to “the tragedy of the
commons” (Hardin, 1968). However, widespread skepticism about self-regulatory institutions
began to change as a result of a series of investigations in the 1980s and 1990s (c.f. Wade, 1988;
Berkes, 1989; Acheson, 1988; Ostrom, 1990).
Elinor Ostrom’s work has had the greatest impact in changing perceptions of the potential
for self-regulation. Through a series of comparative case studies, she demonstrated that
individuals could, in fact, organize institutions to cope with overuse of commonly-held resources
such as fresh water aquifers, fisheries, and forests (Ostrom, 1990). When actors could negotiate,
observe, and enforce compliance with common rules, she argued, self-regulatory institutions
could protect commonly held resources, and the benefits provided by protection of these
common resources could spur participation. In her own assessment, her work helped “shatter the
convictions of many policy analysts that the only way to solve common pool resource problems
is for external authorities to impose full private property rights or centralized regulation”
(Ostrom, 1990). Later in the decade, based on field research conducted by her workshop, she
concluded that “ individuals in all walks of life and all parts of the world voluntarily organize
3
themselves so as to gain the benefits of trade, to provide mutual protection against risk, and to
create and enforce rules that protect natural resources” (Ostrom, 2000: 138).
The notion that self-regulatory institutions can be viable solutions to commons problems
has begun to gain traction in the management literature (King & Lenox, 2000; Rivera and de
Leon, 2004; Jiang and Bansal, 2003). These studies have focused on the determinants of
participation in a self-regulatory institution, though, and have not explored the conditions
antecedent to their formation. As Ingram and Clay (2000: 539) conclude in their review of the
literature on institutions, beyond Ostrom, Ingram and Inman’s (1996) study of hoteliers around
Niagara Falls is perhaps the only other study that conducts the longitudinal analysis needed to
understand the relationship between the existence of a commons problem and self-regulation.
They show that faced with the threat of potential damage to a commonly valued resource,
Niagara Falls, hoteliers were able to form a self-regulatory institution that limited development
and so protected the scenery of the Falls, thereby increasing the survival of nearby hotels.
Ingram and Inman (1996) follow Ostrom (1990) in arguing that the need to protect a
shared physical resource such as water or land can act as the catalyst for effective self-regulation.
Yet, few modern industries are challenged by dwindling stocks of a physical resource openly
shared with rivals, but many modern industries engage in self-regulation. For example, the
Institute of Nuclear Power Operations did not arise in the face of overuse of shared stocks of
uranium in the nuclear power industry, nor did the Beer Institute Code arise in the face of
threatened shortages of communal supplies of barley or hops in the brewing industry. What
might explain the frequent presence of self-regulation in settings such as these?
As we discuss next, firms in an industry share a sort of non-physical commons that can
be damaged by the acts of each firm in the industry. This creates an industry wide commons
4
problem that can develop into a serious threat to the success and survival of member firms
following a major industry crisis. We hypothesize that industry self-regulation may arise to
resolve this type of a commons problem in modern industries.
Industry Commons: A Shared Fate through Shared Sanctions
Interdependence among firms is implicit in the concept of an industry. Firms are rivals
because they have similar and closely substitutable characteristics. As a result, the actions of one
firm are bound to reflect to some degree on all firms within its industry and so shape how
observers assess, reward, and sanction firms within it. King, Lenox, and Barnett (2002) argue
that this interdependence can be viewed as a sort of intangible commons that they term a
“reputation commons.” A firm’s reputation is based on observers’ judgments of the actions of
that firm over time (Fombrun, 1996). If observers can judge the impacts of a firm independent
of the actions of its rivals, no commons exits. However, when one firm’s actions influence the
reputation of another or the reputation of the industry as a whole, a commons arises.
This conflating of observers’ assessments of the characteristics of a firm with those of its
rivals can be favorable if, for example, one firm’s success helps to legitimize an emerging
industry and so eases all such firms’ access to resources (cf., Hannan & Carroll, 1992), but it can
also be problematic. Just as one firm’s success can spill over to other firms, so can the harm
caused by misdeeds. For example, recent news of contaminated spinach harmed the sale of all
salad products – not just the ones where the contamination was found (Galvin, 2007).
Management researchers have established that “industry effects” account for a significant
portion of firm performance (Schmalensee, 1985; Rumelt, 1991; McGahan & Porter, 1997), but
the management literature has been less attentive to the potential for firm misdeeds to harm the
entire industry. Yet, research in financial economics suggests that a negative event attributable to
5
a single firm can indeed have adverse financial consequences for the industry as a whole. Jarrell
and Peltzman (1985) find that a drug recall by one pharmaceutical firm caused a portfolio of 50
rival firm stocks to drop by one percent. They find an even stronger industry-wide effect in the
automobile industry. When Ford or Chrysler initiated a recall, General Motors actually
experienced a larger loss than the recalling firm. Hill and Schneeweis (1983) report a loss in
market value of a portfolio of all electrical utility stocks following the accident at the Three Mile
Island nuclear plant. Mitchell (1989) concludes that the firms in the over-the-counter
pharmaceutical industry lost $4.06 billion following the deadly Tylenol tampering incident.
This shared fate of industry members stems from the broad or arbitrary application of
stakeholder sanctions (Dawson and Segerson, 2005; King et al., 2002). The broad application of
sanctions is more likely when stakeholders are unable to discriminate high and low performer
firms in an industry. For example, unable to ascertain which firms had contributed to toxic waste
dumps, the U.S. government imposed a fee on all chemical producers to fund the Superfund
cleanup effort. In other cases, stakeholders may select individual firms arbitrarily for
sanctioning, thereby putting all firms in the industry at risk. For example, activist discontent
with working conditions in the coffee, athletic shoe, and apparel industries led to high-profile
protests and boycotts against individual firms in these industries – Starbucks, Nike, and Kathy
Lee Gifford – that used suppliers with no worse, and in some cases, superior working conditions
than those employed by their rivals (Malkin, 1996; Hornblower, 2000).
In summary, rival firms share many common characteristics. When one firm succeeds, it
can ease rivals’ access to resources, but when one firm errs, it can harm its rivals. Observers,
either unable to untangle a firm’s performance from that of the industry, or as a result of
increasing their expectations that the focal firm could err in a similar way in the future, may
6
collectively, or even randomly, sanction members of the industry, thereby creating an industry
commons. To test the existence of this industry commons, we thus hypothesize that firms in an
industry will be harmed by the misdeeds of other industry members.
Hypothesis 1: An error at one firm will harm other firms in its industry.
Industry Commons as a Problem
The mere existence of an industry commons is not sufficient to explain self-regulatory
institutions. Rivalry, inertia, and the inherent cost of forming and maintaining such an institution
inhibit firms from coming together (Barnett, 2006). Members of a commons may be motivated
to surmount these barriers, though, when confronted with a crisis. Studies of physical commons
have shown that self-regulation often occurs after overuse threatens the continued viability of a
shared resource. For example, self-regulation in the Maine Lobster industry arose after a
collapse of the fishery caused the closure of some canneries (Acheson & Knight, 2000). Crisis
can help actors overcome cognitive barriers and recognize the existence and importance of a
commons (Weber, Kopelman & Messick, 2004), and it can also change expectations of the value
of taking action (Vasi & Macy, 2003).
An industry commons cannot be physically depleted the same way a fishery can be, but it
can become damaged in a way that significantly harms firms and even threatens the industry’s
ongoing “license to operate.” For example, the crisis at Three Mile Island in 1979 sparked such
deep and enduring public concern about nuclear safety that regulators have not approved any
new nuclear power plants in the US since. Major events like Three Mile Island can be a catalyst
for shifts in stakeholder perceptions of an industry (Rees, 1997, Meyer, 1982; Hoffman &
Ocasio, 2001). Hoffman (1999) argues that stakeholder perceptions of an industry are based on
metaphors. These perceptions can change “suddenly and unpredictably” (pg. 366) as events
7
influence taken-for-granted assumptions and create new metaphors about the industry. These
new metaphors influence the interpretation of future events, and they can cause even minor
events to draw attention and raise the threat of greater sanctions for an industry. Greenwood,
Suddaby, and Hinings (2002) describe a similar process in which “jolts” (Meyer, Brooks, &
Goes, 1990) deinstitutionalize established industry practices and set in motion a process of
“theorization” that determines how observers will view future industry practices.
Consider the airline industry in the aftermath of September 11th, 2001. The events of
September 11th shifted how observers viewed the airline industry, causing many to assess
airplanes not merely as a means of transportation, but also as a means of terrorism. Under this
shifted mindset, observers focused much more attention on airline activity and interpreted new
events in light of their potential to be part of a terrorist plot. As a result, minor breaches of
security that had previously gone unnoticed or unquestioned now drew media attention and
engendered public calls for more stringent security protocols that raised costs and sometimes
lowered demand for the entire industry.
Reports of executives in the petroleum and nuclear power industries validate this
perspective. Following the Exxon Valdez oil spill in 1989, an Amoco executive noted that now
his firm would “have to live with the sins of our brothers. We were doing fine until Exxon
spilled all that oil. Then we were painted with the same brush as them” (quoted in Hoffman,
1997: 189). According to the founding chairman of the Institute of Nuclear Power Operations, in
the aftermath of the Three Mile Island crisis, “it hit us that an event at a nuclear plant anywhere
in our country . . . could and would affect each nuclear plant . . . Each licensee is a hostage of
each other licensee” (Rees, 1994: 2). Therefore, we hypothesize that a major crisis in an industry
8
causes an increase in the severity of shared sanctions that firms suffer following the error of any
firm in that industry.
Hypothesis 2: A major industry crisis causes increased intra-industry spillovers from errors.
Industry Self-regulation as the Solution
From William Forster Lloyd (1833) through Garrett Hardin’s famous 1968 essay, the
idea of “The Tragedy of the Commons” has powerfully influenced scholarly and popular belief
that the resolution of commons problems requires government intervention. The standard refrain
is that the choice is simple: Leviathan or Oblivion (Ophuls, 1973). In recent years, some
scholars have challenged this long-held notion, suggesting that actors who share a common
resource can form self-regulatory institutions to constrain actions that might lead to abuse of the
commons. When the competitive positions of firms in an industry are put at risk due to the
commons-depleting actions of any individual firm within the industry, the members of the
industry have a clear interest in creating an institution that protects the commons. Ostrom (1990)
shows that actors often form institutions to avert the tragedies associated with common
properties, and Ingram and Inman (1996) show that hoteliers around Niagara Falls, driven by the
fear of lost tourism revenue, banded together to form institutions which regulated development
around the Falls.
We extend the logic of self-regulation of physical commons such as water or land to selfregulation of intangible industry commons. Firms share an interest in maintaining their common
reputation so as to reduce their shared risk of sanction. The threat of more stringent regulation
following the Three Mile Island crisis, for example, prompted nuclear power industry executives
to create the Institute of Nuclear Power Operations, a “private regulatory bureaucracy” charged
to “develop standards, conduct inspections, and investigate accidents” (Rees, 1997: 478)
9
Though the literature has not previously addressed the role of industry self-regulatory
institutions in resolving the threat of shared sanctions stemming from a common reputation,
scholars have previously proposed a “functionalist” perspective on institutions. A functionalist
perspective evidences that firms can come together to create institutions to alleviate shared
problems. Such a perspective was originally proposed by scholars such as North (1981) and
Williamson (1985) who suggested that actors might create governing institutions to ameliorate
conditions that impede effective social exchange. They argued that the potential for gain could
encourage economic actors to create new controlling institutions, despite the individual burdens.
However, Granovetter’s (1985) criticism of the empirical basis for such claims dampened
research in this area. He dismissed such studies as “bad functionalism” because they began with
a functionalist conclusion and then inferred “the evolutionary problem that must have existed for
the institution as we see it to have developed” (Schotter, 1981; in Granovetter, 1985: 489).
We hypothesize a functional role for industry self-regulatory institutions, but rather than
inferring the problem that must have existed for such an institution to have arisen, in the prior
hypothesis we outline the common problem that precipitates institution creation. By first
identifying the specific common problem that we hypothesize the institution will ameliorate, we
avoid the trap of bad functionalism.
Hypothesis 3: An industry self-regulatory institution created following a major crisis will
reduce intra-industry spillovers from errors.
Exploring the Mechanisms of Industry Self-regulation
An explanation of self-regulation should include a specification of the means through
which it achieves its function. Previous research has argued that self-regulation helps firms a)
signal superior attributes or b) cooperate with rivals. Self-regulatory institutions that operate as
10
signals help participating firms differentiate themselves from non-participants relative to some
criteria which stakeholders value but cannot observe directly. For example, participation might
be cost-effective only for those with superior quality, and thus participation allows firms to
credibly reveal this hidden quality. In contrast, self-regulatory institutions that facilitate
cooperation help firms coordinate effort toward some shared goal. For example, firms might
pool their resources to lobby government for changes in trade policies.
Scholars predict that self-regulatory institutions that include third-party certification of
compliance with institutional rules may act as signals while those that lack such certification are
more likely to facilitate cooperation (Potoski and Prakash, 2005). Empirical research on several
certification programs provide support for the signaling hypothesis, but research on selfregulatory programs that lack such certification seem to provide evidence that is both
contradictory and in conflict with proposed theoretical models (King & Lenox, 2000; Rivera and
de Leon, 2004; Lenox, 2006).
A common theoretical model for self-regulatory institutions that operate without a
certification mechanism is that they allow firms to forestall stakeholder action by helping firms
coordinate just enough improvement to dissuade stakeholders from taking action to force further
improvements (Dawson & Segerson, 2005). Since stakeholders (including regulators) bear some
cost from acting, this additional improvement can be smaller than the stakeholder would impose,
but it must be higher than would have occurred without the support of the institution. This
model has been very influential in setting trajectories for further research, but it has not yet been
supported by empirical investigations. In conflict with the model’s predictions, several authors
find that participants in such self-regulatory programs do not improve any faster than nonparticipants (King & Lenox, 2000; Howard, Nash, and Ehrenfeld, 2000; Rivera and de Leon,
11
2004). In some cases, empirical studies of different outcomes from self-regulation have revealed
seemingly inconsistent and contradictory findings. King and Lenox (2000) argue that
participants in the chemical industry’s Responsible Care program reduced their toxic emissions
more slowly than non-participants. Yet, Lenox (2006) reports that that average stock values
increased in the chemical industry after the program was formed, and he infers that stakeholders
were rewarding the industry for improvement relative to some desired criteria.
Why would a program that does not seem to provide benefit to stakeholders earn their
favor? One possibility is that self-regulation may allow firms to directly improve the reputation
of the industry or reduce the threat of sanction. For example, by unifying the regulatory
strategies of firms it might reduce the threat that any event would trigger costly sanctions. In this
view, firms are regulating and coordinating their non-market strategy (Baron, 1995), not their
improvement efforts. Another possibility is that firms reduce the propensity for stakeholder
action by cooperating to directly assuage stakeholder concern. For example, the chemical
industry emphasized such an approach when they initiated the Responsible Care program. The
first code promulgated under the program, Community Awareness and Emergency Response,
was not designed to reduce the risk of accidents. Instead, this code required managers of
facilities of participating firms “to identify and respond to community concerns, inform the
community of risks associated with company operations, and have its own emergency plan
integrated and tested with the community’s emergency response plan” (CCPA, 2007).
Giving stakeholders access to additional information may reduce the risk of harm to the
communities surrounding plants by speeding emergency response to accidents. It might also
reduce stakeholder fears of such events, regardless of any actual reduction in the impact of these
12
events. Research shows that the perceived risk of events is elevated when these risks come from
systems about which actors have little information. Slovik and Weber (2002) report that:
The informativeness or signal potential of a mishap, and thus its potential social
impact, appears to be systematically related to the perceived characteristics of the
hazard. An accident that takes many lives may produce relatively little social
disturbance (beyond that caused to the victims’ families and friends) if it occurs
as part of a familiar and well-understood system (e.g., a train wreck). However,
a small incident in an unfamiliar system (or one perceived as poorly understood),
such as a nuclear waste repository or a recombinant DNA laboratory, may have
immense social consequences if it is perceived as a harbinger of future and
possibly catastrophic mishaps.
In summary, scholars expect the design of a self-regulatory program to influence the
pattern of benefits it provides its members and the industry as a whole. If it includes a means of
certifying and communicating compliance to a set of rules, scholars expect a self-regulatory
institution to act like a market signal and provide a private benefit to members only. If it
operates without such a certifying mechanism, then scholars expect it to help its members
achieve some commonly valued goal like forestalling government regulation.
We choose to focus on this latter category because industry-sponsored self-regulation
often lacks a certification mechanism (Howard et al., 2000). Moreover, as discussed above, little
is known about the benefits provided by these programs, and more nuanced understanding of the
pattern of benefits will help reveal the underlying functional mechanisms of these institutions.
Self-regulation that forestalls government sanction by coordinating the non-market strategies of
member firms should reduce the harmful spillover from negative events no matter where they
occur. Self-regulation that coordinates programs to inform stakeholders and assuage their
concerns should reduce the spillover harm when a negative event occurs at a participating firm.
Hypothesis 4a: An industry self-regulatory institution created following a major crisis will
reduce intra-industry spillovers from the errors of any firm in the industry.
13
Hypothesis 4b: An industry self-regulatory institution created following a major crisis will
reduce intra-industry spillovers from the errors of participating firms.
RESEARCH SETTING AND METHODOLOGY
To test our hypotheses, we need an industry that experiences frequent errors of varying
significance, has suffered a major crisis, and has created a self-regulatory institution in the
aftermath of this major crisis. The US chemical industry meets all of these requirements.
Members of the industry suffer multiple accidents each year, and these accidents vary in
significance. Most are small and involve the unplanned release of potentially toxic chemicals.
More serious accidents injure or kill employs or even local citizens. Industry experts and
industry members report that one accident precipitated a major crisis for the industry. On
December 3rd, 1984, methyl isocyanate leaked from a Union Carbide facility in Bhopal, India
and killed between three and ten thousand people. Many thousands more were injured
(Shrivastava, 1987). It remains the most deadly industrial accident on record.
Anecdotal reports from managers in the chemical industry around the time of these
events support the perspective on industry commons we have hypothesized. Numerous
respondents report that the accident at Bhopal created a crisis for the entire chemical industry by
changing how observers viewed the risks of chemical manufacturing. Ronald Lang, then
executive director of the Synthetic Organic Chemical Manufacturers Association, noted, “Bhopal
focuses concern on something that had not been adequately addressed before – the possibility of
catastrophe” (Gibson, 1985a). Another industry leader described the post-Bhopal environment
as “chemophobia” (Gunningham, 1995: 72).
Contemporaneous reports provide evidence that industry participants now had a greater
sense of being part of a commons. Then chairman of Union Carbide, Warren Anderson,
14
remarked, “This is not a Carbide issue. This is an industry issue” (Gibson, 1985a: 21). Others
noted that with such a close focus on the industry, further accidents at any chemical firm would
have significant consequences for all chemical manufacturers. “Every chemical incident
becomes a national story now. A minor spill becomes front page stuff, and that tends to
exaggerate the event and reinforce the public’s concern” (Gibson, 1985b: 90).
Respondents also report that the crisis drove industry members to create a prominent selfregulatory institution: The American Chemistry Council 1 (ACC)’s Responsible Care program.
“More than anything else,” recalls [then] Union Carbide CEO Robert Kennedy,
“it was Bhopal that finally put us on the path that would lead to Responsible
Care.” “Bhopal was the wake-up call,” says [then] Dow Chemical vice president
Dave Buzelli. “It brought home to everybody that we could have the best
performance in the world but if another company had an accident, all of us would
be hurt, so we started to work together” (Rees, 1997: 485).
The first element of the program, Community Awareness and Emergency Response
(CAER), appeared designed to reduce concern about accidents. Responsible Care encouraged
firms to “open the doors and let the fresh air flow through” (Coombes, 1991) so that a skeptical
public would be convinced that the dangers that Bhopal had brought to light were being
rigorously dealt with and that they were not in imminent danger. Responsible Care required
extensive outreach efforts with local officials in communities where plants were located. As part
of CAER, firms conducted thousands of plant tours. The ACC also spent millions on advertising
campaigns to humanize the industry (Heller, 1991).
The program did not include a mechanism for third party certification of compliance with
the rules, and thus critics argued that it was unlikely to help stakeholders accurately determine
those firms with higher performance (Ember, 1995). Others argued, however, that close
connections in the industry could allow members to police adherence to the rules (Rees, 1997).
1
At the time of the events reported in this study, the American Chemistry Council was known as the Chemical
Manufacturers Association (CMA).
15
As discussed earlier, empirical studies suggested that firms did not reduce their pollution any
faster than non-participants, but that the industry still benefited from the existence of the
program (King & Lenox, 2000; Lenox, 2006).
Sample
Our sample includes all firms in the Center for Research on Security Prices (CRSP)
database that reported any operations in the chemical industry (SIC 2800 to 2899) in the
Compustat business segment database between 1980 and 2000 or reported to the Environmental
Protection Agency that they operated a production facility in these sectors. We chose this time
period to allow a nearly five-year pretest window before the Bhopal accident (1980-1984), a
five-year interval between the accident and the creation of Responsible Care (1985-1989), and an
11-year post-test window after the creation of Responsible Care (1990-2000). We chose the
sample to include all firms with any chemical operations, rather than only those with a primary
denomination in the chemical industry, so as to include diversified firms with significant though
non-dominant chemical operations. Our final sample includes 735 unique firms.
Data Sources
We obtained data about our sample firms from the CRSP database, the Compustat
business sector database, and the Environmental Protection Agency’s Toxic Release Inventory
(TRI). Reporting to the TRI began in 1987 and covers facilities with 10 or more employees that
produce, store, release, or transfer more than a threshold amount of any of more than 600 listed
chemicals. We obtained data about errors by performing key word searches of the major
international and U.S. regional newspapers and wire databases within the Lexis-Nexis service.
16
In some cases, we supplemented information from news articles with information from Hoovers
online database and Dialogweb.
Dependent Variable
We hypothesize that an “error” at one firm can “harm” other firms in this industry. To
form our dependent variable we must operationalize both of these terms. Managers in the
chemical industry consider industrial accidents to be serious errors (Greening & Johnson, 1997).
Industry experts further claim that the Bhopal crisis changed the degree to which industrial
accidents harmed the industry (Gibson, 1985b). Thus, we use industrial accidents at chemical
operations as our measure of “error.”
To find industrial accidents, we performed keyword searches using terms such as fire, gas
leak, explosion, chemical spill, chemical release, and chemical discharge. These are terms that
top managers in the chemical industry associate with serious accidents (Greening & Johnson,
1997). Our search uncovered 359 possible accidents. As with any keyword search, however, we
netted numerous inappropriate events such as chlorine burns in swimming pools or ammonia
spills on restaurant floors. We also found numerous accidents related to transportation (e.g. a
truck overturned and exploded at the corner of I-90 and Route 1) and accidents related to
petroleum transport and refining. We included accidents at refining operations, a segment
closely aligned with chemicals (Hoffman, 1997), but we excluded leaks and spills from crude oil
pumping or transport (e.g. the Exxon Valdez accident). Incomplete reporting about key aspects
of accidents further narrowed the sample. In total, we were able to qualify and determine the
date, magnitude, location, and responsible firm for 123 of the raw events.
Problems with the size of events or contemporaneous firm actions caused removal of an
additional 40 accidents. Twenty-five accidents resulted in neither an injury nor death. Our
17
preliminary analysis showed that such accidents were too minor to affect the stock price of even
the firms directly responsible for them, and so we excluded these events from our study of the
spillover effects. At the opposite end of the scale, we removed the Bhopal event due to its
extreme nature. Finally, events should not be confounded with endogenous actions that might
bias coefficient estimates (McWilliams & Siegel, 1997). We excluded fifteen events because
other significant activities were mentioned in the newspaper accounts of the event (e.g.
leadership changes). This left us with 83 accidents. 2
To measure the degree to which these accidents “harm” other firms in the industry, we
evaluate the stock price movements of other firms with chemical operations following each
accident. Market theory suggests that stock prices reflect the best assessment of future cash
flows for firms. If investors think that stakeholder sanction might reduce demand or government
regulation might increase a firm’s costs, then the stock price should fall. The more a firm’s stock
price depends on that of another firm, the more they share a type of industry commons.
Appendix A describes how we capture the harm of each accident with our dependent variable,
cumulative abnormal return (CAR), which is a measure of how much a stock’s value deviated
from its expected value over a particular period of time. To validate that our search uncovered a
set of accidents that might influence stock value, we explored the effect of these accidents on
CAR for the firms directly responsible for the events. As Appendix B shows, five days after
these 83 events the stock price of the firm responsible for the accident fell an average of 1.2%.
Accidents where an employee was killed caused about a 2.7% further reduction in the firm’s
stock price. Thus, the accidents in our sample have a measurable effect on stock prices.
2
Not surprisingly, inclusion of these events reduces the coefficient size for the effect of accidents on at-fault firms,
but it does not change the sign and significance of the reported results.
18
Independent Variables
We hypothesize that one firm’s error can spill over to other firms within the same
“industry.” Scholars commonly use the standard industrial classification (SIC) code as the
definition of an industry. As discussed above, since we are looking at accidents in the chemical
industry, we limit our sample to firms that have at least some operations in the 2800 SIC range
that encompasses chemicals. We also create a more refined measure of operations in the same
industry by creating a binary variable (Same SIC) that captures whether a firm owns a facility
that is in the same SIC code as the one that had an accident 3.
Our second hypothesis predicts that a major crisis will increase this spillover effect. As
previously discussed, industry insiders assert that Bhopal caused a major crisis that affected all
firms with chemical operations. To capture the changes caused by Bhopal, we create a dummy
variable (Pre Bhopal) that captures the time period from the beginning of our sample to
December 3, 1984. It takes on a value of one during this time and a value of zero otherwise.
Our third hypothesis predicts that industry self-regulation decreases this spillover effect.
Aspects of the Responsible Care program began shortly after the Bhopal crisis, but no elements
of the program were promulgated until late 1989. Thus, to capture the post Responsible Care
period, we create a dummy variable, Post RC, which takes on a value of 1 after 1989; else zero.
Finally, in order to explore the mechanisms of industry self-regulation, we must
distinguish Responsible Care participants from non-participants. To capture this effect, we
measure membership in the parent association, the American Chemistry Council. This variable
3 To test the robustness of our findings to various specifications of these variables, we investigated whether a
percentage measure would provide similar results or results with greater explanatory power. We find that our
analysis is robust (sign and significance of variables of interest) to a continuous specification and that a continuous
specification provides no significant increase in explanatory power. It may be that risk is not linear with the scale of
the production enterprise or that our assessments (measured by the fraction of total employees at that facility) do not
precisely measure relative scale across industry sectors. Thus, we use the binary measure.
19
(ACC Member) takes on a value of one if the ACC listed the company as a member in a given
year, and has a value of zero otherwise. From 1990 onward, participation in Responsible Care
was a condition of membership in ACC. We further separate out the unique attributes of the
Responsible Care program by creating a dummy variable (RC Member) that indicates ACC
membership during a year when the Responsible Care program was in existence. We also seek
to understand whether Responsible Care may have had a different effect depending on the
identity of the firm responsible for an accident. Thus, we create variables (Perp ACC and Perp
RC) that capture whether or not the firm that had the accident is a member of the American
Chemistry Council or Responsible Care.
Control Variables
Variation in the magnitude of an event should cause variation in the market’s response,
so we include two measures of each accident’s severity. Emp Killed is a binary variable that
takes a value of one if an employee was killed in the accident. Only about 25% of our accidents
include a fatality, so the binary parameterization is appropriate and truncates little information.
In contrast, we measure the number of employees injured (Emp Injured) using a continuous
measure (log(# injured +1)). We use the log parametric form to reduce the effect of outliers and
to account for possible diminishing effects. Alternative parametric forms (binary and linear)
reduce model fit but do not change the sign and significance of reported results.
The size of both the perpetrator and the recipient firm may influence the market response
to accidents. Larger perpetrators often have better public relations and so may be able to diffuse
public reaction and response. Assets of Perp measures the log of total assets for the firm where
the accident occurred. It is likely that larger firms will be influenced less by accidents at other
firms, either because they are better able to manage their public relations or because they tend to
20
be more diversified. Moreover, the size of a firm influences the variability of its stock
performance (Fama & French, 1992). We measure size of the firm (Assets) as the log of total
assets reported in Compustat for the year the accident occurred.
---INSERT TABLE 1 ABOUT HERE---
We use several approaches to reduce the potential for endogenous managerial choice to
bias our coefficient estimates. We include fixed effects of different types to control for
unobserved (but constant) differences in our sample (e.g., accident industry, spillover industry,
and the year of the accident). To control for unobserved firm-level differences, in some models
we include fixed effects for the firm itself. These fixed effects account for constant attributes of
any firm that might influence spillover effects.
Endogenous choice processes might also bias our sample, particularly if they are based
on changing firm characteristics not captured by our fixed effects. To help account for these, we
also conduct a Heckman two-stage treatment model. The first stage consists of the model
predicting participation in Responsible Care. We base the treatment selection function on a
model of membership proposed by King and Lenox (2000). They find that participation in
Responsible Care was influenced by a firm’s relative emissions, the relative emissions of the
industries in which it operates, the degree to which it is focused in chemicals, its size, and its
reputation. Following their study, we use TRI data to estimate the median emissions for each
industry (4 digit SIC) and create a weighted measure of this value (based on percentages of sales
in this SIC) for each firm 4. The log of this value becomes our measure of the degree to which a
firm operates in sectors with many toxic chemicals (Industry Emissions). We measure the
degree to which a firm operates in chemicals (Chemical Focus) by calculating the percentage of
4
For years 1980-1986, we estimated industry emissions based on the 1987 TRI data.
21
sales from chemical sectors (as estimated from Compustat data). We already have a measure of
the size of the firm (Assets). Due to the limitation of the TRI, we cannot estimate relative
performance prior to 1987. We also cannot measure contemporaneous measures of a firm’s
reputation for all of the firms in our sample. Descriptive statistics for these variables can be
found in Appendix C.
Analytical Methodology
Event study methodologies are commonly used to understand how stockholders interpret
a single event (Brown & Warner, 1980, 1985; Blacconiere & Patten, 1994; Hamilton, 1995;
Patten & Nance, 1998). McWilliams and Siegel (1997) have criticized the use of this method
when firms might be able to alter or time the focal event so that it occurs concurrent with some
other announcement. Accidents, however, are by their very nature unplanned and not amenable
to strategic timing. Thus, such manipulation is not a concern in this study.
Our event study analysis allows us to connect an accident with abnormal stock price
movements. To understand the causes of variance within these movements, we use a linear
regression:
CAR ij = a + BX ij + eij + υ p + u l + ν i + δ t
where CARij is the cumulative abnormal return for firmi 5 days following eventj and Xij is a
vector of independent variables for firmi at the time of eventj. Clearly we cannot measure every
possible factor that might influence the effect of an accident or the spillover from that accident to
other firms. We use a series of fixed effects to try to reduce potential problems from unobserved
heterogeneity. First, we attempt to account for unobservables in the industries in which the
accident occurred. We include a fixed effect ( υ p ) for all p industries in which we have more
than a single accident (16 out of 22 industries). Second, we account for potential differences
22
among industries that were affected by the accident (but not necessarily the one in which they
occurred) by including fixed effects for every two-digit SIC code ( ul ). Alternatively, when we
are considering issues of variable spillover among firms and are interested in the effect of
variables that are not collinear with firm identity, we include both firm (ν i ) and year ( δ t ) fixed
effects.
Clearly, the decision to join the American Chemistry Council and participate in
Responsible Care is endogenous to our analysis. That is, managers make decisions about
participation conditional on the characteristics of their organization. To the extent that we
capture the important firm characteristics through the inclusion of a direct measure or by the use
of fixed effects, our coefficient estimate should be unbiased. These methods will fail, however,
to capture the effect of endogenous choice based on changing and unmeasured factors. To
account for such factors and test the robustness of our analysis, we perform a two-stage
Heckman treatment model. This method uses a first stage model to estimate the potential effect
of the unobserved choice process on residuals, and then a second stage regression that includes a
correction for the estimated distortions of the error terms cause by the endogenous choice
process 5.
RESULTS
---INSERT TABLE 2 ABOUT HERE---
5
Often, treatment models in panels cannot be specified because unobserved determinants of the “stickiness” of
choice processes cannot be identified. In our case, we are conducting robustness analysis and thus are willing to
specify a model that provides the strongest robustness test. As a result, we first specify a model that assumes firms
have great flexibility about participation. That is, we assume that the choice of participation in any year is
independent from previous years. To ensure the robustness of our findings to the opposite extreme assumption, we
then estimated a model that assumes that the choice to participate is made only once based on average firm
characteristics over the panel. Both results were consistent (in sign and significance) with those presented.
23
To explore the effect of spillovers from accidents, we evaluate the abnormal stock market
movements of members of the industry (not including the perpetrator) after an accident. To get a
sense of the average scale of such spillovers, in Table 2 we first specify a simple model without
fixed effects or interaction terms. In support of Hypothesis 1, Model 1 shows that firms that
operated a facility in the same industry in which the accident occurred indeed experienced a
negative spillover. We also find that accidents in which a person was killed or many were
injured had a greater spillover. Note that the effects are smaller than those to the focal firm (see
Appendix B) but are still significant. Following an accident that injured an average number of
employees (~ 5), a chemical firm with operations in the same industry as that of the accident
could expect to lose 0.2% of its stock price. After an accident that caused the death of an
employee, the firm could expect to lose an additional 0.8%.
To explore whether the accident at Bhopal or the formation of Responsible Care
influenced spillovers from accidents, we specify a new model (Model 2) with two dummy
variables that capture the time period before the Bhopal accident and after the formation of
Responsible Care (Pre Bhopal and Post RC). Our results suggest that spillovers from accidents
indeed increased after Bhopal and diminished after the formation of Responsible Care. In the
intervening time period, an average accident could be expected to reduce the stock price of other
firms with operations in that industry by about 1.1%. We find that before Bhopal, this loss was
only about 0.4% and after the creation of Responsible Care it was negligible.
The increased spillovers after the Bhopal accident provide corroborating evidence for the
observations of industry members. Speaking after the Bhopal accident Dan Bishop, then
Monsanto’s director for environmental communication remarked, “Every chemical incident
becomes a national story now. A minor spill becomes front page stuff, and that tends to
24
exaggerate the event and reinforce the public’s concern” (Gibson, 1985b: 90). Industry members
also recognized that further accidents created greater spillover harm: “[T]he damage to Union
Carbide’s credibility has had a spillover effect for the rest of the industry” (Gibson, 1985b: 90).
Industry managers also noted that one stakeholder in particular – insurance companies – formally
implemented this increased perception of risk in a way that affected all firms in the industry,
regardless of their individual characteristics: “Now the Bhopal tragedy has reinforced the new
conservatism of insurance underwriters and, as one broker puts it, given them ‘an excuse to say
no.’…when they see any operations associated with chemicals – even chemical operations
posing no hazard to the public – [underwriters] are ready to paint them with the same brush as
Bhopal’s” (Katzenberg, 1985: 30).
The fact that Models 1 and 2 corroborate perceptions in the industry helps demonstrate
the validity of the findings. Nevertheless, we subjected our analysis to extensive robustness
testing. Neither model includes controls for unobserved industry differences for the industry of
the accident or the main industry of the firms, nor do they account for differences in the
characteristics of the perpetrator or the subject firm. In Model 3, we include additional variables
and fixed effects to account for such differences. We again find consistent results. We also find
that accidents from larger firms have a reduced spillover effect.
Models 1 to 3, in combination with the corroboration of industry insider reports, provide
strong support for Hypotheses 1 to 3. Accidents do cause spillover harm, and consistent with the
logic behind Hypothesis 1, this harm is greater for larger accidents and for firms that were more
similar to the one where the accident occurred. Consistent with Hypothesis 2, this harm appears
to have increased following a crisis in the industry. In support of Hypothesis 3, this spillover
harm decreased after the formation of a self-regulatory institution.
25
In Model 4, we begin our exploration of the mechanism of Responsible Care. To explore
whether Responsible Care provided a general benefit to the industry (Hypothesis 4a) or only
acted when one of its member firms was responsible for an accident (Hypothesis 4b), we include
additional interaction terms to capture spillovers from an ACC member in both time periods. In
support of Hypothesis 4b, we find a positive and significant coefficient for the variable Perp RC,
indicating reduced harm to firms in our sample when an accident occurred at an ACC firm after
the creation of the Responsible Care program. The inclusion of this variable also reduces the
significance of the coefficient for our variable denoting the Responsible Care time period. Thus,
we do not find evidence in support of Hypothesis 4a that spillover harms were reduced for
accidents at any firm, regardless of its membership in Responsible Care.
Our findings that Responsible Care functioned to reduce spillover harm when one of its
members experienced an accident is corroborated by the contemporaneous reports of industry
experts. After touring the site of an Exxon Chemical plant in his city, one city manager noted:
I don’t harbor the fears that I had. I have learned about what they do. I hadn’t
realized the safety precautions, the amount of testing of final products, the
monitoring of air and water that goes on. But I don’t think this will eliminate all
skepticism. It’s ludicrous to think that industry is going to be safe all the time. But
the fact that they have been open and honest is extremely important to me as city
manager (Heller, 1991).
Others reported that when accidents inevitably did occur, Responsible Care coordinated a quick
response. “‘Mutual assistance’ is on all Responsible Care practitioners’ lips, with large firms
helping out small firms an important dynamic” (Heller & Hunter, 1994). Richard Doyle, vice
president of Responsible Care, described such efforts as operating out of a “war room.”
The emergency response effort in the war room also led to an upgrade in
Chemtrec, recognized as the chemical emergency response center in the U.S. It
was established in 1972 to provide timely information and connect emergency
responders with industry experts on the chemicals they were dealing with. After
26
Bhopal, CMA set about upgrading Chemtrec’s operations and improving its
mutual assistance activities (Begley, 1994).
Robustness Testing
To ensure the robustness of our analysis, we specified several alternative models. First,
we used two methods to account for endogenous choice processes based on unobserved fixed
firm differences. As shown in Model 5, a model with firm fixed effects further confirms our
previous findings. We again find that spillover harm increased following Bhopal and that
Responsible Care reduced this harm by reducing the spillover from participating firms. Since
fixed effect (FE) analysis represents a more stringent test, Model 5 provides strong confirmation
of our results.
In Model 6, we perform further robustness testing by correcting for the influence of
endogenous managerial choices. We report the second stage of a Heckman treatment estimation.
As discussed earlier, this model is specified to be sensitive to any changing firm attributes that
might explain membership in the ACC and Responsible Care. The probit for the first stage of
the 1990 analysis is shown in Appendix C. It is consistent with King and Lenox (2000) and
suggests that ACC members tend to be in sectors with more toxic emissions, more focused in
chemicals, and have more assets. We control for the effect of this endogenous choice by
including the Mills Ratio estimated by the probit model for each year as a variable in the second
stage regression. This model provides confirmatory evidence of the robustness of our findings.
In Model 7, we cease our analysis of the effect of time periods so as to allow the use of
year fixed effects to account for any underlying macroeconomic changes that might effect our
results. Once we include these year effects, we must remove the time period dummy variables.
As shown, an analysis with year fixed effects again suggests that Responsible Care provided a
27
benefit to the industry by reducing the spillover effect of an accident at a RC firm. Thus, we
again confirm our support for Hypothesis 4b.
We also conducted additional robustness testing to determine if our time period dummy
variables (Pre Bhopal and Post RC) might be capturing some other temporal effect. First, we
tested whether differences in the frequency of reporting might influence responses to accidents.
To rule out this possibility, we specified models that included measures of the accident previous
to the one under consideration. We included the days since this accident, whether an employee
was killed, and whether an employee was injured. Models with additional variables confirm the
sign and significance of the reported results. We also explored whether our measure Post Bhopal
might simply be capturing a wearing off of stakeholder concern as the accident became a more
distant memory. We tested alternative models with linear and log time since Bhopal (in days).
Coefficient estimates on both variables were not statistically significant. The log form of the
time estimate is highly correlated with the Pre Bhopal variable (ρ = -0.81) and thus it reduces the
significance of this measure in some models. Neither variable provided significant additional
explanatory power.
Throughout our analysis, we explain little of the observed variance in stock prices –
between 1 and 3%. This is not at all surprising. Our method essentially removes fixed
differences in stock prices, leaving only noise and the effect of new information about a firm.
We evaluate the effect of only one type of news, and clearly additional numerous other factors
play an important role in determining stock prices. However, so long as this other news is not
correlated with our events and uncaptured by our variables, our estimates should be unbiased.
As discussed earlier, we carefully screened out events that were contemporaneous with other
28
corporate news. Moreover, the robustness of our analysis to multiple specifications and controls
suggests that we have significant and stable estimates.
In summary, our analysis suggests that Bhopal indeed increased the interdependence of
firms with chemical operations such that an accident at one would have a negative effect on
another. Our analysis also suggests that Responsible Care reduced this spillover effect, but it did
so not by insulating its members from the consequences of accidents at other facilities, but by
reducing the consequences of accidents at Responsible Care facilities.
DISCUSSION AND CONCLUSION
In this article, we explain industry self-regulation where physical commons are absent.
We posit that firms in modern industries share in a non-physical type of commons – what some
have termed a “reputation commons” (King et al., 2002) – that stems from the difficulty that
stakeholders face in distinguishing the relative performance of individual firms and from the
blunt application of arbitrary or industry-wide sanctions. We hypothesize that the risks
associated with this commons can become particularly acute following a major industry crisis.
We further hypothesize that industries create self-regulatory institutions as a means of
ameliorating this heightened threat of shared sanctions.
Employing a longitudinal analysis of our hypothesis, we find that firms in the US
chemical industry did face such an industry commons and that the shared sanctions stemming
from this commons became more severe after the industry suffered a major crisis caused by an
accident in Bhopal, India. Furthermore, we find that this increased risk of shared sanctions
preceded the creation of the industry’s self-regulatory program, Responsible Care, and that
Responsible Care was able to ameliorate industry-wide harm from the errors of individual
chemical firms. Thus, in finding that an aggravated commons problem led to the formation of a
29
governing institution and that it operated to reduce this problem, our study supports the extension
of Ostrom’s (1990) perspective on community self-regulation of common pool resources into
modern industry as a way of explaining industry self-regulation.
We also explore the mechanism through which Responsible Care functioned. We find
that it reduced the industry wide harm from the errors of Responsible Care participants, but we
find no evidence that it buffered Responsible Care participants from the acts of others.
Responsible Care did not solve the industry commons problem for member firms by fencing off
participants from the acts of outsiders. Rather, it fenced off everyone from the actions of
participating firms. Thus, by joining, each firm provided a public benefit to the industry, and the
need to coordinate this benefit may have provided a key motivation for the institution.
In Mending Wall, poet Robert Frost wrote of a type of self-regulation that functions much
like the one we identify. Each winter, storms and hunters knocked down the stone wall that ran
around his property, and every spring Frost would “let [his] neighbor know beyond the hill” that
they need to “meet to walk the line, and set the wall between us once again.” But, as he worked
on the wall one year, he wondered why they were remaking it and asked his neighbor:
He only says, ‘Good fences make good neighbors.’
Spring is the mischief in me, and I wonder
if I could put a notion in his head:
'Why do they make good neighbors? Isn't it
where there are cows?
But here there are no cows.
Before I built a wall I'd ask to know
what I was walling in or walling out,
Our analysis suggests that firms in the chemical industry, like Frost and his neighbor,
share in the construction of a wall between them. We show that they need the wall because each
can indeed harm the other. They need to meet, we find, to “wall in” the effects of inevitable
30
accidents, keeping each more isolated, and so all more protected. Responsible Care is a means
of ensuring that each firm maintains its walls and so protects its neighbors.
As evidenced by Frost’s poem, traditions of mending fences to prevent spillover harm
have an ancient history. Societies all over the world have developed norms to ensure that each
member acts to protect his neighbor. Indeed, the very normalcy of such traditions represents the
culmination of Frost’s poem. Speaking of his neighbor, Frost notes, “He will not go behind his
father's saying, and he likes having thought of it so well, he says again, ‘Good fences make good
neighbors.’” Our analysis suggests that analogs to such traditional responses can be found in
modern industries. Firms unite with rivals to ensure that each buffers the other from a future
problem. Because they are “walling in” their own effect on their neighbors, individual firms
cannot achieve such results independently. Rather, at-risk firms must come together to create an
institution that helps ensure that each protects his neighbor.
Theoretical Implications
Beyond its implications for extending Ostrom’s (1990) community self-regulation
perspective on common pool resources into the literature on industry self-regulation, our study
suggests that management scholars should “endogenize” factors that have been presumed
exogenous to management analysis. Traditionally, firms and their governing institutions have
been viewed as distinct entities, with institutions establishing the “rules of the game” (North,
1990) and firms strategically playing by those rules. Industry self-regulation entails firms
creating an institution that sets and enforces rules governing how member firms are expected to
behave, and so it muddles the divide between those who make the rules and those who play the
game. By demonstrating that firms can come together to create an institution to govern their
own behavior when faced with a common problem, we thus respond to the many recent calls for
31
increased focus on agency in institutional theory (Brint & Karabel, 1991; Dacin, Goodstein &
Scott, 2002; Greenwood & Hinings, 1996; Hirsch & Lounsbury, 1997).
The relative inattention to agency in institutional theory may be attributable, in part, to
measurement difficulties. Prior studies have relied on a variety of indirect proxies to assess
institutional conditions over time, such as counts of press articles, lawsuits, regulatory actions,
and shareholder proxy contests (e.g., Davis & Thompson, 1994; Hoffman, 1997; Miles, 1982).
However, as argued above, in some situations, changing institutional conditions can best be
measured by analysis of variations in stock market values. Such values represent the mean
expectations of investors and so provide a good indicator of how perceptions of a given firm
fluctuate over time. This study, by drawing attention to such a readily available and consistent
longitudinal measure, may facilitate “good functionalism” studies that compare conditions both
before and after a new institution is created. Lack of such studies is “one of the biggest barriers
to creating a richer theory” (Ingram & Clay, 2000: 539) of institutions and so more such studies
are “badly needed” (Ingram & Clay, 2000: 540) to advance institutional theory.
Managerial Implications
Our research suggests that prior skepticism about the viability of industry self-regulation
may be based on an incorrect assumption about the institution’s function, and so previous
empirical studies may have measured the wrong attributes. Our study suggests that the
institution does not function as a signal to observers that member firms are only those with
superior or improving performance. Instead, it serves as a pact to reduce stakeholder concern
through greater outreach and communication, as we show happened in the chemical industry
through its CAER program. Such outreach and communication can have a positive effect on
32
stakeholder relations regardless of the performance of the member firms, and it provides a basis
for distinguishing one firm’s acts from others and so aids in walling in potential spillover effects.
Our research also suggests that managers in an industry can maintain a self-regulatory
institution, even when it provides spillover benefit to non-members. Our findings suggest that
the institution protected all firms from the errors of its members. Thus, participants in the
program provide a public good to the industry. Despite the incentive to free ride, however, firms
agree to participate and (over the years of our analysis at least) the program provides a benefit to
the industry.
For policy makers, our research reveals that private institutions may substitute in part for
public regulation on information disclosure. The need to prevent spillover harm can help drive
the creation of institutions that require the disclosure of information to stakeholders. Thus,
government programs on information disclosure should be analyzed by considering both their
direct effect on firm behavior and their indirect effect on the formation and function of selfregulatory institutions.
Limitations and Future Research
While our study addresses several long-standing issues, it also raises several new ones.
We use stock price movements to measure changes in stakeholder expectations about firms over
time following specific industry events. This methodology does not allow us to observe the
mechanisms that created changes in stakeholders’ expectations of a firm’s future performance.
We do not directly observe, for example, the provision of information from the firm to these
stakeholders. In our particular empirical setting, we note that the CAER program, which is at the
core of Responsible Care, requires members to engage in significant communication with
stakeholders, but we do not actually measure the degree to which members abided by these
33
requirements. In future research, we hope to directly evaluate differences in communication
rates and styles among participants and non-participants.
In our study, we described industry self-regulation as a private decentralized institution,
but it typically involves a central authority. In examples such as that of Responsible Care, there
exists a governing body that oversees compliance with the program’s codes. However,
compliance is often gauged through self-reporting and punishment tends to be limited or nonexistent, given antitrust concerns as well as the institution’s desire to retain as many members as
possible. Thus, even in exemplary and robust instances such as Responsible Care, industry selfregulation tends toward a decentralized model, relying on peer pressure for compliance.
Nonetheless, it would be more accurate to recast industry self-regulation as a hybrid form of
private institution since it contains both centralized and decentralized aspects.
This suggests that Ingram and Clay’s (2000) centralized-decentralized dichotomy for
institutions might better be treated as a continuum, and it raises the question of how institutions
choose to position themselves along this continuum, both initially and over time. Our study
addresses temporal changes, but it does not address changes within the program itself.
Responsible Care started as a primarily decentralized institution but over time has shifted toward
more centralized authority. The program evolved over the 1990s as new codes were hammered
out and promulgated to members. After our study timeframe ended, Responsible Care’s
leadership changed – in part because of several articles which suggested that the program had not
reduced the pollution generated by its members – and reportedly with the new leadership, the
“velvet glove came off” and some members were disciplined for their lax behavior (Reisch,
1998). Recently, it is reported to have changed again, this time seeming to return to a softer
form. Future research should investigate how self-regulatory programs balance centralization
34
and decentralization and the stringency of enforcement in order to attract members without
diminishing the legitimacy of the institution, as well as how this balance changes over time. The
drivers of such changes are poorly understood. Weak enforcement may be necessary to attract
members, yet the appearance that standards are enforced may be essential to maintaining the
legitimacy of the program in the eyes of observers. In future research, we hope to explore the
drivers and mechanisms of these changes.
Finally, our study does not resolve the issue of why firms choose to participate in selfregulatory programs. If participating firms essentially “wall in” their spillovers, safeguarding the
rest of the industry, then the institution provides a pure public good. Given the dominating
incentive to free ride on pure public goods (Olson, 1965), how then does the institution hold
together? Scholars have suggested that the incestuous nature of the chemical industry allows
bilateral sanctions that enforce participation. According to Rees (1997: 489), “the chemical
industry is its own best customer,” and large firms use their power over subordinate suppliers as
“leverage” (Gunningham, 1995: 85) to obtain their compliance. Yet aside from a few anecdotes,
such sanctions remain unobserved. In future research, we hope to further explore the centripetal
forces that hold this and similar institutions together.
Despite these limitations our research makes a significant contribution to emerging
scholarship on self-regulatory institutions and ongoing scholarship on agency in institutional
theory. It shows that exchange problems caused by shared reputation and risk of sanction exist
in modern industries. It shows that a major crisis can intensify problems associated with this
industry commons. Finally, it shows that when faced with exchange problems caused by these
commons, the choice is not between Leviathan or Oblivion (Ophuls, 1973). Industry members
can take matters into their own hands and repair shared problems by forging a new institution.
35
TABLE 1
Descriptive Statistics and Correlations for Spillover Analysis
Variable
Mean
Std.
Dev.
1
2
3
4
5
6
7
8
9
1
CAR (5)
0.141
9.800
1
2
Pre Bhopal
0.037
0.188
-0.007
1
3
Post RC
0.786
0.410
0.050
-0.215
1
4
Same SIC
0.440
0.496
-0.042
-0.090
0.034
1
5
Emp. Killed
0.272
0.445
-0.036
0.172
-0.305
-0.114
6
Emp. Injured
1.444
1.169
-0.050
0.161
-0.109
0.071
0.022
1
7
Assets of Perp
9.520
1.425
0.066
-0.003
0.026
-0.335
-0.091
-0.196
1
8
Perp ACC
0.737
0.440
0.015
0.064
-0.011
0.131
-0.010
-0.147
0.294
1
9
10
11
12
1
Perp RC
0.572
0.495
0.050
-0.126
0.583
0.124
-0.185
-0.205
0.282
0.726
1
10
ACC Member
0.137
0.344
-0.006
0.035
-0.093
0.023
0.040
0.023
-0.009
0.008
-0.048
1
11
RC Member
0.118
0.322
-0.001
-0.039
0.183
0.031
-0.042
-0.001
-0.006
0.009
0.116
0.710
1
12
Assets
4.866
2.357
-0.001
0.018
-0.003
-0.066
0.001
0.000
-0.006
0.009
0.006
0.498
0.472
1
13
Mills Ratio
2.696
1.033
-0.015
-0.029
-0.094
-0.140
0.018
-0.015
0.010
-0.001
-0.052
-0.370
-0.399
-0.649
36
TABLE 2
Spillover Effect of Accidents on Firms Not At-Fault
Model
1
Pre Bhopal
Post RC
Accident in Same
SIC
Emp. killed
Emp. Injured
-0.876**
(0.11)
-0.825**
(0.13)
-0.363**
(0.05)
2
3
4
5
6
0.752*
(0.32)
1.023**
(0.15)
1.010**
(0.33)
0.917**
-0.17
1.143**
(0.34)
0.289
(0.30)
1.225**
(0.349)
0.111
(0.326)
1.202**
(0.423)
0.203
(0.350)
-0.869**
(0.11)
-0.585**
(0.13)
-0.340**
(0.05)
-0.596*
(0.27)
-0.377**
(0.14)
-0.384**
(0.06)
0.369**
(0.05)
-0.601*
(0.27)
-0.414**
(0.14)
-0.360**
(0.06)
0.329**
(0.06)
-0.785*
(0.31)
1.053**
(0.35)
0.275
(0.25)
-0.532*
(0.26)
0.021
(0.03)
-0.585*
(0.272)
-0.404**
(0.145)
-0.353**
(0.056)
0.328**
(0.056)
-0.819**
(0.307)
1.068**
(0.350)
1.225**
(0.349)
0.111
(0.326)
-0.585*
(0.272)
-0.661*
(0.321)
-0.432**
(0.164)
-0.385**
(0.063)
0.343**
(0.063)
-0.933*
(0.363)
1.218**
(0.408)
-0.200
(0.104)
1.202**
(0.423)
0.203
(0.350)
-0.661*
(0.321)
-0.572*
(0.272)
-0.270
(0.163)
-0.243**
(0.072)
0.227**
(0.064)
-1.140**
(0.316)
1.414**
(0.362)
-0.572*
(0.272)
-0.270
(0.163)
-0.243**
(0.072)
0.227**
(0.064)
26139
0.01
30751
0.02 ω
Yes
Yes
Heckman
Yes
Assets of Perp
Perp ACC
Perp RC
ACC Member
-0.037
(0.19)
RC Member
Assets
Mills Ratio
Constant
1.276**
(0.11)
30751
0.01
0.342*
0.18
30751
0.01
44.59**
Observations
30751
30751
30751
R2
0.01
0.01
0.01ω
2
161.55** 13.26**
LR (Chi )
Accident (4 digit)
Yes
Yes
Yes
Industry (2 digit)
Yes
Yes
Yes
Firm
Year
* p < 0.05, ** p < 0.01 ω: within R2 reported for FE analysis
7
Yes
Yes
APPENDIX A
Calculation of Cumulative Abnormal Return (CAR)
To calculate CAR, we first calculated the relationship between the value of each
company’s stock and the market as a whole (measured by the CRSP value-weighted index with
dividends for the entire market)
Rit = ai + BiRmt + eit;
where Rit represents the value of the security i on day t, ai is a constant, Rmt represents the value
of the market portfolio for day t, Bi represents the Beta of security i, and eit represents the error
term. Beta is computed over the period t = -254 to t = -1, where t = 0 is the day of the event.
The abnormal return of a stock is the difference between the actual return of that stock
and its expected return. The abnormal return of security i at time t, ARit , is:
ARit = Rit – (ai + BiRmt)
The cumulative abnormal return for a firm, CARi, is the sum of abnormal returns over the event
window:
CARi = ∑ ARit
To allow for continuous compounding when aggregating the abnormal returns, ln(1+R) is used
in place of R. Thus,
CARi = ∑ [ln(1+Rit) – (a + Bimln(1+Rmt)))]
Key to computation of CAR is determination of the event window, the period of time over
which to cumulate abnormal returns. Event studies commonly begin the event window prior to
the actual event announcement in order to account for information leakage, but dangerous
accidents are, by nature, unanticipated events. Therefore, in this study, if the event occurred
during trading hours on a trading day, the window begins that day; if not, the window begins
with the next trading day.
While it is straightforward to choose the beginning of the event window for this study, it
is less clear when to close the window. The occurrence and magnitude of events sometimes take
several days to become apparent to the market. While news travels fast and markets update their
values nearly instantaneously, many characteristics of major accidents take time to establish. For
example, the enormous toll of Bhopal took many days to unfold, and the magnitude of the Exxon
Valdez oil spill was not immediately evident. Whereas a long event window increases the
likelihood of capturing the full impact of unfolding events, a long event window also increases
the opportunity for intervening events to confound results (McWilliams & Siegel, 1997).
Therefore, researchers should use the shortest possible window that captures the fullest extent of
the event. The bulk of the effects tended to occur within five days after the event. Thus, our
dependent variable in this study is the cumulative abnormal return on the 5th day after an event.
38
APPENDIX B
The Effect of Accidents on Return of At-Fault Firms
Model
Assets
1
Employee Killed
Pre Bhopal (1980-1984)
Post RC (1990-2000)
ACC Member
Constant
-1.222*
(0.53)
N/A
R-squared
N = 83, * p < 0.05
39
2
0.702*
(0.38)
-2.69*
(1.19)
-1.42
(1.90)
0.6
(1.22)
3.914
(2.47)
-7.27*
(3.83)
0.15*
APPENDIX C
Probit Analysis of ACC Participation 6.
Descriptive Statistics for All Years (1980-2000)
Mean
9.576
0.539
4.485
1 Industry Emissions
2 Chemical Focus
3 Assets
Std Dev
2.518
0.480
2.368
1
1
0.099
0.199
2
1
-0.145
N= 5073
Probit Analysis of ACC Membership
All years
0.177**
(0.037)
-0.417**
(0.117)
0.393**
(0.038)
-5.051**
(0.459)
5073
127.16
0.36
Industry Emissions
Chemical Focus
Assets
Constant
N
Chi2
Pseudo R2
1990
0.148**
(0.052)
0.612*
(0.303)
0.401**
(0.055)
-5.448**
(0.636)
276
74.24
0.38
* p < 0.05, ** p < 0.01
6
Separate probit models were estimated for each year from 1980 to 2000 to obtain the Mills Ratio. This was
included in the second stage regression estimation to reduce the effect of non-random treatment. Inclusion of this
term accounts for changes in the expectation of the treatment coefficient but does not correct for heteroskedastic
changes in the error terms that result from non-random treatment. We used a standard White’s method to help
correct for these effects.
40
REFERENCES
Acheson, J.M. 1988. The Lobster Gangs of Maine. Hanover, New Hampshire: University Press.
Acheson, J. M. & J. Knight. 2000. Distribution fights, coordination games, and lobster
management. Comparative Studies in Society and History, 42(1): 209-238.
Barnett, M. 2006. Finding a working balance between competitive and communal strategy.
Journal of Management Studies, 43(8): 1753-1773.
Baron, D. P. 1995. The Nonmarket Strategy System. Sloan Management Review, 37(1): 73-85.
Begley, R. 1994. After Bhopal: A CMA ‘war room,’ new programs, and changed habits.
Chemical Week, Dec. 7, p. 32.
Berkes, F. 1989. Common property resources: Ecology and community-based sustainable
development. London: Belhaven Press.
Blacconiere, W. & D. Patten. 1994. Environmental disclosures, regulatory costs, and changes in
firm value. Journal of Accounting and Economics, 18: 357-377.
Brint, S. and J. Karabel. 1991. Institutional origins and transformations: The case of American
community colleges. In W. Powell & P. DiMaggio (eds.), The new institutionalism in
organizational analysis: 337-360. Chicago: University of Chicago Press.
Brown, S. and J. Warner. 1980. Measuring security price performance. Journal of Financial
Economics, 8: 205-258.
Brown, S. and J. Warner. 1985. Using daily stock returns: The case of event studies. Journal of
Financial Economics, 14: 3-31.
CCPA. 2007. http://www.ccpa.ca/ResponsibleCare/ Accessed July 1, 2007.
Coombes, P. 1991. Union Carbide. Chemical Week, July 17: p. 44
Dacin, M.T. , J. Goodstein & W. R. Scott. 2002. Institutional theory and institutional change:
Introduction to the special research forum. Academy of Management Journal, 45(1): 45-57.
Davis, G. F. and T. A. Thompson. 1994. A social movement perspective on corporate control.
Administrative Science Quarterly, 39: 141-173.
Dawson, N. L. and K. Segerson. 2005. Voluntary Environmental Agreements with Industries:
Participation Incentives with Industry-Wide Targets. University of Connecticut Working Paper.
Ember, L. R. 1995. Responsible Care: Chemical makers still counting on it to improve image.
Chemical and Engineering News, May 29.
41
Fama, E. & K. French. 1992. The cross-section of expected stock returns. Journal of Finance,
47(2): 427-465.
Fombrun, C. 1996. Reputation: Realizing value from the corporate image. Boston: Harvard
Business School Press.
Furger, F. 1997. Accountability and Systems of Self-Governance: The Case of the Maritime
Industry. Law and Policy, 19: 445–476.
Galvin, L. 2007. Outbreak Linked to Spinach Forces Reassessment of Food Safety Practices.
Amber Waves: The Economics Food, Farming, Natural Resources and Rural America, June.
Gibson, W. D. 1985a. After Bhopal: First aid for chemicals’ public image. Chemical Week, Feb.
20.
Gibson, W. D. 1985b. Olive branch activity prevails after Institute. Chemical Week, Oct. 16.
Granovetter, M. 1985. Economic action and social structure: The problem of embeddedness.
American Journal of Sociology, 91: 481-510.
Greening, D. & R. Johnson. 1997. Managing industrial and environmental crises: The role of
heterogeneous top management teams. Business & Society, 36(4): 334-361.
Greenwood, R. & C. Hinings. 1996. Understanding radical organizational change: Bringing
together the old and the new institutionalism. Academy of Management Review, 21: 1022-1054.
Greenwood, R., R. Suddaby & C. Hinings. 2002. Theorizing change: The role of professional
associations in the transformation of institutionalized fields. Academy of Management Journal,
45(1): 58-80.
Gunningham, N. 1995. Environment, self-regulation, and the chemical industry: Assessing
Responsible Care. Law & Policy, 17(1): 57-108.
Gunningham, N. & J. Rees. 1997. Industry self-regulation: An institutional perspective. Law &
Policy, 19(4): 363-414.
Hamilton, J. 1995. Pollution as news: Media and stock market reactions to the TRI data. Journal
of Environmental Economics and Management, 28: 98-113.
Hannan, M. & G. Carroll. 1992. Dynamics of organizational populations: Density, legitimation,
and competition. New York: Oxford University Press.
Hardin, G. 1968. The tragedy of the commons. Science, 162: 1243-48.
Heller, K. 1991. Public outreach: The stakes are high. Chemical Week, July 17.
42
Heller, K. & D. Hunter. 1994. Responsible Care. Chemical Week, July 6.
Hill, J. & T. Schneeweis. 1983. The effect of Three Mile Island on electric utility stock prices: A
note. Journal of Finance, 38(4): 1285-1292.
Hirsch, P. & M. Lounsbury. 1997. Ending the family quarrel: Toward a reconciliation of “old”
and “new” institutionalisms. American Behavioral Scientist, 40: 406-418.
Hoffman, A. 1997. From heresy to dogma: An institutional history of corporate
environmentalism. San Francisco: The New Lexington Press.
Hoffman, A. 1999. Institutional evolution and change: Environmentalism and the U.S. chemical
industry. Academy of Management Journal, 42: 351-371.
Hoffman, A. & W. Ocasio. 2001. Not all events are attended equally: Toward a middle-range
theory of industry attention to external events. Organization Science.
Hornblower, M. 2000. Wake up and smell the protest. Time, 155(15): 58.
Howard, J., J. Nash & J. Ehrenfeld. 2000. Standard or smokescreen? Implementation of a
voluntary environmental code. California Management Review, 42(2): 63-82.
Ingram, P. & K. Clay. 2000. The choice-within-constraints new institutionalism and implications
for sociology. Annual Review of Sociology, 26: 525-546.
Ingram, P. & C. Inman. 1996. Institutions, intergroup rivalry, and the evolution of hotel
populations around Niagara Falls. Administrative Science Quarterly, 41: 629-58.
Jarrell, G. & S. Peltzman. 1985. The impact of product recalls on the wealth of sellers. The
Journal of Political Economy, 93(3): 512-536.
Jiang, R. H. J., & P. Bansal. 2003. Seeing the need for ISO 14001. Journal of Management
Studies, 40(4): 1047-1067.
King, A. & M. Lenox. 2000. Industry self-regulation without sanctions: The chemical industry’s
Responsible Care program. Academy of Management Journal, 43(4): 698-716.
King, A., M. Lenox & M. Barnett. 2002. Strategic responses to the reputation commons problem.
In A. Hoffman & M. Ventresca (eds.), Organizations, Policy and the Natural Environment:
Institutional and Strategic Perspectives. Stanford, CA: Stanford University Press: 393-406.
Lenox, M. J. 2006. The role of private decentralized institutions in sustaining industry selfregulation. Organization Science, 17(6): 677-690.
Lloyd, W. F. 1833 (1968). Two Lectures on the Checks to Population: Augustus M. Kelley.
43
Malkin, E. 1996. Pangs of conscience. Business Week, July 29 (3486): 46-47.
McGahan, A. M. & M. E. Porter. 1997. How much does industry matter, really? Strategic
Management Journal, 18: 15-30.
McWilliams, A. & D. Siegel. 1997. Event studies in management research: Theoretical and
empirical issues. Academy of Management Journal, 40(3): 626-657.
Meyer, A. 1982. Adapting to environmental jolts. Administrative Science Quarterly, 27:515-537.
Meyer, A., G. Brooks & J. Goes. 1990. Environmental jolts and industry revolutions:
Organizational responses to discontinuous change. Strategic Management Journal (special issue):
93-110.
Miles, R. H. 1982. Coffin nails and corporate strategies. Englewood Cliffs, NJ: Prentice-Hall.
Mitchell, M. 1989. The impact of external parties on brand-name capital: The 1982 Tylenol
poisonings and subsequent cases. Economic Inquiry, 27: 601-618.
North, D. C. 1981. Structure and change in economic history. New York: Norton.
North, D. C. 1990. Institutions, institutional change and economic performance. Cambridge:
Cambridge University Press.
Olson, M. 1965. The Logic of Collective Action: Public Goods and the Theory of Groups.
Cambridge, MA: Harvard University Press.
Ophuls, W. 1973. Leviathan or oblivion? In Herman E. Daly (ed.), Toward a Steady State
Economy, San Francisco: Freeman: p. 215-230.
Ostrom, E. 1990. Governing the Commons: The Evolution of Institutions for Collective Action.
Cambridge, UK: Cambridge University Press.
Ostrom, E. 2000. Collective action and the evolution of social norms. The Journal of Economic
Perspectives, 14(3): 137-158.
Patten, D. & J. Nance. 1998. Regulatory cost effects in a good news environment: The intraindustry reaction to the Alaskan oil spill. Journal of Accounting and Public Policy, 17: 409-429.
Potoski, M. & A. Prakash. 2005. Green clubs and voluntary governance: ISO 14001 and firms'
regulatory compliance. American Journal of Political Science, 49(2): 235-248.
Prakash, A. & M. Potoski. 2006. The voluntary environmentalist: Green clubs, ISO 14001, and
voluntary environmental regulations. Cambridge, UK: Cambridge University Press.
44
Rees, J. 1994. Hostages of each other: The transformation of nuclear safety since Three Mile
Island. Chicago: University of Chicago Press.
Rees, J. 1997. The development of communitarian regulation in the chemical industry. Law and
Policy, 19: 477.
Reinhardt, F. L. & R. Vietor. 1994. Starkist. Harvard Business School Case.
Reisch, M. 1998. Industry ponders future of Responsible Care. Chemical and Engineering News,
May 5: 13.
Rivera, J. 2002. Assessing a voluntary environmental initiative in the developing world: The
Costa Rican Certification for Sustainable Tourism. Policy Sciences, 35(4): 333-360.
Rivera, J. & P. deLeon. 2004. Is greener whiter? The sustainable slopes program and voluntary
environmental performance of western ski areas. Policy Studies Journal, 32(3): 417-437.
Rivera, J., P. De Leon, & C. Koerber. 2006. Is greener whiter yet? The sustainable slopes
program after five years. Policy Studies Journal, 34(2): 195-221.
Rumelt, R. 1991. How much does industry matter? Strategic Management Journal, 12(3): 167185.
Schmalensee, R. 1985. Do markets differ much? American Economic Review, 75(3): 341-351.
Schotter, A. 1981. The economic theory of social institutions. New York: Cambridge University
Press.
Shrivastava, P. 1987. Bhopal: Anatomy of a crisis. Cambridge, MA: Ballinger.
Slovic, P. & E. U. Weber. 2002. Perception of risk posed by extreme events. Presented at, Risk
Management Strategies in an Uncertain World Conference, Palisades, New York; April 12-13.
Vasi, I. B. & M. Macy. 2003. The mobilizer's dilemma: Crisis, empowerment, and collective
action. Social Forces, 81(3): 979-998.
Wade, R. 1988. Village republics: Economic conditions for collective action in South India.
Oakland: ICS Press.
Weber, J. M., S. Kopelman, & D. Messick. 2004. A conceptual review of decision making in
social dilemmas: Applying a logic of appropriateness. Personality and Social Psychology
Review, 8(3): 281-307.
Williamson, O.E. 1985. The economic institutions of capitalism: Firms, markets, and relational
contracting. New York: Free Press.
45
Wotruba, T. 1997. Industry self-regulation: A review and extension to a global setting. Journal of
Public Policy & Marketing, 16(1): 38-54.
46