Journal of Environmental Economics and Management 71 (2015) 1–18
Contents lists available at ScienceDirect
Journal of
Environmental Economics and Management
journal homepage: www.elsevier.com/locate/jeem
Stock market and deterrence effect: A mid-run analysis of
major environmental and non-environmental accidents$
Cécile Carpentier, Jean-Marc Suret
Laval University and CIRANO Fellows, Québec, QC, Canada
a r t i c l e i n f o
abstract
Article history:
Received 1 February 2013
Available online 30 January 2015
We analyze the stock market reaction to 161 major environmental and nonenvironmental accidents, reported on the front page of the New York Times for half a
century. To determine if the market induces a real deterrence effect, we extend the event
windows up to one year. On average, the market reacts negatively and enduringly to the
announcement of an accident. However, this average effect is largely driven by the airline
industry and by government interventions. The estimated average compounded abnormal
return following environmental accidents does not differ from zero after one year. This
does not exclude, in severe events affecting large firms, huge losses in equity value, but
the significant negative cumulative abnormal returns estimated immediately after an
environmental accident in previous studies do not persist. Our results suggest that in a
market driven by institutional investors, the deterrence effect is likely to be weak.
& 2015 Elsevier Inc. All rights reserved.
JEL classification:
G14
Q51
Q53
Keywords:
Deterrence effect
Environmental accident
Event study
Mid-run analysis
Stock market
Disaster
Catastrophe
Introduction
The extent to which the stock market can motivate firms to adopt better corporate safety or environmental behavior
remains a fundamental question (Karpoff et al., 2005). If a severe market penalty follows evidence of corporate weaknesses
in controlling hazards, then the market could be seen as a complement or a substitute to the regulatory actions (Dasgupta
et al., 2001). Moreover, mandatory disclosure requirements such as toxic release inventory (TRI) and ‘green labels’ might
become effective regulatory mechanisms for controlling safety or environmental hazards (Konar and Cohen, 1997; Khanna
et al., 1998; Capelle-Blancard and Laguna, 2010; Oberndorfer et al., 2013). To evidence a possible deterrence effect,
researchers have scrutinized the market reaction following ‘negative incidents’ including accidents, lawsuits and
misconduct announcements (Jones and Rubin, 2001; Ambec and Lanoie, 2008). We focus on major accidents that should
significantly modify the agent’s perception of the firm’s compliance with security and environmental standards and that
should cause a drop in market value. In turn, this drop should lead the shareholders to force management to put more effort
into controlling security and environmental hazards.
☆
The authors are grateful to Nicolas Ros for valuable research assistance. The authors thank the Editor, two anonymous reviewers, Carl Brousseau, Eric
De Bodt, Gunther Capelle-Blancard, participants at the SKEMA school of Business and Rennes IGR/IAE 2012 seminars, AFFI 2012 International Conference,
RIODD 2012 conference at Audencia, the 2012 International Symposium on Money, Banking and Finance at University of Nantes, and the IRIAF 2014
conference at Poitiers for helpful comments. The authors assume full responsibility for any errors remaining in the text.
E-mail addresses: Cecile.Carpentier@fsa.ulaval.ca (C. Carpentier), Jean-Marc.Suret@fsa.ulaval.ca (J.-M. Suret).
http://dx.doi.org/10.1016/j.jeem.2015.01.001
0095-0696/& 2015 Elsevier Inc. All rights reserved.
2
C. Carpentier, J.-M. Suret / Journal of Environmental Economics and Management 71 (2015) 1–18
Previous evidence remains limited. Since 1988, seven papers have been devoted to airline and other transportation
accidents and seven to other industries (Table 1). Moreover, results relative to the deterrence effects are still controversial
and incomplete along several dimensions. First, five papers fail to detect a significant effect of accident announcement on
stock prices. Second, previous research focuses on the few days surrounding the announcement. Third, some arguments and
evidence suggest that environmental negative events should have a lesser impact on stock prices than non-environmental
events. Fourth, following some accidents, government actions can induce significant complementary effects on stock prices
that should be considered. This paper completes and extends previous findings along these dimensions.
Several studies fail to report a significant drop in market price following accidents (Mitchell and Maloney, 1989; Knight
and Pretty, 1999; Jones and Rubin, 2001). The mixed results are probably attributable to the heterogeneity of the events
analyzed. Several papers group together very large and publicized major accidents, minor incidents that hardly reach
shareholders, and even product recalls (Knight and Pretty, 1999). To really determine the extent of a deterrence effect, a
focus on major accidents is warranted. We study transportation and industrial accidents large enough to appear on the front
page of an influential newspaper.
Previous research on the impact of accident announcements on stock markets has used the event study approach. The
abnormal return during the few days following the announcement is the difference between the observed returns and a
normal return, estimated using the market model as described by Capelle-Blancard and Laguna (2010). The estimated
cumulative abnormal return over a given event window represents the hypothetical rate of return of an investor who buys
the stock of a firm at the closing price the day before the event and sells the stock at the end of the window. Previous
analyses generally report estimated average cumulative abnormal return for the two days following the accident, ranging
between 1% and 5%. Investors’ wealth is thus negatively affected by the accident (Walker et al., 2006; Capelle-Blancard
and Laguna, 2010; Ho et al., 2013). This observation supports the assertion that financial markets may provide incentives for
firms to change their environmental and safety behavior (Konar and Cohen, 1997; Dasgupta et al., 2001; Engelen and van
Essen, 2011). However, as underlined by Ambec and Lanoie (2008), a real deterrence effect can exist only if the
announcement of a negative event is associated with a persistent effect on stock prices, consistent with an increase in
the cost of equity. If the drop in market value following an accident is limited to a few days, and is followed by a strong
recovery, the deterrence effect should be weak. Institutional investors, whose investment horizon is longer than a year, are
probably not very concerned about such a short-term effect. Every basic book in corporate finance, such as Ehrhardt and
Brigham (2013, p. 9), states that managers should maximize shareholders’ wealth in the long run, and not focus on the
current market price. Accordingly, the shareholders would not really be affected, or be prone to pressure firm management,
if the accident effect on firm value is observed only for a few days. Previous research evidences a stock price reversal
following the drop observed shortly after the accident announcement (Borenstein and Zimmerman, 1988; Jones and Rubin,
2001; Walker et al., 2006), and the lack of significant accident impact after a few weeks or months (Knight and Pretty, 1999;
Capelle-Blancard and Laguna, 2010). Table 1 illustrates the scarcity of results related to this dimension. When provided, the
estimated one-year average cumulative abnormal returns do not differ significantly from zero. Even large-scale events such
as Bhopal or the Exxon Valdez oil spill have not decreased the stock value significantly in the mid-term (Salinger, 1992;
White, 1996). Moreover, the four studies reporting mid-run results fail to address the numerous methodological challenges
Table 1
Stock market reaction to accidents. Event windows for the estimated average cumulative abnormal returns are presented between brackets when they
differ from the classical windows. NA stands for not available.a Env. (acc.) means environmental (accident).
Authors
Sample
[0–2]
[0–10]
6 Months 1 Year
Ho et al. (2013)
Sabet et al. (2012)
Capelle-Blancard and Laguna
(2010)
Walker et al. (2006)
Karpoff et al. (2005)
Walker et al. (2005)
Jones and Rubin (2001)
Knight and Pretty (1999)
Nethercurtt and Pruitt (1997)
Klassen and McLaughlin (1996)
Salinger (1992)
Broder and Morrall (1991)
Mitchell and Maloney (1989)
Borenstein and Zimmerman (1988)
133 Air crashes, 1950–2009
BP Deepwater
64 Chemical disasters, 1990–2005, 10 countries
4.60%nnn
2.62%nnn
1.09%nnn
2.955%nnn
NA
0.60%
NA
NA
2.70%
NA
NA
NA
26 Major railroad events, 1993–2003, US–Canada
478 Env. violations, including accidents 1980–2000
107 Airline disasters, 1962–2003, US
73 Negative env. events, 1970–1992
15 Major corporate catastrophes, 1982–1993
Valujet Flight 592, 1996, US
18 Env. crises, 1989–1990, US
Bhopal, Union Carbide, 1984, India
86 Fatal acc., 1963–1986, US and other
24 Fatal airline crashes, 1964–1987, US
74 Airline crashes, 1960–1985, US
0.27%
1.90%nnn
1%nnn
NA
3.10%nnn
3.18%nnn [0 þ 14]
0.15%
0.44%
NA
6.652%
NA
0.6731%
nnn
1.50%
[0 þ3] NA
NA
31.5%nnn[0 þ20]
1.95%nn
2.57%nn
2.27%
2.51% [0þ 5]
0.873%
0.234%
0.57%
NA
2.93%
NA
0.608%
NA
NA
22%nn
NA
NA
NA
0.28%
NA
5.74%
NA
0.58%
NA
NA
8.90%
NA
NA
NA
a
We exclude the studies of spillover effects of accidents because the deterrence effect is unclear in these cases; see Capelle-Blancard and Laguna (2010
p.194) for a survey. All studies use the market model, but Sabet et al. (2012) add a factor related to oil prices and a GARCH adjustment to consider the
specificities of the Oil and Gas industry. Salinger (1992), who studies the Bhopal accident, adjusts the returns for both the market and a portfolio of
chemical stocks.
nn
Denote statistical significance at the 5% level, respectively.
nnn
Denote statistical significance at the 1% level, respectively.
C. Carpentier, J.-M. Suret / Journal of Environmental Economics and Management 71 (2015) 1–18
3
linked to this horizon and do not consider the numerous other determinants of performance. These limitations justify the
examination of the mid-term effect of major accidents proposed in this paper.
Several studies have concluded that a significant difference exists between environmental and non-environmental
negative events because the former are third-party offenses, where the damaged party does not engage in repeat
contracting with the offending firm (Tibbs et al., 2011). Indeed, some event studies of environmental accidents fail to report
significant negative effects of such accidents and question the deterrence effect for environmental accidents (Jones and
Rubin, 2001; Engelen and van Essen, 2011). The possible difference between the stock prices’ reaction depending on the
environmental dimension is analyzed extensively in this paper. Lastly, the authorities react to some events by grounding
some planes, closing industrial sites or changing regulations. Such actions can induce high direct costs for firms and should
be carefully distinguished from the classical deterrence effect of the market.
The aim of this paper is to examine the various unsolved questions related to the deterrence effect. To reach this
objective, we focus on the mid-term effect of major events over half a century. We use an original sample of 161 major
accidents reported on the front page of the New York Times from 1959 to 2010 and analyze the stock price evolution over
240 trading days. To our knowledge, we present the first comprehensive mid-term analysis of the market reaction following
major accidents, allowing comparison between environmental and non-environmental accidents. To overcome the
methodological intricacies of this type of analysis, we implement two different methodologies and use several benchmarks
and statistical tests to assess the robustness of our results.
We show that on average, the market reacts negatively and enduringly to the announcement of an accident. However,
the average effect is largely driven by two subsamples of events, and we find no evidence of a significant average abnormal
return except for the airline industry and for events that prompt a government reaction. This does not exclude, in severe
cases involving large firms, huge equity value losses, but the significant negative cumulative abnormal return observed
immediately after an environmental accident in previous studies does not persist, even when the worst events and those
that received the most media coverage are selected. Our results are consistent with the proposition that, in the mid-run, the
stock market does not provide a strong deterrence effect for environmental accidents. This is particularly true in a market
dominated by institutional investors with a long investment horizon who are probably not very affected by short-term
fluctuations in stock prices.
The paper is structured as follows. Section 2 describes the background and develops the hypotheses. Section 3 presents
the definitions, the selection process and data features. Section 4 presents the methodological approach. Section 5 discusses
the empirical results. Section 6 concludes the paper.
Background and hypotheses
Why does the mid-term effect matter?
The market can have a disciplinary effect only if the accident announcement depresses the stock prices for a period of
time commensurate with investors’ investment horizon. Individual investors represent a tiny and decreasing proportion of
trades on the NYSE (Evans, 2009). Shares are increasingly owned and traded by institutional investors, who can pressure
firm management if they consider that the previous accident prevention actions and policy are inappropriate. These
investors have an investment horizon of about 15 months (Gaspar et al., 2005). As a result, a short-term drop in prices
followed by a reversal is unlikely to give them a sufficient incentive to pressure firm management. Irrational individual
investors can be the sole losers following an accident, by selling at a low price. Conversely, rational institutional investors
can be the winners by buying at the low price after the announcement. As underlined by Ambec and Lanoie (2008), a shortrun negative price movement does not provide enough substance to create a real deterrence effect. In this paper, we focus
on the one-year horizon. This time frame is commensurate with the investment horizon of institutional investors.
Expanding the analysis to more than twelve months implies a sharp decrease in the number of available observations and a
strong survival bias.1 This is the reason why we use a mid-term analysis.
Why should an accident decrease the market value of equity?
Market value drops following an accident can be traced to one or more of the following factors: direct costs, reputational
effects and irrational investor reaction.
The stock price is assumed to reflect the present value of the expected firm’s cash flows, which can be negatively affected
by the direct consequences of the accident, through material damage, cleaning costs and indemnities. However, insurance
1
A long-run event study based on a small sample can be affected by survival bias, and produce inaccurate results (Kothari and Warner, 1997). Widening
the event windows significantly increases the survival problems in industries where numerous firms delist or file for bankruptcy, as in the airline industry,
which constitutes a large part of our sample (Dempsey, 1991). Moreover, firms should be excluded if simultaneous events occur within the event window
(Dasgupta et al., 2001). The deregulation of the airline industry (1978), the 9/11/2001 attack and the oil crisis (1973, 1979 and 1990), which significantly
influence numerous firms in our sample, are examples of such simultaneous events. However, the observation of a second accident during the years
following the first one is the most frequent reason for deletion. Lastly, several firms reorganize in the years following the accident and should be deleted
from the sample. The longer the event window, the more numerous the cases of simultaneous events and deletions.
4
C. Carpentier, J.-M. Suret / Journal of Environmental Economics and Management 71 (2015) 1–18
covers much of these costs, including most of the civil liabilities. In the environmental domain, third-party liability
insurance compensates for damage caused by the insured to other people, while personal insurance compensates the costs
incurred by the insured in remedying pollution on its own (Richardson, 2001). Accordingly, direct costs of accidents are
generally not material for firms. However, environmental damage insurance is a complex topic (Faure, 2002), and the results
of lawsuits cannot be totally covered.
Lawsuits arising from major accidents are often very long and complex processes resulting in small fines relative to the
firm’s earnings.2 Based on an analysis of damage compensation following the Exxon Valdez and TMI accidents, Lempert
(2009, p. 8) states that “deterrence effects in these cases are likely to be minimal because the ways in which low probability
events cause their harms are difficult to anticipate, and a particular low probability/high consequence harm might not be
expected to occur in the same way again, if it occurs at all.” At the time of the accident, the present value of expected fines is
most often low.
Neither direct costs nor lawsuit outcomes can explain why the stock prices of firms involved in negative events should
decrease; the reputational effect is a more plausible candidate. An accident can negatively impact the perceived quality of
the firm’s outcome and potentially reduce consumers’ propensity to use its products and services (Jones and Rubin, 2001).
According to Servaes and Tamayo (2013), customers take into consideration firms’ corporate social responsibility activities
when making purchase decisions, and they are more likely to purchase goods from firms that are more socially responsible.
However, consumers are often unaware of firms’ corporate social responsibility activities. Accidents that receive heavy
media coverage tend to attract customers’ attention and motivate them to change their purchase preferences. This
ultimately decreases expected cash flows, especially if consumers are easily able to switch to rival firms, as in the air
transportation industry (Bosch et al., 1998). If investors consider the firm’s reputation important, they are likely to change
their portfolio following a major accident. This will imply a decrease in the demand for the stock and in the market value.
For major events, these effects are unlikely to disappear rapidly. This leads to our first hypothesis:
Hypothesis 1. A major accident announcement reduces the firm’s market value in the mid-term.
Not all firms are likely to suffer from a similar reputational effect, mainly because their clients are other firms rather than
individuals. Environmental events often concern firms in the petroleum and chemical industry, which mainly conduct
business-to-business activities with limited direct interaction with individual customers. This can explain why, for negative
environmental events, Harper and Adams (1996), Jones and Rubin (2001) and Laplante and Lanoie (1994) do not detect any
significant drop in value. Hamilton (1995) and Karpoff et al. (2005) find that firms that violate environmental laws suffer
market losses statistically equivalent to the legal penalties imposed. Therefore, firms do not experience reputational loss
when they violate environmental regulations and, according to Karpoff et al. (2005), legal penalties, rather than reputational
penalties, are the primary deterrents to environmental violations. Engelen and van Essen (2011) suggest that the strongest
stock market effects are limited to firms whose customers can transfer to the competitors, as in the aviation and other
transportation sectors. In line with the reputational explanation of the accidents’ impact, we posit the following hypothesis:
Hypothesis 2. The negative mid-term effect of an accident announcement on the firm’s market value is lower for
environmental than for non-environmental accidents.
Behavioral finance proposes that some investors react irrationally to an accident announcement, mainly when it
concerns a product or service they use occasionally. Kaplanski and Levy (2010) affirm that aviation crashes affect people’s
mood and increase their anxiety, which negatively influences investment in risky assets. Their analysis focuses on the shortterm stock market decrease induced by aviation crashes. However, if the crashes have a high psychological resonance, one
can expect to observe a more significant mid-term market reaction following airline accidents than non-airline accidents.
The “switch” effect occurs when air travelers fly with the competitors of the crash airlines (Ho et al., 2013). This effect is
likely to be higher in the airline industry than in other sectors where customers are individuals, because of the higher level
of competition between airliners than between ground transportation firms. The switch effect is probably low for public
utilities where the clients are captive and in industries whose firms mainly trade with other firms. Hence we expect that:
Hypothesis 3. The negative mid-term effect following an accident announcement on the firm’s market value is stronger for
airline accidents than for non-airline accidents.
In a limited number of cases, government agencies act after the accident to protect the public; this can directly hurt the
profits of the firm concerned. For example, following the Chicago crash of American Airlines Flight on May 25, 1979, the type
certificate of the DC-10 was withdrawn by the Federal Aviation Administration, grounding the aircraft indefinitely on June 6,
1979. Government reaction also had a strong impact on stock value after the BP 2010 accident (Sabet et al., 2012) and after
the Three Mile Island accident (Spudeck and Moyer, 1989). Overall, the direct costs of the accident are either insured or
2
The length and unpredictable results of lawsuits following an accident could be illustrated as follows: In June 2008, the US Supreme Court reduced
what had once been a $5-billion punitive damages award against ExxonMobil to about $500 million, related to the Exxon Valdez spill in Alaska in 1989.
Exxon reported total earnings of about $30 billion in 2010. In March 2010 the Paris Court of Appeal decided that Total was not liable for civil damages
following the Erika’s wreckage and pollution, in 1999. Since the 1984 Bhopal accident, legal procedures have been ongoing.
C. Carpentier, J.-M. Suret / Journal of Environmental Economics and Management 71 (2015) 1–18
5
result from the present value of long-term uncertain fines, which are generally small relative to the firm’s earnings. Because
an unusual government action induces real, significant and unexpected costs, we posit that:
Hypothesis 4. The negative mid-term effect following an accident announcement on market value is stronger for accidents
followed by government intervention.
Other accident characteristics
Several characteristics of the firm and of the accident explain the differences between the short-term abnormal returns
following the announcement date (Capelle-Blancard and Laguna, 2010; Ho et al., 2013). These characteristics can also
influence the mid-term effect, but this dimension remains unexplored. Moreover, the characteristics evidenced in shortterm analysis are likely to differ along the main dimensions of our study: for example, fatalities are generally lower for
environmental accidents. We thus complement our analysis of abnormal returns by groups with a multivariate crosssectional model that explains the differences in estimated compounded abnormal returns at the firm level. The variables
described below will be considered control variables in the multivariate empirical tests.
Accidents seem to have a proportionately lower impact for large firms than for smaller ones, probably because large
firms have more resources to face an accident and are generally more diversified than smaller firms (Capelle-Blancard and
Laguna, 2010). Moreover, investor sentiment has a greater effect on smaller firms than on bigger ones (Kaplanski and Levy,
2010, p. 186). The number of fatalities due to the accident, as reported in the days immediately following the event, is
generally positively associated with market value decreases (Knight and Pretty, 1999; Walker et al., 2006; Capelle-Blancard
and Laguna, 2010; Ho et al., 2013). Environmental accidents generally cause fewer fatalities than other accidents. Media
coverage is likely to influence consumers’ and investors’ reaction to the event (Capelle-Blancard and Laguna, 2010). The
relationship between media coverage and market reaction is complex, and both are probably jointly determined by some
characteristics of the accident (Laguna, 2010). In the mid-term, heavily covered accidents are likely to be associated with
larger decreases in market value.
Profitability and previous reputation can also explain some differences between abnormal returns. Our sample is made
up of highly profitable firms (Exxon) and of distressed firms. We assume that the market reaction associated with an
accident is negatively related to firm profitability. A profitable firm can face the costs, fines, loss of clients and reputation
more easily than a distressed one. According to Oberndorfer et al. (2013), a good reputation is a valuable and rare intangible
resource. However, we do not have clear expectations about the link between a firm’s reputation in terms of safety and
environmental awareness and the market reaction to a new accident. The market value decrease should be lower for firms
with bad reputations if the probability of an accident is already incorporated in the risk assessment by the market. The
reaction should be stronger if some large investors revise their firm’s risk assessment upward.
Data
Accident definition
The International Labour Office (1991) defines an industrial accident as “an unexpected, sudden occurrence including, in
particular, a major emission, fire or explosion, resulting from abnormal developments in the course of an industrial activity,
leading to a serious danger to workers, the public or the environment, whether immediate or delayed, inside or outside the
installation and involving one or more hazardous substances. At the very least, it causes substantial injuries, deaths, and
financial costs, as well as serious damage to your reputation.” An aviation accident is defined by the Convention on
International Civil Aviation3 as an “occurrence associated with the operation of an aircraft (…) where a person is fatally or
seriously injured, the aircraft sustains damage or structural failure or the aircraft is missing or is completely inaccessible.” In
addition, something is usually considered to be a major accident if and only if it attracts serious media attention (Shrivastava
et al., 1988). Because such accidents are most likely to have a persistent effect on stock prices, we limited our analysis to this
class of events. We study environmental and non-environmental major accidents.
Event selection
In line with the above definitions, we included only accidents triggered by a physical break, as opposed to events arising
from rumors, fraud or other forms of misconduct. There is no perfect indicator to assess if an accident presents serious
damage to a firm’s reputation. Fatalities can be low in major events (Exxon Valdez), and the costs are generally not specified
in the days following the accident. To be able to select only those accidents that were considered significant by the
community, and to build a relatively homogenous list of events, we have included only those that were reported on the front
page of the New York Times. To compile our sample, we have examined the front pages of this newspaper from January 1959
to December 2010 in detail, using the Wall Street Journal when the Times was on strike. Our selection criteria ensure that
3
Convention on International Civil Aviation: International Standards and Recommended Practices—Aircraft Accident and Incident Investigation, Annex
13, available at http://www.iprr.org/manuals/Annex13.html.
6
C. Carpentier, J.-M. Suret / Journal of Environmental Economics and Management 71 (2015) 1–18
the financial community at large was directly aware of these accidents, given that they appeared on the front page of what it
considers a very influential newspaper. We originally collected 286 accidents. We reduced this list to 247 events by
excluding redundancies and events that were more local to the New York City community, and one case of clear overlapping
of an accident and an exogenous event.4 We dropped 77 events involving private firms or organizations not traded on an
American stock market.5 We also excluded 9 cases of sabotage, hijacking or product tampering, because such events cannot
be influenced by firms. Accordingly, we cannot expect that the market will punish them. Our final list of 161 events
constitutes Appendix A.
Our list includes many of the best known accidents, including the Pan Am/KLM airplane crash in the Canary Islands
(1977); Three Mile Island (1979); the chemical spill in Mississauga, Ontario (1979); the Bhopal tragedy (1984); the ExxonValdez oil spill (1989), and the BP Deepwater Horizon accident (2010). There is no clear-cut distinction between
environmental and non-environmental industrial accidents. Any accident could have a negative effect on the environment,
including an airline crash. However, such an event is generally not considered an environmental accident because of the
small affected area and the temporary effect of the accident. We use the threshold provided by the European Council
Directive (96/82/EC) on the control of major-accident hazards for environmental accidents notifiable to the European
Commission to identify accidents that should be considered environmental. An environmental accident implies a ‘dangerous
substance’ release, and the list of these substances is reported in Schedule 1 of the directive. For air pollution, we consider
significant toxic release based on the quantities of the accidental release. Schedule 6 defines what constitutes an “immediate
damage to the environment.” This includes permanent or long-term damage to terrestrial habitats, to freshwater and
marine habitats or to an aquifer or underground water. We use the area threshold provided in this schedule to determine, in
each case, if the event could be considered as having a significant environmental effect.6 Thirty-eight of the 161 accidents
are considered environmental.
We assume that the market can be immediately informed by media other than newspapers. Accordingly, the event date
is the day of the accident. The only exception is when the accident occurs at a time the US stock market is closed. In such a
case, we use the first trading day following the accident as the event date.
Accidents and firm characteristics
We measure firm size as market capitalization before the accident announcement, expressed in 2010 dollars using the
change in the S&P 500 total index as a deflator. The Log is used to normalize the distribution. We determine the number of
fatalities from the newspaper article reporting the event. We code a binary variable for firms operating in the air
transportation industry. We collected the accounting data from the Research Insight database for 1979 and the following
years, and from Moody’s publications for the previous years. For each firm, we identify the financial year that covers the
largest part of twelve months following the accident.7 We collected the data required to estimate the accounting rate of
return on equity based on net earnings before extraordinary items and ordinary shareholders’ equity.
We assume that a firm has a bad reputation if it has reported one (two) accident(s) in the previous twelve (36) months.8
The bad reputation category includes two firms whose bad reputation was signaled before the accident by the safety rating
provided by specialized sites and found in the press releases available on Factiva. We collected the number of articles
dedicated to each accident in the New York Times, in the two years following the accident. We screened each article to check
if it mentioned the accident and not only the firm. We restricted our compilation to the New York Times because we focus
on press coverage that can indeed influence investors. Moreover, covering a large set of newspapers is not realistic if one
wants to ascertain that each article specifically covers the accident.
Table 2 shows the summary statistics of the whole sample (Panel A) and by environmental issue (Panels B and C). Tests of
differences between both groups are reported in Panel D. As a result of our selection criteria, large firms are involved in our
sample. The median firm market value expressed in 2010 US dollars is about $7 billion, and a significant difference exists
between the firms involved in environmental accidents (median of $20.6 billion) and non-environmental accidents ($6.4
billion). Both groups also differ in terms of profitability: firms involved in environmental accidents are more profitable than
the other firms. Fatalities are significantly lower in environmental accidents. The median number of mentions of the
accident in the reference newspaper is the same range in both groups. However, the average numbers differ significantly
(42.68 vs. 10.78). The much higher coverage of environmental accidents reflects the fact that the worst events, such as Three
Mile Island, Bhopal, Exxon Valdez and BP Deepwater, fall into this category. Overall, those were the only four events in our
4
An accident at Exxon occurred a few days before the intervention of the US Department of Energy at the beginning of the 1979 petroleum crisis, shortly
after the first shock.
5
We use American depositary receipt (ADR) returns, when available, for non-US firms whose stocks are not traded in the US, and we assess the effect of
this choice on our results by a robustness check. The ADR is a negotiable security that represents securities of a non-US firm that trades on US financial
markets.
6
In six cases of blast or fire, we cannot easily determine if the accident can be considered environmental based on available information. These events
are classified as non-environmental in the reported results. When these accidents are considered environmental, we get similar (not reported) results.
7
For fiscal years ending in December, we use the current fiscal year if the accident occurs before June 30 and the next fiscal year for any accident
occurring between July 1 and the end of December.
8
We cannot use the rankings of corporate social responsibility or voluntary participation in programs related to safety or the environment because such
rankings or programs have emerged only in recent decades.
C. Carpentier, J.-M. Suret / Journal of Environmental Economics and Management 71 (2015) 1–18
7
Table 2
Descriptive statistics of the major accidents. M$2010 means millions of 2010 US dollars.
Mean
Median
Sum
Minimum
Maximum
24,999.77
43.44
23.60%
0.40%
8.07%
5.59%
18.31
19.88%
57.14%
7,427.38
15
0
9.84%
0
0
6
0
1
4024,963.17
6,994
38
–
13
9
2,948
32
92
82.10
0
0
590.41%
0
0
1
0
0
442,325.36
550
1
100.00%
1
1
624
1
1
Panel B: environmental accidents (38 accidents)
Market capitalization (in M$2010)
62,071.84
Fatalities
19.76
Return on equity
11.22%
Government intervention
15.79%
Takeover
2.63%
Number of press releases
42.68
Bad Reputation
10.53%
1959–1985
57.89%
20,573.47
0
12.98%
0
0
7.5
0
1
2358,729.93
751
–
6
1
1,622
4
22
155.71
0
21.82%
0
0
1
0
0
279,150.77
410
54.19%
1
1
624
1
1
Panel C: non-environmental accidents (123 accidents)
Market capitalization (in M$2010)
13,546.61
Fatalities
50.76
Return on equity
3.99%
Government intervention
5.69%
Takeover
6.50%
Number of press releases
10.78
Bad reputation
22.76%
1959–1985
56.91%
6,383.49
27
9.17%
0
0
6
0
1
1666,233.24
6,243
–
7
8
1,326
28
70
82.10
0
590.41%
0
0
1
0
0
442,325.36
550
100.00%
1
1
138
1
1
Panel A: whole sample (161 accidents)
Market capitalization (in M$2010)
Fatalities
Environmental accidents
Return on equity
Government intervention
Takeover
Number of press releases
Bad Reputation
1959–1985
Panel D: p-value of the Student’s t-test (Wilcoxon two-sample test) of the difference between the mean (median) of the distributions for
environmental accidents (Panel B) and non-environmental accidents (Panel C)
Market capitalization (in M$2010)
0.0016
o 0.0001
Fatalities
0.0237
o 0.0001
Return on equity
0.0166
0.0776
Government intervention
0.1186
0.0487
Takeover
0.2647
0.3693
Number of press releases
0.0957
0.2041
Bad reputation
0.0560
0.1021
1959–1985
0.9153
0.9169
sample to each generate more than 200 press releases. The proportion of firms with a bad reputation is lower in the
environmental accident group, reflecting the relative weight of air transportation firms in the other group and the high
frequency of airline crashes during the first decades of our analysis.
Methodological approach
Abnormal returns
No consensus exists on the best method to measure and test abnormal returns in the mid- and long-run. We estimate the
average compounded abnormal return using several benchmarks (the buy-and-hold approach), then we also apply the
calendar-time portfolio approach with Fama and French’s (1993) three-factor model and its four-factor extension proposed
by Carhart (1997), which includes an additional momentum-related factor. In line with Jenter et al. (2011), we use daily data
to be able to capture the short-term effect of the events. Even if we focus on mid-term effects, we cannot neglect the fact
that the market reaction is concentrated during the few days following the announcement. Using monthly data, as done in
most long-run analyses of new issues, can mask this effect and the associated volatility in returns.
The buy-and-hold abnormal return approach
The buy-and-hold abnormal return (BHAR) of an event firm i, BHARi.T, is the difference between its observed and
expected (benchmark) buy-and-hold returns. The buy-and-hold approach estimates the total return from a strategy where a
stock is purchased at closing market price on the day before the event date (t¼0) and held for T days (with T¼0 to 240),
against a benchmark. It is defined as follows, where ri,t is the return of the event stock i (with i ¼0 to 161) on day t and rb,t is
the corresponding return for the benchmark portfolio or stock. Because the non-event return cannot be observed, the BHAR
8
C. Carpentier, J.-M. Suret / Journal of Environmental Economics and Management 71 (2015) 1–18
cannot be determined but only estimated. Accordingly:
estðBHARi;T Þ ¼ ∏Tt ¼ 0 ð1 þr i;t Þ 1 ∏Tt ¼ 0 ð1 þr b;t Þ 1
ð1Þ
To aggregate the est(BHARi,T) across the event firms, the estimated average BHAR est ðBHART Þ is the mean of the est
(BHARi,T) for the firms i¼ 1, …, N over a period t¼0, …, T:
XN
wi;0 BHARi;T
ð2Þ
estðBHART Þ ¼
i¼1
PN
where wi,0 is 1/N when abnormal returns are equally weighted, and wi;0 ¼ MV i;t ¼ 1 = i ¼ 1 MV i;t ¼ 1 when they are valueweighted, with firm market values (MVi) estimated at the end of the trading day before the announcement (t¼ 1). The
distribution of BHAR is found to be skewed and the hypothesis that the estimated average BHAR differs statistically from
zero is based on the Johnson skewed-adjusted t-test and on the non-parametric Wilcoxon signed-rank test as suggested by
Ang and Zhang (2015).
Benchmarks
Long-term abnormal returns are very sensitive to the choice of benchmark. Simulations based on large samples indicate
the superiority of the matching firm approach, where each event firm is matched with a non-event firm that shares the risk
characteristics of the event firm, namely size, book-to-market ratio and past return. However, Barber and Lyon (1997)
evidence that matching firm approach can be less powerful than matching by portfolio approach: in small samples similar
to those used in the present study, the use of a matching firm proved to be a noisier way to control for the various risk
dimensions. This is why we also use two other benchmarks.
Matching firm approach. For each month, we formed 20 size portfolios from the CRSP database with an equal number of
firms in each portfolio, based on the market value of equity. For each month and each size portfolio, we then formed five
portfolios based on the ratio of book value of common equity to market value of common equity at the end of the previous
fiscal year. Each of the 100 portfolios has an equal number of firms. We assigned to each event firm its corresponding size
and book-to-market portfolio. The matching firm is the firm in the corresponding portfolio that has the closest prior-sixmonth raw returns to the event firm. We excluded our accident-sample firms from the matched sample for the two years
before and after the event.
Matching by portfolio approach. The return of a portfolio that matches the event firm on size (market equity) and book-tomarket provides the benchmark. The reference portfolios, constructed at the end of each June, are the intersections of 10
portfolios formed on size and 10 portfolios formed on the ratio of book-to-market. We match each event firm with its
corresponding portfolio, and we use the daily returns provided on K. French’s website as a benchmark.
Matching by industry approach. The accidents are associated with a few industries that have evolved very differently during
the 50 years analyzed. The air transportation industry was affected by deregulation (from 1978), a price war (1985–1992)
and the 9/11 attacks. Petroleum firms have been affected by the evolution of oil prices and the 1973 and 1979 oil crises. For
years, the average returns in these industries differed greatly from those of the global index. Sector could be a risk factor that
is not captured by the matching firm approach or by the classical matching by portfolio approach. Major accidents can also
have positive or negative spillover effects on a sector as a whole, which we need to control. Accordingly, we use industry
indices as a benchmark, whereas previous studies are based on a global stock market index, generally the S&P 500. Because
such indices are not available from 1959 to 2011, we constructed our own purged daily weighted indices, based on all firms
included in the CRSP database in each industrial sector. We use these industrial returns as a reference portfolio in the BHAR
approach. Controls for industry-specific factors are also used in several studies in this area (Salinger, 1992; Tibbs et al., 2011;
Sabet et al., 2012).
The calendar time portfolio approach
For each calendar day of the 240 days of the event windows, an event portfolio is formed, consisting of all firms that have
experienced an accident. Day zero for a given stock of the sample of 161 event firms is the event date. Daily return of the event
portfolio is computed as the equally weighted average of daily returns of all firms in the portfolio. Excess returns of the event
portfolio are regressed on the Fama–French daily three factors. Following Tibbs et al. (2011), we estimate the following regressions
obtained with equally weighted portfolios and weighted least square estimations,9 and we test for the significance of the alpha.
Rp;t Rf ;t ¼ αp3 þ βp ðRm;t Rf ;t Þ þ sp SMBt þ hp HMLt þ ep;t
ð3Þ
Rp;t Rf ;t ¼ αp4 þ βp ðRm;t Rf ;t Þ þ sp SMBt þ hp HMLt þ mp UMDt þ ep;t
ð4Þ
9
The weights are proportional to the square root of the number of firms present in each calendar day to such that days with more events are weighted
more heavily. We thus deal with potential heteroskedastic residuals induced by calendar clustering (Ang and Zhang, 2015).
C. Carpentier, J.-M. Suret / Journal of Environmental Economics and Management 71 (2015) 1–18
9
The dependent variable of the regression is the daily excess return of the portfolios (Rp,t Rf,t), which corresponds for a
given day t (with t ¼0 to 240) to the returns of the portfolio of firms affected by an accident (Rp,t) less the risk-free rate
(the daily rate of 91-day Government Treasury bills, Rf,t). The independent variables are the excess market return and two
zero-investment portfolios constructed to mimic the risk factors common to all securities. We use the (size) Small Minus Big
(SMBt) and (book-to-market ratio) High Minus Low (HMLt) factors in keeping with Fama and French (1993). βp, sp, hp
represent the loadings of the portfolio on each risk factor: the market (Rm,t-Rf,t), size and book-to-market ratio.
The parameter (αp3) in Eq. (3) indicates the daily average abnormal return of our accident firms for the three-factor model.
In Eq. (4), Up Minus Down (UMDt) is the performance of high prior return stocks relative to low prior return portfolios. The
momentum factor was proposed by Carhart (1997). (αp4) in Eq. (4) represents the abnormal performance of the portfolio
when the four factors are considered. These factors come from K. French’s website.
Other methodological dimensions
To detect the simultaneous related or unrelated events occurring within the event window, we check the data published by
Value Line, Moody’s, Factiva and the New York Times for the year following each accident. We consider the two oil crises, the
9/11 attack and the subprime crisis major unrelated events that can influence the returns of every firm in our sample. For 13
firms, a second accident occurs during the year following the first accident and is also considered an unrelated event that can
influence the abnormal returns of that firm. Related events ensue from the accidents and include governmental and private
actions, such as takeover announcements following the stock price drop. These three groups of events are treated differently.
First, we delete the estimated abnormal returns for 60 days surrounding the major unrelated events. The results reported are
not affected by this correction, probably because the number of such crises during the years following the accidents is small.
Second, in the case of recurring accidents, we restricted the analysis of the first accident to the period of time that elapsed until
the next one, to eliminate the influence of the second accident on the estimation of abnormal returns. Two methods exist to
deal with this situation. First, one can simply delete the observation for horizons longer than the trimming date. This reduces
the number of available observations at the end of the period under analysis. Second, one can consider that the abnormal
return is null after the trimming date. In this case, the estimated BHAR at the end of the period under analysis is equal to the
one estimated at the trimming date. We used both approaches to test the robustness of our results to this choice. Our results
are very close in each case. We report the results using the second approach because it provides more observations for the
multivariate analysis.10 A small number of firms included in our original sample delist from the stock market following an
acquisition or financial problems. We consider these cases similar to the accident overlap situations.
Third, we study two kinds of related events. The first consists of instances where the government acts in response to the
accident. For example, in several cases of airplane crashes, the Federal Aviation Administration had grounded some firms or
types of planes, incurring high costs for the firms. A dummy variable is set to 1 when such an intervention occurs. Hypothesis 4
is specifically dedicated to the effects of this type of action. When equity value decreases sharply following an accident, the
firm can become a takeover target. We set a dummy variable to 1 for the 9 cases of takeover rumors or decisions occurring
during the year following the studied accident. We also check for the effect of these announcements in the robustness section.
We discuss the results obtained when all accidents are given equal weight in the empirical tests. Because our sample
includes several of the largest firms in the world together with small firms, value-weighted analysis implies exclusion of the
smaller firms from the analysis. We report some value-weighted results in the section devoted to robustness.
Results
Abnormal returns
Table 3 presents the estimated average compounded abnormal returns and their significance tests, for the two methods
and the various benchmarks. Our results are robust to the methodological choices, including benchmark definition, and the
parametric and non-parametric tests provide similar levels of significance. We base our discussion on the Johnson t-tests for
the BHAR approach. Whole sample results (Column 1 of Table 3) indicate that firms underperform significantly for the
period beginning on the event date and ending 240 trading days later, with an estimated average BHAR of 5.71%
(p-value ¼0.0448) using the matching firm approach, and an estimated average BHAR of 9.05% (p-value ¼0.000) using the
matching by industry approach. The estimated average compounded abnormal return is in the same range for each method,
including the calendar time portfolio approach. The p-values indicate significance at the 5% level or less in each case. This
first set of results confirms our first hypothesis that the accidents are followed by significant abnormal returns.
In columns 2 and 3 of Table 3, the sample is split according to the environmental dimension of the accident. Nonenvironmental event firms underperform significantly for the period beginning on the announcement date and ending 240
trading days later, with an estimated average BHAR of 8.69% (p-value ¼0.0083) using the matching firm approach. In
contrast, the market performance of firms involved in environmental accidents does not differ significantly from zero. All
10
Results are available upon request. We replicate the whole study without any trimming for accident overlap and get similar results. Our conclusions
are robust to the method used to control for the accident overlaps.
10
C. Carpentier, J.-M. Suret / Journal of Environmental Economics and Management 71 (2015) 1–18
Table 3
Long-run abnormal returns of accident firms. This table reports estimated average equally weighted BHARs and calendar time portfolio approach over the
240 trading days following the event date. Matching firms-adjusted compounded abnormal returns are calculated as the difference between the accident
firms’ buy-and-hold return and the matched firms buy-and-hold returns. We match firms according to size, book-to-market and prior-six-month raw
returns. Estimated industry-adjusted compounded abnormal returns are calculated as the difference between the accident firms’ buy-and-hold return and
the buy-and-hold return of a value-weighted industry index. Estimated portfolio-adjusted compounded abnormal returns are calculated as the difference
between the accident buy-and-hold return and the buy-and-hold return from an equally weighted portfolio matched on size and book-to-market. The
estimated average abnormal returns in Panel D are obtained by regressing the excess returns of the event portfolio on the Fama-French three factors,
namely the excess market return and the zero-investment portfolios constructed to mimic the size and the book-to-market factors. In Panel E, the
momentum factor is added to the model. The p-values are based on the non-parametric Wilcoxon signed-rank test for non-zero median (non-parametric),
on the Johnson t-tests adjusted for skewness for non-zero mean differences (Johnson test) and on Student’s t-tests (t-test). The full sample includes 161
accidents from 1959 to 2010. Env means environmental, Gov Int. means government intervention.
Whole sample
Non-Env.
Env.
Airline
Non-Env. non-airline
Gov Int.
No Gov Int.
13
148
21.7630
0.0942
0.0608
4.2987
0.3147
0.1444
19.9794
0.0803
0.1010
8.0958
0.0000
0.0000
Panel C Buy and hold benchmark approach, Portfolio-adjusted matched returns, Fama–French reference portfolios
Estimated average BHAR, in %
5.2327
8.2285
4.4639
10.7607
2.5641
p-Value (non-parametric)
0.0150
0.0006
0.1100
0.0010
0.3954
p-Value (Johnson test)
0.0294
0.0058
0.2504
0.0108
0.3556
15.9535
0.1909
0.1932
4.2911
0.0341
0.0762
Panel D Calendar time portfolio approach Factor model, 3 factors Fama French
Annualized abnormal return, in %
5.8438
10.0586
5.4114
p-Value (t-test)
0.0141
0.0004
0.1546
16.1680
0.0000
1.2712
0.7424
18.6077
0.0213
4.5645
0.0679
Panel E Calendar time portfolio approach Factor model, 4 factors Fama French
Annualized abnormal return, in %
7.1521
11.2651
4.1166
p-Value (t-test)
0.0027
0.0001
0.2792
16.5054
0.0000
0.0530
0.9891
17.9097
0.0270
6.1600
0.0139
N
161
123
38
85
38
Panel A Buy and hold benchmark approach, Matching firms-adjusted compounded abnormal returns
Estimated average BHAR, in %
5.7088
8.6860
3.9277
11.4130
2.5861
p-Value (non-parametric)
0.1074
0.0173
0.3199
0.0146
0.6334
p-Value (Johnson test)
0.0448
0.0083
0.5209
0.0119
0.4742
Panel B Buy and hold benchmark approach, Industry-adjusted compounded abnormal returns
Estimated average BHAR, in %
9.0553
10.5014
4.3746
14.0531
p-Value (non-parametric)
0.0000
0.0000
0.2992
0.0000
p-Value (Johnson test)
0.0000
0.0000
0.2292
0.0000
2.5568
0.2140
0.3055
other methods/benchmarks provide consistent results that confirm our second hypothesis that the negative long-run effect
of an accident announcement on stock performance is lower for environmental than for non-environmental accidents. We
cannot exclude the possibility that the lesser impact of environmental accidents could be traced to the size of the affected
firms, or to lower fatalities; we will test this possibility using multivariate analysis. However, the most likely explanation is
that environmental accidents are less likely than airline crashes to induce a switch effect.
In columns 4 and 5, we present the results of the analysis for the subsamples of airline and non-environmental non-airline
(NENA) accidents. Airline crashes clearly drive the market reaction for non-environmental accidents, with an estimated average
BHAR of 11.41% (p-value¼0.0119) using the matching firm approach. We do not observe any significant underperformance of
firms involved in NENA accidents. All other methods or benchmarks provide consistent results. For airline accidents, the calendar
time portfolio approaches provide estimations of significant abnormal returns in the vicinity of 17%. These results confirm our
third hypothesis: that the negative mid-term effect of an accident on stock prices is stronger for airline than for non-airline
accidents. This may result from a change in the transportation habits of individual consumers, consistent with a reputational
effect at the consumer level, or to the behavioral explanation provided by Kaplanski and Levy (2010).
In columns 6 and 7, we study the impact of government intervention. The results should be analyzed with caution
because the sample of government interventions comprises only 13 observations. These interventions are associated with
an estimated average BHAR of 21.76% (p-value ¼0.0608) using the matching firm approach. The estimated average BHAR
does not differ from 0 for the group of events without an intervention. The level of significance of the tests for both groups is
sensitive to the estimation method. For government interventions, two of the approaches fail to provide a significant result,
and three other models provide p-values equal to or lower than 6%. On average, the long-run negative effects on market
value are stronger when the government acts in response to an event. However, the statistical test fails to be significant in all
cases, due to the inclusion of a strong outlier associated with a takeover, which we analyze in the robustness section.
Graphical illustration and cost estimations
The average abnormal returns discussed in the previous section provide a global picture, but do not illustrate the
evolution of the stock performance during the studied year. We illustrate these changes using equally weighted estimated
average BHAR computed using industrial indices. Fig. 1 shows the continuous decrease in firm market value for the whole
sample and for the non-environmental subsample. Environmental accidents are associated with a decrease in market value
that reverses after approximately 120 days. Fig. 2 illustrates the influence of the airline industry accident. Whatever the time
C. Carpentier, J.-M. Suret / Journal of Environmental Economics and Management 71 (2015) 1–18
Environmental
accidents
Non- environmental
accidents
11
All accidents
-2%
0
7
14
21
28
35
42
49
56
63
70
77
84
91
98
105
112
119
126
133
140
147
154
161
168
175
182
189
196
203
210
217
224
231
238
0%
-4%
-6%
-8%
-10%
-12%
Fig. 1. Estimated average BHAR of 161 major accidents, by environmental issue, 0 to 240 days. Estimated average BHAR are equally weighted and computed
using industrial indices.
Non- environmental
all accidents
Environmental
accidents
Airline
-2%
0
7
14
21
28
35
42
49
56
63
70
77
84
91
98
105
112
119
126
133
140
147
154
161
168
175
182
189
196
203
210
217
224
231
238
0%
-4%
-6%
-8%
-10%
-12%
-14%
-16%
Fig. 2. Estimated average BHAR of 161 major accidents, by environmental and airline issue, 0 to 240 days. Estimated average BHAR are equally weighted
and computed using industrial indices.
Government
intervention
all accidents
No
government
intervention
Government
intervention
airline
234
225
216
207
198
189
180
171
162
153
144
135
126
117
99
108
90
81
72
63
54
45
36
27
9
18
-5%
0
0%
-10%
-15%
-20%
-25%
-30%
-35%
-40%
-45%
Fig. 3. Estimated average BHAR of the 161 major accidents, by government intervention and airline issue, 0 to 240 days. Estimated average BHAR are
equally weighted and computed using industrial indices.
elapsed since the events, the average compounded abnormal return of airline accidents is below that of non-environmental
accidents, indicating that most of the estimated average BHAR observed in this subsample can be traced to airplane crashes.
Fig. 3 highlights the effect of government intervention depending on the sector. The worst abnormal returns are obtained in
the airline industry, when the government acts following an accident. The combined effect of both characteristics is huge,
with an average compounded abnormal return estimated at about 40%.
Even if some groups of accidents induce statistically non-significant abnormal returns, they can be associated with large
losses in value, likely to influence investors’ perception and, in turn, managers’ willingness to comply with safety and
12
C. Carpentier, J.-M. Suret / Journal of Environmental Economics and Management 71 (2015) 1–18
Table 4
Decrease in equity values following major accidents. For each firm, the loss associated with the accident is the product of the pre-accident market value and
the estimated compounded industry-adjusted abnormal returns over the 240 trading days following the event date. Estimated industry-adjusted
compounded abnormal returns are measured as the difference between the accident firm’s buy-and-hold return and the buy-and-hold return of a valueweighted industry index. Losses are in millions of 2010 US dollars. Env means environmental, Gov Int. means government intervention.
N
Total loss
Mean loss
Standard deviation
Number of losses
% of losses
Maximum loss
Whole sample
Non-Env.
Env.
Airline
161
267,548
1662
9,259
104
65
24,325
123
79,082
643
3,729
82
67
19,103
38
188,466
4960
17,615
22
58
24,325
85
45,737
538
1,250
61
72
3,309
Non-Env. non-airline
38
33,346
878
6,497
21
55
19,103
Gov Int.
13
119,072
9,159
25,750
9
69
11,510
No Gov Int.
148
148,476
1,003
5,812
95
64
24,325
environmental standards. In such a context, where investors’ perceptions and moods seem to be central (Kaplanski and
Levy, 2010), the analysis of the estimated losses is relevant. They are presented in Table 4. For each firm, the loss represents
the reduction in market value associated with the accident, when the changes in the industry index are considered.
To measure the loss, we multiply the pre-accident market equity value by the estimated BHAR after 240 trading days. We
calculate the total losses at $268 billion for the whole sample, expressed in 2010 dollars.11 The average market loss is
estimated to be about $1.7 billion per accident. Not all events are followed by losses. The proportion of losses related to the
number of events is 65% for the whole sample, but only 58% for environmental accidents. However, the average loss
following environmental accidents is more than seven times higher than those observed after non-environmental accidents.
Despite a stronger decline in stock prices, the losses are smaller following non-environmental accidents because the
involved firms are approximately one quarter of the size of those involved in environmental accidents (Table 2). The average
equity loss is nine times larger for accidents followed by government intervention than for other accidents. The small
sample of 13 accidents with government interventions induces a total cost of $119.07 billion, while the other 148 accidents
generate a total cost of $148.48 billion. This illustrates the strong effect of these interventions on shareholders’ wealth.
Complementary robustness checks
Table 5 presents three different robustness checks. First, we estimate average value-weighted BHAR (Panel A). This
weighting scheme gives more influence to events occurring in larger firms and produces larger and more significant
negative average abnormal returns. The estimated average BHAR for environmental accidents is significant and reaches
11.86% with the first approach (matching firm) and 19.75% with the second one (matching by industry). The other
methods of abnormal return estimation provide similar results (not reported). The comparison of Tables 3 and 5 indicates a
strong effect of event firm size on the results, particularly for environmental accidents. This can be traced to the fact that
several of the worst environmental accidents of the last half century are associated with very large market capitalizations,
such as Exxon (5 accidents) or BP (2 accidents). Owing to the mechanics of the value-weighting scheme combined with our
sample’s characteristics, the value-weighted results are mainly driven by a small subsample of very large firms. This is why
we mainly use equally weighted schemes in this study.
The small differences in prices between ADRs and stocks are not likely to affect returns in the long run (Gagnon and
Karolyi, 2010). However, the inclusion of these types of security deserves analysis. Our second robustness check consists in
measuring the average compounded abnormal returns without the ADRs (Panel B). For the events followed by government
intervention, this exclusion changes the estimated average BHAR, which becomes non-significant. Four of the 13 cases of
government intervention have targeted foreign firms that have listed ADRs.12 Removing these four observations, which
include the 2010 BP Oil Spill, changes the results for this small subsample. However, the effect of firm nationality appears to
be stronger than that of ADR characteristics.
Third, we exclude the 9 takeover targets (Panel C). Results are unchanged. These robustness analyses illustrate the
complexity of the interaction effects addressed in the section devoted to multivariate analysis.
Multivariate analysis of abnormal returns
We perform ordinary least square (OLS) regressions for estimated BHARs to determine if environmental and nonenvironmental accidents differ when the other characteristics of the events and of the firms are considered in the model
(Table 6). We have to deal with potential multicollinearity problems. For example, fatalities are lower and market
11
The estimation of the losses depends on the empirical method used to estimate the compounded abnormal returns, but our estimations with other
methods (not reported) produce results in the same range.
12
ADR firms are associated with 7% of our accident sample, but with 30.8% of government interventions. The question of the government’s reaction
depending on the nationality of the accident firm deserves attention, but has been left for further research.
C. Carpentier, J.-M. Suret / Journal of Environmental Economics and Management 71 (2015) 1–18
13
Table 5
Robustness check. Panel A reports estimated average value-weighted BHAR over the 240 trading days following the event date, when the abnormal returns
are weighted by the market value before the accident. Estimated matching firms-adjusted compounded abnormal returns are calculated as the difference
between the accident firms’ buy-and-hold return and the matched firms buy-and-hold returns. We match firms according to size, book-to-market and the
prior-six-month raw returns. Estimated industry-adjusted compounded abnormal returns are calculated as the difference between the accident firms’ buyand-hold return and the buy-and-hold return of a value-weighted industry index. Panel B reports estimated average equally weighted BHAR over the 240
trading days following the event announcement date, excluding ADRs. Panel C reports estimated average equally weighted BHAR over the 240 trading days
following the event date, including ADRs but excluding event firms that become a takeover target following the accident. The p-values are based on the
non-parametric Wilcoxon signed-rank test for non-zero median (non-parametric) and on the Johnson t-tests adjusted for skewness for non-zero mean
differences (Johnson test). The full sample includes 161 accidents from 1959 to 2010. Env means environmental, Gov Int. means government intervention.
Whole sample
Non-Env.
Env.
Airline
Panel A: value-weighted abnormal returns
N
161
123
38
85
Buy and hold benchmark approach, Matching firms-adjusted compounded abnormal returns
Estimated average BHAR, in %
9.7868
6.7104
11.8591
9.9550
p-Value (Johnson test)
0.0000
0.0032
0.0073
0.0171
Buy and hold benchmark approach, Industry-adjusted compounded abnormal returns
Estimated average BHAR, in %
12.9382
2.8297
19.7479
8.7518
p-Value (Johnson test)
0.0000
0.0845
0.0000
0.0043
Panel B: equally weighted abnormal returns, without ADRs
N
149
116
33
79
Buy and hold benchmark approach, Matching firms-adjusted compounded abnormal returns
Estimated average BHAR, in %
5.6191
8.3258
3.8955
10.9963
p-Value (non-parametric)
0.0561
0.0134
0.5628
0.0190
Buy and hold benchmark approach, Industry-adjusted compounded abnormal returns
Estimated average BHAR, in %
8.8131
10.7119
2.1386
14.3411
p-Value (non-parametric)
0.0000
0.0000
0.5614
0.0000
Panel C: equally weighted abnormal returns, without takeover
N
152
115
37
78
Buy and hold benchmark approach, Matching firms-adjusted compounded abnormal returns
Estimated average BHAR, in %
6.1154
8.6621
1.8003
11.0060
p-Value (non-parametric)
0.0296
0.0095
0.7705
0.0192
Buy and hold benchmark approach, Industry-adjusted compounded abnormal returns
Estimated average BHAR, in %
9.9072
11.4023
5.2602
15.0096
p-Value (non-parametric)
0.0000
0.0000
0.1482
0.0000
Non-Env. non-airline
38
Gov Int.
13
No Gov Int.
148
5.6883
0.0690
25.4085
0.0002
1.2495
0.5350
0.9640
0.6705
29.3547
0.0024
3.9667
0.0007
37
9
140
2.6238
0.4799
16.7826
0.2418
4.9014
0.1042
2.9631
0.2472
13.5255
0.3420
8.5102
0.0000
37
12
140
3.7210
0.2895
26.8605
0.0200
4.3372
0.1338
3.7978
0.0639
25.2575
0.0420
8.5914
0.0000
capitalization is higher for environmental accidents than for other accidents, which is why we present several versions of
the model. We also removed two outliers and checked our results with robust regression methods. Our sample presents
several industry and time clusters, because airline accidents occur mainly during the first decades of the studied period. In
such a pattern, the residuals may be correlated across sectors or across time, and OLS standard errors can be biased
(Petersen, 2009). To overcome this problem, we calculated two-way clustered standard errors based on industry and decade
(Petersen, 2009; Gow et al., 2010). We present the results of different models explaining BHAR estimated with the matching
by industry approach.13 Our models compare well with the other cross-sectional studies of long-run abnormal returns
following major events.14 The adjusted R2 of our models stands between 11.95% and 15.16%.
Several accidents have very few or no fatalities, but others caused hundreds of deaths. The distribution of fatalities is thus
highly asymmetric; we have tested several functional forms for this explanatory variable. None of these forms provide a
significant relationship between the number of deaths and the estimated BHAR. We report the results using a Log
transformation (model 1) and a dummy that takes a value of 1 for the 10th percentile of the death distribution, beginning
with 111 fatalities (models 3 and 4). We do not detect any significant relation even when the number of press releases is
excluded from model 1 (not reported). Neither do we detect a significant effect of firm size on the estimated BHAR. The
firm’s reputation before the event is also not significantly associated with the BHAR.
We show that the market effect of the accident does not differ according to the environmental dimension when media
coverage, firm size and fatalities are considered. When introduced, the dummy associated with government intervention
plays a highly significant role in the model, and its coefficient is negative ( 16%), which is consistent with the differences
reported in Table 3. The two-way clustered standard errors based on sector and decade cannot be estimated in this case
because some clusters are empty due to the low number of government interventions. The coefficient of the dummy
associated with the environmental dimension is not significant in any model. When the other characteristics of the
accidents and of the firms are considered, environmental accidents do not differ from other accidents. However, the dummy
13
All reported results have been tested for the undesirable effect of collinearity using the procedures described by Belsley et al. (2005).
In their analysis of the consequences of corporate misconduct, Murphy et al. (2009) report adjusted R2 ranging from 2.4% to 20%. Those reported by
Andre et al. (2004) following mergers are lower than 7%.
14
14
C. Carpentier, J.-M. Suret / Journal of Environmental Economics and Management 71 (2015) 1–18
Table 6
Determinants of the long-run estimated abnormal returns following major accidents. The dependent variable is the estimated BHAR over 240 trading days.
Estimated BHARs are computed using industrial indices. Very large accidents is a dummy set to 1 if the number of deaths is equal or higher than the 90th
percentile of distribution (111). Nb means number. We use White corrected standard errors (t stat). CL-2 t statistic is the regression coefficient scaled by the
coefficient standard error corrected with industry and decade clustering (Petersen, 2009). The full sample includes 161 accidents from 1959 to 2010. We
removed two leveraged points, and four observations are missing due to unavailable return on equity, which reduces the number of observations to 155.
Model 1
Intercept
Market cap. (K$2010, Log)
Fatalities (Logn)
Very large accidents
Return on equity
Gov. intervention
Takeover
Bad reputation
Nb of press releases (Log)
Environmental accident
Airline
Number of observations
Adjusted R square
p-Value
Model 2
Model 3
Model 4
estimated
parameter
CL-2 t stat
estimated
parameter
t stat
estimated
parameter
t stat
estimated
parameter
CL-2 t stat
0.01629
0.00401
0.00284
0.07
0.28
0.56
0.10274
4.92 nnn
0.09407
0.00872
0.53
0.79
0.14591
0.01016
0.62
0.77
0.28171
0.16239
0.21994
0.04272
2.81 nnn
2.04 nn
1.94 n
0.79
0.09168
0.27576
0.15253
0.23102
0.02417
1.30
2.80 nnn
2.05 nn
2.29 nn
0.45
0.05983
0.28275
1.35
7.62 nnn
0.01165
0.08543
155
0.1516
o .0001
0.29
2.14 nn
0.24391
0.02777
0.02655
0.00276
0.07996
155
0.1364
0.0002
2.45 nn
0.73
2.18 nn
0.08
2.72 nn
0.28683
0.23915
0.04419
0.03452
0.04074
155
0.1195
0.0005
6.10
nnn
2.43 nn
1.07
3.12 nnn
0.82
nnn
0.03619
155
0.1313
o .0001
0.96
nnn
nnn
nnn
n
Denote statistical significance at the 10% level.
Denote statistical significance at the 5% level.
nnn
Denote statistical significance at the 1% level.
nn
associated with the airline industry is significant (model 4). This is consistent with our analysis of the estimated average
BHAR (Table 3) where neither environmental nor NENA accidents are followed by significant market reactions. This confirms
that airline accidents induce a different market effect, when all other characteristics are included in the model.
The difference in media coverage significantly explains the differences observed in estimated BHAR. As expected, the
return on equity positively influences the BHAR and the coefficient is strongly significant. This indicates that major events
are likely to have the worst consequences in financially weak firms. The dummy takeover is positively and significantly
associated with the estimated compounded abnormal return. The multivariate analysis confirms that when the other
variables are kept constant, environmental accidents differ significantly from airline crashes.
Conclusion
This paper examines the mid-term stock market effect of 161 major accidents occurring during the last half-century.
Thirty-eight accidents are considered environmental. Our results indicate that the market reacts negatively and enduringly
to an announcement of an accident. The estimated average compounded abnormal return of the firms involved in large
accidents reach 5.23% to 9.05% after one year, depending on the estimation method. On average, major accidents reduce
firms’ market value by about $1.7 billion. The total decrease in the market capitalization of the event firms is approximately
$268 billion. Such effects can constitute a real deterrence effect, even if shareholders invest with a long horizon. This result
contrasts with the non-significant effect of negative events after one year reported in previous studies. This can probably be
traced to the selection of the worst accidents of the last half-century.
The overall result should be qualified. The average compounded abnormal return estimated for the sample of
environmental accidents does not differ from zero, except when we use value-weighted returns that increase the influence
of a small group of very large firms. Overall, our results are consistent with the proposition that, in the mid-term, the
deterrence effect of the stock market for environmental problems is weak. In fact, the average effect of major accidents is
largely associated with two sub-categories of events: airplane crashes and accidents followed by strong government
intervention. In the airline industry, an accident induces a large negative return and significant average compounded
abnormal returns, reaching 17% with some estimation models. Non-environmental non-airline accidents are followed by
non-significant abnormal returns. Overall, we find no evidence of a strong market deterrence effect, except for the airline
industry and for events that prompt government reaction.
The proposition that the stock market community motivates managers to enhance the safety of their products or of their
production processes is confirmed, but only if one considers industries as a whole. Left alone and without government
intervention, stock markets generally do not punish the stockholders of firms involved in environmental accidents. This is
perhaps one of the explanations for the exponential growth in man-made disasters since 1970 documented by Coleman
(2006).
C. Carpentier, J.-M. Suret / Journal of Environmental Economics and Management 71 (2015) 1–18
15
Appendix A
See Table A1.
Table A1
List of major industrial accidents and main features.
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
Event date
Firm name
Type of accident
No. of fatalities Dummy¼ 1 if environmental accident Industry
1959-09-30
1960-03-18
1961-09-18
1962-11-23
1962-12-03
1963-02-12
1963-06-03
1963-12-09
1964-02-25
1964-07-10
1964-11-23
1965-02-09
1965-08-17
1965-09-17
1965-11-09
1965-11-12
1965-12-06
1966-08-08
1967-03-09
1967-03-30
1967-06-26
1968-05-06
1968-10-28
1968-12-13
1968-12-26
1969-01-07
1969-01-20
1969-09-09
1970-06-22
1970-09-09
1971-01-11
1971-01-25
1971-02-03
1971-06-08
1971-09-07
1972-03-06
1972-05-02
1972-10-30
1972-12-11
1972-12-21
1973-01-02
1973-02-20
1973-06-04
1973-06-11
1973-07-23
1973-07-24
1973-07-31
1973-08-14
1973-08-17
1973-11-05
1974-01-18
1974-01-31
1974-04-23
1974-09-11
1975-02-27
1975-06-25
1976-01-05
1976-04-27
1976-08-30
1976-11-22
1977-03-28
1977-03-28
1977-04-25
Braniff Airways
Northwest Airlines
Northwest Orient Airl
United Air Lines
Eastern Air Lines
Northwest Airlines
Northwest Airlines
Pan Am World Awys
Eastern Air Lines
United Air Lines
Trans World Airl.
Eastern Air Lines
United Air Lines
Pan Am World Awys
American Airlines
United Air Lines
Eastern Air Lines
Braniff Airways
Trans World Airl.
Delta Air Lines
Mohawks Airlines
Braniff Airways
Northeast Airlines
Pan Am World Awys
Allegheny Airlines
Allegheny Airlines
United Air Lines
Allegheny Airlines
Witco Chemical corp.
Transamerica Corp.
American Airlines
Standard Oil
Thiokol Chem corp.
Allegheny Airlines
Alaska Airlines
Mohawks Airlines
Sunshine Mining
Illinois Cent. Ind.
UAL
Delta Airlines
Eastern Airlines
Consol Ed. NY
Imperial Oil
Penn Central Co.
Pan Am.
Ozark Airlines
Delta Airlines
Mallinckrodt Chem
Consol Ed. NY
Pan Am
Shell Oil
Pan Am
Pan Am
Eastern Airlines
AT&T
Eastern Airlines
Exxon
American Airlines
Atlantic Richfield
Warner Lambert
Pan Am
KLM
Phillips Petroleum
Air crash
Air crash
Air crash
Air crash
Air crash
Air crash
Air crash
Air crash
Air crash
Air crash
Air crash
Air crash
Air crash
Air crash
Air crash
Air crash
Air crash
Air crash
Air crash
Air crash
Air crash
Air crash
Air crash
Air crash
Air crash
Air crash
Air crash
Air crash
Refinery fire
Air crash
Air crash
Oil spill
Blast at arms plant
Air crash
Air crash
Air crash
Mine fire
Train crash
Air crash
Air crash
Air crash
Blackout
Ship crash
Train crash
Air crash
Air crash
Air crash
Chemical plant explosion
Dam break
Air crash
Ship crash
Air crash
Air crash
Air crash
Fire
Air crash
Oil tank explosion
Air crash
Atom waste blast
Factory blast
Air crash
Air crash
Oil slick
33
63
37
17
25
43
101
81
58
39
50
84
30
30
58
43
4
42
25
19
34
85
32
51
20
11
38
83
3
11
2
0
24
28
111
17
91
44
45
10
99
0
5
1
78
38
89
2
2
3
16
97
107
70
0
113
0
37
0
0
550
248
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
1
0
0
1
1
0
0
0
0
0
0
0
0
0
1
0
0
0
0
1
0
0
1
0
0
0
0
0
1
0
0
0
0
0
1
Airline
Airline
Airline
Airline
Airline
Airline
Airline
Airline
Airline
Airline
Airline
Airline
Airline
Airline
Airline
Airline
Airline
Airline
Airline
Airline
Airline
Airline
Airline
Airline
Airline
Airline
Airline
Airline
Chemical
Airline
Airline
Petrol
Chemical
Airline
Airline
Airline
Others
Railroad-Transp.
Airline
Airline
Airline
Electricity
Petrol
Railroad-Transp.
Airline
Airline
Airline
Chemical
Electricity
Airline
Petrol
Airline
Airline
Airline
Others
Airline
Petrol
Airline
Petrol
Chemical
Airline
Airline
Petrol
16
C. Carpentier, J.-M. Suret / Journal of Environmental Economics and Management 71 (2015) 1–18
Table A1 (continued )
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
Event date
Firm name
Type of accident
No. of fatalities Dummy ¼1 if environmental accident Industry
1977-07-15
1978-03-01
1978-03-17
1978-05-09
1978-08-01
1978-09-25
1978-12-29
1979-01-08
1979-03-28
1979-04-09
1979-05-25
1979-10-31
1979-11-12
1979-11-12
1980-03-27
1980-07-28
1981-02-11
1981-02-13
1981-04-08
1982-02-16
1982-07-13
1983-01-07
1983-01-24
1984-07-24
1984-10-08
1984-12-03
1984-12-21
1985-08-05
1985-08-12
1986-01-06
1986-05-19
1987-08-17
1988-01-04
1988-02-22
1988-05-04
1988-07-07
1988-09-01
1989-02-24
1989-03-28
1989-07-19
1989-09-21
1989-10-23
1989-12-29
1990-01-02
1991-02-04
1991-03-04
1991-04-08
1991-07-31
1992-01-20
1992-03-23
1992-07-20
1992-11-03
1992-12-03
1992-12-21
1993-09-22
1994-01-07
1994-03-02
1994-03-24
1994-07-05
1994-09-09
1994-11-01
1994-12-14
1995-02-16
1995-08-21
1995-08-28
1995-12-20
1995-12-21
1996-02-20
1996-05-13
1996-07-08
Consol Ed. NY
Continental Airlines
Exxon
National Airlines
Occidental Pet.
P S A Inc.
UAL Inc.
Gulf Oil
GPU
Seaboard Coast Line
American Airlines
Western Airlines
Canadian Pacific
Dow Chemical
Phillips Petroleum
IC Industries
Hilton
Ralston Purina
Consol Ed. NY
Mobil Corp
Pan Am
Texaco
Greyhound
Unocal
American Cyanamid
Union Carbide
Utah Power & Light
Delta Airlines
Union Carbide
Kerr McGee
Norfolk Southern Corp
N W A Inc
Ashland Oil Inc
American Airlines
PEPCON
Occidental Pet.
Delta Airlines
U A L Corp.
Exxon
U A L Corp.
US Air Group
Phillips Petroleum
Consol Ed. NY
Exxon Corp
U.S. Air Group
United Airlines
Atlantic Southeast Airlines
C S X Corp
Peoples Energy
US Air Group
Textron Inc
Consol Ed. NY
Repsol S A
KLM
C S X Corp
Exxon
Continental Airlines
Panhandle Eastern
US Air Group
US Air Group
American Airlines
A M R Corp
Consol Ed. NY
Atlantic Southeast
Textron
Tower Air
A M R Corp
C S X Corp
Valuejet
Delta Airlines
Power failure
Air crash
Ship crash
Air crash
Toxic release
Air crash
Air crash
Tanker blast
Radiation release
Chemical spill
Air crash
Air crash
Toxic release
Toxic release
Oil platform collapse
Chemical car derailing
Fire
Chemical leak
Chemical spill
Oil-drilling rig sinking
Air crash
Tank explosion
Bus crash
Explosion
Leak of noxious fumes
Gas leak
Fire
Air crash
Toxic cloud leak
Nuclear leak
Train derailing
Air crash
Oil spill
Air crash
Plant explosion
Oil rig blast
Air crash
Air crash
Tanker oil spill
Air crash
Air crash
Factory blast
Blast
Oil leak
Air crash
Air crash
Air crash
Train crash
Gas explosion
Air crash
Air crash
Blast
Tanker oil spill
Air crash
Train wreck
Oil spill
Air crash
Pipeline explosion
Air crash
Air crash
Air crash
Air crash
Cable fire
Air crash
Air crash
Air crash
Air crash
Train crash
Air crash
Air crash
0
2
0
1
0
144
10
50
0
0
273
73
0
0
120
0
5
0
0
84
153
1
3
16
0
410
27
135
0
1
4
156
0
18
2
167
13
9
0
111
2
23
2
0
34
25
23
8
4
27
7
2
0
56
40
0
0
1
37
132
68
15
1
8
5
0
160
12
110
2
0
0
1
0
1
0
0
1
1
1
0
0
1
1
0
1
0
1
1
0
0
1
0
0
1
1
0
0
1
1
0
0
1
0
1
1
0
0
1
0
0
1
0
1
0
0
0
0
0
0
0
0
1
0
0
1
0
1
0
0
0
0
0
0
0
0
0
0
0
0
Electricity
Airline
Petrol
Airline
Petrol
Airline
Airline
Petrol
Electricity
Railroad-Transp.
Airline
Airline
Railroad-Transp.
Chemical
Petrol
Railroad-Transp.
Others
Others
Electricity
Petrol
Airline
Petrol
Railroad-Transp.
Petrol
Chemical
Chemical
Electricity
Airline
Chemical
Chemical
Railroad-Transp.
Airline
Petrol
Airline
Chemical
Petrol
Airline
Airline
Petrol
Airline
Airline
Petrol
Electricity
Petrol
Airline
Airline
Airline
Railroad-Transp.
Electricity
Airline
Others
Electricity
Petrol
Airline
Railroad-Transp.
Petrol
Airline
Electricity
Airline
Airline
Airline
Airline
Electricity
Airline
Others
Airline
Airline
Railroad-Transp.
Airline
Airline
C. Carpentier, J.-M. Suret / Journal of Environmental Economics and Management 71 (2015) 1–18
17
Table A1 (continued )
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
Event date
Firm name
Type of accident
1996-07-18
1996-11-25
1997-01-10
1997-07-31
1997-12-29
1998-06-22
1999-03-16
1999-06-02
2000-02-01
2001-03-19
2001-11-12
2002-04-19
2002-07-22
2003-01-29
2004-04-19
2004-12-22
2005-03-24
2005-08-03
2006-01-03
2007-07-19
2008-02-08
2008-09-15
2009-02-13
2009-06-01
2010-02-08
2010-04-05
2010-04-21
2010-09-10
Trans World Airlines
Air crash
C S X Corp
Train crash
Comair Holdings
Air crash
Federal Express
Air crash
United Airlines
Air accident
Greyhound
Bus crash
CSX Corp
Train crash
American Airlines
Air crash
Alaska Group
Air crash
CSX Corp
Train crash
A M R Corp
Air crash
CSX Corp
Train crash
Consolidated Ed NY
Power outage
West Pharma. Services Inc Blast
CSX Corp
Train crash
Newmont Mining
Toxic release
BP
Refinery explosion
Air France
Airplane burning
I C O Inc
Blast
Consolidated Ed NY
Steam pipe explosion
Imperial Sugar
Refinery blast
Union Pacific
Train crash
Continental Airlines
Air crash
Air France
Air crash
Siemens
Blast
Massey Energy
Blast
BP
Oil spill
Pacific Gas & Electric
Blast
No. of fatalities Dummy¼ 1 if environmental accident Industry
220
1
29
0
1
7
13
9
88
1
265
6
0
3
0
0
15
0
13
1
13
18
49
228
5
25
11
4
0
0
0
0
0
0
0
0
0
0
0
0
0
1
0
1
1
0
0
0
1
0
0
0
0
0
1
1
Airline
Railroad-Transp.
Airline
Others
Airline
Railroad-Transp.
Railroad-Transp.
Airline
Airline
Railroad-Transp.
Airline
Railroad-Transp.
Electricity
Others
Railroad-Transp.
Others
Petrol
Airline
Chemical
Electricity
Others
Railroad-Transp.
Airline
Airline
Others
Electricity
Petrol
Electricity
References
Ambec, S., Lanoie, P., 2008. Does it pay to be green? A systematic overview. Acad. Manage. Perspect. 22 (4), 45–62.
Andre, P., Kooli, M., L’her, J.-F., 2004. The long-run performance of mergers and acquisitions: evidence from the Canadian stock market. Finan. Manage. 33
(4), 27–43.
Ang, J., Zhang, S., 2015. Evaluating long-horizon event study methodology. In: Lee, C-F., Lee, J.C. (Eds.), Handbook of Financial Econometric and Statistics,
Springer, pp. 383–411.
Barber, B.M., Lyon, J.D., 1997. Detecting long-run abnormal stock returns: the empirical power and specification of test statistics. J. Finan. Econ. 43 (3),
341–372.
Belsley, D.A., Kuh, E., Welsch, R.E., 2005. Regression Diagnostics. John Wiley & Sons, Inc..
Borenstein, S., Zimmerman, M.B., 1988. Market incentives for safe commercial airline operation. Am. Econ. Rev. 78 (5), 913–935.
Bosch, J.-C., Eckard, E.W., Singal, V., 1998. The competitive impact of air crashes: stock market evidence. J. Law Econ. 41 (2), 503–519.
Broder, I.E., Morrall, J.F., 1991. Incentives for firms to provide safety: regulatory authority and capital market reactions. J. Regul. Econ. 3 (4), 309–322.
Capelle-Blancard, G., Laguna, M.-A., 2010. How does the stock market respond to chemical disasters? J. Environ. Econ. Manage. 59 (2), 192–205.
Carhart, M.M., 1997. On persistence in mutual fund performance. J. Finance 52 (1), 57–82.
Coleman, L., 2006. Frequency of man-made disasters in the 20th century. J. Contingencies Crisis Manage. 14 (1), 3–11.
Dasgupta, S., Laplante, B., Mamingi, N., 2001. Pollution and capital markets in developing countries. J. Environ. Econ. Manage. 42 (3), 310–335.
Dempsey, P.S., 1991. The disintegration of the US airline industry. Transp. Law J. 20 (1), 9–46.
Ehrhardt, M., Brigham, E., 2013. Corporate Finance: A Focused approach. Cengage Learning.
Engelen, P.-J., van Essen, M., 2011. Reputational penalties in financial markets: an ethical mechanism?. In: Vandekerckhovee, W., et al. (Eds.), Responsible
Investment in Times of Turmoil, Springer, pp. 55–74.
Evans, A.D., 2009. A requiem for the retail investor? Virginia Law Rev. 95 (4), 1105–1129.
Fama, E.F., French, K.R., 1993. Common risk factors in the returns on stocks and bonds. J. Finan. Econ. 33 (1), 3–56.
Faure, M., 2002. Environmental damage insurance in theory and practice. In: Swanson, T. (Ed.), Research in Law and Economics, vol. 20. , Emerald Group
Publishing Limited, pp. 283–328.
Gagnon, L., Karolyi, A., 2010. Do international cross-listings still matter. In: Beck, T., Schmukler, S., Claessens, S. (Eds.), Evidence on Financial Globalization
and Crises, Elsevier North-Holland Publishers.
Gaspar, J.-M., Massa, M., Matos, P., 2005. Shareholder investment horizons and the market for corporate control. J. Finan. Econ. 76 (1), 135–165.
Gow, I.D., Ormazabal, G., Taylor, D.J., 2010. Correcting for cross-sectional and time-series dependence in accounting research. Acc. Rev. 85 (2), 483–512.
Hamilton, J.T., 1995. Pollution as news: media and stock market reactions to the toxics release inventory data. J. Environ. Econ. Manage. 28 (1), 98–113.
Harper, R.K., Adams, S.C., 1996. CERCLA and deep pockets: market response to superfund program. Contemp. Econ. Policy 14 (1), 107–115.
Ho, J.C., Qiu, M., Tang, X., 2013. Do airlines always suffer from crashes? Econ. Lett. 118 (1), 113–117.
International Labour Office, 1991. Prevention of Major Industrial Accidents.
Jenter, D., Lewellen, K., Warner, J.B., 2011. Security issue timing: what do managers know, and when do they know it? J. Finance 66 (2), 413–443.
Jones, K., Rubin, P.H., 2001. Effects of harmful environmental events on reputations of firms. Adv. Finan. Econ. 6, 161–182.
Kaplanski, G., Levy, H., 2010. Sentiment and stock prices: the case of aviation disasters. J. Finan. Econ. 95 (2), 174–201.
Karpoff, J.M., Lott, J.R., Wehrly, E.W., 2005. The reputational penalties for environmental violations: empirical evidence. J. Law Econ. 48 (2), 653–675.
Khanna, M., Quimio, W.R.H., Bojilova, D., 1998. Toxics release information: a policy tool for environmental protection. J. Environ. Econ. Manage. 36 (3),
243–266.
Klassen, R.D., McLaughlin, C.P., 1996. The impact of environmental management on firm performance. Manage. Sci. 42 (8), 1199–1214.
Knight, R.F., Pretty, D.J., 1999. Corporate catastrophes, stock returns, and trading volume. Corp. Reputation Rev. 2 (4), 363–398.
18
C. Carpentier, J.-M. Suret / Journal of Environmental Economics and Management 71 (2015) 1–18
Konar, S., Cohen, M.A., 1997. Information as regulation: the effect of community right to know laws on toxic emissions. J. Environ. Econ. Manage. 32 (1),
109–124.
Kothari, S.P., Warner, J.B., 1997. Measuring long-horizon security price performance. J. Finan. Econ. 43 (3), 301–339.
Laguna, M.-A., 2010. Unexpected Media Coverage and Stock Market Returns: Evidence from Chemical Disasters (Working Paper). Université Paris Dauphine
(Available at).
Laplante, B., Lanoie, P., 1994. The market response to environmental incidents in Canada: a theoretical and empirical analysis. South. Econ. J. 60 (3),
657–672.
Lempert, R., 2009. Low Probability/High Consequence Events: Dilemmas of Damage Compensation. University of Michigan Legal Working97 (Available at).
Mitchell, M.L., Maloney, M.T., 1989. Crisis in the cockpit—the role of market forces in promoting air travel safety. J. Law Econ. 32, 329–355.
Murphy, D., Shrieves, R.E., Tibbs, S.L., 2009. Understanding the penalties associated with corporate misconduct: an empirical examination of earnings and
risk. J. Finan. Quant. Anal. 44 (1), 55–83.
Nethercurtt, L.L., Pruitt, S.W., 1997. Touched by tragedy: capital market lessons from the crash of Valujet Flight 592. Econ. Lett. 56 (3), 351–358.
Oberndorfer, U., Schmidt, P., Wagner, M., Ziegler, A., 2013. Does the stock market value the inclusion in a sustainability stock index? An event study analysis
for German firms. J. Environ. Econ. Manage. 66 (3), 497–509.
Petersen, M.A., 2009. Estimating standard errors in finance panel data sets: comparing approaches. Rev. Finan. Stud. 22 (1), 435–480.
Richardson, B.J., 2001. Mandating environmental liability insurance. Duke Environ. Law Policy Forum 12, 293–319.
Sabet, S.A.H., Cam, M.-A., Heaney, R., 2012. Share market reaction to the BP oil spill and the US government moratorium on exploration. Aust. J. Manage. 37
(1), 61–76.
Salinger, M., 1992. Value event studies. Rev. Econ. Stat. 74 (4), 671–677.
Servaes, H., Tamayo, A., 2013. The impact of corporate social responsibility on firm value: the role of customer awareness. Manage. Sci. 59 (5), 1045–1061.
Shrivastava, P., Mitroff, I.I., Miller, D., Miclani, A., 1988. Understanding industrial crisis. J. Manage. Stud. 25 (4), 285–303.
Spudeck, R.E., Moyer, R.C., 1989. A note on the stock market’s reaction to the accident at three mile island. J. Econ. Bus. 41 (3), 235.
Tibbs, S.L., Harrell, D.L., Shrieves, R.E., 2011. Do shareholders benefit from corporate misconduct? A long-run analysis. J. Empir. Legal Stud. 8 (3), 449–476.
Walker, T.J., Pukthuanthong, K., Barabanov, S.S., 2006. On the stock market’s reaction to major railroad accidents. J. Transp. Res. Forum 45 (1), 23–39.
Walker, T.J., Thiengtham, D.J., Lin, M.Y., 2005. On the performance of airlines and airplane manufacturers following aviation disasters. Can. J. Admin. Sci. 22
(1), 21–34.
White, M.A., 1996. Investor Response to the Exxon Valdez Oil Spill. University of Virginia.
Purchase answer to see full
attachment