referee report

User Generated

xreeltbfxl

Economics

Description

We will present two papers each class, two reports of the two papers are due at
the start of each class. All the reports should be printed, one and half spacing, no more than two
pages. In the report, you should include the following sections:
• Motivation
• Research question
• Background and literature
• Data
• Methodology
• Findings and policy implication
• Your comments and critiques
Use your own understanding to write the report, do not copy and paste.


In addition, I need an insightful question about this paper,

Unformatted Attachment Preview

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
User generated content is uploaded by users for the purposes of learning and should be used following Studypool's honor code & terms of service.

Explanation & Answer

hello t...


Anonymous
Nice! Really impressed with the quality.

Studypool
4.7
Trustpilot
4.5
Sitejabber
4.4

Related Tags