Shareholders and the environment: a review of four decades of academic research

We provide a synthesis of four decades of empirical research regarding the reaction of shareholders to environmental events. This literature is at the crossroads of finance, environmental economics, management and corporate social responsibility (CSR). To set the stage, we first provide an account of the Brumadinho ecological disaster that occurred in Brazil on January 25th, 2019. Second, we provide a critical review of more than 100 event studies. These papers cover a diverse set of events, such as industrial accidents, public disclosure programs, legal actions following environmental violations, changes in environmental regulation, environmental news, and corporate initiatives. This review makes four contributions. First is the synthesis of a large strand of literature in a structured setting, so as to be readily handled by both experts and non-experts. Second is the observation that stock market penalties in the event of environmental concerns are likely to be quite low: on average there is a (temporary) drop in the excess stock market return to events that are harmful to the environment of about 2% and the median is −0.6%. Third is to highlight the limits of CSR as a business strategy towards a sustainable society. Fourth is to provide an open access bibliographic database.


Introduction
Over the last four decades, environmental policy and regulation underwent a complete revolution under the aegis of environmental economists. The paradigm shift has consisted of 'harnessing market forces to protect the environment' , to use the title of a 1989 report (Stavins 1989). The origin of this report was a bipartisan initiative from the US Congress entitled Project 88, which gathered 50 people from academia, industry, government and environmental non-governmental organizations (NGOs), and was directed by Robert Stavins. The ambition was 'to find innovative solutions to major environmental (…) problems' (Stavins 1989, p 5); Cropper and Oates (1992) consider this moment as 'the most exciting time (-and perhaps a critical juncture-in the evolution of economic incentives for environmental protection' (p 676). Not only does this seminal report warn against climate change with a strikingly visionary perspective, it also paves the way 'towards a new area' of environmental policies based on market-oriented instruments: 'Although conventional regulatory policies have often worked well, they have also tended to pit economics and environmental goals against each other. These goals should complement one another in the long run' (Stavins 1989, p 34). This report was not the first to highlight the potential of market incentives, but until then, these ideas were mostly confined to academics. The report was embraced by most decision-makers. In its wake, Thomas Tietenberg summarized the policy context as follows: 'The change in attitude has been triggered by a recognition that this former adversary, the market, can be turned into a powerful ally' (Tietenberg 1990, p 17). He made this view the pillar of his textbook on natural resource and environmental economics that has been used to train legions of students around the world (Tietenberg and Lewis 2018).
More than 30 years later, these ideas still play a major role in environmental protection policies 4 . In addition, the debate continues. In a column published in 2019 targeted at young climate activists, Daron Acemoglu uses precisely the same argument: 'Markets need not stand in our way. On the contrary, they could be a powerful ally' (Acemoglou 2019). Actually, it is striking to see this debate reappear, whereas it seemed settled for good. In this regard, Joseph Stiglitz provides a meaningful snapshot of the issues at stake. He argues that between 1989 and 2019 we went from 'The End of History' to 'An Age of Discontent' (Stiglitz 2019). Indeed, after the initial promising experiences, there was disappointment regarding the capacity of markets to address environmental problems and to ensure social welfare. For Stiglitz, it is no longer a question of markets being the solution to environmental problems, but of environmental problems helping us to become aware of the problems posed by free markets: 'If the 2008 financial crisis failed to make us realize that unfettered markets do not work, the climate crisis certainly should' (Stiglitz 2019). Moreover, it should be noted that the reference to the financial crisis is anything but trivial; it has reinforced distrust of anything that might look like market's control over society (Zingales 2015). At this stage, it is important to note that most environmental economists who advocate market-based instruments are anything but blind supporters of free markets. It is quite the opposite, since the idea is to use market schemes to address market failures. Along the same lines, discontent about the market does not mean that markets should be abolished: 'We need to exploit the benefits of markets while taming their excesses, making sure that markets work for people, and not the other way around' (Stiglitz 2019).
The aim of this paper is to question the place and the role of financial markets today in the framing of environmental policy. This requires a better understanding of the relationship between shareholders and the natural environment. Many current environmental policies are based on Pigouvian taxes (a tax aimed at the producer of polluting products) or tradeable permits, without necessarily having a direct link with financial markets. These instruments already are well investigated (see Bergquist et al 2013, Schmalensee and Stavins 2019, Aldy et al 2020. In contrast, despite there being numerous studies regarding the financial markets' response to environmental events, there is no generally accepted understanding of the reaction of shareholders, the owners of the companies' stock, to these events. Therefore, we feel it timely and important to consolidate this research, to specify the underlying mechanisms, and to provide a meaningful interpretation of the results. In this regard, we review four decades of academic research devoted to the reaction of shareholders to environmental news. Be aware that our study is not a conventional meta-analysis as we want to provide an assessment of how and why academic research in this specific arena did develop over the course of four decades. Thus, ours is a synthesis of the literature and not a meta-analysis. To set the stage, we start with an account of the Brumadinho ecological disaster of January 25, 2019. The rupture of a Brazilian mining dam, operated by the multinational company Vale, resulted in the death of hundreds of people and the pollution of an immense territory. The example illustrates the problematic relationship between environment, firms and shareholders. It highlights the limits of the voluntary approach to environmental protection, as well as how financial markets actually value environmental damage. Next, we detail the relationship between environmental events and the response from shareholders. To this extent, we focus on event studies, which are highly homogeneous in terms of their methodology (Salinger 1992, MacKinlay 1997, Kothari and Warner 2006. We compose a narrative of the empirical literature based on event studies to assess the impact of environmental events on firms' stock market value. We document more than 100 papers over a period of four decades. Here, the aim is to explore whether it is wise to entrust financial markets with the responsibility of disciplining companies for their environmental performance. We highlight three streams in the evolution of academic research: accidents, regulation, and social responsibility. The streams cannot be clearly differentiated from one each other, but the focus of attention shifts in each of them. The first stream of studies accompanies the occurrence of several dramatic industrial accidents. These studies sit in the finance literature driven by the efficient market hypothesis (EMH), which assumes that all new information is immediately reflected in asset prices. They provide a first estimate of the losses incurred by shareholders. Most of the literature from the 1980s can be grouped in this stream. In the second stream, it is predominantly environmental economists that contribute to and in fact take over the analysis of the potential impact of environment incidents. They propose to use market forces to incentivize firms to protect the environment. This literature suggests that the traditional 'command-and-control' instruments should be complemented by market-based incentives when public disclosure is warranted. Hence, it is up to the shareholders to sanction firms, instead of public authorities. The law and economics literature shows that in some cases market penalties may be higher than fines. Most of the studies in this stream appear in the 1990s. The third stream takes off in the new millennium. Now, the focus is more business-oriented as the concept of market efficiency gives room to that of corporate social responsibility (CSR), and the tone of the debate becomes more positive. Further, there seemingly are less dramatic accidents than previously occurred, and the studies extend to all kinds of events that could have an environmental impact. The research in this stream also benefits from the advent of big data techniques and from reporting and monitoring innovations.
Our review of the academic literature on the interaction between the natural environment and shareholders highlights the intellectual processes by which environmental policy has moved from a regulatory approach, based on norms and taxes, to a market driven perspective that is based on the following conjecture: disciplining firms is not (only) the task of governments, but more broadly that of stakeholders (consumers, employees, creditors, shareholders). This conjuncture is based on two presuppositions: (a) government policy is prone to failures; (b) stakeholders have effective tools for action. We feel this approach provides a better understanding of how the literature regarding the stock markets' response to environmental events is framed and how it has evolved over time. While the environmental economics literature initially was motivated by the concept of market failure, recent approaches have been mainly justified by government failure. What is at stake is the balance between the two: market failures that legitimise public intervention, particularly in the area of environmental protection, are well known. These are mainly: public goods, market control, externalities, and imperfect and/or asymmetric information. Government failures are also well identified. These range from opportunism and short-termism on the part of elected representatives, lack of financial or human resources on the part of regulatory authorities, to pressure from vested interests. This is complemented by globalisation of supply chains as well as the lack of cooperation between states, which weakens the attempt of domestic regulation. Faced with this, it is easy to understand the temptation to involve other stakeholders. As nowadays reputation is one of the key assets of companies, anything that can damage this reputation can be used as a means of pressure-and therefore as a means of action. Boycotts and 'naming & shaming' campaigns can effectively dissuade consumers from buying a company's products. A good reputation can encourage employees or potential recruits to seek a position in a more responsible company, and to guide the capital of investors.
However, to be effective, information must be reliable, valid, relevant, timely, transparent, and public. Currently, it seems this goes missing in the case of most non-financial information as adequate reporting, monitoring and auditing standards are still emerging and evolving. Then, how to assure that the information is useful for all participants? Should disclosure of information be voluntary, encouraged, constrained? Should formats be harmonized or left to competition? This also raises questions about the actual effectiveness of environmental policy based on such disclosures. Are stakeholders capable of assimilating and assessing firms' non-financial conduct? Are the stakeholders (consumers, employees, investors) representative of the general interest? Do stakeholders indeed use their leverage to punish non-virtuous companies and to promote others? The answers to these questions are important for the analysis of environmental policies and strategies. They cover a broad spectrum of theoretical and empirical studies, to which the event studies synthesized in this paper are closely linked.
The key issue addressed in event studies is the penalty imposed by shareholders (or, where applicable, the premium granted) regarding non-financial corporate conduct. Our synthesis reveals that the penalty is often quite low (on average, the short-term response seems to be about −2% but with wide fluctuation variation). More importantly, we want to point out that such a penalty is far from sufficient to discipline firms. Apart from a few extreme cases, the market value of firms is only slightly affected by the public dissemination of environmental information. As such, it is highly unlikely that investors bring about the changes that would align corporate conduct with concepts as planetary boundaries or sustainable development goals (see also Bebbington et al 2020).
In addition, our analysis provides an alternative interpretation of CSR. At first glance, since the ambition is to temper shareholder primacy, it is tempting to interpret CSR as a way to curb the race for short-term profits, to dampen the almighty power of multinational corporations, and ultimately to soften the neoliberal model. However, it seems that CSR does not oppose the pro-market ideology, but rather complements it (e.g. Kinderman 2012, Roulet andTouboul 2015). In this paper, we interpret CSR as the belief in the proper functioning of markets and that markets can be entrusted with the responsibility to sanction and reward. In this regard, we find support for the notion of 'delegated penalty' , in resonance with the concept of 'delegated philanthropy' (Bénabou and Tirole 2010). With delegated philanthropy, CSR spending represents firms' investments to the benefit of stakeholders that would otherwise be paid by the shareholders in full. However, with delegated penalty, CSR represents an assessment of firms' social and environmental policy by shareholders that would otherwise be made by public authorities. Overall, we conclude that the market disciplining approach to environmental protection is limited and we advocate a critical look at the societal value of CSR practices. As such, our conclusions are in line with those of Freeman and Liedtka (1991, p. 93), who warn against a CSR concept that would 'promot[e] incompetence by leading managers to involve themselves in areas beyond their expertise-that is repairing society's ill' 5 .
The paper is organized as follows. Section 2 details the example of the Brumadinho accident. Section 3 provides a typology of environmental event studies and briefly presents our narrative. Furthermore, this section informs about how to account for the impact of environmental events on shareholders, the conceptual underpinnings and the methodology of event studies and the sources used to detect the events. The following sections discuss the empirical papers. Section 4 focuses on high-profile industrial accidents and their impact on shareholder value. Section 5 investigates with the stock market response to policy events. Section 6 addresses the impact of CSR events on shareholder value. Section 7 concludes the paper. An open access bibliographic database is provided as supplementary material online. It holds all the information used in the paper and allows researchers to work with it 6 .

Environmental disaster, stock markets, and deterrence
On 25 January 2019, a terrifying mudslide, launched at over 70 km h −1 , carried 13 million cubic meters of toxic industrial waste (equivalent to 5000 Olympic swimming pools). It resulted from the rupture of the Brumadinho dam in Brazil. This accident killed more than 250 people (as of January 2020) 7 . It was both an ecological disaster and a human tragedy. The images immediately made it to social networks and 24/7 news channels. There is nothing natural about this disaster: It is the direct result of the mining activities of the Brazilian multinational company, Vale, the world's largest producer of iron ore. The Brumadinho dam disaster vividly illustrates the current societal and academic debates about CSR. We use it to describe the complex relationships between environment, firms and shareholders. The accident displays the limitations of the market disciplining approach to environmental protection, as well as how financial markets value environmental damage. We conclude that the 5 It should be noted that Milton Friedman himself considers that if the social responsibility of managers is to make profits, their action must be bound by rules: the responsibility of corporate executives is 'to make as much money as possible while conforming to their basic rules of the society, both those embodied in law and those embodied in ethical custom' (Friedman 1970). 6 https://adriendesroziers.com/data-and-databases/. 7 www.wsj.com/articles/brazils-vale-vowed-never-another-damcollapse-then-an-even-worse-one-11577809114.
Brumadinho dam disaster casts doubt on the societal value of CSR practices.
To what extent is Vale responsible for the accident and its consequences? According to its president, Fabio Schvartsman, Vale followed the safety recommendations of international experts 8 : 'i am not a mining technician. I followed the technicians' advice and you see what happened. It did not work (..). We are 100% within all the standards, and that did not work.' Schvartsman also promises 'to go beyond any national or international standards.' This seems a highly challenging promise: should we trust companies that promise to go beyond international standards? Would it not be more appropriate to tighten the standards and to enforce them? 'To go beyond the standard' is exactly what Project 88 (see previous section) considered 30 years before this accident as a novel tool to protect the environment: 'The report's recommendations are designed to increase environmental protection and economic productivity by providing incentives for businesses and individual to go beyond what regulators can require' (Stavins 1989, p 2).
Like all multinationals, Vale publishes annual CSR reports. In 2017 9 , Vale prides itself on being in full compliance with the global reporting initiative, supported by the United Nations Global Compact programme. The report also highlights the protection of 8500 km 2 of land around production sites, the recycling of 82% of its wastewater, and several partnerships with indigenous populations. However, such efforts appears meaningless in the perspective of the tragic human loss, the extent of the area ravaged by the mudslide, the contamination of the Paraopeba River, which feeds about 50 towns in the state of Minas Gerais downstream and over a million inhabitants, and the irreparable damage to the ecosystem on which the Pataxó tribe depends 10 . The Brumadinho event challenges the reputation of the concept of CSR. Encouraging companies to be more concerned about stakeholders is obviously commendable as this allows for a broader interpretation of value maximization (Hart and Zingales 2017). However, in practice, CSR too often is an instrument of greenwashing, which casts doubt on the role of CSR in the arsenal of social and environmental measures (Lyon and Maxwell 2011). Further, it is unlikely that shareholders are able and willing to act in the interest of stakeholders (Tirole 2006).
The Brumadinho disaster is not the first accident which involves Vale. It echoes a similar tragedy that occurred a 100 km away at the Samarco mineco-owned by Vale and BHP, an Australian mining company-just three years before. Then, toll was 'only' 19 deaths, with several million cubic meters of mud submerging the local town of Mariana and dumped over 650 km into the Atlantic Ocean 11 . This accident was, until then, considered the most serious ecological accident in Brazil's history. When Mr. Schvartsman was appointed chief executive at Vale in May 2017, he mentioned the tragedy to the staff 12 : 'We must all adopt a motto: 'Mariana never again' (…). This is the last time that this company is involved, directly or indirectly, in an ecological and social disaster on the scale of Mariana' . Legal proceedings were initiated against the firm and its managers, but Vale still denies the toxic nature of thegigantic mudslide. Vale reached an agreement with the government for a fine of around 6.8 billion reais ($1.8bn) 13 . CSR might compensate for the weaknesses in regulations and controls, and it might be that financial markets discipline companies. Several scholars have advocated this view to legitimize CSR, particularly in developing countries (for example, Lanoie et al 1998, Dasgupta et al 2001. The idea is appealing: if regulatory bodies cannot effectively control and sanction companies, they should ensure that disciplining comes from the market. In this logic, it is up to shareholders-who are supposed to have more influence, incentives, and impact than public authorities-to put pressure on company managers to care about environmental and social consequences of their activities. However, this idea only works if the market sanctions are strong enough. So, what about the stock markets' response to the Brumadinho dam disaster? This is reflected in figure 1, which shows the closing price of Vale's deposit receipts at the NYSE and the transaction volume in 2019. Vale is cross-listed on several stock exchanges: São Paulo, New York, Madrid and Paris. On the NYSE, following the accident, the stock lost 8% in market value on January 25, and 18% on January 28. Figure 1 also shows the transaction volume is seven times higher than usual. The loss for shareholders due to the accident is estimated around 60 billion reals. This loss is supposed to take into account the costs that the company will incur. Six months after the accident, Vale publishes its financial results with a 7% increase of the total turnover ($9 bn). This increase was due to a surge in the price of iron (+51% Q2:2019) caused by a shortage of supply, which in fact resulted from the Brumadinho dam disaster. In July, Vale's stock price returned to the same price range as it was before the disaster. From this, we conclude that the Brumadinho accident did not substantially or structurally affect the financial value of Vale or the wealth of its shareholders. Only those who sold their shares in the wake of the disaster will have been likely to have made a loss.
The state's justice system in Minas Gerais immediately blocked 11 billion reals ($2.9bn) in Vale's accounts, half of which was to compensate victims and the other half to cover the cost of environmental damages 14 . The cost to the company is of course higher, as it has to include legal costs, and the firm might also increase the safety of its other sites, etc. However, is this sufficient to compensate for the loss of all stakeholders, and to restore the environmental damage? Capelle-Blancard and Laguna (2010) suggest that market sanctions are too low. In particular, they observe sanctions are far from posing a serious threat to the managers, do not allow for full compensation of the damage to the planet, and do not provide sufficient incentives to improve the safety of industrial sites in most cases. In addition, there is no resolution mechanism regarding the transfer of the market value lost to relatives and communities and to restore natural habitats. Because the effects are external to the firm and its owners, the private and social accounts are unbalanced.
The Brumadinho dam disasters highlights the issues at stake. An international company, complying with the highest standards and committed to social responsibility, faces a major accident with more than 250 casualties and vast and heavy pollution. The authorities imposes a fine, the shareholders face a (temporary) loss on their wealth. But after a couple of months, it is business as usual.

Four decades of research
In the past four decades, a large number of event studies on environmental events have been undertaken, and we establish that this literature is quite homogenous. We identify 139 studies with a very steady publication rate, as can be seen in figure 2 15 .
Although there are no important methodological differences between the studies, the nature of the events investigated has substantially changed over the period studied. For the purpose of our 14 www.reuters.com/article/us-vale-sa-disaster-damages/brazilcourt-freezes-further-13-billion-in-vale-accounts-over-damburst-idUSKCN1PL0IJ. 15 We have searched for studies that use event study methodology to assess the impact of environmental news; therefore, we have excluded studies, which consider other methodologies to examine the relationship between CEP and CFP, and studies which consider other extra-financial issues, not strictly environmental. The search for articles was carried out using the usual search engines in economics (Econlit, SSRN, JSTOR, ScienceDirect), successively with the JEL codes G14 and Q5, or the keywords 'event studies' , 'environment' , 'industrial accidents' , 'environmental disclosures' , 'environmental regulation' or 'CSR news' . We have also used the references of the selected articles. The list of papers (with the abstract and its DOI) is available in the Online Appendix. narrative, we draw a typology based on three different types of negative ('eco-harmful') or positive ('ecofriendly') events: industrial accidents with environmental concerns (high profile single-event studies or studies with a sample of accidents), environmental regulation (public disclosures, legal actions following environmental violation, and (changes in) environmental policy-including events related to new emission trading schemes), and CSR news (various environmental news, corporate practices). We use this categorization to detail the empirical findings and analysis. We follow a chronological reading of the literature to discuss how the debate has evolved over time. As such, we put the development of this event study literature in a socio-economic and academic context.
In this regard, table 1 provides a condensed overview of the main categories of events studied, the main papers, as well as the main academic outlets. Initially, studies simply show that firms causing environmental disasters encounter substantial financial losses to their shareholders. Since the 1980s, dozens of articles have examined the consequences of the main industrial catastrophes, whether it was nuclear accidents Three Mile Island (TMI, Chernobyl, Fukushima), oil spills (Exxon Valdez, BP DeepWater Horizon), or workplace accidents (Bhopal, Rana Plaza, Brumadinho). In general, it shows that the initial market response is substantial. Therefore, it is no surprise to find that it was mainly finance scholars who took up this issue (e.g. Bowen et al 1983, Hill andSchneeweis 1983), often influenced by the efficient market theory, which was at its zenith at that time. Most of these papers are case studies, but there are also some global impact assessments with more than one event (Capelle-Blancard and Laguna 2010, Carpentier andSuret 2015, 2021).
The event studies on the impact of industrial accidents nurtured the idea that financial markets could be a powerful tool to overcome struggles in implementing environmental regulations and to incentivize companies to adopt eco-friendly behaviour. Gradually, such environmental regulation has shifted from command-and-control policies to corporate disclosure strategies. An example is the toxic release inventory (TRI) promoted in the US by the Emergency Planning and Community 'Right to Know Act ' of 1986' of . Hamilton (1995 examine the first publication of the TRI and show that the most polluting firms experienced negative abnormal returns (ARs) after the disclosure of their performance in this regard. Several studies followed (e.g. Konar and Cohen 1997), supporting Hamilton's findings. This gave rise to the idea that financial markets should be entrusted with the task of disciplining and sanctioning firms, especially in developing countries with poor regulatory quality (e.  With the course of time, the literature gradually started paying attention to increasingly diverse kinds of environmental news. Highly relevant in this regard are the papers of Laplante and Lanoie (1994) and Klassen and McLaughlin (1996) that examine both negative and positive environmental events. From there on, the literature gradually moves beyond the scope of environmental economics to a more business-oriented approach, with a specific focus on CSR. In particular, voluntary firm initiatives, investments for environmental protection, certifications and green awards have been studied (e.g. Gilley et al 2000, Jacobs et al 2010. Finally, the expansion of environment, social, governance (ESG) information, combined with advances in ('big') data processing has provided opportunities to deepen the knowledge of the integration of environmental factors in shareholders' strategy (Flammer 2013, Krueger 2015, Capelle-Blancard and Petit 2019). The remainder of this paper informs how and why the literature about shareholders and environment has moved from finance to environmental economics, law and economics, management, and business ethics. These fields, which are generally compartmentalized, offer complementary perspectives. In this regard, we think the academic journal where the study has been published is a useful indicator of these evolutions. The first studies on the impact of environmental events on stock markets in the 1980s were published in finance journals (e.g. The Journal of Finance, The Journal of Financial and Quantitative Analysis). Then, from the mid-1990s onwards, the most influential papers were published in environmental economics and ecological economics academic journals (e.g. Journal of Environmental Economics and Management, Ecological Economics). In addition, several influential papers were published in law and economics (e.g. Journal of Law and Economics). Finally, in the 21st century, the literature has become more widespread in the academic business literature (e.g. Management Science, Academy of Management Journal, Journal of Business Ethics). In the next three sections, we consecutively present the results of the main studies for all three types of events (i.e. accidents, environmental policy regulation, and CSR in sections 4-6 respectively).

Shareholders and unexpected environmental events
Unexpected events can influence the value of the firm. This value usually is calculated as the net present value. Net present value relates the revenues and costs with regard to the firm's operations over the project's lifetime to the discount rate. This discount rate usually is made up of the risk-free market rate and a mark-up to reflect industry and/or firm-specific risk. The event may have an impact on one or several of these constituting elements.
Most of the valuation effect results from changes in costs and risks. The environmental events can result in costs to the firm, as showcased in the Brumadinho dam accident. The range of potential costs in relation to adverse environmental events is very broad. For instance, after an accident it includes damages to fixed assets (including property), losses of inventories, raw materials and finished products, as well as business interruptions. Those costs are obvious and much-publicized. However, as firms carry insurance against several of these, not all of them will be considered as relevant by investors. The decline in firm value is partly related to costs that are uninsured. The motives for corporate insurance purchases include risk aversion (that of the manager and/or the shareholders), taxes, bankruptcy costs and underinvestment (Mayers andSmith 1982, Smith andStulz 1985). As such, environmental events also affect the risk perception of the firm, or even that of the industry as a whole. Further, many risks are still unknown and/or cannot be insured. In addition, the events can influence the firm's reputation: Investors might come to regard the firm as more risky, consumers might see it as less responsible, and prospective employees might view it as less attractive as an employer. Reputation is a relative matter, as investors, consumers and workers will compare between different opportunities to allocate their resources.
In practice, firms seldom are fully insured against the wide range of costs related to adverse events. This is because of several reasons. First, the cost to cover each possible state of nature is prohibitive. Second, the probability of some type of events can be very low, and firms are likely to be prone to cognitive bias (risk underestimation). Third, insurance is poorly designed to cover for penalties and fines from enforcement actions (Shavell 2007). Fourth, firms are not willing to be fully insured against clean-up costs or third party claims (legal expenses and tort liability) due to incidental pollution, as liability is often limited in insurance contracts (Biais et al 2010). Governments, at the expense of taxpayers, are most of the time committed to be the insurer of last resort. In addition, risk can materialize beyond the horizon of the firm, investors, and governments.
In case of environmental events, stakeholders are likely to update their beliefs about the safety and responsibility of a particular plant, firm or the specific industry. For example, insurance companies may increase insurance premiums and require more stringent safety standards after a particular incident. Public authorities may reinforce the regulations and raise the cost of regulatory compliance for firms. In addition, customers and suppliers may want to steer clear from doing business with these firms. Further, financial analysts and investors may regard the expected earnings of firms riskier than before. Then, their market valuation can be affected by investment portfolio rebalancing, due to a shift in systematic risk (i.e. the sensitivity of the firm's stock to changes in the overall financial market) or investor taste. As a result, firms may increase their efforts to enhance a better image towards meeting the preferences of consumers, investors, and of the community and environmental pressure groups. Kotchen and Moon (2012) argue that it is especially firms and industries that face controversies who invest most in CSR to accommodate the demands of stakeholders (see also Heal 2008).

How to measure costs and benefits for shareholders? The event study methodology
From the discussion above, the question arises how to measure costs and benefits for shareholders. In this regard, a substantial number of studies use the event study methodology. This methodology is developed to investigate the reaction of the stock market to news or events concerning a firm, an industry or the complete stock market (Brown and Warner 1980, MacKinlay 1997. The theoretical underpinning of event studies is the EMH (Fama et al 1969), which states that when new information becomes available it is fully taken into account by investors assessing its current and future impact. Under this hypothesis, changes in stock prices (i.e. stock returns) reflect the discounted value of current and future firm performance. Thus, the stock price of a particular firm is assumed to reflect the discounted value of current and future cash flows of the firm. From a more tangible perspective, market prices are considered to result from the behaviour of heterogeneous shareholders' reaction when new information comes available. Then, unexpected changes in stock returns caused by a specific event can inform about investors' expectations regarding the impact of this event on the value and viability of the firm (Malkiel and Fama 1970). Therefore, if different categories of events have different impacts on companies' stock returns, it informs to which event type firms are most sensitive in terms of their market value and their viability.
An event study measures the impact of a specific event on the value of a firm using financial market data (MacKinlay 1997). The usefulness of such a study results from the fact that, given rationality in the marketplace, the valuation effect of an event immediately reflects in stock returns (MacKinlay 1997). Using stock prices, it is possible to measure the economic impact of an event over a short time period (Brown and Warner 1985). The general flow of analysis in an event study is as follows (based on MacKinlay 1997, pp 14-15): '… [first] define the event of interest and identify the period over which the security prices of the firms involved will be examined-the event window.
[…] The period of interest includes the day of the announcement of the event and the day after. This captures the price effects of announcements, which occur after the market closes on the announcement day. […] After identifying the event, it is necessary to determine the selection criteria for the inclusion of a firm in the event study. […] The appraisal of the event's impact requires a measure of the abnormal return.
[…] The abnormal return is the actual ex post return of the security over the event window minus the normal return of the firm over the event window. The normal return is defined as the expected return without conditioning on the event taking place…' .
The event study methodology has been widely used in finance to determine shareholders' reaction to all kind of unexpected news (see Sprecher and Pertl 1983, Brown and Warner 1985, MacKinlay 1997, Kothari and Warner 2006 16 . Numerous event studies have been devoted to the impact of the announcement of shares distributions, earnings, dividends, mergers and acquisitions, technological accidents, product recalls, massive layoffs, etc 17 . To determine if an event has a significant impact on a firm's stock market return, one needs to test whether that the firm's ARs during the event window (the period closely surrounding the actual event) are significantly different from zero. These ARs are the difference between the actual stock returns to the returns that would have been generated in the absence of this event (i.e. the expected or 'normal' returns), where expectations are based on recent market behaviour of the firm's stock. Thus, investors' reaction to a given information release (i.e. a news item) can be measured by comparing the observed return during a given time period following the event to the expected return in the absence of such an event. If investors react (un)favourably to an event, we may expect (negative) positive significant ARs. While the event study methodology is quite homogeneous, several approaches can be implemented and may vary according to the length of the event windows, the length of the estimations windows, the method to aggregate returns (average abnormal returns, AARs; cumulative average abnormal returns, CAARs; buyand-hold abnormal returns, BHARs) and the model used to compute the counterfactual returns (constant returns, market model, multi-factor model, etc) Warner 1980, 1985). Once the ARs are assessed, several researchers then run multivariate regressions to relate cross-sectional differences in the loss incurred by shareholders (as measured by ARs and the associated market value loss) to the event's or firm's features.
The strength of the event study approach is that it is based on the overall assessment of many investors who quickly process all available information in assessing each firm's market value. All these investors have (pecuniary) incentives to make the best decision. However, a general limitation of studying financial 16  markets that also relates to event studies is the Grossman and Stiglitz (1980) paradox which states that there are few incentives for investors to actually generate information as their efforts will spread out immediately to those who do not invest in information production. However, freely available information does not suffer from this paradox. Moreover, the event study methodology must be used only on financially material and easily determinable events in order to reduce the probability of confounding biases. This relates to the possibility to capture ARs from other, unrelated, events or specific news to firms present in the sample period. This also has ramifications for the use of the methodology; it only makes sense to use it in the case of a short event window around the moment the news arrives at the market (MacKinlay 1997).
We will present several synoptic tables in the next three sections. We provide a breakdown by categories of event (industrial accidents, public disclosures, legal actions, environmental regulations, environmental news, green awards and certifications, and corporate environmental practices), and summarize, for each study: the sample, the nature of the events, and the main results. The results are displayed as AAR, CAAR, BHAR or as the coefficient associated to the event in the multivariable regression.

How to collect and identify environmental events?
The main sources of information used in event studies are newspaper articles, official publications, company press releases, and information from extrafinancial rating agencies or NGOs (see the overview in table 2).
Newspaper articles can be used to assess firms' corporate social performance, but they cover a very diverse set of news. The news can relate to the amount and frequency of pollutant releases, the announcement of collective redundancy plans, the number of legal disputes, lawsuits or fines for non-compliance with environmental rules, etc. Using newspaper articles is interesting to identify which information has a significant impact on firms' stock market value, but it requires an extensive data collection process and the heterogeneity of the information can be problematic. Further, what appears in newspapers is filtered by the editors and journalists, who respond to what is happening and filter on the basis of what they deem interesting and relevant to their audiences. In addition, one should keep in mind that news is a product as well: Media have an interest in attracting audiences as this feeds their business model. As a result, this type of information is likely to be biased.
Several studies focus on the public disclosure by authorities, such as the TRI by US firms (Hamilton 1995, Konar and Cohen 1997, Khanna et al 1998. Every year, to ensure public information under the 'Right-to-Know Act' , US companies must report their polluting emissions (chemical and toxic products) to the authorities. The advantage of such public disclosure is access to homogeneous and detailed information. However, these data do not cover all pollutants and do not provide information about companies' environmental policies and efforts. In addition, only US firms report this information. Further, several online portals provide accidents databases, which are usually constructed and maintained public institutions. Examples are EMDAT (Centre for Research on the Epidemiology of Disasters), INRS (French national research institute for safety), MARS (Major Accident Reporting System), or ZEMA (Zentrale Melde und Auswertestelle für Störfälle und Störungen in verfahrenstechnischen Anlagen).
Compared to these information sources, the information and data provided by the firms themselves are neither regular, objective nor uniform, but are nevertheless a widely used source of information. More and more companies report about their social or environmental performance in their annual reports. Such information has taken on a strategic dimension (Aerts et al, 2008, Lyon andMaxwell 2011). For several years, the amount of information disseminated by companies has exceeded what is required by law. Financial communication has become essential for companies (and public administrations). The objective of companies for doing so is twofold. First, it serves to convey a positive image of the issuer to the financial community and beyond. Secondly, it makes it possible to better manage crises (announcements of results lower than expected, threat of takeover, technological accidents, etc). For the firm, the decision to disclose or not to disclose certain (non-mandatory) information can be formalized simply through a cost-benefit arbitrage model, the main benefit being the reduction of costs associated with information asymmetries and agency conflicts, being conflicts between owners and managers (see Healy and Palepu 2001). The latter are due to the subjective nature of the announcements made by the companies. A major concern of course that this allows the companies to mark their own homework, which carries the risk of greenwashing.
In addition, to assess companies' policies in terms of social and environmental responsibility, one can also use information generated by third-party organizations such as NGOs or CSR rating agencies. The former usually have a non-financial agenda and aim to highlight the importance of particular societal interests. Most NGOs focus on specific issues and generally are very transparent about the information the produce. However, this information usually is not very systematic and not available at a regular basis. This contrasts with the commercial CSR rating agencies. Next to the conventional rating agencies (Moody's, Standard and Poor's), CSR rating agencies have emerged and they are specialized in producing an assessment of the (relative) commitment of companies to improve their environmental and social performance. Their information aims at being synthetic and homogeneous between different companies and industries. The other side of the coin is the processing of 'primary' information, which is not always transparent (Berg et al 2019). In addition, the focus of these agencies is on processes and not on performance. This results from the lack of standardized and verifiable environmental accounting standards and metrics. As they rely to a large extent on unvalidated and unaudited firm information, the same problems as discussed in the previous paragraph prevail in the case of ratings. From the perspective of the user of the information, in particular the shareholder, we think five properties are highly relevant. These are availability (is the publication easily accessible?), richness (is the information detailed?), parsimony (is the information easy to gather and process?), reliability (is the source objective and reliable?) and time delay (is the information known immediately?). Table 2 suggests that none of the information sources is superior to the others in all respects. Hence, it depends on the purpose of the use of information which sources are most appropriate.
Regardless of the topic, the number of events considered in each event study gradually increased over time. This is to be expected as more incidents are reported with the course of time. This is due to the improvement of monitoring technology and changes in reporting requirements in the past decades. As a result, environmental incidents are easier to be detected and reported. Figure 3 shows the number of events (in logs) for each study used in our analysis by event type, ranked by chronological order (namely by the year of publication of the articles). The rise in the number of events in figure 3 might only reflect the increase in events reported by the studies; not the actual development in the number of environmental accidents that took place since 1980.

How much?
The aim of event studies in general is to provide a quantitative assessment of the penalty or reward associated to a particular type of event. In contrast, the aim of this paper is to give an overview and reflection about the topics studied in this event study literature and not to provide a conventional metaanalysis. Nevertheless, we have collected all the estimates of those studies to provide insights regarding the empirical findings in this literature. In this regard, it is important to acknowledge that the studies rely on highly different events, samples, countries, periods, methodological choices, and are difficult to compare. Figure 4 presents for each study, detailed in the sections hereafter, the average CAAR, along with the median as well as with the minimum and maximum estimates. It is based on all the CAARs we have collected for the meta-analysis, by event type. Figure 4 shows a strong dispersion of the estimates, in particular for the studies with the largest impact (in absolute value). On average, the shortterm drop in market value consecutive to eco-harmful events is almost 2%, whereas the 'reward' for ecofriendly events is 0.2%. We also included the median results. With eco-harmful events, the median of the stock market responses is −0.6%. With eco-friendly events, the median response is 0%. As we aim at a synthesis of the literature in the past four decades, we will not dig deeper into the studies that are used to compile this Figure. We put together a database that can be used for such research purposes as well as for education. We feel that including a full-fledged meta-analysis in this review would take too much space and would damage the flow of our analysis. Instead, in sections 4-6, we will analyse the three main streams identified in the literature regarding the interaction between shareholders and the natural environment, namely high-profile industrial accidents (section 4), policy events (section 5), and the impact of CSR events on shareholder value (section 6).

Industrial accidents with environmental concerns
In this section, we reflect on several studies about high-profile industrial accidents with environmental concern. These events result from an accident, negligence or incompetence, or from any combination of these factors and do not have a natural cause. They generally either directly or indirectly resulted in a substantial number of deaths and/or injuries, as well as significant material and environmental damage, and the consequences were both immediate and delayed. In particular, aligned with the chronological order, we address the TMI nuclear meltdown, the Chernobyl nuclear disaster, the Bhopal chemical explosion, the Exxon-Valdez oil spill, the BP Deepwater Horizon spill, and the Fukushima Daiichi nuclear disaster 18 . In several cases, the market value of the firm involved in the accident experienced a significant drop. This relates to an increase in expected costs and risks (table 3). However, as there is mean reversion in investment returns (Mukherji 2011), it is not clear whether such a drop always persists and/or if it is sufficient to provide effective incentives in term of safety. 18 Previous studies on the effect of bad safety records on equity returns especially focus on civil airplane crashes. Borenstein and Zimmerman (1988) find that equity value losses are related to the number of fatalities and serious injuries, but not to the expected drop in sales. Mitchell and Maloney (1989) find that equity value losses are statistically significant only when crashes are due to pilot errors, by contrast with defects in construction. Broder and Morall (1991) show that stock market reactions vary by which agency has regulatory jurisdiction over the accident and where ex-ante inspection is lax or nonexistent, investor reaction is stronger. Bosch et al (1998b) show that the airline industry as a whole incurs losses, while direct competitors of firms responsible for crashes may even benefit from an equity value premium following accidents, see also Walker et al (2005) for a study with a sample including September 11, 2001. In addition, several studies examine the impact of natural (or, at least, not directly causes by a specific firm) environmental disasters: earthquakes (Shelor et al 1990, Scholtens and   To allow for a comprehensive assessment of the findings in the literature, we also transformed to an inverse hyperbolic sine (a.k.a. area hyperbolic sine). Hence the y-axis is depicted as CAAR and as ASINH(CAAR). The figure shows the average as well as the median CAAR (in %), along with the minimum and the maximum estimates (the y-axis is cut off to avoid extreme value for the sake of readability). The studies are presented by year of publication.   the accident at TMI permanently changed the nuclear industry 19 . Bowen et al (1983) examine the market value of 83 electric utility companies (General Public Utilities, the owner of TMI facility is excluded) following this accident. The sample is partitioned into several groups. They consider successively utilities with plants build by Babcock & Wilcox, i.e. contractors for the TMI facility (11 cases), firms with where nuclear facilities generated more than 20% of their overall capacity (21 cases), firms with less than 5% of their capacity from oil, no dependence on gas and a nuclear capacity above 10% of overall capacity (14 cases), and a control group of non-nuclear, non-oil and non-gas firms (14 cases). They show that firms with nuclear power plants and utilities planning to build these experienced a drop in equity returns following the TMI accident. The impact on firms with plants build by Babcock & Wilcox or with nuclear capacity above 20% of total capacity of the utility was nearly identical, while the control group appeared to be almost unaffected. Rothwell (1989) examined 37 nuclear reactor failures over the period 1978-85. He finds that the daily return for the owner of a failed reactor dropped by 0.24% after two trading days but increased by 0.24% four trading days later, both compared with a portfolio of nuclear utilities. In addition, Bowen et al (1983) provide evidence of a positive shift in the systematic risk and the total risk for 21 nuclear-intensive firms after the TMI accident. However, given that the control group also exhibited an increase, they do not attribute the magnitude of the shift to the accident only. Hill and Schneeweis (1983) confirm the negative impact of the TMI accident. Nuclear utilities (with at least 10% installed nuclear capacity) and non-nuclear utilities experienced a significant decline in their equity returns of −6.1% (during the first month of the event) and −5.4% (six months after the event) respectively following the TMI accident, with an effect significantly greater for the former sample of firms 20 . 20 In addition, Barrett et al (1986) study the impact on the risk premium attached to 76 electric utilities bonds. They show that the average bond risk premium increased after the TMI accident. The impact was larger for firms with nuclear capacity, but it did not relate to the extent of a firm's commitment to nuclear power. Pettway (1981), Fraser and Kolari (1983), Zimmerman (1983) and Chen (1984) report increases in systematic risk, as measured by the beta, while Brooks and D'Souza (1982) provide evidence of a positive shift in total risk for nuclear firms after the TMI accident. Spudeck and Moyer (1989) examine the determinant of the systematic risk for a sample of 62 electric utilities. Their results suggest that some of the market reaction heretofore ascribed to the accident resulted instead from regulatory activity occurring before the accident, which is consistent with Bowen et al (1983).

Bhopal
On 3 December 1984, the Union Carbide plant in Bhopal, India, leaked 27 tons of methyl isocyanate, a highly poisonous gas. About 4000 people were killed and another 200 000 were injured. Bhopal is frequently considered as the world's worst chemical disaster 21 . According to Salinger (1992), the primary financial claim regarding Union Carbide was made by the government of India on the behalf of the victims for $3 billion; the lawsuit was settled in February 1987 for $470 million 22 . Consequently, the market value of Union Carbide fell, from $3.5 to $2.5 billion; a decrease of 28%. During the first month, the AR was −31.5%, but a year after the disaster, the cumulative AR had become insignificant and in March 1986 the CAR was +31% after a failed takeover bid and a restructuring and recapitalization of the firm.
Blacconiere and Patten (1994) examined the stock returns of 47 firms in the chemical industry following this disaster. For Union Carbide, the drop in market value went on for more than one month and the chemical industry as a whole was pulled down during ten days after the disaster (−2.4%). Firms with more extensive environmental disclosure prior to the accident experienced a somewhat lower negative loss of their market value (0.6% and −5.34%, respectively). Blacconiere and Patten (1994) suggest that investors interpreted extensive disclosure as a good signal, giving rise to the possibility for firms to disclose good news and suppress bad news. Kalra et al (1995) confirm this result for Union Carbide, and find stronger contagion effects to firms producing hazardous chemicals which are more similar to Union Carbide (−1.9%) rather than other non-hazardous chemical firms (−1.2%). Nonetheless, they notice that those companies' returns have rebounded fast and surpassed their pre-crisis levels 18-19 days after the event.
On the 20th anniversary of the Bhopal disaster, a new event caused some turmoil for shareholders. In February 2001, the US based company Dow Chemical had acquired the shares of Union Carbide, which became a fully owned subsidiary. On December 3, 2004, two US artist-activists (the Yes Men) succeeded in being invited by the BBC pretending to be Dow Chemical representatives 23 . They announced that Dow Chemical would accept full responsibility for the Bhopal disaster, and would agree to clean up the site and compensate the victims for an amount of $12 billion. This hoax resulted in a 4% drop in Dow's share price, wiping $2 billion off the market value of the company; the shares rebounded after the BBC 21 Broughton, E. The Bhopal disaster and its aftermath: A review. Environ Health 4, 6 (2005): https://doi.org/10.1186/1476 -069X-4-6. 22 Amnesty International, Clouds of injustice: Bhopal disaster 20 years. 23 https://yeslab.org/tags/bhopal. issued a disclaimer stating that Dow had not taken this responsibility.

Chernobyl
On April 26, 1986, one of the reactors of the Chernobyl station, located in what is now Ukraine, suffered a catastrophic steam explosion that resulted in a fire, a series of additional explosions, and a nuclear meltdown. The surrounding geographic area was contaminated, and several hundred thousand people were evacuated. Two hundred people were hospitalized immediately, of whom 31 died. Further, the nuclear meltdown produced a radioactive cloud which spread all over Europe. The accident raised concerns about the safety of the nuclear power industry, even in developed countries 24 .
Fields and Janjigian (1989) investigate the effect of this accident on 89 US public electric utility stock prices and find that shareholders suffered significant negative ARs. Moreover, utilities using nuclear power experienced greater losses than those without (respectively −6.4% and −4.17% up to three days after the event). These results are highly similar to those reported for the TMI accident, but there is no evidence of a shift in either the systematic risk, the total risk, or the market risk. Kalra et al (1993) consider 69 public power utilities listed on the NYSE or the AMEX. Their sample is split into three groups: nuclear capacity of 20% of total capacity or more, nuclear capacity of less than 20% and no-nuclear. In particular, they document a small and transitory effect for the first and the last groups. Indeed, 30 trading days after the Chernobyl meltdown, the cumulative average AR was +0.9% for the non-nuclear and −1.3% for the nuclear firms. The impact was most pronounced for the firms with nuclear capacity between 1% and 20%. This surprising result may be explained by the fact that those firms were in the process of building new nuclear capacity 25 .

Exxon Valdez
On March 24, 1989 the Exxon corporation tanker Valdez ran into a submerged reef, spilling some 250 000 barrels' of crude oil into the sea. The Exxon Valdez oil spill was one of the largest environmental disasters ever to occur at sea at that moment, seriously ravaging plants and wildlife 26 . According to Jones et al (1994) Mansur et al (1991) study the market response to six events associated with the Exxon Valdez oil spill for 14 major oil firms listed on the NYSE. The only event that caused significant ARs is the announcement (ten days after the disaster, on April 10) that Exxon and Alyeska Pipeline might face substantial punitive damage claims, which were not covered by insurers (−1.69%). In particular, the market did not significantly respond to the disaster, when Alaska initiated a criminal inquiry (on March 30th), or when oil spill response capabilities at the time of the accident were deemed to have been inadequate (on April 7th). Mansur et al (1991) suggest that investors were able to discriminate among oil companies based on their exposure to the Trans-Alaska Pipeline since the more exposed firms experienced greater losses than firms less exposed. However, their results were not statistically significant.
Likewise, Herbst et al (1996) show that for the five consecutive trading days after the Valdez accident, there were no significant ARs for Exxon. Actually, the decrease in stock price was progressive (see also Katsikides et al 2016). Only after ten trading days (two weeks), the cumulative AR was significantly negative for Exxon (−6.33%). Herbst et al (1996) do not provide evidence of any intra-industry effect since none of the five largest petroleum companies experienced any loss. According to Herbst et al (1996), the market properly assessed the probability that Exxon would have to pay the full cost of clean-ups plus punitive damages, since the stock price decrease was in line with the present value of these total damages. Moreover, there was no significant change in the volatility of Exxon's market value, compared to the overall stock market.
White (1996) confirms that only a limited market impact was to be observed for the first 20 trading days after the oil spill and reports that the stock price decrease was progressive with a significant negative cumulative ARs even after 120 days. White (1996) reports negative but not significant effect for Alyeska consortium or Exxon competitors. In addition, he documents that companies from other sectors rated independently by the council on economic priorities (CEP) as having good environmental performance experienced a positive impact on their market value (11.2% up to 90 days after the event).

Deepwater Horizon
On April 20, 2010, the Deepwater Horizon rig operated by the oil company BP exploded, spilling out nearly five million barrels of oil into the Gulf of Mexico and killing 11 workers. This accident is one of the worst oil spills in history, with long lasting and dramatic consequences both for the environment and for the local fisheries and tourism industries in Louisiana and Florida 28 .
In the aftermath of the explosion, BP incurred an AR drop of approximately −6% over the first week. Actually, the stock market response to the catastrophe was not immediate. The largest drop in market capitalisation was observed after April 28 when (following unsuccessful attempts to secure the source of the leak) authorities declared that the oil spill was of national significance and President Obama made a clear statement that BP would be 'ultimately responsible for funding the cost of response and clean-up operations' 29 . Sabet et al (2012) show that the main marketmover was the Gulf oil exploration moratorium on deep water drilling that resulted from the concern about the explosion (see also Katsikides et al 2016).
In the first few months after the disaster, BP lost half its stock market value, which represents tens of billions dollars. In addition, the accident and the subsequent regulatory efforts impacted not only BP, but also its subcontractors and BP's competitors (Koda 2016). However, Heflin and Wallace (2017) show that among the firms drilling in US waters, those with greater environmental disclosure suffered less 30 .

Fukushima
On March 11, 2011, a magnitude 9.0 Richter earthquake unleashed a sever tsunami in Japan. It disabled the power supply and the cooling of the Fukushima Daiichi nuclear power plant, operated by TEPCO, a Japanese utility. Three core reactors melted in three days and were stabilized in the following two weeks thanks to recycled water. Although no deaths have been reported for radiation sickness from the nuclear incident, the natural catastrophe led to a toll of about 19 000 human casualties 31 .
Kawashima and Takeda (2012)  findings indicate that firms which have been affected directly by the earthquake and plants that have relied more on nuclear power faced significant negative ARs (−37.03% and −9.99% respectively up to two days after the event). However, they do not find any evidence of the market sanctioning firms having a similar nuclear reactor as the Fukushima Daiichi plant. They establish an increase in systematic risk and total risk as the market beta and the variances increased following the event. Overall, these findings indicate that shareholders were concerned about the future cost of energy, in particular nuclear power, which could significantly increase due to new safety policies.
Ferstl et al (2012) consider stock returns in France, Germany and the US from nuclear and alternative energy firms. Their results show that nuclear stocks in Japan underwent a significant and prolonged market penalty (−15.3% up to 24 days after the event), while in Germany and France, similar companies experienced the same drop (−6.3% and −6.4% respectively up to four days after the event), but one that faded away more quickly. However, they do not find any significant stock price impact for US firms. Moreover, they find alternative energy companies benefited from the event in terms of positive returns (mainly in Europe) but this effect did not last. Ferstl et al (2012) point out that shareholders might expected a change toward alternative energies for Germany and France. Nonetheless, the effect was stronger in Germany compared to France (39.6% and 13.4% respectively up to four days after the event). The German government announced a phasing out of all nuclear energy generation three days after the Japanese disaster. Confirming these findings, Betzer et al (2013) examine the spillover effect of the Fukushima event in Europe. They find that the abrupt decision of the German government to stop nuclear energy production was the main factor behind the shareholders reactions; not the dramatic Fukushima Daiichi event. No German stock reacted significantly during this first tragic event while European stocks have not reacted to any of the two events. However, they results indicate that German renewable energy firms have strongly benefited from their policy announcement change on the short run while nuclear and conventional stocks were significantly penalized (11.07% and −3.27% respectively on event day). In addition, Betzer et al (2013) highlight that even if those initial market reactions were strong, both industries' stocks reached a plateau for 20 days. Finally, Lopatta and Kaspereit (2014) consider a worldwide sample of 52 utilities relying to various degree on nuclear power. Their results show that firms with a higher share of nuclear power in their total capacity have been impacted more strongly by the event (−9.2% up to six days after the event while all stocks dropped by 2.81%). They also note that all utilities with nuclear capacity were affected.

Others
In 1996, a Canadian industry, Placer Dome, with mining operations in the Philippines, suffered a massive failure of the Makulapnit Dam, which dumped four million tons of sewage into the Boac River. Magness (2009) analyse the contagion effects of the Placer Dome event on the Canadian mining industry. She observes a significant drop of 2.6% in share prices for the Canadian mining industry one day after the event. However, she finds that firms having a more environmentally committed management have been affected to a lesser extent, while financial disclosures have negatively affected the firm's market value. Kowalewski andŚandewski (2020) have also considered the mining industry and find that financial markets react more strongly to the natural rather than human-made disasters occurring in mines.
The British Buncefield depot caught fire on December 11, 2005, and this raged for three days after several explosions. Although it caused no human casualties, the event has affected around 400 businesses nearby and released a non-toxic black smoke across the area. At the source of the accident, four major oil companies (Total, Texaco, BP and Shell), owner of the Buncefield depot, were considered responsible. Capelle-Blancard and Laguna (2008) note that on the day of the explosion, the four owners suffered on average only a slight financial loss of −0.6% despite the seriousness of the event.

Sample of accidents
In contrast with the previous single-event studies 33 , Capelle-Blancard and Laguna (2010) examine the stock market reaction to separate accidents in chemical plants and refineries worldwide between 1990 and 2005. Their sample includes 64 major events (a quarter of them resulted in a toxic release, and half of them caused at least one death or serious injury), but not as high-profile as those mentioned previously in this section. They establish that firms' market capitalization decreases on average with 1.3% the two days following the disaster, and they show that the loss is related to the number of casualties and to pollution. Similarly, Makino (2016) examines chemical accidents (18) that occurred in Japan between 2005 and 2012 and finds that the financial markets have reacted more strongly to serious accidents. Carpentier and Suret (2015) examine the stock market reaction to 161 industrial accidents (both environmental and non-environmental), reported on the front page of The New York Times from 1959 to 2010. They show that the decline in firm value is mostly related to government interventions and 33 The first event studies based on several accidents were carried out by Laplante and Lanoie (1994) and Klassen and McLaughin (1996), but in tandem with other positive environmental events (e.g. green awards). These papers are discussed in the next section. rather low for accidents with environmental concerns compared to airplane crashes. In addition, they show that the CAARs following environmental accidents do not differ from zero after one year.
The recent literature also examines the impact of specific accidents, such as oil spills (Grand and D'Elia 2005, Feria-Domínguez et al 2016, Katsikides et al 2016) or mining tragedy (Kowalewski,Śpiewanowski, 2020). Overall, the market reaction appears to be limited, and these market incentives are highly unlikely to encourage managers to change fundamentally their environmental safety policy or to consider divestment by financial investors.

Environmental regulation
In this section, we reflect on the studies that investigate the stock market response to environmental regulation. Over the past four decades, the principles of environmental protection policies have evolved from legal command-and-control regulation (the 'first wave'), to market-based instruments (the 'second wave') and to disclosure strategies (the 'third wave') (Tietenberg 1998, Kotchen 2013. We first discuss studies about the impact of public disclosure, then how stock markets react to other public information regarding legal actions in relation to environmental regulation, and lastly studies that focus on environmental policy change, especially the launch of emission trading schemes.

Public disclosures
Disclosure strategies involve public and private attempts to improve information about pollution. The idea is to stimulate market dynamics, in order to induce firms to self-regulate and adopt environmentally friendly strategies (Tietenberg 1998, Khanna 2001. The growing popularity of these strategies among domestic and international regulators, which comes at a low cost for regulators, raises the question of whether information disclosure is effective in generating pressures from shareholders on firms to improve their environmental performance. Disclosure may have much less systematic effects on firm valuation than accidents. First, there can be a positive effect because the firm is becoming more transparent. This reduces the risks of the firm as such from the perspective of the shareholder. There also can be a negative market response when the information being disclosed suggests additional risks might be incurred or costs can be expected. An example of such disclosure policy is the TRI, which has been promoted in the US by the Emergency Planning and Community 'Right to Know Act' of 1986, is the most important case of environmental policy based on public disclosure. In a seminal article, Hamilton (1995) studies the stock market response to the first public announcement of the TRI in 1989 for 436 firms. The sample is composed up to 75% of manufacturing firms (chemicals, paper, primary metals, petroleum and textiles). All firms were listed at the NYSE or AMEX. Hamilton finds that the stock market value on average dropped by $4.1 million the first day (−0.284%) of firms reporting TRI releases (see table 4).
Since the study of Hamilton (1995), many papers have been written and continue to appear dealing with this topic. The upper panel of table 4 reports some of these studies. For example, Konar and Cohen (1997), using a sample of 130 firms, show that investors respond to information published by the media on the TRI database, rather than to the TRI itself. They also find that firms that experienced the largest stock price decline subsequently reduced their emissions more than their industry peers did. Khanna et al (1998) extend Hamilton's analysis over several years (1989)(1990)(1991)(1992)(1993)(1994) for a sample of 91 chemical firms. They find that firms whose toxic release rates increased between 1991 and 1994 suffered significant negative ARs (ranging from 0.23% to −0.71%). Arora (2001) investigate the market response to the TRI in 1997 for 645 firms. Arora (2001) first estimates with a logistic regression the probability of engaging in pollution prevention activity. Then, firms are classified according to the difference between actual and predicted performance. Firms that fail to meet expectations experienced a drop in their stock market value, while firms that exceed their expected level of prevention activity are not rewarded (Arora 2001). Bui (2005) is skeptical about the relevance of stock market appreciation regarding the TRI (see also Ananathanarayanan (1998) and Cram and Koehler (2001) for methodological issues). She studies a sample of 106 petrochemical firms over the period 1989-1998 and finds that the observed fall in toxic releases due to the TRI has been overestimated. She controls for the fact that all reporting firms face a common event window: the TRI events are identical for each company in each year. The assumption of zero covariance across ARs is therefore almost surely violated. Bui (2005) concludes that the negative ARs around TRI reporting dates for petroleum companies are insignificant. She reports the results for two types of companies: Companies are classified as reporting 'bad news' ('good news') if their TRI releases are larger (lower or equal) in the current year than in the previous year. Then, she finds there is no evidence that companies reporting bad news had statistically significant negative ARs during the event window, or had returns that were statistically significantly different from firms reporting good news. Bui (2005) also compares ARs of firms reporting the TRI, with those of non-reporting firms, and find no evidence of significant ARs differences There also is a broad literature that relates firms' longterm financial performance with environmental performance. For example, Konar and Cohen (2001) show that poor environmental performance (as measured by TRI level of emissions in 1988, and the number of environmental lawsuits against the firm in 1989) has a significant effect on the intangible-asset value and Tobin's q of publicly traded firms that belong to the S&P 500, after controlling for market share, R&D expenditures, advertising expenses, etc Cormier and Magnan (2007) investigates the impact of voluntarily disclosed information of environmental reporting. They control for the endogeneity between a firm's decision to disclose information, its exposure to media and its stock market value. Disclosure by German firms of environmental information has a positive effect on investors' appreciation of the relationship between its reported earnings and value. It shows French and Canadian firms are not affected. Further, for all three countries, firm size and media exposure are strongly related to environmental reporting. disclosure reduces the impact of the TRI announcements, limiting the scope of environmental disclosure under a largely voluntary regime. Actually, the first event study regarding environmental public disclosure did not consider the wellresearched TRI. Instead, Shane and Spicer (1983) consider the release of eight major studies conducted by the US CEP on firm's environmental performances in four industries (Oil, Power, Paper and Steel) over the period [1970][1971][1972][1973][1974][1975] (see bottom panel of table 4). A total of 72 firms are considered. Information is from The Wall Street Journal and The New York Times. Shane and Spicer (1983) show that firms experience large abnormal negative returns during two days prior to the newspaper reports on the release of the CEP studies. Firms revealed to have low pollution-control performance rankings have significantly more negative ARs than companies with high rankings (high pollution control firms have on the event day 2.08% higher returns than low pollution control firms).
Other studies in this line of research, not relying on TRI data, are Lanoie et al (1998) who base their study on a list of polluters published twice a year by the Ministry of the Environment of British Columbia (ME-BC) in Canada over the period 1990-1992. The sample consists of 19 publicly listed Canadian firms. The firms are put into two categories: firms not complying with an environmental standard or permit and firms with environmental performance near, or going well beyond the regulatory threshold. It shows that firms do not incur statistically ARs. However, firms that appear in the list of polluters more than once incur higher losses than firms appearing only once. Further, Gupta and Goldar (2005) examine the Green Rating Project by a leading environmental NGO in India and funded by the United Nations Development Programme through the central Ministry of Environment and Forests. They examine three industries, pulp and paper, automobile and chloralkali firms, and find that the announcement of weak environmental performance leads to large negative ARs up to ten days (−0.12%). Dasgupta et al (2006a) examine the value impact of public disclosure of environmental performance with regulated facilities in South Korea (SK). Since the mid-1980s, the ME-SK has published a list of facilities in violation with existing Korean environmental laws and regulations (emission standard, equipment failure, etc). They find that the average reduction in market value (−9.7% 35 ) is found to be higher than the estimated changes in market value for similar events in Canada and the US, and of a similar magnitude as observed changes in other developing countries (Argentina, Chile, Mexico, and Philippines). Dasgupta et al (2006b) also show that the larger the extent of coverage by newspapers, the larger the reduction in market value. In addition, Cañón-de-francia et al (2008) consider the first publication of the European Pollutant Emission Register (EPER) and its impact on the market value of 64 Spanish firms between 2003 and 2004. They show a negative impact for EPER listed firms, and a positive impact for those which might have been considered as potential polluters but which do not exceed the legal thresholds and consequently not listed in the EPER.

Legal actions
Lawsuits and judicial actions following environmental violations date back to the 1970s (Grad and Rockett 1970) and enforcement is part of the strategic plan of most of the environmental public agencies 36 . The eventual fines have obviously a negative impact 35 It should be noted that this strong decrease in market value might be due to some extreme events, as one event in their sample caused a CAR higher than 100%. 36 See www.epa.gov/enforcement/basic-information-enforcem ent. on firm value. However, it depends on several factors, such as type of opponent, firm characteristics (Bhagat et al 1998), the reputational damage (Alexander 1999) and in case fines are less than expected, stock returns could be positive. We provide an overview of this literature in table 5. Muoghalu et al (1990) examine 128 initial lawsuits and 74 case settlements involving a fine related to the Resource Conservation and Recovery Act and the Superfund Acts in the US over the period 1977-1986(See also Little et al 1995. The source of information mainly is The Wall Street Journal. They find abnormal losses at the time of announcements of lawsuits (an average loss of $33.3 million in market value, −1.23%), but no significant ARs at the time of settlement. They also find that losses due to lawsuits filings are much weaker for the petrochemical industry than for other industries. Closely related is the study of Laplante and Lanoie (1994), who investigate the impact on firms' market value of 9 lawsuits and 13 suit settlements announcement in Canada. To identify the events of interest, they use Canadian print media (mostly The Financial Post and The Globe & Mail). In contrast with Muoghalu et al (1990), they find that the stock value declines only the day of suit settlements. Laplante and Lanoie (1994) conclude that weak regulatory enforcement and the lax response of the Canadian legislative authorities compared with the US ones may explain these divergent results.
Badrinath and Bolster (1996) examine the stock market response to legal actions of the US Environmental Protection Agency (EPA) over the period 1977-1991. Their sample consists of 704 actions (385 violations on the filing date and 319 cases on the settlement date) regarding firms publicly listed on the NYSE, the AMEX, and the NASDAQ. Badrinath and Bolster (1996) find a 0.43% loss in violator firm value during the week of settlement and establish that the market penalty is unrelated to fine size. Further, they show it is more pronounced for violations under the Clean Air Act, for repeat violators, and for more recent EPA actions. Bosch et al (1998a) show that firms may benefit by cooperating with the EPA, since compliant strategies reduce wealth losses. Karpoff et al (2005) study the impact on stock market excess return in relation to 478 environmental violations by US publicly traded companies over the period 1980-2000 (see also Karpoff et al 1998). The sample is obtained from a search in 'The Wall Street Journal Index' . In contrast to Badrinath and Bolster (1996), they find that companies violating environmental laws incur equity value losses of a similar magnitude as the amount of legal penalties (fines and clean-up costs) eventually imposed on them. Hence, it seems that US stock markets do not impose a reputation penalty on firms for being responsible for harmful effects. Karpoff et al (2005) argue that the violations may produce ill will, but do not affect the  (2017), relying on 83 260 lawsuits filed in the US (including 877 environmental litigation), suggest that the reputational loss due to corporate misconduct is lower than previously expected and that the loss in market value can be mainly attributed to the media coverage, the expectation of subsequent litigation, and the defendant's willingness to settle. Philipich (2018) examine the impact for the coal-burning industry in the US of the EPA Clean Power Plant introduced in 2015 and the legal efforts to challenge it. He reports a negative market reaction to the issuance of the final version of the Clean Power Plan (but not for the draft version) and a positive on following a Supreme Court ruling which forced the EPA to alleviate some rules.

Policy (change)
In table 6, we provide an overview of studies that investigate the stock market response to (changes in) environmental policies. In this regard, the early studies primarily covered specific laws. For example, Blacconiere and Northcut (1997) consider the impact of the establishment of the Superfund Amendments and Reauthorization Act of 1986 on 72 chemical firms. Next to the disclosure of mismanagement, the main purpose of this amendment was to toughen up the Comprehensive Environmental Response, Compensation, and Liability Act of 1980. It included a minimum clean-up requirement and specified that most clean-up agreements with polluters would have to be entered in federal courts. Blacconiere and Northcut (1997) show that shareholders reacted negatively to the implementation of this act. However, they responded much less pronounced in the case of firms communicating more on their environmental initiatives in their 10 K reports.
Diltz (2002) examines how electric firms' shareholders react to 20 different stages in the transition of the 1990 Clean Air Act Amendments. This program of amendments sought to reduce toxic air emissions    The IPPC transposes a European directive of 1996, which consists of standardizing measures between EU member states regarding pollutant emissions and the evaluation of companies' environmental performance. They found that investors reacted negatively to the law (−1.29%), but firms investing more in research and development (R&D) were less affected. This might results from the fact that R&D investment enhances firm capacity to adapt to new regulation.
Next to specific environmental policies, the event study literature also started to consider a wider variety of announcements like speeches made by politics regarding the environment or international climate policies. As to the former, Ramiah et al (2015) consider the election on November 4th, 2008 of the president Barack Obama who has centred his political campaign on energy policy, climate change. Ramiah et al (2015) find that the biggest polluters suffered negative ARs. Similarly, Schütze et al (2018) consider Donald Trump's speech regarding the US withdraw from the Paris Agreements (−0.9% and −0.27% respectively for green and brown firms up to two days after the event) 37 . Focusing on international policies, Schütze et al (2018) study the environmental decisions by the socalled Conference of the Parties (COP)s from 2009 to 2016. They find that the Cancun and Doha summits benefited environmentally friendly businesses, while brownfield firms were not impacted. The Paris summit penalized polluting firms, but did not result in significant ARs for green businesses. They interpret their findings as a change in market behaviour: the financial markets seem to have switched from rewarding green firms to penalizing brown ones and they seem to have started to view fossil fuel industries as more financially risky. Jiang and Luo (2018) focus on China's nonratification of the Copenhagen Climate Summit from 2009 to 2011. They find that delayed legislation on carbon emissions has increased the share price of the state-owned firms. Relatedly, Johnston et al (2008) consider the emission allowances of SO 2 rights and analyse both the asset and option value. They show that stock markets positively value firms having an excess in allowances. The main reason is because these rights can be sold or inventoried for future use. However, they do not provide strong evidence regarding the option value of the emission permits. Johnston et al (2008) find positive returns for firms purchasing rights, but these were not statistically different from non-purchasing firms (i.e. control group). Therefore, they do not know if the market really priced the real option value of the emission allowances.
Inspired by the US experiences, the EU attempted to tackle greenhouse gas (GHG) emissions with a similar market emission system implemented in January 2005, but with much less success. The EU's Emissions Trading System (ETS) has been motivated by the Kyoto Protocol. It encompasses 27 countries and has been implemented in four phases; it excluded several industries and was quite lenient in granting emission permits 38 . For the first phase, Bushnell et al (2013) show that carbon intensive firms were the most penalized, while less regulated industries have benefited from this downturn (see also Jong et al 2014). Jong et al (2014) arrive at different results. Firstly, they discover a significant share price increase for lower carbon intensive firms and the ones holding more permits. Secondly, they do not find a significant market reaction to ETS trade activities or from the passthrough of carbon leakage.
They argue this can be due to carbon leakage, which occurs when firms reallocate their activity into countries where the restrictions on carbon emissions are weaker ( For the Australian ETS, Luo and Tang (2014) examine how firms reacted to seven key carbon 38 From 2005-2007, the first phase was a 'learning by doing' phase enabling regulators, governments and firms to adapt themselves and to estimate a carbon price for phase II. It only covered CO2 emissions of power generators and energy-intensive industries, most of the allowances were given for free and financial penalties were given for not complying. From 2008 to 2012, phase II imposed lower caps in allowances (6.5% lower compared to 2005), reducing the level of free allocation (to around 90%), increasing noncompliance fines (from 40€ in 2005-100€ per tonne), authorising firms to buy international credits, increasing the number of sectors covered by including airplane, and accounting for NO2. From 2013 to 2020, phase III strengthened the system by targeting a 20% reduction of the GHG emissions. It introduced a single EU-wide cap, the auctioning of the allowances, the harmonization of member states allocations rules, and the increase in the number of states subject to the ETS. Finaly, from 2021 to 2030, phase IV aims to cut emission by 43% of 2005 levels for 2030. It mainly consists in reducing allowances to 2.2% as of 2021 and strengthening the Market Stability Reserve. legislative events permitting the setting up of the market in 2011 (see also Chapple et al 2013). They find higher negative markets reaction for material and financial sectors, and only significant ARs for firms' direct carbon exposure (covered by the tax) and not to indirect emissions (not covered by the tax) 39 .
To wrap up, three major aspects of this regulatory literature can be noted. First, it aims at analysing the markets' reaction to announcements that frame the entire regulatory process. Indeed, those event studies cover: upstream, day-to-day and downstream regulatory tools. Upstream, they analyse the implementation or change of policies. Closer to firms' daily activity, they study various regulatory tools that aim to motivate companies to adopt more responsible practices, such as public disclosure or ETS mechanisms. Downstream, they consider repressive legal measures in the event of non-compliance. Therefore, this literature analyses all the longitudinal aspects of the regulatory process. Another important aspect of this evolution is the gradual shift from a national (public disclosure, legal actions and certain policies) to an international perspective (COP and ETS) of environmental issues. It can be seen that studies on emissions of polluting or toxic products with localized effects (TRI) have gradually given way to studies on GHG emissions that need to be tackle proposing global solutions. Finally, in the light of this literature, stock market reactions appear to be relatively consistent since public disclosure, legal sanctions and announcements of policy changes mostly induce negative ARs. Nevertheless, most of these effects are small, meaning that they do not seem to result in sufficient financial incentives for firms to improve their environmental performance. This does not mean that it is necessary to fundamentally improve those tools since they already do have wealth effects. However, in our opinion, regulators should hold on to this toolbox and significantly impose stricter standards and sanctions at a faster pace.

CSR
This section reviews the event studies about firms' social responsibility, in particular environmental corporate news and corporate environmental practices. With the course of time, academics have started to consider an increasingly wider set of environmental events. Moreover, it shows that the papers have 39 Also focusing on the Australian market, Ramiah et al (2013) examine the introduction of 19 prior bills between 2005 and 2011 regarding the reduction of carbon pollution (including related to ETS) and renewable energy policies. They find strong market reaction in particular once the Australian government decided not to commit to the Copenhagen Accord objectives. They record a 31% decrease in cumulative abnormal returns for the alternative energy sector. become less and less interested with the impact of 'events' strictly speaking (such as accidents, oil spills, lawsuits, new regulations), but more and more by the impact of 'announcements' , sometimes without a specific tangible outcome. The pioneering papers (Laplante andLanoie 1994, Klassen andMcLaughlin 1996) include a broad set of both negative and positive environmental news. These studies were followed by a large number of papers focusing on corporate disclosures, with-unsurprisingly-a positive tone. Lastly, the most recent approaches take advantage of the abundance of data available to conduct more detailed studies on the type of news (or ads) that can affect stock prices.

Environmental news
The studies discussed in the previous two sections primarily focus on one specific type of events. However, several studies also consider the impact on shareholder wealth on various environmental events, often less dramatic, and using large sample of both negative and positive events. Table 7 provides and encompassing overview of the main features of studies addressing the stock market response to various environmental news regarding corporations.
As far as we are aware of, Laplante and Lanoie (1994) were the first to conduct a study on both negative and positive environmental events. They study the impact of 47 events published in two Canadian newspapers from 1982 to 1991. They consider a various set of eco-harmful events (accidents and legal actionssee also tables 3 and 5), but also eco-friendly investments. Overall, they found significantly negative ARs for firms on the announcement of both suit settlements and on investment date.
In the same spirit, Klassen and McLaughlin (1996) assess firms' corporate environmental performance by using green awards, and several environmental crisis. They consider companies listed on the NYSE or the AMEX over the period 1985-1991 and examine both positive and negative events. For a sample of 22 negative events (oil spill, gas leak, explosion and other incidental pollution-see also table 3), they find statistically significant ARs of −0.8%. For their sample of 140 positive events (announcement of an environmental award), they find statistically significant ARs of 0.82%, that is an average increase of over US$80 million per event. They interpret this smaller reaction, in absolute value, as 'scepticism' from shareholders toward the firm's true commitment to environmental projects. Nonetheless, this finding demonstrates that financial markets may positively value strong environmental commitments. Rao (1996) uses a sample of 14 firms, which were the subject of an article in The Wall Street Journal over the period 1989-1993 that denounced unethical corporate conduct towards the environment. The results suggest that the stock market responded very much since average AR reached −5.3% on the day of announcement 40 . Dasgupta et al (2001) collected environmental news from newspapers in Argentina (La Nacion), Chile (El Mercurio), Mexico (Excelsior) and the Philippines (The Manila Bulletin). They show that stock markets do react negatively 41 to citizens' complaints targeted at specific firms (see also tables 5 and 8), and positively to the announcement of rewards and recognition of superior environmental performance. Lorraine et al (2004) study the impact of 23 fines for environmental pollution and nine various positive environmental news from 1993 to 2000 to assess ecofriendly corporate performance extracted from both UK public agencies and newspapers. Their sample size limits their findings but overall, they find a significantly negative market reaction to fine settlements until one week after the event release. Regarding positive news, they obtained a small but significantly positive market reaction. Hsu and Wang (2013) constructed two textual indicators to observe how markets react to news related to GHG emissions of 1345 firms from 1989 to 2008. Their first indicator corresponds to the total number of positive and negative terms in newspapers. The second indicator is the total number of negative terms minus the positive ones divided by the total number of positive and negative words. Hence, they show that shareholders react favourably to negative media reports about companies' exposure to climate risk. They conclude that shareholders view climate change investments as an expensive cost. However, this reaction is weaker for firms that pollute the most and for the ones, which have poor environmental performance. Jones and Rubin (2001) study a broad set of environmental events. Their sample is composed of 98 negative environmental events reported in The Wall Street Journal between 1970 and 1992 in which electric power companies or oil firms with listed stocks were involved. The events selected by Jones and Rubin (2001) did have to meet two criteria. First, the event must have had a negative environmental impact as the result of the actions of a producing division of the firm. Second, the event must not have affected the quality of the firm's physical product. They selected events that produce ill will, but do not affect the quality of the firms' final products nor break implicit labour or supply contracts. Overall, they find  that the capital market response was not statistically significant. Some papers also examine the case of environmental frauds, in particular in the automotive industry. In September 2015, the EPA revealed that Volkswagen fraudulently modified its diesel engines in order to comply with emission regulation. This environmental fraud resulted in the recall of 600 000 vehicles in the United States, and lawsuits against the company. Fracarolli Nunes and Lee Park (2016) examine 33 US auto-manufacturers after the scandal and report strong negative spillovers. Wood et al (2018) consider a broad range of 41 automanufacturers failures between 1984 and 2016. They report an average loss of 1% and an increased suspicion after the Dieselgate event. They claim that the stock market's response should encourage companies to adopt more ethical behaviour, at least as part of their macro-marketing strategy. However, it is not clear whether the penalty is severe enough.
More recent studies with various CSR events rely on much larger sample sizes: hundreds or even thousands (environmental and non-environmental) events which allow for the examination of their influence on (almost) day-to-day basis 42 . Flammer (2013) has a sample of 117 eco-friendly events and 156 ecoharmful events extracted from The Wall Street Journal over the period 1980-2009. She finds that announcements of eco-harmful corporate behaviour lead to negative ARs (−0.65%), and that eco-friendly corporate initiatives generate positive ARs (0.84%). Her results suggest that companies have been increasingly penalized for irresponsible behaviour towards 42 Other examples of studies which examine the impacts of CSR events, with a focus on environment, are Deák and Karali (2014) The data are extracted from the KLD database. The results confirm that negative news is followed by a stock price decrease, while the impact of good events is positive only in cases of poor stakeholder relations. In the content analysis, they show that investors react more strongly to ESG news containing more economic and legal information. Capelle-Blancard and Petit (2019) investigate the stock market's reaction to about 33 000 ESG news flashes from 2002 to 2010 provided by Covalence-EthicalQuote, which targets 100 multinational companies. The interesting feature of this paper is that their extensive database allows examining the impact of day-to-day environmental news. They find that firms coping with negative ESG events experience a low but statistically significant drop in their market value, namely −0.1%. In contrast, companies coping with positive events do not experience any significant change in their market value. In addition, it seems that stock market penalties do not vary significantly over time. Their results also indicate that existing firm reputation mitigates the losses and that market participants only react to information disclosed by the media and not to firm's own press releases. Finally, they establish that losses are larger when there is a cultural proximity between shareholders and the event, and when the content of the environmental news has a quantitative, an economic or a legal dimension.

Corporate practices
While most of the debate focuses on the damage that companies cause to the environment, some companies can be proactive and aim to limit their negative impacts and reduce their ecological footprint.  , and green philanthropy. We complement this classification by including rankings and ratings. These are both delivered by third parties and permit to adopt an assessment of the firms' perceived 'greenness' . However, rankings (often) classify best performers while ratings are more inclusive. Nonetheless, they both slightly depart from green certifications since they do not request firms to follow a series of norms to keep being accredited. Please be aware that we also include self-reported green/environmental annual statements in the CEIs. We report the key features of studies after environmental awards, rankings, ratings and certifications in  Klassen and McLaughlin (1996) and Dasgupta et al (2006b) were the first to consider green awards delivered to firms with good environmental performances. Dasgupta et al (2001) refer to this through the terminology of 'explicit government recognition' . They find they associate with positive wealth effects. In the same vein, rankings are classifications of top environmental performers made by third parties. In this regard, Yamashita et al (1999) focus on firms' environmental conscientiousness scores published in Fortune magazine in 1993. The article reported '10 Leaders' , '10 Most Improved' and '10 Laggards' according to their environmental performance from 130 of the US's largest manufacturing companies. It shows that firms did not experience any statistically significant ARs, albeit that the signs are those expected: 'The Leaders' and 'The Most Improved' companies experienced positive ARs while the 'The Laggards' had ARs on the event day. Takeda and Tomozawa (2008) analyze the stock price reaction during 1998-2005 of the annually published environmental management ranking of the Nikkei newspaper. They find negative ARs from 1999 to 2000 and positive ones after 2003. To explain this market reversal, they consider that the ratification of the Kyoto protocol and the establishment of the Ministry of the Environment have been strong signals, which are thought to have incentivized this market sentiment reversal. However, using the same database but controlling with a GARCH model for the heteroskedasticity of the residuals, Yamaguchi (2008) shows 43 Relatedly, several papers examine the impact on firm's market value of being include in (or exclude from) socially responsible (SR) indices. However, since most of these SR indices are based on ESG criteria, and do not focus on green indices, we did not consider these studies in our sample. See, for instance, Capelle-Blancard and Couderc (2009)   that the market reacts positively for the top ranked firms, while it penalizes the lowest ranked ones. The Newsweek's 'Green Score' , which publish several rankings evaluating US or the world largest companies' green performance from 1 to 100 (best performer), is also extensively used. In 2010, the Green Score consisted of a weighted sum of three subscores: 'Environmental Impact Score' , 'Green Policies Score' (GPS,) and 'Reputation Survey Score' (RSS) with a weight of 45%, 45% and 10% respectively 44 . Murguia and Lence (2015) find an increase in the market value for the world's 100 highest rated firms. This improvement is especially pronounced for the top 50 firms, non-US companies, and the firms from less polluting sectors. By considering the 2009 Newsweek green ranking of the largest 500 US firms, Cordeiro and Tewari (2015)  Finally, green ratings are wide-ranging third party assessment of firms' perceived 'greenness' . Cellier and Chollet (2010) and Dorfleitner et al (2018) examine the impact of the publication of those ratings on firms returns. They take a 'best-in-class' approach by comparing the financial performance of the best and worst eco-friendly firms. Their finding show 44 The EIS examines historical environmental performance such as GHG or SO2 emissions or water use. The GPS is an index provided by MSCI to assess the firms' managerial performance on five major issues: climate-change policies and performance, pollution policies and performance, product impact, environmental stewardship, and management of environmental issues. Finally, the RSS is an opinion survey of CSR professionals, academics and other experts. that the worst environmental performers have higher returns.

Certification
Several types of environmental certification aim at improving 'green' management practices. They usually support the implementation of green supply chain. The International Standardization Organisation (ISO) has been the most commonly covered certification. Actually, the literature studies particular ISO certification regarding the '14000' category because this evaluates firms' green manage- Furthermore, Bouslah et al (2010) cover three types of forest certifications and the ISO 14001 certification announcements in the US and Canada. For the short run, they find insignificant market reactions for forest certifications. However, for the long run (for which event studies are not very suitable though), they arrive at a negative impact. Their results also show that financial markets seem to react differently according to the issuer of the certification. In particular, shareholders seem to penalise certifications delivered by private entities, while they do not significantly respond in case NGOs are involved in the certification process. Zhang et al (2017) use the announcement of the clean development mechanism projects (CDM) certifications in China. The CDM is a project under the Kyoto Protocol, which allows signatory developed countries to purchase certified emission reductions from developing ones. To do so, high-income countries can buy the certificates on financial markets or invest in projects aiming at reducing developing countries' GHG emissions. Zhang et al (2017) find that financial markets assess CDM certification announcements positively, but this effect tends to decrease over time.

Environmental reports
Environmental reports refer to the publication by the firms themselves of environmental information in their financial reports. In particular, several scholars have scanned companies' 10-K, 10-Q or sustainability reports for their influence on stock market returns. For example, Yusoff et al (2006) use this methodology on 152 Malaysian firms and find that the disclosure in 2003 of environmental reports did not significantly enhanced the companies' share prices. However, he finds that profit and size of the firms are relevant motives for firms to disclose environmental reports. Likewise, Nossa et al (2009) analyse the balance sheets of 100 largest companies quoted on the Sao Paulo Stock Exchange from 1999 to 2006 to construct a social and environmental indicator. They did not find any significant market reaction. Reddy and Gordon (2010) study the publication of sustainability or environmental reports from firms in Australia and New Zealand from 2003 to 2009. They find a positive and significant market reaction to sustainability reports being published. However, they do not find any specific market reaction in case the companies publish environmental reports. Wang et al (2019b) examine several types of announcements and reports relative to firms' environmental projects. They find a weak but positive share price increase when a third party actually recognize the greenness of the firms. Nonetheless, financial markets do not react to announcements of any action or reports published by the firms (Wang et al 2019b). They also shows that more socially responsible and transparent firms regarding their CSR and environmental projects have higher returns. Nonetheless, the results also suggest that once confounding news is accounted for, the market reaction is firstly induced by the firm's financial performance, and then by its environmental efficiency. Further, Wang et al (2019b) find that financial markets put a higher weight on forward-looking announcements, even though this kind of action does not always translate into concrete projects.

VEP
VEPs, which are programs followed discretionary by companies to go beyond compliance, in order to exhibit their commitment to environmental issues. For example, the US EPA has launched the national environmental performance track (NEPT) Program in 2000. The goal of this program is to improve and assess firm's environmental performance through a series of indicators relative to climate change, clean water and land preservation. It also incentivizes firms to publish annually their progress to achieve their objectives. Yu (2012) analyses how shareholders interpret firms' announcements when signing up to the NEPT program and finds a significant and positive market reaction. He notes that companies that have invested heavily in human capital have been the main beneficiaries of these announcements. Once part of the program, these investments allow attracting highly qualified people capable of increasing the company's productivity, which can be beneficial for the company's image.
Keele and DeHart (2011) conduct a highly similar study regarding another institution launched in 2002, namely the US EPA Climate Leaders program. This program is to help firms keep track of and manage their GHG emissions. Using the announcement of the partnership between the companies and the program, they find insignificant results on the announcement day in the short run and significantly negative returns with a wider event window of two or three days. Fisher-Vanden and Thorburn (2011) study the market reaction to the integration and to subsequent announcements related to these programs. They consider the Climate Leader and the Ceres program. Ceres is a project created in 1989, which follows general principles regarding ecological awareness and accounting. In this regard, Fisher-Vanden and Thorburn (2011) find significant negative returns for both joining the Climate leader program and announcing their objective to reduce their GHG emissions. Regarding the Ceres program, they obtain insignificant ARs. Overall, Fisher-Vanden and Thorburn (2011) establish that trying to tackle environmental issues and climate change through environmental programs will not be effective in the US, because GHG emission reduction are considered to conflict with profit maximization 45 .
VEP studies are also used to consider GHG emissions reduction, for example along the carbon disclosure project (CDP). Inspired by the Kyoto Protocol of 1997, GHG emissions, in particular CO 2 , have been targeted for their effect on climate change. In this regard, the CDP has been launched in 2002, in response to the request of a consortium of 300 institutional investors, who asked the world's 500 largest companies to voluntarily disclose their GHG emissions and their objectives to reduce these. However, such a program is not mandatory and its impact is likely to be limited. Kim and Lyon (2011) find no effect on stock market value associated to the participation in the CDP, until Russia's ratification of the Kyoto Protocol, which caused the Protocol to go into effect. Dam and Petkova (2014) consider the participation of firms in the environmental supply chain sustainability programs of the CDP and find a low but 45 Scholtens and Dam (2007) study the effect of banks signing up to the Equator Principles. This VEP stipulates social and environmental requirements regarding international project finance. They establish that banks who sign up are not penalized by their investors. significant market penalty. Griffin et al (2017) consider a sample of nearly 2000 SEC 8 K filings that refer to GHG emissions and do not find any difference in the market response for CDP disclosers compared to CDP non-disclosers, which suggest that disclosure to the CDP makes no difference to the way shareholders assess the impact of GHG emissions on firm value. Lee et al (2013) show that markets sanction Korean companies that disclose their carbon emissions through the CDP. Foster and Gutierrez (2013) claim that mandatory environmental standards and voluntary certification programs are less substitutes than complements. To support their argument, they consider four Mexican firms for which a Clean Industry Certificate was granted between 2003 and 2007. They find a positive reaction of the stock market to the news about the certificate being granted.

Green investment
Without being part of any VEP or having committed to a specific program based on a certification, firms can also discretionarily decide to invest in green projects. In this regard, Gilley et al (2000) distinguish process-and product-driven environmental initiatives. Process-based initiatives are an attempt at greening the production process; they include recycling, waste management, and improved production design and delivery system practices. As a result, they limit supply chain costs, improve input management and business organization, reduce the use of hazardous materials, and decrease the likelihood of accidents. Since these investments are internal and are not widely communicated, they have less effect on reputation. Product-driven initiatives occur when firms improve or create goods and services that explicitly account for environmental externalities. These improvements are less significant due to less environmental impact than process-based innovations and can be used for greenwashing (Bénabou and Tirole 2010). For example, Hall and Rieck (1998) test the market reaction of donations, social policy, recycling (so process-driven initiatives), and environmentallyfriendly products. They find that shareholders value both positively environmentally-friendly activities and recycling. A positive impact is also found by Nagayama and Takeda (2006) Gilley et al (2000) find no market reaction to green process-driven investments, but a significant increase of the share price for productdriven initiatives. In contrast, Knowles-Mathur and Mathur (2000) find no significant market reaction to green products, recycling, or appointments of environmental policy managers, and find that green promotion drive downward share prices. Mixed results are also found by Callado-Muñoz and Utrero-González (2008) for the Spanish market. Halme and Niskanen (2001) examine green investments by the Finnish forest industry from 1970 to 1996 and report an instantaneous negative impact, but followed by a price recovery. Unlike previous findings based on developed countries, Lam et al (2016) find that in China, shareholders react negatively to CEI announcements.

Conclusion
Our study reviews four decades of academic research after the response of shareholders to environmental accidents, policies, and responsibility. This is motivated by the debate about whether to regard the financial market as a friend or a foe in the struggle against environmental degradation (Tietenberg 1990, Acemoglu, 2019. Our review shows that academic research does not occur in isolation but speaks to and with society and reflects upon and engages with the transformations it undergoes. For this 40 year period, we detected three different streams in the literature (finance, regulation, CSR), that derive inspiration from each other and result in a better understanding as to how financial markets respond to environmental performance of firms and industries. The first stream reflects on the occurrence of several dramatic industrial accidents. It relies on the finance literature driven by the EMH and provides a first estimate of the losses incurred by shareholders. The second stream, motivated by the potential magnitude of these losses, sees environmental economists taking over this topic. They propose using market forces to incentivize firms to protect the environment. This literature suggests that the traditional 'command-and-control' instruments should be complemented by market-based incentives when public disclosure is warranted. Hence, it is up to the shareholders to sanction firms, instead of public authorities. The law and economics literature shows in some cases market penalties may be even higher than fines. The third stream starts in the new millennium. Here, the topic becomes more business-oriented and the concept of market efficiency gives room to that of CSR. Further, the tone of the debate becomes more positive. In addition, the focus expands to a multitude of environmental corporate practices. Novel big data techniques support this line of research. Thus, in the 40 year period reviewed, the perspective changes from financial markets to policy, and from policy to business, as such reflecting broader societal processes.
Our literature review provides a structured setting. It is not only a narrative, but also provides analysis and additional materials, which we hope will be useful to the scientific community for further research, strategy and policy. In this regard, the references collected here are available in the form of an online open access bibliographic database 46 . We aim to update this database. This material is designed to be used as a source and point of departure for other synthesis studies, such as meta-or bibliometric analysis, and intended to help especially young researchers progress and improve their research design and to contribute to the academic literature and societal debate.
In this review, we observe that with many environmental events, there is a response from stock market participants and that the value of the firm is affected. In particular, this is the case with ecoharmful events. Here, we find that the AARs are −2.0% in the event window. However, the median of these returns is −0.6%. For eco-friendly events, the response is even more modest. They yield on AARs of +0.2%, but the median is 0%.
The event study methodology is well-suited to establish the short-term response of the participants, but not very useful to assess the long-term impact. However, whether the response from the stock market participants is adequate to bring about change is much more difficult to assess. This is because environmental accidents are economic externalities, which by definition are hard to price. Further, applying the methodology assumes that all financial market participants have thorough understanding of the ramifications of these occurrences. This is hard to believe given the huge complexity of ecosystems, both on a local and global scale. We establish that there is a significant response, but it is small from an economic point of view. We cannot even be sure whether the direction of the market response is accurate due to the lack of appropriate metrics. The latter relates to the incongruence between management sciences and natural sciences regarding the use of denominators; the former keeps to monetary units, whereas the latter requires a smorgasbord of metrics and indicators. In the literature studied here, it is monetary units that signal the appreciation of market participants. We feel this a very poor proxy for the actual impact on the natural habitats of plants and wildlife, on ecosystems, and on health, wellbeing and livelihoods of people affected by a case such as the Brumadinho dam disaster. With the development of the study of ecosystem services and willingness-to-pay instruments, the literature tries to put a price tag on these items. This is highly useful, but we feel that using monetary values alone is not helpful to address the degradation of ecosystems as it presumes the existence of trade-offs. Monetizing and making trade-offs is the hallmark of economics and management, but we do not think it useful to transpose this tool to the natural environment and to social relations. This concern is supported by our analysis, which shows that 46 Bibliographic database: https://adriendesroziers.com/data-anddatabases/.
there is no convincing evidence that financial markets can be entrusted with the responsibility of disciplining companies for their environmental performance. This results from the fact that there is insufficient theory and data as to the effects of externalities in relation to firm and industry performance and the role of financial markets and institutions in this regard. Therefore, we think multidisciplinary, and preferably interdisciplinary, research and theory is required to progress and to allow for addressing and internalizing the economic externalities.
In the beginning of our review, we questioned the place and role of financial markets in the framing of environmental policies. Our analysis does not provide support for the notion that financial markets can be put central in this regard. The often very limited market response to environmental and social harm suggests that financial market participants are not adequately equipped to gauge the impact of market externalities. Therefore, it does not seem wise to entrust them with the responsibility of disciplining companies' environmental performance. We hope and expect our review will stimulate the research about the effects of economic externalities. In particular, we encourage researchers to include non-financial performance in the analysis and to investigate the long-term effects of such externalities. Furthermore, we think it is important to relate this research to the context of the problems analysed and to acknowledge that research interests as well as methodologies used are subject to broader societal challenges.
To wrap up, it shows that we end with more questions than we started with. We establish that the financial market reaction may complement regulation but that it cannot be seen as an appropriate substitute. It shows that financial markets find it hard to translate environmental concerns to financial consequences. As a result, the market disciplining approach to environmental protection is limited, let alone that it is a force to the good. Therefore, we advocate a critical look at the societal value of CSR practices and the role of financial markets.

Data availability statement
All data that support the findings of this study are included within the article (and any supplementary files).