The contagion effect of environmental violations The case of Dieselgate in Germany

We examine how environmental violations affect the stock returns of the violating firm and how these financial implications then spread to industry peers. Volkswagen's diesel emissions scandal (Dieselgate) and the German automotive industry serve as a seminal case for the examination. Research often limits examinations of corporate environmental scandals to the primary event announcement. Yet the Dieselgate scandal exhibits a processual character that requires the examination of multiple events over time. We identify 10 Dieselgate events and employ event study methodology to detect abnormal stock reactions. Based on agency and signaling theory, the results indicate that Dieselgate has harmed the stock returns of Volkswagen and its industry peers substantially. Surprisingly, Volkswagen suffered financial damage only upon the initial event of Dieselgate. Subsequent events had significant effects only on industry peers. These findings contribute comprehensively to the research of environmental misconduct and provide valuable implications for practitioners.

. The EPA stated that the NO x emissions of these particular diesel vehicles were 10-40 times higher than allowed (Barrett et al., 2015). Although Dieselgate represents the latest automotive environmental violation scandal, using defeat devices for emission testing has a history. The EPA enforced penalties against other automotive firms (e.g., Honda and Ford) for emission manipulation in the past (Schaeffer, 1998), making the fallout of Dieselgate repetitive. To conclude, Dieselgate represents a corporate scandal combining dismal ecological performance with fraudulent characteristics, which provides broader insights into the theoretical understanding of corporate environmental violations. Although the emergence of management's unethical behavior exhibits links to agency problems, a firm's unethical behavior might signal common business practices of misconduct within the industry to the stakeholders. This is particularly so for interrelated and similar industries such as the German automotive industry. In these settings, the impact of major corporate scandals may not be limited to the guilty firm, and the risk of industry peer contagion increases (Laufer & Wang, 2018).
However, we know little about Dieselgate's financial effects on VW and its industry peers after new information became available.
All prior event studies on Dieselgate limit their examination to the EPA announcement on September 18, 2015. Yet Dieselgate comprises several subsequent events; some of them are still ongoing (e.g., diesel vehicle bans in cities). Moreover, the German automotive industry provides a special setting for the analysis of Dieselgate. Germany's car manufacturing industry is closely linked, and stakeholders overlap heavily, making Dieselgate a German automotive rather than a purely VW problem. Then, it represents one of Germany's biggest economic sectors, employers, and institutions (The Economist, 2018). Additionally, Germany's automotive industry espouses strong interlocks among and across supply chains (Barthel et al., 2015). Nonetheless, current studies do not analyze Dieselgate's impact on stock returns in a manner that accounts for both further events and the contagion effect. Thus, we conjecture that the impact of environmental violations deserves special attention among German car manufacturers. This motivates us to pose the following research question: RQ: How do Dieselgate announcements affect the stock returns of VW and its industry peers (contagion effect)?
Methodologically, we conduct an event study (Brown & Warner, 1985, 1980MacKinlay, 1997). This methodology allows us to evaluate the impact on stock returns of Dieselgate events on both an individual and a group level. Thus, we extend the understanding of Dieselgate's financial impact in two ways: First, we use Dieselgate events in direct relation to VW (individual events) to measure the comprehensive reaction of VW's stock. Second, we use Dieselgate events affecting the overall automotive industry in Germany (group events) to determine the stock price reaction of industry peers individually and for a group. We select events based on how meaningful the media portrays them, as media has a strong impact on the public perception and the value relevance of a scandal (Carberry, Engelen, & van Essen, 2018;Clemente & Gabbioneta, 2017;Xu, Zeng, Zou, & Shi, 2016).
Our findings suggest large and highly significant, negative abnormal stock returns for VW on the initial EPA announcement event. Deviating from our prediction, the following events remain statistically insignificant for VW, indicating that the markets have anticipated and priced the full extent of VW's misconduct.
However, the contagion effect analysis of Daimler, BMW, and the car manufacturing group as a whole (portfolio of VW, Daimler, and BMW) reveals major significant and negative abnormal stock returns for subsequent events.
Our study makes four substantial research contributions to the literature on the financial effects of corporate scandals: First, we provide evidence for a strong horizontal contagion effect at the same level of the supply chain. Second, our analyses demonstrate that the full extent of the contagion effect becomes more visible by considering subsequent events over a longer timeline. Third, we combine information economic theoretical frameworks, which help to better understand the stock market reaction to VW and the contagion effect. We conjecture that two related theoretical perspectives are necessary to grasp the financial impact of the corporate scandal itself (agency theory) and the financial contagion effect on industry peers (signaling theory). Fourth, we show that the "guilty by association" effect holds in the specific case of Dieselgate, where the heavy industry peer contagion effect (sum of stock losses for Daimler and BMW for the specification with the highest significance, respectively) exceeds the financial loss of VW by 54.85%. For practitioners, our study holds two important contributions: First, we illustrate that violating environmental regulations to obtain business advantages might not payoff and the downsides might be overwhelming. Second, we conclude that firms that are too interwoven and similar in their business models are subject to becoming "guilty by association" and, thus, should actively ensure a differentiation from industry peers to avoid financial contagion.
The remainder of this paper is organized as follows: Section 2 provides the background of our analysis including the theoretical foundation of the respective stock market reactions as well as the relevant literature, both synthesized to derive the research hypotheses.
Section 3 specifies our event study methodology and data. Section 4 presents the empirical findings. Section 5 provides a discussion, critical acknowledgments, and a conclusion.

| BACKGROUND
To derive the research hypotheses of this study, we develop a framework that combines two related theories from the school of information economics: agency theory and signaling theory.
2.1 | The direct effect of Dieselgate: An agency theory-based hypothesis Agency theory assumes that interests and utilities between the principal and the agent, assigned to act on behalf of the principal, are not necessarily aligned, which may lead to agency problems such as moral hazard (Eisenhardt, 1989;Fama, 1980;Jensen & Meckling, 1976;Ross, 1973). Corporate misconduct meant to obtain a competitive advantage over industry peers can be attributed to this notion (Carson, 2003). Industry peers' inability to replicate the emission values of VW diesel vehicles helped VW to penetrate the US market aggressively and to become one of its leading diesel vehicle vendors (Barrett et al., 2015;Fracarolli Nunes & Lee Park, 2016).
The embedded quest for aggressive growth within the corporate culture, in line VW's strategy to become the leading automotive firm (Armstrong, 2017), exhibits clear characteristics of shareholder primacy, a corporate maxim well discussed by law scholars (Lee, 2005;Smith, 1998). In settings of shareholder primacy, all corporate actions target the maximization of shareholder value that might imply diminished moral responsibility and "short-termism" (Burkert & Lueg, 2013;Smith & Rönnegard, 2016;Stout, 2013). Supporting the shareholder primacy perspective, a strong performance-driven compensation component for the management fostered VW's quest for aggressive growth. This further created incentives for short-term orientation and unethical behavior to maximize shareholder value and personal compensation (Li, McMurray, Xue, Liu, & Sy, 2018).
When tests detected the fraud in 2014 and the EPA initiated the enforcement procedure against VW, investors anticipated the potential financial losses for VW and reacted accordingly. However, the release of value-relevant information did not end with the EPA announcement. During the ongoing course of Dieselgate, the media has disclosed new and relevant information over time and portrayed Dieselgate events prominently and, thereby, VW's misconduct, responsibility, and its consequences credibly. By that, these information become particularly value relevant (Carberry et al., 2018). This myopic perspective on shareholders, which led to Dieselgate, had a significant impact on other stakeholders whose demands fell behind in VW's growth strategy. Hill and Jones (1992) provided an extensive framework in their stakeholder-agency theory, in which stakeholders assume the principal's role. Thereafter, every single group of stakeholder has specific demands on the management, which, if satisfied, lead to superior business performance because of better access to stakeholders' resources. The inversion of this argument implies that neglecting the demands of stakeholder groups might induce restricted access to resources. Applying this stakeholder-based framework to VW emphasizes that the fraudulent software implemented did not only hurt the shareholders but many other stakeholders as well.
Customers had to deal with the issue of resolving the cheating software as well as with the decreased market value of their vehicles (Markowitz, Chapman, Guckian, & Lickel, 2017 Employees had to worry about their jobs for the same reason (Müssgens & Peitsmeier, 2016). The government saw their environmental regulations disregarded as well as major public health and environmental damage as a result of the excess NO x emissions (Chossière et al., 2017;Dey, Caulfield, & Ghosh, 2018;Holland, Mansur, Muller, & Yates, 2016;Oldenkamp, van Zelm, & Huijbregts, 2016;Tanaka, Lund, Aamaas, & Berntsen, 2018). These consequences led stakeholders to penalize VW with calls for boycotts (customers), penalty fines (government), and other means of stakeholder activism with implications for the stock price. Thus, the stock market reaction to VW in response to the announcement of Dieselgate should be analyzed in a (stakeholder-) agency context.
In accordance with this theoretical framework, researchers provide extensive empirical evidence on the stock market reaction to the announcement of corporate misbehavior, bad environmental performance, and environmental regulation violations. Gunthorpe (1997) found negative abnormal returns (ARs) for firms that have announced that they engaged in any kind of unethical behavior or that they are under investigation for it. Specifying the notion of unethical behavior to environmental issues, Gupta and Goldar (2005) found that the announcement of bad environmental performance in terms of the Green Rating Project by India's Centre for Science and Environment triggered a negative stock price reaction. Hamilton (1995) and Khanna, Quimio, and Bojilova (1998) complemented the examination of environmental issues by looking at the public disclosure of toxic release information. They concluded that the announcement of toxic waste releases negatively affects the stock returns of the firms involved. Klassen and McLaughlin (1996), Dasgupta et al. (2001), and Flammer (2013) examined further corporate environmental misbehavior announcements such as spills and contaminations and consistently derived a negative reaction by investors to these announcements. As an extension of the examinations of environmental misbehavior, Lanoie, Laplante, and Roy (1998) integrated the legal dimension of violating environmental regulations and investigate the information release that a firm is listed on the "out of compliance" and "of concern" list of polluters in Canada. They report a negative stock market reaction, which is even more pronounced for large polluters. Dasgupta et al. (2006)  For VW, they concluded significant financial losses as a result of the EPA announcement, which are robust against variations in event windows and security types. We complement their research approach by considering a longer timeline (i.e., multiple events) in our examination. On a broader scale, Wood et al. (2018) examined the abnormal stock reaction to 41 car manufacturers' environmental failure announcements (i.e., unethical behavior, deception, and failure to meet standards) for an international set of firms. They found highly significant, negative, mean ARs for the announcement of a car manufacturer's environmental failure with results being robust against variation in estimation models. Thereby, stock losses due to environmental failure announcements following Dieselgate are stronger than those resulting from prior announcements as Dieselgate damaged consumer confidence in car manufacturing substantially and increased investors' risk aversion to environmental issues. We address this increase in investors' risk aversion to environmental issues in car manufacturing in our in-depth analysis of VW and illustrate how investors value VW's stock throughout Dieselgate.
Thus, as expected, the results suggest that the stock market reacts negatively to the Dieselgate announcement by the EPA. From the agency theoretical context, the ongoing emergence of new and relevant information during Dieselgate, and the extensive empirical evidence, we derive our first research hypothesis: H1. The announcement of Dieselgate individual events is associated with negative ARs for VW.

| The contagion effect of Dieselgate: A signaling theory-based hypothesis
Intentional environmental fraud and the corresponding stock market reaction to the fraudulent firm result from existing agency conflicts such as shareholder primacy inside the firm. Additionally, understanding the contagion effect from VW to industry peers requires the consideration of a complementary theory building on different aspects of the same information asymmetry. According to signaling theory, two parties hold different information bases and, typically, one party has an information advantage over the other. The information sender (signaler) decides how to communicate the information to the recipient. The recipient then interprets this information, processes it, and reacts upon through feedback or other means (Connelly, Certo, Ireland, & Reutzel, 2011;Spence, 1973).  employed the signaling theory to derive a theoretical framework that explains the contagion effect of environmental violations. According to them, the environmental violation announcement of one firm reveals the environmental risks of the whole industry as its members share the same (or very similar) technical conditions and the production output. Therefore, the announcement passively signals the inherent industry risk to stakeholders, making them reassess their assumptions about the attractiveness of the industry and the corresponding resource distribution.
Signaling has direct implications for Dieselgate. Dieselgate revealed issues involving compliance with environmental standards in the United States, which had led to environmental fraud to overcome them. More precisely and in line with the notion of moral hazard, VW simply was not able to meet the environmental standards in the United States with its technic and without exceeding the budget and, therefore, decided for fraud to gain a competitive edge over industry peers (McGrath Goodman, 2015). The public announcement of the EPA passively signals the risk of this environmental fraud within the industry to the stakeholders, causing them to question the integrity of the German automotive industry in general. Laufer and Wang (2018) showed that crisis contagion is most likely when firms are similar and share the country of origin, industry, and organizational type (profit orientation, ownership structure, etc.), as well as positioning strategy (high-end vs. low-end orientation). Ouyang, Yao, and Hu (2020) specified this finding for the context of environmental misconduct and illustrated that stakeholders tend to categorize firms by similarity.
Looking at the case of Dieselgate, most of the crisis contagion criteria fit German car manufacturers. They have the same country of origin as well as industry, organizational structure (Barthel et al., 2015;Fasse, 2019), and, to a large extent, positioning strategy (GTAI, 2018; The Economist, 2018); thus, the risk of contagion from VW to other German car manufacturers is high because of their perceived similarity.
This financial contagion is, most likely, a consequence of investors' learning. As pointed out by Bebchuk, Cohen, and Wang (2013), investors tend to adapt to changes over time within a learning procedure and take this learning into account in their investment decision. As above mentioned, Dieselgate passively signaled the risk of misconduct and a corresponding EPA enforcement for other German automotive firms emphasized by the strong interlocks between German automotive firms. This passive signal then triggered a learning process at the investors who saw themselves exposed to both the risk of Daimler and BMW being involved in the scandal and the risk of financial losses. This, in turn, led them to sell their stake in these firms. Figure 1 illustrates our underlying theoretical framework composing of agency and signaling elements.
Several authors provide empirical evidence for a contagion effect from corporate scandals or incidents. In 2010, BP's Deepwater Horizon oil platform caught fire, which led to a massive oil spill in the Gulf of Mexico. Event studies on the Deepwater Horizon oil spill by Humphrey, Carter, and Simkins (2016) and Sabet, Cam, and Heaney (2012) illustrate how the incident affects the overall oil and gas industry. Even though the incident does not represent intentional environmental fraud, the studies provide insights into the existence of a contagion effect from environmental incidents.
The Deepwater Horizon incident, with its far-reaching consequences on drilling in the Gulf of Mexico, impacts not only the stock of BP and other firms directly involved in the oil spill but also that of unrelated oil and gas exploration, drilling, equipment, and services firms. The market is, however, able to differentiate between firms as the spill affects oil and gas firms not involved in offshore drilling (e.g., pipeline companies) less heavily. Regarding intentional environmental misconduct,  specifically examined the contagion effect from environmental violations using 59 announcements across industries by China's Ministry of Environmental Protection for their event study. They concluded a negative intraindustry contagion effect to 282 industry peers, which is more pronounced for firms in environmentally insensitive business areas (e.g., coal mining). Jin, Cheng, and Zeng (2020) dealt with a similar topic, examined environmental misconduct in environmentally sensitive industries (extractive, chemical, steel, and building materials industries) in China, and derived significant negative reactions to their public announcement by China's Ministry of Ecological and Environmental Protection for the misconducting firm. More interestingly, however, they revealed a notable spillover effect from the misconducting firm to its industry.
Hence, an industry peer contagion effect is observable for various incidents and scandals irrespective of the type (environmental scandal, accounting scandal, etc.).
Several authors detect a contagion effect from Dieselgate using the EPA announcement event. Fracarolli Nunes and Lee Park (2016) examined 33 automotive firms in a US-based event study. Based on the EPA announcement event, they divided their sample of firms into car manufacturing and supplier companies. They found large and significant abnormal stock losses for two US car manufacturing firms and eight out of 27 suppliers. Both findings are robust to variation in event windows and so confirm the existence of a financial contagion effect from VW to car manufacturers and suppliers in the United States, which is however limited to the EPA announcement.
Providing further evidence for the contagion effect from the EPA announcement, Griffin and Lont (2018)  H2a. The announcement of Dieselgate group events is associated with negative ARs for Daimler.
H2b. The announcement of Dieselgate group events is associated with negative ARs for BMW.
F I G U R E 1 This figure displays our theoretical framework. It illustrates how Volkswagen is interwoven with its industry peers, how Dieselgate passively signaled the risk to the investors of Daimler and BMW and triggered a learning process, which, in turn, led to the sellout of Daimler's and BMW's stocks H2c. The announcement of Dieselgate group events is associated with negative ARs for the group of car manufacturers (Daimler, BMW, and VW).

| Events study methodology
According to Fama (1970), in conditions of semistrong information efficiency, stock prices adjust to the announcement of publicly available, relevant information (e.g., stock split announcements; Fama, Fisher, Jensen, & Roll, 1969). Exploiting this information efficiency to test our research hypotheses, we apply event study methodology (Brown & Warner, 1985;MacKinlay, 1997) based on daily stock returns (Brown & Warner, 1980).
In line with the underlying hypotheses, we divide this study into two parts: The first represents the event study on VW based on individual events (H1), and the second represents the event study on Daimler (H2a), BMW (H2b), and the group of car manufacturers (H2c) based on group events. We apply the event study methodology proposed by MacKinlay (1997) and accordingly define event windows, estimate normal stock returns, calculate ARs, and test statistical significance.
We first calculate stock returns for the market index and the firms based on the stock prices using the following formula: where R it represents the stock return for i on day t and P it represents the stock price of i on day t. To estimate the expected stock returns, we employ the widely used market model using the broad German "Prime All Share" index as the underlying market: where E(R it ) represents the expected stock return for i on day t; R mt , the market return on day t; β i , the beta factor (risk); and ε it , the disturbance term. As proposed by MacKinlay (1997) where AR it refers to the AR for i on day t. The ARs, being the residuals between expected and realized stock returns, display returns that one cannot explain using the market model and, thus, are a result of the event announcement (Martin Curran & Moran, 2007). Then we accumulate these ARs over multiple days to produce cumulative abnormal returns (CARs) for evaluating the time series for i using the following equation with t 1 ,t 2 being the event window boundaries: Cumulative abnormal portfolio returns (CAPRs) enable conclusions on the average effect of an event on the examined portfolio. Therefore, we compound each security's ARs within the regarded portfolio and calculate their mean (Kothari & Warner, 2008): where CAPR t refers to the CAPRs and N to the number of securities We test the CARs for statistical significance using parametric and nonparametric test statistics. To overcome the overrejection of the null hypothesis due to event-induced variance and cross-sectional correlation, we employ the t-statistic by Boehmer, Masumeci, and Poulsen (1991) adjusted by Kolari and Pynnönen (2010). For nonparametric testing, we apply the generalized rank test by Kolari and Pynnönen (2011), which offers advantages in testing aggregate data (such as CARs) as it is robust against the autocorrelation of ARs and event-induced volatility.

| Data and event description
For this study, we obtain stock price and market capitalization data for the German "Prime All Share" index, VW, Daimler, and BMW from Thompson Reuters' Eikon. We select Daimler and BMW as they represent VW's major industry peers in Germany, which are listed on the same stock exchange.
We hand-collect event data using the continuous Dieselgate  Note: Table 1 depicts the event data and assigns an event ID to each event. We assign events to event panels according to their implications. We assign events 1 and 2 to both event panels, even though they only refer to VW, as they enable conclusions on how Dieselgate immediately spilled over to VW's industry peers. We obtain the announcement date and the event description from Handelsblatt's Dieselgate chronicle.
Dieselgate events following the EPA announcement did not yield any significant results, indicating that these events are not relevant for VW. Despite their financial implications, neither the penalty fee announcements (IDs 3 and 9) nor the announcement that the fraudulent software represents a material defect (ID 10) leads to any abnormal stock returns. According to Bhattacharya, Daouk, Jorgenson, and Kehr (2000), this surprising finding may have different reasons: Markets may be inefficient; markets are efficient but, the news is not value relevant; or markets are efficient, and the news is value relevant, but the market already anticipated these events and priced them beforehand. As the market processes the information of the Dieselgate announcement and reacts accordingly, the nonefficient hypothesis does not appear to be fully convincing. On the basis of the potential financial losses for VW, we assume that these events certainly are value relevant.  Note: Table 2 illustrates Volkswagen's cumulative abnormal returns (CARs) for multiple event window specifications generated by the Dieselgate individual events. We accumulate abnormal returns over the defined event windows to generate the respective CARs (first column of each event window). We employ two test statistics for significance testing. KP represents the parametric t-test by Boehmer et al. (1991) adjusted by Kolari and Pynnönen (2010). GR represents the nonparametric generalized rank test by Kolari and Pynnonen (2011  Note: Table 2 illustrates Volkswagen's cumulative abnormal returns (CARs) for multiple event window specifications generated by the Dieselgate individual events. We accumulate abnormal returns over the defined event windows to generate the respective CARs (first column of each event window). We employ two test statistics for significance testing. KP represents the parametric t-test by Boehmer et al. (1991) adjusted by Kolari and Pynnönen (2010). GR represents the nonparametric generalized rank test by Kolari and Pynnonen (2011). The p values (second and third columns of each event window) are stated in parentheses. *** Statistical significance at the 1% level.   Table 3 illustrates Daimler's, BMW's, and the group's cumulative abnormal returns (CARs) for multiple event window specifications generated by the Dieselgate group events. We accumulate abnormal returns over the defined event windows to generate the respective CARs (first column of each event window). We employ two test statistics for significance testing (second and third columns of each event window). KP represents the parametric t-test by Boehmer et al. (1991) adjusted by Kolari and Pynnönen (2010). GR represents the nonparametric generalized rank test by Kolari and Pynnonen (2011 (Lueg & Radlach, 2016).

| H2c: The group effect
This combination manifests itself in incentives for moral hazard (Eisenhardt, 1989). Thereafter, the management exploits information asymmetries at the expense of the principal, acts unethically (e.g., by violating environmental regulations), and, thus, obtains a business advantage ultimately resulting in higher management compensation (Li et al., 2018). This agency conflict cannot be limited to a pure shareholder-management relationship but rather has to be extended to a stakeholder-agency approach (Hill & Jones, 1992). Dieselgate affects several stakeholders negatively, which leads to pressure on VW's stocks. The emergence of Dieselgate comes with a passive release of information on environmental risks to the stakeholders (Lueg, Krastev, & Lueg, 2019;, which enhances a learning process at the investors and triggers the financial contagion effect.  Table 3 illustrates Daimler's, BMW's, and the group's cumulative abnormal returns (CARs) for multiple event window specifications generated by the Dieselgate group events. We accumulate abnormal returns over the defined event windows to generate the respective CARs (first column of each event window). We employ two test statistics for significance testing (second and third columns of each event window). KP represents the parametric t-test by Boehmer et al. (1991) adjusted by Kolari and Pynnönen (2010). GR represents the nonparametric generalized rank test by Kolari and Pynnonen (2011 Fourth, our findings suggest that being "guilty by association" pronounces when firms are interwoven and have many similarities in their business models.

| Practical contributions
For practitioners, our analyses provide two important insights into the business advantages from environmental violations and the financial impacts once they are uncovered. become to their competition (e.g., by operating the same business model) and the stronger the interlinkage between them is, the higher the probability will be that an industry peer's scandal will affect them.
This, in the second step, implies the contagion effect whereby related industry peers might even suffer stronger financially damaging consequences than the fraudulent firm. Therefore, firms should avoid extensive overlaps in business models and interrelations to assure an effective differentiation should an industry peer be involved in a devastating corporate scandal. A differentiation by explicitly stipulating the environmentally responsible principles in corporate strategy might be a good mean to protect oneself from the scandal-driven contagion effect (Lueg, Lueg, Andersen, & Dancianu, 2016). Thus, the case of Dieselgate provides important lessons for firms regardless of the industry and illustrates that violating environmental regulations to obtain business advantages should be omitted by firms as legal and financial consequences can be devastating. The violating firm, as well as its industry peers, might have to deal with a long-term reputational loss that has the potential to transform formerly highly reputable firms into despised entities in the society.
F I G U R E 2 This figure denotes cumulative abnormal returns (CARs) on the y-axis and the events on the x-axis. We display significant CARs centered in colored and filled bars and insignificant CARs in dashed, uncolored bars. For all events and firms, we select the most significant CARs, respectively. For significant CARs, we display the absolute loss in market capitalization in parenthesis below or above each bar. We assign the events to the respective hypotheses and provide a short explanation of the findings [Colour figure can be viewed at wileyonlinelibrary.com]

| Limitations and future research
The interpretation of our findings is subject to five limitations: These include the limited generalizability to other scandals, our theoretical framework, the potential bias arising from the interdependence of German car manufacturers and the involvement of Daimler and BMW in the Dieselgate scandal, the circumstance that our subsequent events do not represent surprises to market participants, and the negligence of long-term effects as well as investor characteristics.
First, it is questionable whether one can apply our findings to other cases in which the fraudulent firm has rather unrelated industry peers, distinct stakeholders, and in which they serve different customer needs (i.e., different business models). As pointed out, we derived our findings using German car manufacturers in the analysis of Dieselgate, which are deeply interrelated, have a large overlap in stakeholders, and have many similarities with regard to organizational type, market positioning, and so forth. Thus, we conjecture that one can generalize our findings but only to firms with similar business models.
Second, we employ an information economics perspective (agency and signaling) in the analysis of Dieselgate. Our underlying theories, building on how Dieselgate revealed environmental risks to the stakeholders, are able to provide a theoretical explanation for the stock market reaction to VW and the contagion effect. However, diminished legitimacy might be the pivotal issue in other scandals, for example, in the fashion industry (Lueg et al., 2015). As shown by Jonsson, Greve, and Fujiwara-Greve (2009), corporate scandals imply legitimacy losses for the firm involved, which eventually spill over to industry peers when the firms are similar but not necessarily interrelated. Hence, a legitimacy theoretical lens might be more suitable to explain the "undeserved losses" when examining the contagion effect of scandals for similar but unrelated firms.
Third, our statistically significant, negative stock returns for the German automotive industry could partially be a consequence of the interdependence of German car manufacturers and the involvement of Daimler and BMW in Dieselgate itself. In our analyses, we stick to the legal perspective that all three car manufacturers are independent, legal entities. From a business administration perspective, one might argue that the long-term cooperation between German car manufacturers (Barthel et al., 2015) blurs the legal boundaries and reveals discernible interdependence. This might, partly, explain the contagion effect as VW's problems automatically become a problem for the German industry peers through interdependence. Besides, legal authorities later found other German car manufacturing firms guilty of violating environmental regulations. This is truer for Daimler (Delamaide, 2018) than for BMW (Handelsblatt, 2019b), although both are subject to legal prosecution for irregularities with their diesel vehicles. Thus, the involvement of both firms in Dieselgate might have implications for our findings on the contagion effect. However, we favor the interpretation that a substantial part of the contagion effect is rather built on the "guilty by association" effect: The individual analyses of BMW and Daimler reveal that both firms were immediately targeted for the first two Dieselgate events when, at that time, nobody associated them with the scandal. Furthermore, we checked that none of the allegations against Daimler and BMW took place at the same time as any of our defined events. Hence, we conjecture that the risk of a bias coming from scandal involvement is relatively small.
Fourth, following the premise of market efficiency strictly, one might argue that our subsequent events do not hold any new information and that considering them is thus unnecessary. However, we counterargue that, indeed, these events do not represent real surprises as media already portrayed them and they followed the initial announcement; still, our findings provide evidence that investors, in line with our learning argument, had problems to fully grasp the potential consequences for their firms right from the beginning. This, most likely, led them to rethink their investment decision over time as they could not evaluate their risk of being dragged along upfront. Hence, we strongly argue that our subsequent events are necessary for the analysis of Dieselgate.
Fifth, we critically assess that we limit ourselves to short-term effects and do not examine the potential recovery following the financial fallout for VW and its industry peers. Finally, we do not distinguish between different groups of investors and assume homogeneity. Future research should clarify how VW and other automotive firms performed in the long-run following Dieselgate and if different investor groups reacted differently to the scandal.