Focusing Attention on Misconduct

Research in behavioral finance (Barberis & Thaler, Reference Barberis and Thaler2003; Hirshleifer & Teoh, Reference Hirshleifer and Teoh2003) with its connections to more established work in cognitive psychology (Kahneman & Tversky, Reference Kahneman and Tversky1979; March, 1978), emphasizes that investors, like all individuals, face constraints on their ability to access and process information. In the case of misconduct, these limits can be exacerbated by additional information asymmetry created by the firm under scrutiny. Investor reactions to misconduct, therefore, will first depend on whether investors notice it (Barnett, Reference Barnett2014), i.e., whether it falls “within their strategic field of vision and overcome other stimuli” (Bundy et al., Reference Bundy, Shropshire and Buchholtz2013: 359).

Since the media play an agenda-setting role by focusing the public’s attention on specific issues (Carroll & McCombs, Reference Carroll and McCombs2003; McCombs, Reference McCombs2005), the volume of coverage they devote to an act of misconduct can make it more difficult for investors to ignore it. Indeed, prior research has found that the volume of media coverage has an important effect on investor reactions to IPOs (Pollock & Rindova, Reference Pollock and Rindova2003) and earnings announcements (Bushee et al., Reference Bushee, Core, Guay and Hamm2010). In addition, a large literature in management and organizational theory has demonstrated the critical role of the media in exposing corporate misconduct (Desai, Reference Desai2011; Greve et al., Reference Greve, Palmer and Pozner2010; Jonsson & Buhr, Reference Jonsson and Buhr2011). As the volume of media coverage of misconduct increases, therefore, it will focus the attention of more investors and stakeholders, making it more difficult for them to ignore it, and raise their expectations about the likelihood that stakeholders will notice it and impose reputational penalties. Moreover, the firm will become more vulnerable to the dynamics of stigmatization that can damage firms more seriously (Carberry & King, Reference Carberry and King2012; Desai, Reference Desai2011; Devers et al., Reference Devers, Dewett, Mishina and Belsito2009), and invite broader investor and stakeholder attention. We therefore predict that as media coverage of an act of misconduct increases, investor attention will become more focused on the act. Therefore:

Enhancing the Clarity and Credibility of Claims about Misconduct

Although the volume of media coverage will likely focus investor attention, the media can also influence investor interpretations and reactions by how they report the misconduct. Corporate behavior that can be classified as misconduct only becomes misconduct through a process of social construction (Palmer, Reference Palmer2012) in which observers assess and agree about the behavior under scrutiny. More specifically, corporate behavior will not be considered misconduct “unless judgments … are made in comparison to laws that have been written, ethical principles that have been promulgated, [or] norms of social responsibility that have been elaborated” (Greve et al., Reference Greve, Palmer and Pozner2010: 78). The media help to construct specific corporate behaviors as misconduct (Breit, Reference Breit2009) through the types of information they include in their coverage and the way in which they frame the information (Ahmad, Han, Hutson, Kearney, & Liu, Reference Ahmad, Han, Hutson, Kearney and Liu2016; Pollock & Rindova, Reference Pollock and Rindova2003; Rhee & Fiss, Reference Rhee and Fiss2014).

To theorize which information reported by the media will be important, we draw upon signaling theory, which was originally developed by Spence (Reference Spence1973) to explain how job seekers send signals to potential employers regarding their employability within labor markets characterized by information asymmetry. Management scholars have more recently applied signaling theory to explain variation in investor behavior in a number of different contexts, including stock option backdating (Janney & Gove, Reference Janney and Gove2011), the hiring of consultants (Bergh & Gibbons, Reference Bergh and Gibbons2011), CEO certifications of financial statements (Zhang & Wiersema, Reference Zhang and Wiersema2009), and the valuation of IPOs (Certo, Reference Certo2003; Park & Patel, Reference Park and Patel2015). This work has relaxed one of the original assumptions of signaling theory that signals are sent intentionally by one market actor to others. As Bergh and Gibbons (2011: 546) have shown, under conditions of information asymmetry, even if signals are not sent intentionally by a firm, investors will still look to “observable and alterable characteristics” to make sense of investment opportunities.

Signaling theory focuses attention on how observers use specific pieces of information to develop broader assessments of organizations and organizational actions (Spence, Reference Spence2002). In the case of misconduct, since firms are motivated to maintain information asymmetry, media coverage will be one of the primary transmitters of signals to investors. When the media provide information that allows investors to more clearly interpret the behavior as misconduct, investors will view the claims as more credible, making it less likely that they will ignore the claims and more likely that they will expect stakeholders to inflict reputational damage. This will reduce variation in investor interpretations and responses, increasing the percentage of investors who will sell their shares, which will be reflected in a decrease in the stock price, i.e., negative abnormal returns.^{Footnote 1} In contrast, “when signal quality is difficult to judge, investors treat signals as ambiguous” (Epstein & Schneider, Reference Epstein and Schneider2008: 197). In the case of misconduct, ambiguous signals will mean that investors will be less able to interpret the information and more likely to ignore it, leading to more heterogeneity in responses that diffuse the overall reaction of the stock market.

The ultimate legitimacy of claims about misconduct, therefore, depend on observers having clear and credible information that it was misconduct that occurred and that the company is responsible for it. We propose that three specific types of information will function as signals to facilitate these judgements.

First, for investors and other stakeholders to consider and label corporate behavior as misconduct, it must be understandable within their existing cultural schema and accepted norms of ethical corporate behavior. Although most firms now operate globally and can engage in misconduct in different countries, laws, ethical principles, and norms are importantly constructed at the national level. We therefore argue that when the media reports misconduct as occurring within the country in which investors are located, this will provide a clearer and more credible signal to investors that misconduct has occurred. This claim is supported by broader work in finance that has demonstrated that investors exhibit a “home country” bias, i.e., they tend to invest in stocks of firms that are headquartered in their own countries (Barberis & Thaler, Reference Barberis and Thaler2003; Huberman, Reference Huberman2001). One of the primary explanations for the home country bias is that domestic firms are more familiar to investors (Barberis & Thaler, Reference Barberis and Thaler2003). An important corollary is that not only are the portfolios of most investors concentrated in firms in their own countries, but the distribution of ownership of individual firms is concentrated among domestic investors, as is the case in Western Europe (Huizinga & Nicodème, Reference Huizinga and Nicodème2006). As the global reach of many firms has increased, however, acts of misconduct occur all over the world, with national culture and social institutions influencing the likelihood of which types of misconduct will occur (Bame-Aldred, Cullen, Martin, & Parboteeah, Reference Bame-Aldred, Cullen, Martin and Parboteeah2013; Martin, Cullen, Johnson, & Parboteeah, Reference Martin, Cullen, Johnson and Parboteeah2007). Since most of a firm’s investors will be located in the country where a firm is headquartered, however, acts of misconduct that occur in this country will be more legally, ethically, and normatively interpretable for a larger proportion of investors than acts that occur in other countries. When the media reports that an act has occurred within the home country of the firm and investors, therefore, this will enhance the overall clarity and credibility of the signal for a larger proportion of investors. We therefore predict:

Simply because the alleged behavior is more understandable within the institutional and cultural environment of investors, however, does not necessarily mean that there will be agreement that misconduct actually occurred and that the firm is responsible for it. Two other specific types of information, however, can function as signals to facilitate these judgements. For example, although observers will judge misconduct in comparison to laws, ethics, and norms, the clearest way in which misconduct is defined as such is by regulatory officials with the authority of the state. These are the key “line drawers, monitors, and enforcers” (Greve et al., Reference Greve, Palmer and Pozner2010), who determine which “behaviors are wrongful” (Palmer, Reference Palmer2012: 78). Therefore, if the media reports in its initial coverage that the act is under investigation, this will enhance the clarity and credibility of the signal that it was misconduct that occurred and that the firm is responsible for it (Lange & Washburn, Reference Lange and Washburn2012). Paruchuri and Misangyi (Reference Paruchuri and Misangyi2015), for example, found that government investigations of misconduct mitigated negative spillover effects to innocent industry peers because investigations increased investor perceptions of the perpetrating firm’s responsibility.

The media can cover acts of misconduct at different stages of the social construction process. The first stage is the rumor stage. The next stage is the initiation of a formal investigation by a regulatory agency, followed by the completion of the investigation, which can end in a number of ways, including exoneration, the imposition of fines, or the filing of additional legal charges that lead to proceedings by a court or other adjudicatory body. Although some studies have found that investor reactions to misconduct may vary across different stages of social control (Christensen et al., Reference Christensen, Paik and Williams2010; Karpoff & Lott, Reference Karpoff and Lott1993; Muoghalu et al., Reference Muoghalu, Robison and Glascock1990), they have focused primarily on investor reactions to a single act of misconduct at different stages, rather than how reactions to different acts that are exposed at different stages may vary. If an act of misconduct is under formal investigation at the time it is exposed, investors will see the claims of responsibility as more credible than an act exposed in the rumor phase. In later stages, such as the end of an investigation or the court phase, investors are not likely to be receiving a great deal of new information (Agguzoni, Langus, & Motta, Reference Agguzoni, Langus and Motta2013; Muoghalu et al., Reference Muoghalu, Robison and Glascock1990). This leads to our next hypothesis:

Finally, attribution theory, which was developed in social psychology (Berscheid, Graziano, Monson, & Dermer, Reference Berscheid, Graziano, Monson and Dermer1976) and has recently become more broadly used in the management literature (Lange & Washburn, Reference Lange and Washburn2012), allows us to identify another type of information that will clarify that misconduct occurred and that the firm is responsible for it. More specifically, one of the most important factors that facilitates blame attribution is a clear link between an act and undesirable effects (Hamilton, Reference Hamilton1980). Although perceptions of undesirable effects depend in part on the “magnitude of consequences” of the misconduct (Brown et al., Reference Brown, Buchholtz and Dunn2016), there is often uncertainty about severity when misconduct is first reported in the media. Information about the potential or actual consequences of misconduct is not always reported, and when it is, it is usually in the form of estimates because the final quantifiable outcomes are often not known until long after the misconduct has been initially exposed (Hirsh & Cha, Reference Hirsh and Cha2015). However, we argue that any concrete information about the potential, estimated, or initial impacts of misconduct that the media report will facilitate blame attribution. Such information might include the estimated value of misstated earnings, the number of shares traded on insider information, or the amount of an alleged bribe. In addition, this information will enhance the clarity and credibility of claims that it was misconduct that occurred because the consequences of the behavior will be more real and concrete to investors. Hence:

Focus of Blame and Restorative Capacity

While certain information reported in the media will likely enhance the clarity and credibility of the claims about misconduct, investor expectations about whether a firm will incur reputational penalties will also be influenced by their assessments about the likelihood that the firm will repeat the behavior in the future (Connelly, Ketchen, Gangloff, & Shook, Reference Connelly, Ketchen, Gangloff and Shook2016; Karpoff, Reference Karpoff, Pollock and Barnett2012). If the misconduct appears to be an aberration and easily solved, investors will be less likely to expect that stakeholders will impose reputational costs. If specific individuals are blamed, for example, the firm can more easily address the problems (or at least give the appearance of doing so) by firing these employees (Devers et al., Reference Devers, Dewett, Mishina and Belsito2009). Indeed, firms will often assign blame to specific people to divert attention away from assessments that implicate the entire firm (Paruchuri & Misangyi, Reference Paruchuri and Misangyi2015), and firms will often fire CEOs after misconduct has been exposed (Connelly et al., Reference Connelly, Ketchen, Gangloff and Shook2016). If the media attributes blame for the misconduct to one person or a small group of people, therefore, investors will be more likely to see them as “bad apples” (Felps, Mitchell, & Byington, Reference Felps, Mitchell and Byington2006) and view the misconduct as an isolated incident that can be easily rectified (Janney & Gove, Reference Janney and Gove2011). If so, investors will be less likely to expect that stakeholders will inflict reputational damage by altering the terms of or withdrawing from exchange with the firm. An exception, however, is that if top executives are implicated, this suggests deeper organizational problems than if lower level employees are implicated (Brown & Treviño, Reference Brown and Treviño2006).

If an act of misconduct is blamed on the corporation in general, however, this signals that the act more likely stems from “moral collapse” (Shadnam & Lawrence, Reference Shadnam and Lawrence2011), i.e., enduring, systemic problems relating to organizational structures (Palmer & Yenkey, Reference Palmer and Yenkey2015), or an ethical climate and culture (Greve et al., Reference Greve, Palmer and Pozner2010; Martin & Cullen, Reference Martin and Cullen2006; Treviño, Weaver, & Reynolds, Reference Treviño, Weaver and Reynolds2006) that promotes misconduct and normalizes it (Martin, Johnson, & Cullen, Reference Martin, Johnson and Cullen2009; Vaughan, Reference Vaughan1997). As Palmer (2012: 37) observes: “collective wrongdoing wreaks more damage on organizations and societies than individual wrongdoing, because the more people involved in a wrongful course of action, the more it can accomplish.” We therefore predict that if media coverage assigns blame to the corporation, this signals that the misconduct was the result of systemic organizational problems, as opposed to a small number of individuals who deviated from otherwise sound internal policies and procedure, and investors will be more likely to expect reputational damage. Hence:

Finally, our framework has focused on theorizing the effects of information reported in the media about acts of misconduct. Since this information will be new to stock market participants, it is likely to be reflected in market reactions. Although firms that engage in misconduct will themselves vary along a number of dimensions, since firm-level information will not be new to investors, the efficient market hypothesis would not predict that this information would influence investor behavior. However, we argue that a firm’s restorative capacity will moderate the predicted relationship between the focus of blame and investor reactions to misconduct. We define restorative capacity as a firm’s capacity to engage in actions to restore its legitimacy (Pfarrer, Decelles, Smith, & Taylor, Reference Pfarrer, Decelles, Smith and Taylor2008) and repair trust (Gillespie & Dietz, Reference Gillespie and Dietz2009) with internal and external stakeholders following misconduct. As described in the frameworks developed by Pfarrer et al. (Reference Pfarrer, Decelles, Smith and Taylor2008) and Gillespie and Dietz (Reference Gillespie and Dietz2009), effectively restoring legitimacy and repairing trust require both verbal responses through which a firm acknowledges the wrongdoing, accepts responsibility, and offers appropriate explanations, and substantive actions through which the firm makes good-faith efforts to investigate and report the causes of the misconduct, and make changes to the organizational practices, policies, incentives, and cultural elements that contributed to the misconduct. Legitimacy restoration and trust repair also require outreach and dialogue with external stakeholders to rebuild trust and restore relationships. The actions needed to effectively restore legitimacy, repair trust, and reintegrate an organization with its stakeholders, therefore, are extensive and involve a number of different organizational systems (Gillespie & Dietz, Reference Gillespie and Dietz2009), and thus require significant resources, managerial skills, and careful planning (Bertels et al., Reference Bertels, Cody and Pek2015; Gillespie et al., Reference Gillespie, Dietz and Lockey2014). This suggests that a firm’s restorative capacity will importantly depend on committed leaders who understand the challenges and complexity of restoration and reintegration, and are willing to devote the required resources and time to it.

Investors will be most concerned about restorative capacity when the corporation itself is blamed for the misconduct, which signals that the misconduct was the result of more systemic organizational problems that are more difficult to address. When the corporation is blamed, but investors believe that a firm has more significant restorative capacity, they will be more likely to believe that a firm will be able to address the underlying problems and less likely to expect the firm to incur reputational penalites. Although corporations will sometimes demonstrate their restorative capacity by taking swift action and providing detailed plans for addressing misconduct when it is exposed (Suchman, Reference Suchman1995), such action and plans often only emerge (and restorative capacity revealed) over time. Investors, in other words, are “generally aware that a failure has occurred before they encounter or have to interpret a firm’s signals aimed at restoring investor confidence” (Connelly et al., Reference Connelly, Ketchen, Gangloff and Shook2016: 2138).

Investors might, however, form initial assessments of restorative capacity based on information that indicates that the leadership of the firm has the commitment, ability, and resources to address the causes of the misconduct. One indicator of restorative capacity is whether the firm has strong governance mechanisms in place, which can help to focus corporate managers on the challenging work of restoring legitimacy and repairing trust. Since the long-term benefits of this work may not be obvious, however, corporate managers may prefer to engage in the least costly and quickest courses of action, which may not be the most effective (Gillespie & Dietz, Reference Gillespie and Dietz2009). For example, defensive actions that deny or downplay misconduct, or that decouple espoused actions from their implementation, likely mitigate the effectiveness of efforts to restore legitimacy (Gillespie et al., Reference Gillespie, Dietz and Lockey2014; Goodstein et al., Reference Goodstein, Butterfield, Pfarrer and Wicks2014). In addition, it may be difficult for potentially complicit managers to adopt the type of “open and objective approach recommended for reintegration” (Gillespie et al., Reference Gillespie, Dietz and Lockey2014: 397). Strong governance mechanisms, however, can help to ensure that the firm engages in the necessary restorative actions for the long-term benefit of the firm. Independent board members, for example, provide more objective oversight of corporate managers because they are not employees of the firm and do not have direct exchange relationships with the firm (Paruchuri & Misangyi, Reference Paruchuri and Misangyi2015). In the case of misconduct, independent directors will play an essential role in overseeing restorative and remedial actions, such as the implementation of new monitoring and control systems (Palmer, Reference Palmer2012). The independence of the board, therefore, will be an important indicator of a firm’s restorative capacity when the corporation is blamed for misconduct. We therefore predict:

Measuring Cumulative Abnormal Returns

Stock market reactions were measured as cumulative abnormal returns around the announcement of the misconduct using an event study approach (MacKinlay, Reference MacKinlay1997). This analytical approach examines whether the average abnormal returns (AAR) on the event day (announcement of the misconduct in the media) are equal to zero (null hypothesis) versus the alternative hypothesis of a non-zero average abnormal return. The average abnormal return on the event day (AAR_E) is the aggregation of N individual stock abnormal returns (AR_i,E) aligned in event time (MacKinlay, Reference MacKinlay1997):

(1)$$AA{R_E} = {1 \over N}\sum\limits_{i = 1}^N {A{R_{i,E}}}$$[]

We examined the abnormal return pattern on the event day, and on two trading days before and after the reported misconduct, resulting in a 5-day event window [-2,+2]. As robustness checks, we also calculated abnormal returns over event windows [-1,+1] and [-1,0], where 0 indicates the event day, -1 the trading day before, and +1 the trading day after the event day (McWilliams & Siegel, Reference McWilliams and Siegel1997).^{Footnote 2} Abnormal stock returns (AR_i,t) were measured as the difference between the actual return on the event day (R_i,t) and the expected return E[R_i,t], which what one would expect in the absence of the event or misconduct:

$$A{R_{i,t}} = {R_{i,t}} - E\left[ {{R_{i,t}}} \right]$$

We used the market model to estimate the expected return on stock prices.^{Footnote 3} In the market model, abnormal returns are calculated as $A{R_{i,t}} = {R_{i,t}} - \left( {{{\hat{a}}_i} + {{\hat{b}}_i} \cdot {R_{m,t}}} \right)$, where ‘^’ denotes the OLS-estimates from the market model ${R_{i,t}} = {a_i} + {b_i} \cdot {R_{m,t}} + {e_{i,t}}$, with R _i,t the return of stock i in period t, R _m,t the market index return in period t, a _i and b _i the intercept and slope coefficient, respectively, of the market model (stock-i-specific and time-independent parameters), and e _i,t the random disturbance term of the market model for stock i in period t. The parameters of the market model are estimated over a period from –130 to –11 trading days before the event day. The 120-day estimation period corresponds to calendar time period of half a year before the misconduct. Such an estimation period is typical in the event study literature (Engelen & Kabir, Reference Engelen and Kabir2006).

The significance of the mean abnormal returns is tested using the standard test statistic assuming cross-sectional independence (Brown & Warner, Reference Brown and Warner1985), which standardizes abnormal returns for each stock by its standard deviation ($\hat s_i $) calculated from the estimation period:

$$t - test = {{\sum\limits_{i = 1}^N {SA{R_{i,E}}} } \over {\sqrt N }} \sim t\left( {N - 1} \right)$$

with

$$SA{R_{i,E}} = {{A{R_{i,E}}} \over {{{\hat{s}}_i}}}.$$

Explaining Cumulative Abnormal Returns

Although the event study approach is one of the most commonly used methods in the finance literature (Kothari & Warner, Reference Kothari, Warner and Eckbo2007), it does not allow researchers to examine the effects of different variables together in a multivariate framework. We therefore use standard OLS regression models to test our hypotheses, regressing our independent variables on cumulative abnormal stock returns (CARs) (Homburg, Vollmayr, & Hahn, Reference Homburg, Vollmayr and Hahn2014; Paruchuri & Misangyi, Reference Paruchuri and Misangyi2015). The regression approach provides a more robust framework for isolating the effects of specific variables than the bivariate, event study framework. In all regression models, we use robust standard errors, with observations clustered by firm because some firms in our sample were connected to more than one act of misconduct.

Independent Variables

Data for our independent variables were collected and coded from media coverage of the misconduct. Although we searched all media sources within each country to identify the acts of misconduct, to collect the data for our independent variables, we focused on a more narrow set of 20 general and business newspapers. For the UK, this included The Financial Times, The Times and The Independent; for the Netherlands: NRC Handelsblad, Het Financieele Dadblad, Algemeen Dagblad and De Volkskrant; for Belgium: De Tijd and L’Echo; for Germany: Börsen Zeitung, Handelsblatt, Allgemeine Zeitung, Berliner Zeitung, Die Zeit and Die Welt; and for France: Le Monde, Le Figaro, Les Echos, Le Journal des Finances, and La Tribune. We used these newspapers as proxies for all media coverage because they have the largest circulation in each country. We coded our independent variables using all articles published on the day that the media initially exposed each act of misconduct. We read and manually coded a total of 937 news articles with a total of 359,695 words. Stock price data, market return data, and other firm-level control variables were obtained from Thomson Reuters Datastream. Data on board independence was obtained from corporate annual reports. We now discuss how we measured and coded each independent variable in more detail.

Control Variables

We collected data on additional variables to capture the potential impact of other factors on negative abnormal returns. The first was a measure of past misconduct by the firm. Since past misconduct is not new information to investors, the efficient market hypothesis would predict that it would not influence investor reactions to new misconduct. However, Lange and Washburn (Reference Lange and Washburn2012) suggest that observers with existing negative beliefs about a company might be more likely to blame a company for new acts of misconduct than those by companies for which they had positive beliefs because existing beliefs often serve as an anchor for decision making (Tversky & Kahneman, Reference Tversky and Kahneman1974). Other evidence suggests that investors might give firms without past misconduct the benefit of the doubt, with a more positive reputation serving as a buffer (Dorobantu et al., Reference Dorobantu, Henisz and Nartey2017; Janney & Gove, Reference Janney and Gove2011). To control for the possibility that past misconduct influenced investor reactions, we created a variable for each firm in our sample that counted the number of articles published during the five years before the event that included the firm’s name and any of the aforementioned keywords relating to misconduct. We counted back five years starting from the day before the beginning of each event window.

In addition, we wanted to ensure that the volume of media coverage of the misconduct predicted to influence investor reactions in Hypothesis 1 was not capturing the effects of the general media coverage that firms receive. We therefore included a variable of nonnegative media coverage that was a continuous measure of the number of articles in our sample of newspapers that included the name of the firm during the five years before the misconduct was reported. We counted back five years starting from the day before the beginning of each event window. We subtracted from this total the number of articles that covered any previous misconduct in order to avoid collinearity issues with the variables measuring past misconduct.

Since investors might be more concerned with different types of misconduct, e.g., financial fraud (Paruchuri & Misangyi, Reference Paruchuri and Misangyi2015), we controlled for the type of misconduct. Using the newspaper articles, we classified each act using a ten-category typology: insider trading, corruption and bribery, accounting fraud, tax fraud, other fraud, price fixing (i.e., monopolistic practices), employee discrimination, human rights violations, environmental violations, and miscellaneous transgressions. To control for country, sector, and year effects, we used dummy variables for each country (5), each sector (19) and each year (11), with one of the countries, one of the sectors, and one of the years used as reference categories. We also controlled for standard predictors of stock market value. We measured firm age as the number of years since the founding of the company (Van Essen, Engelen, & Carney, Reference Van Essen, Engelen and Carney2013), firm leverage as the value of debt divided by the value of total assets (Kedia & Rajgopal, Reference Kedia and Rajgopal2011), and the market-to-book ratio (Rajan & Zingales, Reference Rajan and Zingales1998). These data were obtained from Thomson Reuters Datastream. Table 1 provides descriptive statistics for all variables and Table 2 presents the correlation matrix for all variables.

Table 1: Descriptive Statistics of Variables

Note. Years are dummy variables representaing the year in which the misconduct was reported.

Table 2: Correlation Matrix

Note. N = 423; * p < .05.