Strategic financial communication
during corporate crises.
Master’s Thesis
Strategic financial communication during corporate crises:
The influence of media news coverage on Volkswagen’s stock price during the emission crisis.
Franca Sophia Volpert 11811277
Graduate School of Communication Master’s Program Communication Science
Track: Corporate Communication Supervisor: dhr. dr. P.H.J. (Pytrik) Schafraad
Abstract
This study sheds some light on the relationship between news coverage and the stock market, which has largely been ignored within the field of crisis communication, yet. A manual content analysis of three major U.S. news outlets was conducted (N = 246) covering articles about the Volkswagen Group’s emission crisis for a period of 17 month. Aggregated on a weekly level, the media variables were by means of an OLS regression model related to the opening price of Volkswagen’s common stock of the following week while controlling for the opening price of the previous week as lagged dependent variable. Drawing on agenda setting theory, this study detected a negative relation between media salience and Volkswagen’s stock price. With regard to the predictive power of different crisis response strategies, the findings suggest that compensation payments to the authorities exert a positive influence on the stock price. Moreover, a negative moderating effect of a critical firm-specific tone was revealed between the relationship of compensation payments to authorities and the stock price. However, an influence of different levels of attribution of responsibility on the stock price could not be identified. Hence, this empirical study contributes to the body of research in several ways: First, it expands first-and second-level agenda-setting theory in regard to investors, an understudied, yet important stakeholder group. Second, this study extends recently conducted studies on this relationship by relating it to crisis communication research and thus seeks to generate new insights and perspectives in the field. Lastly, the findings inform communication managers in stressing the importance of media relations in times of crises and give some further insights on how to successfully manage financial performance.
Key words: crisis communication, SCCT, strategic financial communication, financial
Introduction
During organizational crises, stakeholder relationships and the organization’s
reputation are under pressure (Van der Meer, Verhoeven, Beentjes, & Vliegenthart, 2016) and the financial performance is put at risk (Kleinnijenhuis, Schultz, Utz, & Oegema, 2015). Those given facts render the affected organizations newsworthy and therefore generate much publicity in the media, where the majority of stakeholders learns about an organizational crisis (Coombs, 2007a). This assigns the media a critical role in crisis situations, since the way it is reported on a crisis has great influence on how stakeholders perceive the information (Carroll & McCombs, 2003; Hallahan, 1999). In this vein, it is often built on agenda setting theory (AST) in order to substantiate how media coverage affects stakeholders (Carroll & McCombs, 2003). Communication research has generated manifold insights on the interplay between corporations, the news media and the public during organizational crises (e.g. Van der Meer, Verhoeven, Beentjes, & Vliegenthart, 2014, 2017) as well as the effects of different
situational crisis factors on organizational reputation (Coombs & Holladay, 2002).
On the other hand, remarkably little research has been carried out on examining the impact of news media coverage on financial performance and shareholders, even though information distribution through the media is crucial on the markets (Strycharz, Strauss, & Trilling, 2018). Answering to the rising demand to integrate the fields of economics and communications (Kleinnijenhuis, Schultz, Oegema, & van Atteveldt, 2013), recently,
communication science researchers included increasingly business-relevant outcome factors, such as stock prices in their studies (cf. Fang & Peress, 2009; Van der Meer & Vliegenthart, 2018). While quite a few studies within the field of communication research focused by now on the impact of mere media salience on the stock market (Fang & Peress, 2009; Strauß, Vliegenthart, & Verhoeven, 2018), others took a more comprehensive view and investigated attributes of the media coverage and its effects on stocks, such as emotions and sentiments (Strauß, Vliegenthart, & Verhoeven 2016; Tetlock, 2007) or corporate topics (Strycharz et al., 2018).
However, research that includes financial outcomes into the field of crisis
communication is to my best knowledge scarce (for exceptions see Chen, Ganesan, & Liu, 2009; Kleinnijenhuis et al., 2015). In accounting for these shortcomings and drawing on agenda setting theory, this research aims to investigate whether and how the presence of crisis response strategies in news coverage relates to the stock price. Thereby, it was made use of the strategies developed by Coombs (2007a) within the Situational Crisis Communication
Theory (SCCT). Furthermore, this study explores if media salience and firm-specific tone affect the stock of a corporation facing a crisis. With respect to the most important situational factors within a crisis (Coombs, 2007a), it is lastly examined whether different levels of attribution of responsibility influence the stock price. The latter has to my best knowledge not been investigated so far and draws attention to another important dimension when assessing interrelations between the media and stock markets during corporate crises.
The case that was chosen to conduct this research is the Volkswagen Group’s emission crisis breaking in September 2015. As being one of the largest of the last decades, the crisis has triggered major consequences for the whole automotive sector. This selection is further motivated by the fact that this crisis has an outstanding long crisis course characterized by many turns and especially dramatic stock price fluctuations (Stewart, 2015). The news coverage is analyzed by means of a manual content analysis of three major U.S. news outlets and is related to the corporation’s common stock at NASDAQ stock market exchange.
Addressing the outlined gaps in research and trying to generate some insights on how to address explicitly investors in corporate crisis situations, this research seeks to answer the following research questions.
RQ1: How does media salience in a corporate crisis relate to the corporation’s stock price? RQ2: How do different levels of attribution of responsibility in a corporate crisis relate to the stock price?
RQ3: How does the presence of crisis response strategies relate to the stock price and are there differences between different strategies?
RQ4: Does the firm-specific tone moderate the relationship between crisis response strategies and the stock price?
Theoretical Framework
The financial market and the media
The efficient market hypothesis (EMH) states that markets are efficient as prices “fully reflect” the available information and further assumes that investors react homogenously to new information (Malkiel & Fama, 1970, p. 384). The basic assumption is the rational, profit-maximizing individual who absorbs all available information and engages in rational
calculations in order to reach individual, self-serving goals (Davis, 2006a). However, several behavioral finance as well as social sciences researchers have questioned the EMH arguing
that participants in the stock markets might also be influenced by psychological factors, such as emotions or mass behavior (Davis, 2006a; Strycharz et al., 2018). Moreover, the flow of information is an “intrinsic dimension of the financial markets” (Thompson, 2013, p. 208). Therefore, the mass media are a strong indicator for and also reflector of the consensus opinion (Davis, 2006b; Thompson, 2013). Building mainly on agenda setting theory, social sciences researchers hence claim that news media coverage influences stock market
movements to a certain extent (Pollock & Rindova, 2003; Strauß et al., 2016; Tetlock, 2007). Furthermore, as outlined by Scheufele, Haas and Brosius (2011) it must be considered that there are different kind of investors with different access to corporate information. While professional or institutional investors draw on a broader range of resources, including regular ‘Analyst and Investor Calls’ where they are directly informed by representatives of the firm about current (financial) issues or strategic decisions (Strauß, 2018), small investors mainly rely on mass media coverage (Jang, 2007; Scheufele et al., 2011). These different levels of information status further trigger herd behavior, meaning that investors do not act fully rational, but comply with the consensus belief on the market (Bikhchandani & Sharma, 2000; Lux, 1995). Moreover, scholars argue that professional investors still consult media coverage to anticipate what private investors will do, which can be described by the third-person-effect (Davison, 1983; Fang & Peress, 2009; Van der Meer & Vliegenthart, 2018). The impulse to follow the herd in case of uncertainty is additionally motivated by the fear to be the only one taking the wrong trading decision which can be traced back to an intrinsic human need for conformity (Bikhchandani & Sharma, 2000; Davis, 2006a).
Investor agenda-setting effects
In order to explain the link between media news coverage and stock market
movements, the present research draws on agenda setting theory (AST) (Carroll & McCombs, 2003). AST determines that the mass media have an agenda-setting function, meaning that issues salient in news coverage are transferred to the public’s agenda (McCombs & Shaw, 1972). Initially developed within the context of election campaigns, Carroll and McCombs (2003) applied AST to business communication showing that media visibility of organizations can have an influence on the public image of a firm. This core idea is described as first-level agenda-setting (McCombs, Shaw, & Weaver, 2014). An explanation for why agenda-setting effects occur and investors might consider news media when taking investment decisions is the concept of ‘Need for Orientation’(NFO), which turned out to be a strong predictor of first-level agenda-setting effects (Chernov, Valenzuela, & McCombs, 2011; McCombs et al.,
2014). The extensive media coverage during a corporate crisis is expected to send a signal to investors to take a closer look at the occurrences and reassess their investment.
Furthermore, not only the salience but also how corporate issues are presented in the media can influence stakeholder’s evaluations. This is reflected in second-level AST, defined as “the impact of the media agenda on the public agenda regarding the salience of the
attributes of these objects” (McCombs et al., 2014, p. 782) and comprehended by two
dimensions: substantive and evaluative attribute agenda-setting effects (Carroll & McCombs, 2003; Kiousis, Mitrook, Wu, & Seltzer, 2006). Thereby, the substantive dimension describes the agenda of attributes. According to Carroll and McCombs (2003) “specific substantive traits or contemporary issues associated with a firm” (p. 45) that are relatively salient in the media are transferred to the image the public holds about a corporation. In other words, news media articles highlight certain attributes and aspects of a corporate crisis. Drawing on this notion, it is firstly assumed that if the media strongly discuss the attribution of responsibility, investors adapt the presented perceptions and interpretations. Further, this aspect of AST substantiates the assumption that different crisis response strategies conveyed in news coverage might influence investor’s trading behavior during an organizational crisis.
Additionally, second-level agenda-setting is concerned with an evaluative level (Carroll & McCombs, 2003), interpreted by Deephouse (2000) as ‘media favorability’. This refers to the idea that the media do not only provide information, but moreover assessments (Deephouse, 2000). More precisely, news media coverage conveys a specific tone (Carroll & McCombs, 2003). Consequently, the evaluative level of AST implies that if the media report very critical about the crisis management of a firm, it is likely that shareholders develop a critical image of the organization as well (Deephouse, 2000). This in turn might induce investors to reconsider their investment in that firm, strengthened by the assumption that media coverage conveys the consensus belief on the market (Davis, 2006b).
First-level agenda-setting: Media salience
The key question in the literature on the interrelation between media coverage and the stock market is whether media salience on corporate news affects firm-level returns. While some scholars did not find evidence for the relationship between media coverage and stock markets at all (Campbell, Turner, & Walker, 2012), others have found that corporate media salience affects the financial performance (Fang & Peress, 2009; Strauß et al., 2018). However, previous research focuses on different financial outcomes, such as the trading volume, stock indexes or individual stocks and detected effects in different directions. Several
researchers conclude that newspaper media coverage has a positive influence on closing prices (Strycharz et al, 2018) and the transaction volume (Alanyali, Moat, & Preis, 2013). Furthermore, economic tweets from news agencies on intra-day short-term intervals influence the fluctuation of the Dow Jones Industrial Average positively (Strauß et al., 2018). On the contrary, Van der Meer & Vliegenthart (2018) recently detected a negative influence of media salience on corporate stocks, confirming previous work by Fang and Peress (2009).
Nevertheless, research on the influence of media coverage on the stock markets in times of crises is rather scarce. An exception is a study by Kleinnijenhuis et al. (2015) who detected agenda-setting effects of newspaper and online news coverage on financial markets by focusing on the BP oil spill crisis. More specifically, media salience about the incident had a negative effect on the BP share price. Drawing on these former findings and first-level AST, it is assumed that the salience of a corporate crisis in the media is an indicator of changes in the stock. Since a crisis is most likely associated with negative news and in line with Van der Meer & Vliegenthart (2018) who argue with a general negativity among economic news, it is hypothesized that news coverage affects the stock price negatively.
H1: The higher the salience of the corporate crisis in the news, the higher the chance of a negative effect on the corporation’s stock price.
Substantial attributes (1): Responsibility on an organizational and individual level
Attribution theory (Weiner, 1985, 1986), which underlies the SCCT, states that the “threat of a crisis is largely a function of crisis responsibility” (Coombs, 2007b, p. 136). This means that in organizational crises the attribution of responsibility determines how severe the crisis is perceived by stakeholders and therewith how long-lasting and profound reputational and financial damage are (Coombs & Holladay, 2002; Ma & Zhan, 2016). Despite the key role attribution of responsibility holds within the SCCT framework (Coombs, 2007a), research investigating whether there is a difference between individual or organizational attribution of responsibility in regard to stakeholder’s perceptions is scarce (for exceptions see An & Gower, 2009; Verhoven, Van Hoof, Keurs & Van Vuuren, 2012). However, when corporations face a crisis, it is not unusual that executives, often the CEO itself, apologize, take responsibility and step back in order to reduce the organization’s reputational damage as it was also the case during the Volkswagen crisis. Moreover, Verhoeven et al. (2012) found in an experimental design that an organization’s reputation is damaged far greater in a crisis situation than the individual reputation of a CEO. Organizational reputation in turn is closely
intertwined with financial performance (Wei, OuYang, & Chen, 2017). This points to the assumption that there might be indeed a difference regarding the financial damage between a scenario where the whole corporation is blamed and one, where individuals are held
responsible for the misconduct. Further, it is even conceivable that there is a difference between blaming executives or lower level employees: Executives represent the organization and are therefore closely linked to it (Zerfass, Verčič, & Wiesenberg, 2016). Employees in contrast can be seen as more independent and even decoupled from the organization. This rationale is supported by the conceptualization of diminishing and denying crisis response strategies that are advised to be matched with crisis of the victim type and aim to reduce the organizational attribution of responsibility (Coombs, 2007a). Hence, by blaming a few employees, the organization presents itself as a victim. Based on this logic and the evaluative level of AST, the following explorative hypotheses will be tested.
H2: The higher the attribution of responsibility to the whole organization in a corporate crisis, the higher the chance of a negative effect on the corporation’s stock price.
H3: The higher the attribution of responsibility to individual executives in a corporate crisis, the higher the chance of a positive effect on the corporation’s stock price.
H4: The higher the attribution of responsibility to individual employees in a corporate crisis, the higher the chance of a positive effect on the corporation’s stock price.
Substantial attributes (2): Crisis response strategies
A large body of research in crisis communication assesses crisis response strategies and their effect on stakeholder’s perceptions (Benoit, 1997; Coombs, 2000; 2007a; Coombs & Holladay, 2002). The primary theoretical framework is the Situational Crisis Communication Theory (SCCT) (Coombs, 2000, 2007a). Although being criticized by some scholars (Austin, Liu, & Jin, 2012), the model is still considered as a solid foundation for empirical research in the field (cf. Ma & Zhan, 2016; Claeys, Cauberghe, & Vyncke, 2010; Zhou & Shin, 2017). SCCT is a reconceptualization of Benoit’s (1997) Image Restauration Theory which accounts for the strong similarities between the defined strategies. The main goal of crisis response strategies is the protection of the reputation, more precisely it is aimed to influence
stakeholder’s attributions and perceptions and to minimize a negative affect (Coombs, 1995). SCCT takes crisis responsibility as a link between the specific crisis situation and an
appropriate crisis response and draws thereby on an ‘accommodative-defensive’ continuum in order to cluster the strategies (Coombs, 2007a). Thus, the more accommodative the crisis
response strategies are, meaning that an organization addresses stakeholders in an open and transparent manner, the more they perceive the organization as taking responsibility (Coombs, 2007a; Holladay, 2009).
Moreover, Benoit (1997) emphasizes that organizations passing through a crisis need to address various stakeholders, such as customers, governmental regulators or shareholders. Those audiences have different goals and values, indicating that not every stakeholder group can be coincidently satisfied. Therefore, organizations need to prioritize audience groups before choosing a strategy (Benoit, 1997). Following this notion, a study by Lamin and Zaheer (2012) demonstrated that the general public and investors have in fact diverse interests and consequently various perceptions and evaluations of strategies. Furthermore, the study revealed that investors solely consider a decoupling strategy positively in the context of a company defending legitimacy threats. Moreover, Chen et al. (2009) have shown that during a product-recall crisis, proactive strategies affect the firm value more negatively than a passive conduct. The study argues that this is due to the fact that the crisis and consequences are considered as more severe when a proactive strategy is applied, which in turn implies further financial losses. Additionally, Marcus and Goodman (1991) determined in an earlier study that accommodative and defensive signals sent by the management during a crisis lead to different shareholder reactions.
Considering these findings, it seems to be useful to divide between accommodative and defensive strategies when assessing investors’ perceptions. In this regard, the substantial dimension of second-level agenda-setting (Carroll & McCombs, 2003) becomes relevant in order to explain the impact of different crisis response strategies on investors’ trading behavior. Aiming to shed some light on if and how the presence of different crisis response strategies in the media affect the stock market and extending on the findings by Chen et al. (2009), the following hypotheses are posed:
H5: The higher the amount of accommodative crisis response strategies in a corporate crisis, the higher the chance of a negative effect on the corporation’s stock price.
H6: The higher the amount of defensive crisis response strategies in a corporate crisis, the higher the chance on a positive effect on the corporation’s stock price.
Evaluative attributes: Firm-specific tone
In communication science, second-level agenda-setting effects (Carroll & McCombs, 2003) have also been detected in terms of the relationship between media sentiment and
market reactions (Kiousis et al., 2006). It is argued that the media affect investors’ opinion not only due to the pure salience of topics, but especially by presenting information in a certain way, referring to the evaluative dimension of AST (Scheufele & Tewksbury, 2007; Strycharz et al., 2017). This argument finds support by behavioral finance research, stating that
investors’ decisions are not solely based on information, but are also substantially led by emotions and moods (Bollen, Mao, & Zeng, 2011; De Long, Shleifer, Summers, &
Waldmann, 1990). The relationship between tone in news coverage and the stock market has been researched with different definitions so far, reaching from emotionality or media sentiment (Strycharz et al., 2018) and tone (Ahmad, Han, Hutson, Kearney, & Liu, 2016) to media pessimism or optimism (Tetlock, 2007). Moreover, the findings were mixed. Whereas some scholars detected a downward pressure on the market price when the media pessimism is high (Tetlock, 2007) or the firm-specific tone negative (Ahmad et al., 2016) others did not find any influence of tonal media coverage on stock prices at all (Strauß et al., 2016). The scholars argue that it might be that the effects are actually reversed, hence, that the stock market rather influences the media than vice versa.
It must be noted that research investigating in particular firm-specific tone is scarce, most likely due to the fact that most firms do not appear frequently in the media (Ahmad et al., 2016). According to Ahmad et al. (2016), a negative firm-specific tone leads to variations in the stock returns, however, the relationship varied over time and between different firms. Likewise, Strycharz et al. (2018) recently detected that high emotionality predicted stock market fluctuations for two out of three analyzed firms. This research aims to further investigate this relationship in the context of a corporate crisis, where the obstacle of scarce data does not apply. Distinguishing between critical and supportive media news coverage, it is explored whether the way how the firm’s crisis responses are portrayed in the media affects the stock price. Explicitly this relation is investigated as it is assumed that the evaluation of the crisis responses by the media has a strong influence on how investors perceive the actions taken by the corporation (Carroll & McCombs, 2003). The media are seen as indicator for the consensus opinion (Davis, 2006b), wherefore a critical assessment of the response strategies might lead to investors reallocating their assets accordingly. Hence, drawing on the evaluative dimension of second-level agenda setting, the following hypotheses will be tested.
H7: A critical firm-specific tone in the news coverage moderates the relationship between crisis response strategy and the stock price negatively.
H8: A supportive firm-specific tone in the news coverage moderates the relationship between crisis response strategy and the stock price positively.
The presented hypotheses lead to the conceptual model of this research portrayed in Figure 1.
Figure 1. Conceptual Model.
Method
Case Selection
The Volkswagen Group’s emission crisis, belonging to the preventable crisis type (Coombs, 2007a), was chosen for the present research, since it is considered as one of the most far-reaching global corporate crises of the last decades. The car manufacturer
intentionally installed cheating software into their diesel vehicles which led to illegal levels of pollution (Erwing, 2015). Criminal investigations were conducted against several involved managers and one of the largest consumer settlements in U.S. history took place with a payment of 15 billion dollars in fines (Bomey, 2016). Further, the Volkswagen shares plummeted more than 30 percent (Stewart, 2015), which indicates that the markets reacted very sensitively to the occurrences. Finally, the fact that the crisis has already been covered in numerous other research projects (cf. Zhang, Marita, Veijalainen, Wang, & Kotkov, 2016), further implies that the case is applicable to crisis communication research.
Data Collection
News coverage. Following previous research (Scheufele et al., 2011; Strauß et al.,
2016), this study performed a manual quantitative content analysis on newspaper articles from H1 H5, H6 Stock price Media salience Crisis response strategy Firm-specific tone H7, H8 Attribution of responsibility H2, H3, H4
the leading U.S. news outlets The New York Times, USA Today and The Washington Post. The selection is motivated by the large national circulation of those newspapers in the United States (Misachi, 2017). Furthermore, The New York Times is according to financial journalists “perceived as the ‘fifth estate’ and ‘an independent monitor’ of the financial markets” (Strauß, 2018b, p.12) and therefore expected to have major influence on the consensus opinion and consequently the trading decisions of investors.
The articles were retrieved from the LexisNexis database by searching for the keywords ‘Volkswagen’ or ‘VW’ in the headlines and ‘diesel’ in the whole article. This approach guarantees that the articles included in the sample are actually relevant and the crisis is not only mentioned as a side issue. The timeframe was set between the 20th of September
2015 until the 31st of January 2017. Due to limited resources, it was decided to focus on the
initial crisis phase, starting with the first rumors about the manipulation and ending with the court settlement between the Volkswagen Group and the U.S. justice department in January 2017. The initial sample consisted of 288 mostly extensive news articles, however, following the exclusion of some irrelevant and duplicate articles during coding procedure, the final sample consisted of 246 articles (The New York Times n = 169, USA Today n = 54, The
Washington Post n = 23).
Stock market. In order to investigate whether news media coverage leads to
agenda-setting effects on the stock market, it is required to retrieve the data from a trading market within the same country as the news media coverage. Since the price fluctuation between Volkswagen Group’s common stock (VLKAF) is generally in synergy with the preference stock (VLKPF) only one has been examined. Thus, the weekly opening price of the common stock (VLKAF) was retrieved from the second-largest American stock exchange NASDAQ via Yahoo Finance. It must be noted that the majority of Volkswagen stocks are traded at
Frankfurt Stock Exchange in Germany. However, due to arbitrage trading stock prices adjust
between the different stock markets permanently (Shleifer & Vishny, 1997), wherefore these circumstances do not affect the present research.
Procedure and Measurement
The codebook (see Appendix C) underlying this research consists of five different blocks. First, some general information such as the publication date or the length of the article are captured. The following three sections focus on the media attention, the attribution of responsibility and the crisis response strategies. Finally, the firm-specific tone is covered.
Media salience. The salience of the crisis in the media was measured by the number of
articles per week. This variable was not captured within the codebook, but during the data preparation phase. First, on the basis of the recorded date, articles were assigned to weeks. Second, during the data aggregation, a new variable describing media salience by means of the number of articles per week was created.
Attribution of Responsibility. In order to record the attribution of responsibility, it was
distinguished between an individual and an organizational level. The individual level was further differentiated in executives and employees at lower levels. All three variables were measured on a dichotomous scale whether they are present or not (0/1).
Crisis response strategies. In order to construct the codebook for the present study, the
strategies from Coombs (2000, 2007a) and Benoit (1997) were clustered to the two ends of the ‘accommodative-defensive’ continuum. This leads to eight accommodative strategies, which are on the one hand confession, apology and corrective action. The latter was further divided into the three subcategories product recalls, clarification and rectification.
Furthermore, the bolstering strategy and four compensation strategies were used, divided into the target groups customers, authorities, dealers and shareholders. On the other hand, the strategies minimization and denial, further divided into deny intention and deny knowledge of
executives as well as scapegoat are clustered as defensive crisis response strategies. It must be
noted that the scapegoat strategy was redefined for the purpose of this study, stating that a small group of employees is blamed for the crisis. Additionally, the variables no statement,
withhold information and impede investigations were added to the proposed defensive
strategies in order to capture if the corporation actually acts, e.g. other than promised in official statements or does not react at all. Each strategy is operationalized on a dichotomous scale, in particular whether they are present or not (0/1).
Firm-specific tone. The firm-specific tone was measured on a 5-point Likert scale
reaching from very supportive (1) to very critical (5). This operationalization is expedient since it offers the possibility to record if the article is completely neutral. However, before aggregating the data, the variable was recoded creating dummies for each point on the scale in order to avoid a ‘canceling out’ effect.
Intercoder Reliability
Test coding and intercoder reliability testing was conducted with the help of a second coder, who was familiar with crisis communication research. Since not all variables showed sufficient reliability coefficients after a first round of coding, adjustments were made and
more detailed instructions as well as examples were added to the codebook. Krippendorff’s Alpha was eventually calculated by coding a random sample (n = 32) covering all crisis phases, which is approximately ten percent of the whole sample (Neuendorf, 2002). After this second round of reliability testing, almost all Krippendorff’s Alpha coefficients ranged from .77 to 1, which indicates sufficient (> .67) to very good reliability (> .90) (Krippendorff, 2004) (all reliability measures are reported in Appendix 1). The variables withhold
information and deny intention reached due to their rare presence only a reliability coefficient
of .66, which is considered as not sufficient. However, as Lombard, Snyder-Duch and Bracken (2002, 2003) state, for explorative research, which holds true for the present study, lower coefficients can still be accepted. Therefore, the overall reliability of the codebook was regarded as decent.
Data Aggregation
Following past research (Van der Meer & Vliegenthart, 2018), all data were aggregated from the unit of single articles to a weekly level in order to create richer data points constructed through several observations. The disadvantage of this procedure
introduced by Vliegenthart (2014) is that inferences about particular articles are not possible. However, this is out weighted by the fact, that time series analysis with aggregated data facilitates to see general patterns and allows for stronger claims about causality (Vliegenthart, 2014). Further, in regard to a crisis that passes through different crisis stages, data aggregation is especially useful since new upcoming issues are very likely to be discussed for several days in the media. The weekly level seemed appropriate, since especially during the initial crisis phase new information were revealed frequently. Therefore, a rather low data aggregation level was required in order to fulfill the premise of causality and to be able to test agenda-setting effects.
The data aggregation was conducted by calculating the average of the variables, particularly by adding all articles of a week (from Monday to Sunday) and dividing it through the total amount of articles. This procedure allows to compare weeks with a different amount of media salience. Weeks in which no articles were published were treated as zero values in the aggregated dataset. Moreover, the variable measuring media salience, operationalized as articles per week, was created during this procedure.
Data Lagging. In order to conduct the time series analysis and meet the causality
claim, the media data needed to be lagged in time in relation to the stock price data points (Vliegenthart, 2014), meaning that the media variables of the aggregated weeks are related to
the stock market opening prices of the following weeks. Consequently, the media variables of week 1 were related to the stock price of week 2 within the analysis.
Method of Analysis
In order to investigate whether and how media salience and characteristics influence the stock price of Volkswagen during the crisis multiple analyses were conducted. First, all media variables were explored using descriptive statistics and visual graphs for the analysis over time. Second, Pearson’s correlations as well as linear OLS regression models were conducted to investigate whether there are linear associations and to determine the explained variance of the stock price by the media variables. Furthermore, in order to assess whether the inclusion of more variables in the model improves the explained variance of the stock price, subsequently a hierarchical OLS regression, in particular a lagged dependent variable model was employed. Thereby, the variables, whose regression model with the stock price were (almost) significant, were included while controlling for the stock price of the week before as a lagged dependent variable. In order to test the moderation effect of the firm-specific tone, further multiple regression analyses were performed using PROCESS by Andrew F. Hayes (Hayes, 2017). Due to space restrictions, only significant findings will be reported in detail. Statistically not significant results will be reported in the Appendix. The assumptions were tested for all significant regression models. This included tests for homoscedasticity, the Durbin-Watson statistic, a test for normality distribution as well as multicollinearity for the hierarchical OLS regression model.
Results
Presence of media variables
Descriptive statistics (Table 2 in Appendix B) demonstrate that on average three articles per week were published focusing on the Volkswagen emission crisis
(M = 3.37, SD = 4.96). However, there are large differences within the crisis phases: Whereas during the outbreak of the crisis up to 33 articles were released in one week, there were several weeks at a later stage with no media coverage at all. With regard to attribution of responsibility, most frequently the whole corporation is considered as responsible for the installation of the defeat devices (M = .41, SD = .40), followed by individual executives (M = .32, SD = .39). The most commonly used crisis response strategy by Volkswagen is a defensive one, namely withhold information (M = .29, SD = .24), closely followed by deny
knowledge of executives (M = .22, SD = .32) and scapegoat (M = .20, SD = .31). The fact that
also accommodative strategies, in particular product recall (M = .20, SD = .32), and compensation strategies such as compensation for customers (M = .23, SD = .38) are frequently used, illustrates that the corporation did not follow one clear overall strategy. However, when inspecting the media variables over time (Figure 2; Figure 3 and 4 in
Appendix B), it becomes apparent that the defensive strategies are rather present in the initial crisis phase, whereas the accommodative strategies have been increasingly applied at a later crisis stage.
Intercorrelation of media variables
The correlation matrix (Table 3 in Appendix B) implies that many media variables are strongly correlated. In line with the expectations, different accommodative strategies are used simultaneously. To give some examples, this holds true for obviously closely related
strategies such as apology and confession, r = .53, p < .001 or recall and compensation for
customers, r = .72, p < .001. Likewise, defensive crisis responses are present at the same time: withhold information shows significant moderate and strong correlations with minimization, r = .55, p < .001, deny intention, r = .60, p <.001 as well as deny knowledge of executives, r = .44, p < .001. The latter is in turn strongly correlated with the scapegoat strategy,
r = .71, p < .001. Moreover, the use of the strategies by the organization is closely intertwined
with the attribution of responsibility as it is outlined in the news coverage. If lower level
employees are blamed, often the strategies scapegoat, r = .30, p < .05, deny knowledge of executives, r = .45, p < .001 and deny intention, r = .34, p < .005 are present coincidently.
Relationships between media variables and stock price
In order to conduct the analysis as a first step linear regression analyses were
employed to evaluate which media variables contribute to the prediction of the stock price.
Media salience. The regression model with the stock price as dependent variable and
the media salience as independent variable is significant F (1, 71) = 10.72, p < .01. Therefore, the model can be used to predict the stock price, although the strength of the model is only weak, explaining 13 percent of the stock price fluctuation (R2 = .13).
Attribution of responsibility. Three linear regression models were run with the
variables measuring the attribution of responsibility as independent and the stock price as dependent variable. The regression model testing the effect of responsibility of the
The model explains 8,9 percent of the variance of the stock price, R2 = .09. In regard to attribution of responsibility on an individual level, only the regression model with the
employees as independent variable was significant F (1, 71) = 4.96, p < .05, although the
effect was rather weak, R2 = .06.
Crisis response strategies. In order to assess if and which crisis response strategies
predict the stock price, likewise several linear regression models were conducted entering each crisis response strategy separately as independent and the stock price as dependent variable. None of the models is significant. However, the strategy compensation for
authorities is nearly significant, F (1, 71) = 3.31, p = .073, even if the model is very weak, R2 = .03. Additionally, two separate linear regression models were run entering respectively all summarized and as average computed accommodative and defensive crisis response strategies as independent variable. However, both models are not significant, wherefore further analyses with these constructs were not employed.
Figure 2
Time series for key variables that were included in the final research model.
Weekly opening price (VLKAF), NASDAQ stock market exchange
Responsibility: Employees Compensation for authorities
Responsibility: Corporation Media salience
OLS regression modelling
Finally, a hierarchical OLS regression model was employed in order to investigate whether the inclusion of more news media variables in the model improves the explained variance of the fluctuation of the stock price. All media variables that were revealed as significant predictors in the previous conducted linear regression models were included in blocks to the final model. Thus, the variables media salience, attribution of responsibility on
the corporation, attribution of responsibility on employees and compensation for authorities
were incorporated. Although, the linear regression model with the latter is not significant, the variable was entered due to the potential explanatory power in a hierarchical model.
Additionally, the lagged stock price (one week) was added as a control variable.
Table 4 summarizes the results including the improvement of the model by adding more media variables. The final model (Model 4) is statistically highly significant, R2 = .59,
F (5, 67) = 19.13, p < .001, explaining 58,8 percent of the variance of the stock price.
However, this very high explanatory power is mainly due to the addition of the stock price of the previous week as control variable which leads to a strong statistically significant increase compared to the previous model (Model 3), ΔR2 = .29, ΔF (1, 67) = 46.85, p < .001.
Nevertheless, solely the media variables (Model 3) still explain 30 percent of the variation of the stock price, R2 = .30, F (4, 68) = 7.29, p < .001. Within the final model the variables media
salience, b* = -.28, t = -3.36, p < .001, 95% CI [-.92, -.24], compensation for authorities, b* = .21, t = 2.50, p < .05, 95% CI [1.69, 15.03], as well as the lagged stock price, b* = .55, t = 6.85, p < .001, 95% CI [.37, .67], are significant predictors of the stock price.
The robustness of the model is satisfying. The residuals are normally distributed, and homoscedasticity is assumed. Further, the independence of observations is acceptable, as deduced by a Durbin-Watson statistic of 2.05. Moreover, it was tested for multicollinearity. The correlations are lower than .70 and the variance inflation factors (VIF) are in norm, reaching from 1.06 to 1.18, which points that there is no problem with collinearity (Hair, Black, Babin, & Anderson, 2014).
Table 4
Summary of hierarchical OLS regression model.
Variable Model
1 2 3 4
B SEB B SEB B SEB B SEB
Constant 146.50*** 1.39 147.05*** 1.82 146.87*** 1.75 70.99*** 11.17 Media Salience -.76** .23 -.71** .23 -.75*** .22 -.58*** .17 Responsibility: Corporation -4.50 2.87 -5.92* 2.82 -.4.09 2.20 Responsibility: Employees 8.60* 4.01 6.689 3.93 5.28 3.05 Compensation for authorities 10.92* 4.30 8.36* 3.34 Lagged stock price .52*** .08 R2 .13 .23 .30 .59 ΔR2 .13 .10 .07 .29 F 10.72** 7.02*** 7.29*** 19.13*** ΔF 10.72** 4.62* 6.45* 46.85*** Note. N = 73
*Significance at p < .05 **Significance at p < .01 ***Significance at p < .001
Moderating effect of firm-specific tone
In order to answer RQ4 and to assess whether the firm-specific tone moderates the relationship between the crisis response strategy and the stock price, an extended analysis was employed using PROCESS (Hayes, 2017). Thereby, the single crisis response strategy that has shown to have a significant effect, namely compensation for authorities, was included as independent variable while the stock price was entered as dependent variable. While
controlling for the other media variables and the lagged stock price as covariates, three multiple regression models were conducted using the recoded dummy variables very critical,
critical and neutral tone. Due to the fact that the supportive and very supportive tone are
(almost) not present, the moderation could not be tested.
All three regression models are significant (see Table 5 in Appendix B), however, only the model with the critical tone shows a significant interaction effect with the crisis response strategy, F (7,65) = 14.86, p < .001. The independent variables predict 62 percent of the total variance of the stock price, R2 = .62. Confirming the results of the previously employed
regression model, media salience, compensation for authorities as well as the lagged stock
price, entered to the model as control variable, are significant predictors of the stock price.
Moreover, the interaction effect significantly predicts the dependent variable, b = -23.25,
t = -2.15, p < .05, 95% CI [-44.88, -1.63].
Summarizing the findings
Based on these results, H1 can be supported: When the media salience concerning the Volkswagen crisis increases, the stock price decreases during the observed crisis period. Therefore, media salience can be used to predict stock market fluctuations within crisis situations on a weekly aggregated level. With regard to the attribution of responsibility, H2, H3 and H4 need to be rejected. When controlling for the other media variables, neither the attribution of responsibility on an organizational nor on the two individual levels do significantly predict the stock price. Furthermore, with respect to the crisis response strategies, H6 needs to be clearly rejected. None of the defensive strategies, used in this research did show a significant effect on the stock price. However, inferences about the effect of accommodative crisis response strategies must be made in a more differentiated manner: At least one of the accommodative strategies, namely compensation for authorities, did show a significant effect on the stock price. Nevertheless, H5 needs to be rejected since the
direction of the effect is the opposite than predicted: An increase in the respective defensive response strategy leads to an increase in the stock price as well. Finally, in terms of the firm-specific tone, H8 cannot be answered at all due to the absence of a supportive tone in the sample. However, H7 can be partially supported: While a very critical tone did not show an interaction effect, a critical tone in the news coverage had indeed a moderating effect on the relationship between the crisis response strategy and the stock price. This points that the way a crisis response strategy is presented in the news does indeed affect investors’ perceptions.
Figure 5. Overview significant results.
Discussion
This study aimed to investigate whether media salience and characteristics of news media coverage improve the prediction of corporations’ stock prices during a crisis.
Particularly, this study explored if attribution of responsibility, crisis response strategies and firm-specific tone can be used to predict stock market fluctuations and thereby extends past research that mainly focused on media salience and emotionality in non-crisis situations (Fang & Peress, 2009; Strauß et al., 2018; Tetlock, 2007). This has been done by means of a case study of the Volkswagen Group’s emission crisis. Thereby, news articles of three major U.S. news outlets have been content analyzed and associated with the Volkswagen Group’s common stock (VLKAF) at NASDAQ stock exchange.
The results reveal that media salience negatively influences the stock price. Hence, investors seem to judge the severity of the crisis by how much attention the media as a reflector of the consensus opinion (Davis, 2006b) devote to the issue. This finding answering to RQ1 confirms earlier findings from research investigating the relationship in non-crisis situations (Fang & Peress, 2009; van der Meer & Vliegenthart, 2018) as well as during the BP oil spill crisis (Kleinnijenhuis et al., 2015). The negative influence can on the one hand be justified by the general negativity bias related to economic news in the media (Van der Meer & Vliegenthart, 2018), intensified by the uncertainty on the markets (Strauß et al., 2016). Additionally, a crisis already implies negative news and it is very likely that the additional media salience about the firm is devoted to the misconduct. This reasoning accounts for the fact why other studies found a positive effect of media salience on the stock price in non-crisis situations (Alanyali et al., 2013; Strycharz et al., 2018). However, as this study did not
b = - 23.25 b* = - .28 Articles per week Stock price b* = .21 Compensation for authorities Critical tone Lagged stock price b* = .55
systematically cover all characteristics of the media coverage, such as media frames (cf. Nijkrake, Gosselt, & Gutteling, 2015), we can only suspect which media contents are accountable for the downward pressure on the stock. With regard to the specific crisis situation, it would be conceivable that the profound discussion of the legal and therewith significant financial threats as well as the long-lasting speculations about the responsibilities might have strongly contributed. This rationale revives the logic of SCCT that takes the crisis responsibility as indicator about the severity of the reputational and hence financial damage (Coombs, 2007a).
With regard to RQ2, the results do not suggest that different levels of attribution of responsibility do affect the stock price. However, the attribution of responsibility on the corporation as well as the employees are almost significant and might therefore still
contribute to stock price predictions, even if not evident in this case study. As a matter of fact, it must be noted that the directions of those insignificant effects actually follow the predicted logic: While in case the whole organization is regarded as responsible the stock price
declines, in contrast, when only a small group of employees is blamed, the stock price rises. This can be obviously explained by the notion that in the latter case, the issue can be solved much easier, faster and especially more cost-effective. Further, in regard to the reputational damage which goes along with the financial performance of a corporation (Wei et al., 2017), the accusation of a few employees constitutes the whole organization as a victim. This might benefit the organization, since crisis of a victim type are considered to be less threatening (Coombs, 2007b).
With respect to the crisis response strategies Volkswagen applied, this study did only identify one accommodative strategy that influenced the stock price: The announcement of compensation payments to the authorities appeared to have a positive effect on the stock price. This finding contradicts the predicted direction, since it was assumed that concessions are rather seen critical by investors due to the fact that they result in increased costs. Due to the explorative character of this research, the results cannot be related to many past findings. However, Marcus & Goodman (1991) equally concluded that accommodative signals by the management of a corporation after a crisis serve stakeholder’s interests. Thus, in order to make sense of this result, one could argue that the payment of fines to the authorities mark an end of the crisis and are therefore appreciated by investors, who strive to reduce uncertainties (Avery & Zemsky, 1998; Thompson, 2013). Following this notion, it is logical, that
compensation payments for customers on the contrary do not affect the price. First, those payments are more unpredictable due to the high number of affected cars and the uncertainty
how to fix the defects. Second, a settlement with the authorities might have a more ultimate character due to the regulatory power they hold. However, none of the other accommodative and defensive strategies examined within this research did show a significant effect on the stock price. This lack of effects can be explained by taking the application of the strategies by Volkswagen over time into account. The corporation did not follow one stringent strategy, on the contrary, Volkswagen rather used opposed strategies at the same time and repeatedly during different crisis stages. This obscure pattern of conduct and the many turns during the crisis course complicates it for investors to get a clear picture about the occurrences.
Therefore, the rather limited findings in regard to RQ3 can be traced back to these circumstances.
Moreover, with regard to the firm-specific tone, the effect of a supportive tone could not be tested due to the absence within the sample. However, this research found a
moderating negative effect of a critical tone on the relationship between the compensation strategy for authorities and the stock price. As past studies have argued, this might be due to a general skepticism and negativism in the media (Strauß et al., 2016; van der Meer &
Vliegenthart, 2018), strengthened by the fact that a crisis is automatically associated with negative issues. In the present case, compensation payments, generally seen as positive reputational actions (Coombs, 2007a), conveyed with a critical tone in the news coverage seem to trigger skepticism. The findings on an aggregate weekly level are in line with research by Ahmad et al. (2015) who found that an increase of firm-specific negative tone leads to a lower firm-level stock return on the next day. Despite not focusing on firm-specific but general tone, multi-directional influence of sentiments and tone on financial outcomes were likewise detected in several previous studies (Strycharz et al., 2018; Tetlock, 2007). However, the results also contradict previous findings by Scheufele et al. (2001) or Strauß et al. (2016) who did not detect a moderating effect of emotions in the media on a daily
aggregated level. Drawing conclusions from these findings, it must be reasoned that a critical tone does not show consistent effects on the corporation’s stock price, since an effect of very critical tone could not be identified. Due to these ambiguous findings, inferences must be made very carefully, and the results can be rather seen as a first indication than a clear comprehension.
All in all, the findings suggest that media salience and characteristics can help to predict the fluctuation of stock prices during a crisis. However, only very few in the present study examined characteristics of the news media coverage appear to influence the stock price. In this sense, it is important to emphasize the explorative character of this study, which
intended to extend crisis communication research by focusing on corporate financial
performance. Following this notion, the results are quite promising and moreover the amount of explained variance exceeds many studies exploring the interdependence between media coverage and stock prices within non-crisis situations (cf. Strycharz et al., 2018; van der Meer & Vliegenthart, 2018). This makes certainly sense considering that investors are particularly sensitive during a crisis. According to investor surveys, the ‘quality of management’ is besides ‘financial status’ one of the most important factors when judging about firms (MORI, 1998, 2000 as cited in Davis, 2006b). Hence, during a crisis when a firm’s survival is
potentially at stake (Coombs, 2007a), the management’s performance is exceedingly subject of attentive observations. Additionally, uncertainty facilitates herd-like tendencies (Avery & Zemsky, 1998), meaning that investors increasingly monitor how the market reacts, e.g. by considering media coverage and as a result response collectively. On these grounds, agenda-setting effects seem to intensify compared to regular times, which accounts for the stronger effect sizes. However, the importance of the financial status emphasizes that news media only help to a certain extent to explain market volatility, since there are obviously a variety of other, mostly financial factors such as balance sheets that exert influence.
Concluding, the findings clearly demonstrate with regard to strategic financial communication that media relations matter – especially during times of crises that are characterized by high uncertainty. By providing information and evaluations about
corporations, media hold the power to form expectations and shape opinions which in turn affect the environment and conditions under which firms compete (Rindova & Fombrun, 1999; Pollock & Rindova, 2003).
Limitations and Future Research
This study comes along with a number of limitations. First, this study only examines one specific organizational crisis, and therefore, generalizations are hard to draw.
Comparative research is called for to explore if the findings in regard to the Volkswagen case also applies to other corporate crises.
Another major limitation concerns the media sample. Due to availability reasons, only general news outlets have been incorporated. However, future research should equally include financial newspapers such as The Wall Street Journal (cf. Alanyali et al., 2013; Tetlock, 2007) when exploring investor agenda-setting effects. Additionally, as past research suggests (Bollen et al., 2011; Strauß et al., 2017; Zhang, Fuehres, & Gloor, 2011) the integration of
online news media sources should be considered to obtain a more comprehensive
understanding of stock market reactions. This is required based on the logical reasoning that globalization and digital technologies increase the pace information are spread via (social) media. Thus, print news coverage is possibly already outdated when published (Davis, 2006a; Van der Meer & Vliegenthart, 2018). Concurrently, the stock markets nowadays are
characterized by algorithm based high-frequency trading (Kleinnijenhuis et al., 2013). However, the choice of the media and the time lag must always be aligned. The causality claim only holds true if the cause precedes the consequences (Vliegenthart, 2014). In this sense, one can question the methodological decision to aggregate the data. However, it was argued that a crisis passes through different stages and this procedure allows to limit the amount of zero values, which could distort the results. With regard to future research, it should yet be considered to define crisis stages in order to facilitate the detection of patterns in the use of crisis response strategies.
The selection of the news outlets was based on the logic of AST and the consensus market opinion, that is expected to be reflected in the leading national newspapers. However, in fact we are neither able to identify the individual news consumption of investors, nor do we know on which specific media information investors base their trading decisions (Scheufele et al., 2011). Besides, the fact that many shares are part of a portfolio and therefore not traded independently is a general difficulty in assessing isolated media effects on the stock markets (Scheufele & Haas, 2008). The situation is aggravated by the notion that there are different kind of investors as outlined by Scheufele et al. (2011). This uncertainty is a ‘black box’ and seems to be the major challenge for studies investigating media effects on the stock market (Scheufele et al., 2011). Capturing the latter notions, future research is encouraged to include a broader range of news sources in order to take different kind of investors and their possibly different habits of news consumption into account.
Lastly, as already indicated, this study shows some weaknesses in regard to the measurement of the firm-specific tone conveyed in the media. The results are rather
ambiguous and likewise are the findings from previous studies. The various definitions and operationalizations that have been applied by scholars to measure this construct, including automated analysis tools such as dictionaries (cf. Tetlock, 2007; Strauß et al., 2016; Strycharz et al., 2018) imply that an accurate approach has not been found so far. Therefore, future research is encouraged to systematically investigate the (moderating) influence of tone and emotionality in the media coverage on the stock market.
Implications
This study has contributed to communication science literature in several ways: First, from a theoretical perspective, this study showcased how to integrate crisis and financial communication research by drawing on the SCCT framework and relating it to a firm’s financial performance. Thereby, AST was extended to the so far scarcely investigated stakeholder group of investors, contrary to the extensive body of research on the public’s perception of crisis communication (e.g. Coombs & Holladay, 2002; Claeys et al., 2010). The results suggest that especially first-level agenda-setting effects occur during a corporate crisis. Media salience on a corporate crisis leads to a downward pressure on the firm-level stock. Moreover, also second-level agenda setting effects have been detected on a substantial as well as evaluative dimension, demonstrating the validity of all aspects of AST for investors.
Additionally, the well-investigated situational crisis factor attribution of responsibility
(Coombs, 2007a) was examined on a new level, differentiating between an organizational and two individual levels, in particular between executives and employees. In this sense, this study will hopefully guide future research by pointing out a new dimension of crisis communication research and showcasing how to approach the topic empirically.
Moreover, several practical implications can be deducted from the results of this study. First, the general importance of media relations when managing financial performance in times of crises must be stressed. As other researchers suggested, the presentation of crisis responses in media news coverage might appear to be more credible than information a corporation provides via their own channels (Pollok & Rindowa, 2003; van der Meer & Vliegenthart, 2018). However, this does not mean that communication managers should constantly be in touch with journalists. Instead, the fact that media salience influences the stock price negatively during a crisis points that practitioners are advised to carefully consider timing and evidence base. This counts especially for providing statements of the CEO or executives due to the fact that this kind of information usually lead to high media awareness (Zerfass et al., 2016). Consequently, constant media monitoring and evaluations are required. Thereby, communication managers should especially pay attention to a media discourse about legal consequences, since results of this study reveal that strategies related to legal issues affect the stock price.
Furthermore, with regard to the choice of crisis response strategies, the results indicate that accommodative strategies do not necessarily negatively affect the stock price as assumed. Quite the contrary: Accommodative strategies can have a positive impact if they indicate at
the same time that the issue will be solved near-term and the firm can return to normal levels. Lastly, this study implicates that communications managers should prioritize when facing a corporate crisis which stakeholder group is most important and incorporate this notion into choice of crisis response strategies (Benoit, 1997; Lamin & Zaheer, 2012).
References
Ahmad, K., Han, J., Hutson, E., Kearney, C., & Liu, S. (2016). Media-expressed negative tone and firm-level stock returns. Journal of Corporate Finance, 37, 152-172. doi: 10.1016/j.jcorpfin.2015.12.014.
Alanyali, M., Moat, H. S., & Preis, T. (2013). Quantifying the relationship between financial news and the stock market. Scientific reports, 3, 3578. doi: 10.1038/srep03578
An, S. K., & Gower, K. K. (2009). How do the news media frame crises? A content analysis of crisis news coverage. Public Relations Review, 35(2), 107-112. doi:
10.1016/j.pubrev.2009.01.010
Austin, L., Fisher Liu, B., & Jin, Y. (2012). How audiences seek out crisis information: Exploring the social-mediated crisis communication model. Journal of Applied
Communication Research, 40(2), 188-207. doi: 10.1080/00909882.2012.654498
Avery, C., & Zemsky, P. (1998). Multidimensional Uncertainty and Herd Behavior in Financial Markets. The American Economic Review, 88(4), 724-748.
Benoit, W. L. (1997). Image repair discourse and crisis communication. Public relations
review, 23(2), 177-186. doi: 10.1016/S0363-8111(97)90023-0
Bikhchandani, S., & Sharma, S. (2000). Herd behavior in financial markets. IMF Staff papers,
47(3), 279-310. doi: 10.2307/3867650
Bollen, J., Mao, H., & Zeng, X. (2011). Twitter mood predicts the stock market. Journal of
computational science, 2(1), 1-8. doi: 10.1016/j.jocs.2010.12.007
Bomey, N. (2016, October 25). Judge approves $15B Volkswagen settlement. USA Today. Retrieved from: https://www.usatoday.com/story/money/cars/2016/10/25/volkswagen-settlement-approved/92719174/
Campbell, G., Turner, J. D., & Walker, C. B. (2012). The role of the media in a bubble.
Explorations in Economic History, 49(4), 461–481. doi: 10.1016/j.eeh.2012.07.002
Carroll, C. E., & McCombs, M. (2003). Agenda-setting effects of business news on the public’s images and opinions about major corporations. Corporate Reputation Review,
6, 36–46. doi: 10.1057/palgrave.crr.1540188
Chen, Y., Ganesan, S., & Liu, Y. (2009). Does a firm's product-recall strategy affect its financial value? An examination of strategic alternatives during product-harm crises.
Chernov, G., Valenzuela, S., & McCombs, M. (2011). An experimental comparison of two perspectives on the concept of need for orientation in agenda-setting theory. Journalism
& Mass Communication Quarterly, 88(1), 142-155. doi: 10.1177/107769901108800108
Claeys, A. S., Cauberghe, V., & Vyncke, P. (2010). Restoring reputations in times of crisis: An experimental study of the Situational Crisis Communication Theory and the moderating effects of locus of control. Public Relations Review, 36(3), 256-262. doi: 10.1016/j.pubrev.2010.05.004
Coombs, W. T. (1995). Choosing the right words: The development of guidelines for the selection of the “appropriate” crisis-response strategies. Management communication
quarterly, 8(4), 447-476. doi: 10.1177/0893318995008004003
Coombs, W. T. (2000). Designing post-crisis messages: Lessons for crisis response strategies.
Review of Business, 21(3/4), 37.
Coombs, W. T. (2007a). Protecting organization reputations during a crisis: The development and application of situational crisis communication theory. Corporate reputation
review, 10(3), 163-176. doi: 10.1057/palgrave.crr.1550049
Coombs, W. T. (2007b). Attribution theory as a guide for post-crisis communication
research. Public Relations Review, 33(2), 135-139. doi: 10.1016/j.pubrev.2006.11.016 Coombs, W. T., & Holladay, S. J. (2002). Helping crisis managers protect reputational assets:
Initial tests of the situational crisis communication theory. Management Communication
Quarterly, 16(2), 165-186. doi: 10.1177/089331802237233
Davis, A. (2006a). Media effects and the question of the rational audience: lessons from the financial markets. Media, Culture & Society, 28(4), 603-625. doi:
10.1177/0163443706065035
Davis, A. (2006b). The role of the mass media in investor relations. Journal of
Communication Management, 10(1), 7-17. doi: 10.1108/13632540610646337
Davison, W. P. (1983). The third-person effect in communication. Public opinion quarterly,
47(1), 1-15. doi: 10.1086/268763
De Long, J. B. D., Shleifer, A., Summers, L. H., & Waldmann, R. J. (1990). Noise trader risk in financial markets. Journal of Political Economy, 98(4), 703–738. doi:
10.1086/261703
Deephouse, D. L. (2000). Media reputation as a strategic resource: An integration of mass communication and resource-based theories. Journal of management, 26(6), 1091-1112. doi: 10.1177/014920630002600602
Erwing, J. (2015, September 21). Volkswagen Stock Falls as Automaker Tries to Contain Fallout. The New York Times. Retrieved from:
https://www.nytimes.com/2015/09/22/business/international/volkswagen-shares-recall.html
Fang, L., & Peress, J. (2009). Media coverage and the cross-section of stock returns. The
Journal of Finance, 64(5), 2023-2052. doi: 10.1111/j.1540-6261.2009.01493.x
Hair, J. F., Black, W. C., Babin, B. J., & Anderson, R. E. (2014). Multivariate data analysis
(7th ed.). Essex, UK: Pearson.
Hallahan, K. (1999). Seven models of framing: Implications for public relations. Journal of
public relations research, 11(3), 205-242. doi: 10.1207/s1532754xjprr1103_02
Hayes, A. (2017). Introduction to Mediation, Moderation, and Conditional Process Analysis:
A Regression-Based Approach. New York: Guilford Press.
Holladay, S. J. (2009). Crisis communication strategies in the media coverage of chemical accidents. Journal of Public Relations Research, 21(2), 208-217, doi:
10.1080/10627260802557548
Jang, W. Y. (2007). Media exposure or media hype: Evidence from initial public offering stocks in Taiwan. Journal of Media Economics, 20(4), 259-287. doi:
10.1080/08997760701668151
Kiousis, S., Mitrook, M., Wu, X., & Seltzer, T. (2006). First-and second-level agenda-building and agenda-setting effects: Exploring the linkages among candidate news releases, media coverage, and public opinion during the 2002 Florida gubernatorial election. Journal of Public Relations Research, 18(3), 265-285. doi:
10.1207/s1532754xjprr1803_4
Kleinnijenhuis, J., Schultz, F., Oegema, D., & Van Atteveldt, W. (2013). Financial news and market panics in the age of high-frequency sentiment trading algorithms. Journalism,
14(2), 271-291. doi: 10.1177/1464884912468375
Kleinnijenhuis, J., Schultz, F., Utz, S., & Oegema, D. (2015). The mediating role of the news in the BP oil spill crisis 2010: How US news is influenced by public relations and in turn influences public awareness, foreign news, and the share price. Communication
Research, 42(3), 408-428. doi: 10.1177/0093650213510940
Krippendorff, K. (2004). Content analysis: An introduction to its methodology. Thousand Oaks, California: Sage.
