1
Facebook as a platform for crisis communication:
Damage control during the Dieselgate crisis
Master Thesis Corporate Communication
Student name: Dominic Kortram
Student number: 10972501
Supervisor: James Slevin
2
Abstract
Facebook and other social media are increasingly more important platforms to conduct crisis communication. A crisis situation generates a need for information as different parties try to make sense of the sense of the situation. Limited studies have been conducted researching crisis communion performed through social media. This study focuses on the way Facebook can be used to repair an organizational reputation damaged by a crisis. More specifically, the study examines how crisis frames used on Facebook by different parties and comments posted underneath a crisis message can influence people's perception of an organization. A 2 X 2 (crisis frame X type of comments) experiment is conducted. The results reveal that the crisis frame applied and the tone of the comments play a role in the CSR evaluation of an organization.
Key words: social media, Facebook, crisis communication, NWOM, framing
Introduction
Facebook changes the playing field for crisis managers. Facebook enables critical stakeholders a place to phrase their voice. Additionally, crisis managers compete with other parties on social media to establish how a crisis is framed, which in turn have an effect on how an organization's reputation is perceived. The main purpose of this research is to examine how crisis frames and Facebook comments underneath a crisis message on Facebook influence people's perception of an organization's reputation. More specifically, this research focuses on the way Facebook can be used to repair an organizational reputation following a crisis.
A crisis situation creates a need for information. Stakeholders try to make sense of the crisis situation and information helps them do so. The organization which is involved in the crisis is not the only party trying to provide this information. News media are another channel through which crisis information is distributed. As such, news media are one of the core variables crisis managers should pay attention to (An, 2011). Different parties have different concerns during a crisis. This is often reflected in the information they communicate surrounding that crisis. This concept of highlighting certain information at the cost of other information is called framing (McDonald, Sparks, & Glendon, 2010). In term, different frames lead to different perceptions of the crisis by stakeholders. Social media are an increasingly more important source of crisis information (Schultz, Utz, & Goritz, 2011), and increasingly more important to establish a crisis
3 frame (van der Meer & Verhoeven, 2013). As such, it is an important factor in crisis communication. Crisis communicators often overlook the role of social media (Liu & Kim, 2011). Proper crisis communication has the potential to help members of an organization and other stakeholders make sense of the situation (Schultz, Suddaby, & Cornelissen, 2014). However, it is not only the organization who tries to tell the story of the crisis. The media also try to tell their versions of the story, which may or may not incorporate the entire story as it was intended by the company. Furthermore stakeholders are largely dependent on the media for news stories about crises (Coombs T. W., 2007). In fact, the media can package and frame the issue according to their own agenda and guidelines. This poses a challenge for crisis communicators as they want their crisis frame to be the accepted version of the truth. Social media offer crisis managers a platform to communicate with their stakeholders directly. However, news media can also inform stakeholders about a crisis through social media. Thus understanding how differently framed messages reach and influence the perception of stakeholders is valuable for crisis communicators. Additionally, contemporary crisis communication often takes place on social media platforms. These platforms offer both risks and opportunities of their own. They allow social media users to phrase their opinion on the crisis. Understanding how negative or positive Facebook comments influence the reputational evaluation of an organization in a crisis situation will help crisis communicators improve their strategy.
In order to research this, a case study in the form of an experimental survey was conducted in which the sample was split into four different groups. Each group was exposed to a crisis communication article from either the NY Times or Volkswagen accompanied by distinctly positive or distinctly negative Facebook comments as stimulus. After the exposure to the stimulus the attitude towards to Volkswagen brand was measured. Prior to conducting the survey crisis messages were analyzed in determine which frames are being used by online social media platforms by the aforementioned parties.
Literature review
The goal of this literature review is to provide a systematic overview of the current state of crisis communication research. This literature review will provide a description of the relevant concepts related to crisis communication used in this study. Furthermore, this literature review will discuss several crisis communication frameworks. In addition, I will discuss several crisis communication studies and crisis communication studies that focus specifically on social media. Then, I will
4 discuss what gaps still remain unexplored in crisis communication research and finish with the research question and hypotheses. Like previous authors, I will use crisis and organizational crisis conversely for the sake of readability of this thesis.
Crisis
A crisis is an unexpected, non-routine occurrence that leads to uncertainty. A crisis jeopardizes or is perceived to jeopardize an organization's high priority objectives (Patriotta, Gond, & Schultz, 2011). Crises can have a significant disruptive effect on societies (van der Meer T. G., Verhoeven, Beentjes, & Vliegenthart, 2014). A crisis not only rattles the social order but it can also potentially affect the interaction between stakeholder and organization in case of an organizational crisis (Dowling, 2002). According to Coombs (2012) an organizational crisis is the perception of an unforeseen occurrence that jeopardizes critical expectancies of stakeholders associated to health, safety, environmental and economic issues and has the potential to bring substantial harm to an organization's performance and generate unfavorable results. An organizational crisis tends to threaten an organizations reputation and legitimacy (Patriotta, Gond, & Schultz, 2011; Weiner, 2006). Crises commonly concern multiple stakeholders and crises can affect different groups of stakeholders in diverse manners (McDonald, Sparks, & Glendon, 2010).
The more an organization is held responsible for a crisis the more reputational damage occurs as a consequence. Additionally, if a crisis is perceived to be severe the damage done to the reputation is greater (Claeys, Cauberghe, & Vyncke, 2010). A reputation is a valuable, intangible asset for an organization which can drive financial success of an organization. Fombrun describes it as a "perceptual representation of a company's past actions and future prospects that describe the firm's appeal to all of its key constituents (1996, p. 165)". A reputation develops over time through information that stakeholders obtain about an organization via interaction with the an organization, through news media or second-hand information (Coombs T. W., 2007). As such, crisis managers are not the only ones reporting on the crisis. Journalists from different media parties may also play an important role in spreading news about a crisis. The role the media fulfill during a crisis situation should not be neglected. Media may pick up on an issue based on its news value. Additionally, news which has a negative connotation tends to be more salient than positive news (Schultz, Castello, & Morsing, 2013). This means that crisis news tends to have a high news value (An & Gower, 2009) and is likely to obtain a lot of news media attention.
5 "Stakeholders depend on media-filtered information to determine crisis cause." (McDonald, Sparks, & Glendon, 2010, p. 269). This filtering process is commonly referred to as framing. The concept of framing will be discussed at a later point in this literature review. Another key factor that shapes the reputational threat is the type of crisis. A prior history with crises also plays a role in shaping reputational threat as an extensive crises history can be proof that the organization has an ongoing problem.
According to Coombs (2007) there are three factors that play a key role in shaping the reputational threat. The first one is the initial crisis responsibility. The initial crisis responsibility is the extent to which stakeholders hold an organization accountable for a crisis. A second element determining the reputational threat is crisis history. Finally, the organization's reputation prior to the crisis plays a role as well.
Understanding how to best perform crisis communication is important as every organization goes through a crisis at least once in their existence (Spillan, 2003). Sohn & Lariscy (2012) researched how a good corporate reputation can proof to be a double edged sword in times of crisis. According to their research a good reputation can buffer against reputational damage in times of crisis but it can also backfire and cause increased damage to the organizational reputation. They argue that it depends on what type of crisis the organization is dealing with. During an ability crisis, a crisis in which the competence of an organization is discredited, a good prior reputation can buffer against reputational damage. In times of a morality crisis, a crisis in the CSR domain on the other hand, a good prior reputation can backfire as stakeholders start to reevaluate the organization's integrity. In this case having no prior reputation to speak of can be more beneficial than having a good reputation.
Framing
It is impossible for people to fully understand the world. As such, there is a constant struggle for people to make sense of the world around them and process new information in an efficient manner (Scheufele & Tewksbury, 2007). Thus people are constantly selecting certain information at the cost of other information. This process is known as framing. News media too employ frames to build stories. Entman (2003) defined a news frame as "selecting or highlighting some facets of events or issues, and making connections among them so as to promote a particular interpretation, evaluation and/or solution (p. 417)." Coombs (2007) explains that the frame building processes are paramount importance for shaping an organizational reputation in
6 times of organizational crises. Frames can help define a problem and shape public opinion (Knight, 1999).
Van der Meer et al. (2014) explain that issues can be viewed from various angles, as such, when an organizational crisis occurs different frames are likely to emerge between different domains. These frames help different parties make sense of what is happening. Because different parties have disparate interests, different accounts of the crisis emerge. The same information is used but the varying domains differ in terms of where they place their emphasis in their stories. "How an issue is characterized in news reports can have an influence on how it is understood by audience (Scheufele & Tewksbury, 2007, p. 11)".
Framing plays an important role in crisis development (Schultz, Kleinnijenhuis, Oegema, S, & van Atteveldt, 2012). Several frames are very prominent in news surrounding crisis situations. Research found that the human interest frame, attribution of responsibility frame, economic frame, conflict frame and morality frame are frequently used in news media surrounding a crisis (Semetko & Valkenburg, 2000; An, 2011; An & Gower, 2009). Up until now PR and crisis research has devoted little attention to how crises are framed by prominent actors (van der Meer & Verhoeven, 2013). "The public perceives not the objective fact of a crisis event but the fact constructed by the media or news releases from the party in crisis. Framing or describing of a crisis may well influence the public's evaluation of organizational responsibility for the crisis event (Cho & Gower, 2006, p. 420)." The public perception of crises are overlooked (Cho & Gower, 2006).
Cho & Gower (2006) investigated the effects of different frames applied to the news coverage of a corporate crisis. They found that different frames lead to different perceptions of crisis responsibility and blame. Additionally, their study found that the more intentional the act on part of the company is perceived the more blame will be ascribed to the company.
Some research has investigated frame alignment during an organizational crisis (Snow, Vliegenthart, & Corrigal-Brown, 2007; Liu & Kim, How organizations framed the 2009 H1N1 pandemic via social and traditional media: Implications for US health communicators, 2011; Schultz & Raupp, The social construction of crises in governmental and corporate communications: An inter-organizational and inter-systematic analysis, 2010). These researches show that although frames may differ in the initial phase of the crisis. Over time, however, the frames across different domains become more alike. Van der Meer & Verhoeven (2013) studied how the framing of the initial phase of organizational crises differed between news media and
7 the public. They found that social media, especially twitter, played a crucial role in establishing a crisis frame for the public. Their research showed that the framing differed initially between the public and the media. However, as more information became available the public crisis frame conformed to the media crisis frame and social media was mainly used to share information. A study by van der Meer et al. (2014) researched how crisis frames develop over time between different domains. The study focused on the frame building process in times of organizational crisis between PR, news media and the public. Over time, as more information becomes available, frames undergo a process called frame crystallization. This process entails that frames become less ambiguous and increasingly more aligned and even overlap instead of remaining isolated. Frame crystallization does not imply total fusion of different frames. Rather, this study found, that crisis frame alignment occurs over time in the domains of PR, media and the public. The alignment occurs over three phases. In the first phase the three domains made sense of the crisis situations in a different ways thus leading to different frames. In the second phase the first sign of frame alignment occurs. The initially produced frames interact. The domains collectively make sense of the complex crisis situations. In the final phase the frames employed by the different domains tend to differentiate again. The reason given for this is that the need for a congruent information supply is fulfilled and the different domains again emphasize their own agenda through frame selection. This result shows that there are certainly different frames being employed by different parties.
Crisis research frameworks
Given the impact a crisis can have on an organization it comes as no surprise that organizational crises are the topic of extensive research. "Crisis communication research mainly deals with the interrelationships between crisis situations, communication strategies and crisis perceptions" (Schultz, Utz, & Goritz, Is the medium the message? Perceptions of and reactions to crisis communication via twitter, blogs and traditional media, 2011, p. 20). According to Avery et al. (2010) there are two major streams of research with regards to crisis management and communication in public relations. The first one is the theory of image restoration coined by Benoit (1995; 1997) and the situational crisis communication theory (SCCT) coined by Coombs (Coombs T. W., Protecting Organization Reputations During a Crisis: The Development and Application of Situational Crisis Communication Theory, 2007). This theory section will be focusing on the latter as contemporary models of crisis communication are often based on the
8 SCCT model. These present-day models have more emphasis on the medium of crisis communication as will become evident through this literature review.
The role of crisis managers is to strategize a crisis communication strategy that will minimize the damage done to the organizational reputation (Cho & Gower, 2006). The strategy picked depends on the type of crisis at hand. The situational Crisis Communication Theory (SCCT) was developed by Coombs to investigate which crisis response strategy managers should apply in specific crisis situations to restore an organization's reputation in the best possible way (Claeys, Cauberghe, & Vyncke, 2010). Coombs (2007) argues that crisis managers can profit from understanding how crisis communication can be utilized to safeguard an organization's reputation during a crisis. Coombs has developed a framework that helps anticipate how stakeholders will react to a crisis and, as such, assess how much of a reputational threat the crisis possesses. The SCCT model contains three different crisis type clusters. In the first cluster, the victim cluster, the organization is seen as a victim of the crisis and the organizational responsibility attribution is low. The second cluster, the accidental cluster, contains crises which are caused by accidents. As such, the level of responsibility attributed to the organization is low in this cluster. The third cluster, preventable cluster, involves crises for which the organization is perceived as being responsible leading to a high attribution. Once the reputational threat has been established, the SCCT offers crisis managers a theoretical foundation to base their crisis response strategy on. This part of the model consists of primary and secondary response strategies. The first crisis response strategies are the deny response strategies. Deny strategies claim that no crisis exists or that the organization is not involved in the crisis whatsoever. A second set of strategies is the diminish strategy. Diminish strategies argue that the crisis is not as bad as it is made out to be. This strategy can attempt to downplay the crisis itself or the organizational involvement of the crisis. Third come the rebuild strategies which involve offering excuses and compensating affected stakeholders. Rebuild strategies lead to more effective reputation restoration than the other two strategies. However, they are generally more costly. Coombs & Holladay (2002) argue that a crisis response strategy should be selected based on the potential extent of reputational damage a crisis may cause. Coombs also offers a secondary response strategy which consist of bolstering. Bolstering attempts to increase positive reputational perceptions by, for example, reminding stakeholders of good work conducted in the past. This secondary response strategy should, according to Coombs, not be used as a standalone strategy. In an earlier research Coombs & Holladay (1996) already found that when crisis communication responses match the crisis type in terms of responsibility attribution it leads to a more positive evaluation of reputation than when an organization does not respond to a
9 crisis or responds using the wrong strategy. Some authors suggest combining different crisis response strategies from the same crisis response cluster or different crisis response strategies from different crisis response clusters (Benoit, 1997; Coombs T. W., 2007). According to these authors combining of crisis communication strategies can lead to more effective image restoration after a crisis. The SCCT model suggests using multiple crisis strategies from the same cluster.
Previous research using the SCCT model
Claeys, Cauberghe & Vyncke (2010) measured the impact of crisis type and crisis response strategy on perceptions of corporate reputation using the SCCT framework developed by Coombs(2007). In order to test this they conducted a survey exposing participants to either one of three crisis types (victim crisis, accidental crisis or preventable crisis) and either one of three crisis response strategies (deny, diminish or rebuild). Their results show that preventable crises have the most negative impact on organizational reputation. The rebuild strategy leads to the least most positive perception of organizational reputation. Additionally, they found that the more severe a crises judged to be the more negative the organizational reputation is evaluated. Interestingly they found no significant interaction effect between the type of crisis and the crisis response strategy on corporate reputation. This means that, according to their research, matching crisis type to crisis response strategy does not lead to more positive perception of firm reputation than if the crisis type and the crisis response strategy are misaligned (Claeys, Cauberghe, & Vyncke, 2010). This finding is in line with the work of McDonald, Sparks, & Glendon (2010). They found that confessing or taking responsibility for a crisis is a successful crisis account regardless of what caused the crisis. Contrary to their expectations they found that confession reduces responsibility judgments the most effectively. It is useful for reducing anger and negative word of mouth (NWOM) about the company by stakeholders. Additionally, it increases sympathy, loyalty and boosts the attitude towards the company. Interestingly, confessing to a crisis also leads to a lower responsibility judgment among stakeholders. In contrast, denial, excuses or justification of responsibility, which all aim at lowering responsibility judgment, actually increase negative stakeholder reactions and increase the responsibility judgment among stakeholders.
10 Crises management and social media
Social media play an increasingly more important role. Not just in society but also in the social construction of crises (Schultz, Utz, & Goritz, Is the medium the message? Perceptions of and reactions to crisis communication via twitter, blogs and traditional media, 2011; Utz, Schultz, & Glocka, 2013). According to Coombs (2008) crises that occur online are usually more unpredictable and faster developing than crises that emerge offline. This is predominantly because of properties of the internet which allow for timely, interactive communication (Seltzer & Mitrook, 2007). Social media especially are good tools for building relations and are more dialogic, faster and more interactive than traditional media (Schultz & Wehmeier, 2010). Furthermore, information on the internet can be updated quicker than traditional media. In addition, people tend to spend more time online in times of crisis. During a crisis people want to obtain information about what exactly is going on. Social media offer a convenient platform in this regard as online platforms offer the opportunity to verify if friends and family are unharmed (Liu, Austin, & Jin, How publics respond to crisis communication strategies: The interplay of information form and source, 2011). As such, it comes as no surprise that internet usage, and social media usage specifically, peaks during a crisis event.
Social media offer the possibility to engage with stakeholders in a more human, conversational voice (Keheller, 2009). These make social media excellent for relationship building and communicating more directly with stakeholders without the intervention of third party, gate keeping journalists. Social media such as blogs, twitter and Facebook offer organizations the opportunity to communicate to their stakeholders directly. As such, social media are considered as efficient reputation repairing tools by organizations in crisis situations.
Some authors have pointed out the lack of attention being given on (social) medium in classic crisis communication models (Utz, Schultz, & Glocka, 2013; Liu, Austin, & Jin, How publics respond to crisis communication strategies: The interplay of information form and source, 2011). To overcome the lack of social media orientation of other crisis communication models, Liu, Austin & Jin (2011) developed the Social Media Crisis Communication model (SMCC). This model explains how crisis information source and form determine how an organization ought to respond on social media to a crisis. This model builds on the SCCT model and adds to it by including that the way in which the crisis message (source and format) is conveyed affects public responses. The model lists three publics that produce and consume crisis information. These publics can be individuals or organizations. (1) influential social media creators, (2) social media followers and (3) social media inactives. The third group is not active on social media but can still
11 consume crisis information posted on social media indirectly through alternative media or word of mouth.
Jin & Liu (2010) found that public acceptance of a crisis is influenced by the source (third party or organization) and the format (traditional media, social media, word of mouth) through which an organization communicates about the crisis. The findings show that it is important for organizations to align crisis information form and source when communicating regarding the crisis. Another research by Liu, Austin & Jin (2011) yielded similar results. Liu, Austin & Jin (2011) studied the social-mediated crisis communication model (SMCC) in an empirical research. They studied the effect of both crisis communication form and crisis communication source on public approval of crisis response strategies and public emotions. Their results indicate that it is important for organizations to choose a matching form and source through which to communicate in times of crisis as this leads to more public approval of the crisis.
Schultz et al. (2011) analyzed the effects of crisis communication conducted through blogs and twitter in comparison to traditional newspaper articles in an experimental research design. they found that individuals exposed to more crisis communication through social media tend to rate the organizational reputation higher after a crisis than people exposed to crisis communication through traditional media. A different study by Utz, Schultz & Glocka (2013) builds on the work of Schultz et al. (2011). Utz, Schultz & Glocka (2013) found, much like Schultz et al. (2011) that crisis communication via social media (Facebook and Twitter) is better for an organization's reputation. Utz, Schultz & Glocka (2013) also found that the medium effects are stronger than the effects of the crisis type. They found that providing crisis information through social media is good for the reputation regardless of the type of crisis at hand. However, they do not downplay the role of traditional media. They argue that traditional media still play an important role during crisis communication as journalists are still considered credible gatekeepers. As a consequence people still tend to talk more about news obtained through traditional media as opposed to news obtained through social media.
Schultz et al. (2011) also researched the secondary crisis reaction in their study. Secondary crisis communication is concerned with how much stakeholders talk about the crisis with family and friends. they found that the medium through which crisis communication is conducted matters more than the message that is being communicated. Their results indicate that crisis communication is best conducted through twitter. People who read the organizational tweets were less likely to boycott the organization and less likely to engage in NWOM. Schultz Utz & Goritz (2011) found that the medium through which crisis communicators share information with
12 recipients influences how likely people are to talk about the crisis. The results indicate that newspaper articles score highest in secondary crisis communication meaning that people tend to talk about crises they read about in newspapers more than they tend to talk about crises they read about on twitter or blog posts. In a different research, Utz et al. (2013) found that secondary crisis response is influenced by both the medium through which a crisis is communicated and the type of crisis. Their results also indicate that crisis communication through Facebook resulted in less secondary crisis reactions. These are interesting findings as social media make it near-effortless to share information with friends and family. One explanation for this unexpected finding that Schultz et al. (2011) offer is that newspapers may be perceived as more reliable and neutral than corporate blogs and corporate tweets.
From these researches it becomes evident that social media plays an important role in crisis communication. Social media play an important role in establishing a crisis frame in the initial phase of a crisis (van der Meer & Verhoeven, 2013). Crisis communication through social media offer both opportunities and challenges. Organizations themselves providing crisis communication is better for the perception of the organization (Liu, Austin, & Jin, 2011). Crisis communication through social media leads to a better reputational perception than traditional media (Schultz, Utz, & Goritz, 2011). Additionally, people tend to engage more in NWOM when receiving crisis communication through traditional media (Utz, Schultz, & Glocka, 2013).
However, these researches overlook two key factors, both of which are brought about by social media. The first factor is that traditional media can also have a presence on social media. Current research, considers traditional media and social media as two distinct platforms through which crisis communication can be conducted, however in reality, traditional media often have a social media page through which they share news articles as well. The gatekeeping function that journalists offer is not lost when media parties communicate through social media. Another factor which has not been considered by previous researchers is that social media offer individuals the opportunity to respond to crisis communication conducted through social media. Research has considered the secondary crisis response consisting of offline word of mouth (Schultz, Utz, & Goritz, 2011). However, the online word of mouth has not been considered in previous researches. On Facebook the opinions of people responding to social media become part of the social media message. These opinions of individuals may have an impact on how a crisis is understood.
13 Crisis research shortcomings
Avery, Lariscy, Kim & Hocke (2010) quantitatively studied 66 articles on crisis communication using Coombs' SSCT model and Benoit's image restoration theory. They found that the topic of crisis communication is widely established within the PR research field. They argue however that although the amount of research is plentiful, the field would benefit from more prescriptive rather than descriptive research. Current research lacks criticism of the model and more attention should be paid to pre-crisis stages. These findings are summed up by Coombs (2009) comment on crisis research "Much of the existing reputation repair research has generated more speculation about what should be done rather than testing of actual prescriptive claims" (p. 113). Throughout the literature it becomes apparent that some aspects in crisis communication research have not been researched exhaustively. Claeys, Cauberghe & Vyncke (2010) argue that only few prior researches that investigate crisis response strategies rely on an experimental design. In general a lot of studies in this field are conducted through a case-study. As such, there is a lack of experimental approaches in this field (Schultz, Utz, & Goritz, 2011; Claeys, Cauberghe, & Vyncke, 2010). Furthermore, the exact dynamics of how social media affects crisis communication is a branch of crisis communication research which has largely been neglected (Liu, Austin, & Jin, 2011; Liu & Fraustino, 2014). Existing crisis research largely fails to address the effects of different social media on stakeholder perceptions in crisis situations (Schultz, Utz, & Goritz, Is the medium the message? Perceptions of and reactions to crisis communication via twitter, blogs and traditional media, 2011). Instead, the focus lays more on the interplay between crisis type and crisis communication strategy (Utz, Schultz, & Glocka, 2013). Lastly, Cho & Gower (2006) urge researchers to conduct more framing studies to better understand crisis communication.
Research Question
The research at hand aimed to address some of the shortcomings in the field of crisis research. The goal of the study is to learn how crisis communication conducted through Facebook and Facebook comments posted underneath the crisis message influence people's perception of an organization damaged by a crisis. This will give a better understanding of the exact dynamics of how social media influence crisis communication (Liu, Austin, & Jin, 2011; Liu & Fraustino, 2014). This will be tested through an experimental survey giving heed to the urge by Claeys, Cauberghe & Vyncke (2010) and Schultz, Utz, & Goritz (2011) to test crisis response strategies
14 experimentally. Finally this research gives heed to the urge by Cho & Gower (2006) to conduct more framing studies. In order to conduct this study an actual crisis scenario has been selected. In May 2014 researchers discovered that Volkswagen cars were equipped with a cheat device that could cheat emission tests. As a consequence, Volkswagen finds itself in a crisis situation causing the company to lose billions. Additionally, the reputation of the German car manufacturer is damaged as a consequence of the crisis as well. The Volkswagen 'dieselgate' scandal offers a valuable opportunity for conducting a case study as different organizations, including Volkswagen, have framed the crisis in various ways. Additionally, a lot of crisis communication was conducted using Facebook, provoking many sympathizers and critics to phrase their opinion on the scandal. As the crisis is no longer a prominent topic in the media, the reporting on the crisis has dropped significantly, now is a suitable time to test the effect of actual crisis messages on Facebook and Facebook comments on the perception of audience members. A more extensive description of the Volkswagen scandal, an analysis of the scandal using Coombs SCCT model, as well as an analysis of various communication channels used by Volkswagen throughout the 'Dieselgate' crisis can be found under Appendix A.
The following research question has been developed to further investigate the Volkswagen crisis using an experimental survey.
RQ: How do differently framed Facebook crisis news items from different sources accompanied by Facebook comments influence the public's perception of the Volkswagen crisis?
A number of hypotheses have been developed to help answer this research question.
H1: People exposed to the negative comments perceive Volkswagen as less trustworthy than people exposed to positive comments.
H2: People exposed to the negative comments perceive Volkswagen as less reputable than people exposed to positive comments.
H3: People exposed to the negative comments perceive the CSR reputation of Volkswagen as less reputable than people exposed to positive comments.
H4:People exposed to the Volkswagen frames have more trust in Volkswagen than people exposed to either of the NY times frames.
15
H5: People exposed to the Volkswagen frames are more positive towards Volkswagen's reputation than people exposed to either of the NY times frames.
H6: People exposed to the Volkswagen frames are more positive towards Volkswagen's CSR reputation than people exposed to either of the NY times frames.
H7: People exposed to the Volkswagen news frame with positive comments will score the highest on the perceived trust scale and people exposed to the NY Times news frame with negative comments the lowest.
H8: People exposed to the Volkswagen news frame with positive comments will score the highest on the perceived reputational scale and people exposed to the NY Times news frame with negative comments the lowest..
H9: People exposed to the Volkswagen news frame with positive comments will score the highest on the CSR scale and people exposed to the NY Times news frame with negative comments the lowest.
Methodology
In order to test the hypotheses and answer the research question posed at the end of the theoretical section, data has been collected by means of a survey. The full survey can be found under Appendix B. The survey used in this thesis is cross-sectional, meaning that there is no later follow-up questionnaire.
Data Collection
The data for this research was collected using an experimental survey. Volkswagen is an internationally operating multinational this crisis has affected a lot of people across a lot of countries. Furthermore, both the NY Times and the Volkswagen had international Facebook pages. As such, the sample population for this research was not necessitated participants to be from certain countries. However, the respondents were required to be over the age of 18.
To gather participants a convenience sample was used. The reason for this sampling method was that it has proven to be a cost efficient and quick way to recruit participants. The survey was spread through social media and was conducted on a voluntary basis. Participants were thus not
16 compensated for partaking in the survey. The data was collected between May 20th and May 31st. In total 128 adults across 9 different countries participated in the survey. 128 participants started the survey. 99 participants finished the survey. This means that the dropout rate for this survey was relatively high with a dropout rate of 22.65%.
Survey's which were not a 100% completed were excluded from the analysis. The vast majority of the participants was Dutch (84.8%) followed Canadians who accounted for 6.1%. Ages of the participants ranged between 19 and 69. The mean age of the sample was 30.41 (SD=13.53). 53.5% of the participants were female and 46.5% were male.
24.2% of the participants owned or has owned a Volkswagen car in the past. 75.8% of the participants has never owned a Volkswagen. Additionally, two questions were asked to see how familiar people were with the Volkswagen crisis. From these questions it became evident that 7% of the participants had not heard of the Volkswagen crisis prior to this study and 93% had. Most people were to some degree knowledgeable about the Volkswagen crisis. 9.1% of the sample answered that they were not well informed at all concerning the Volkswagen crisis. 40.4% answered that they were slightly well informed, 34.3% answered moderately well, 15.2% very well. and 1% claimed to be extremely well informed.
Design
In order to measure the effect of a crisis news frame and Facebook comments on the perceived reputation of an organization in a crisis situation an online survey-embedded experiment was conducted. This experiment used a 2 (crisis frames from different sources) x 2 (types of comments) design. However, before the survey could be distributed, the crisis news frames that were being employed by the media and Volkswagen had to be identified. Before conducting the survey a content analysis has been conducted.
Participants taking the survey were exposed to one of four different stimuli. The stimuli consisted of either a New York Times Facebook post and accompanying article or a Volkswagen Facebook post and accompanying article. In order to create four different stimuli the comment section under the Facebook posts had been manipulated. As such, both media fragments had a version in which the comments were adapted to be negative towards Volkswagen and a version in which the comment section was very positive towards the car brand.
Participants were randomly assigned to one of the four conditions. However, due to some participants dropping out, respondents were not equally divided among the four conditions. 26
17 people saw the NY Times post with positive comments. 22 people were shown the NY times post with negative comments. 25 people were exposed to the Volkswagen post with positive comment and 26 people were presented to negative comments alongside the Volkswagen post. The two articles selected as stimuli were the product of extensive analyses of the news messages concerning the crisis from both the New York Times and Volkswagen. From the content analysis, which has been conducted prior to designing the survey, it became apparent that a number of distinct elements were reoccurring in the media messages of both parties. These elements built into two distinct frames which were inductively identified using a media packaging approach based on the work of Gamson and Modigliani (1989) and Dai and Hyun (2010). The media packaging approach has been used to reconstruct the frames used by the coverage of the Volkswagen crisis by the New York Times and Volkswagen itself. Gamson and Modigliani (1989) argue that writers of news stories make use of framing devices. These framing devices can be verbal or visual. They determine how the audience will think about a certain issue. Gamson and Modigliani (1989) identified a list of 5 framing devices. Van Gorp (2010) later added 9 more items to the list bringing the total to 14 framing devices. Examples of framing devices are metaphors used when talking about a certain issue, the themes that are underscored, the picture that accompanies a news article or sources of information used in the news piece. When applying this list of framing devices to the news items covering the 'Dieselgate' crisis it became evident that Volkswagen and the New York times frame the topic rather differently.
The New York Times took a very factual approach to the Volkswagen crisis. As such this frame has been dubbed the "scandal facts frame" They provided lots of information concerning the car types and the number of cars affected. This frame emphasized unresolved issues, such as, people who have a faulty car and want Volkswagen to buy it back, or the fact that, although people had been stepping down, no one at Volkswagen has as of yet been held accountable. The New York Times and their "scandal facts frame" also emphasized the company losses in terms of reputation, car sales, share prices and potential legal fees.
Volkswagen, on the other hand, applied the "road to recovery frame" in which the Volkswagen company expressed regret for the events that have taken place, but also places a lot of emphasis on the fact that the company tried to move forward and attempted to come up with a technical solution as soon as possible. At the same time the Volkswagen company underscored the fact that the cars were still technically sound. With the "road to recovery frame" Volkswagen placed the responsibility of the crisis with themselves and emphasized that they wanted to get to
18 the bottom of this issue. Additionally, they made it very clear that they wanted to help their victimized customers in the best possible way, and restore their trust as soon as possible. The comments, which were the second part of the stimuli, were selected on the language in which they were written, tone of voice, whether they were understandable and positivity or negativity towards Volkswagen. A total of six comments were taken from Volkswagen's actual Facebook page. Three comments were distinctly negative with regards to Volkswagen. The other three comments were unmistakably positive towards Volkswagen. These comments were then manipulated and placed under the Facebook posts using Photoshop.
Measures
The survey measured a number of variables. Most of these were measured on a 7-point Likert scale unless mentioned otherwise. Additionally, some existing scales were adapted to a 7-point Likert scale from a 5-point scale. The reason for this is because the 7-point scale provides a balanced answering opportunity and is capable of measuring more nuanced differences than a 5-point Likert scale.
The first variable measured in this survey is news media exposure. This measure served as a covariate during the later analyses. This variable was measured using an adapted version of the scale by Trilling & Schoenbach (2013). The questions in this scale ask respondents how often in a typical week they consume different types of news media. This scale measured how often people expose themselves to news from different sources. Because (social) media is often the prime source of corporate information people that exposed themselves to more news media were expected to be better informed, and were thus expected to be more aware of the Volkswagen crisis. This scale included items such as: "How many days in a typical week (Monday to Sunday) do you read a newspaper?" and "How many days in a typical week (Monday to Sunday) do you listen to newscasts on the radio?" A factor analysis conducted on the seven items of this scale showed that traditional media and new media load onto two distinct factors. Four items that made up traditional media news exposure and three made up new media news exposure. The internal consistency for traditional media exposure was questionable (α=0.694). However, when the item "how often in a typical week do you obtain news through
teletext?" was removed the internal consistency increased to an acceptable level (α=0.706). As
such, this item was deleted from the scale. The internal consistency for new media news exposure was α=0.560. This means that the three items making up this scale had questionable internal consistency.
19 The next part of the survey consisted of one of the four stimuli discussed above. The questions following the stimuli were set out to measure the perception of the Volkswagen brand. These questions were adopted from three different surveys.
The first scale stemmed from the the RepTrack™ Pulse developed by Ponzi, Fombrun, & Gardberg (2011). This scale sets out to measure the corporate reputation. The items were adapted to apply to the Volkswagen brand. This scale includes items such as "Volkswagen has a good overall reputation."
A factor analysis of these items showed that all 4 items were correlated. The internal consistency for corporate reputation was high (α=0.905). If any item was taken out of this scale the internal
consistency would be lower, except for the item "Volkswagen has a good overall reputation". If this item were to be deleted the internal consistency would be α=0.915. However, having more items in a scale is in this case preferable to having a slightly higher internal consistency. As such, all four items were kept in this scale.
The next questions were concerned with measuring corporate trust in Volkswagen. The items from this survey originate from the work of Tong (2013). The items were adapted to apply to the Volkswagen crisis. It included items such as: "Volkswagen is a company that I can trust" and "Volkswagen is providing accurate information concerning the crisis." A factor analysis showed that "Volkswagen is a company that I can trust" also loaded strongly on the same factor as all of the corporate trust items. This meant that a strong correlation exists between these items. The internal consistency for the trust scale was high (α=0.802) with seven items included in the
scale. If any items were to be taken out of the scale the scale would be weaker. As such, all items were kept. The high Chronbach's Alpha score indicated that the scale was reliable and measured what it sets out to measure.
The third set of questions aimed to measure people's attitudes towards the Corporate Social Responsibility (moral and ethical aspects) of Volkswagen. The original set of questions for this last scale were retrieved from (Fatmaa, Rahmanb, & Khanc, 2014). However, the original scale sets out to measure perceived CSR activities addressing customers, employees, shareholders, the environment and society in general. A number of these variables were selected for this study in order to shorten the overall length of the survey. As such, the variables measuring the CSR perception towards employees were taken out of the scale. Additionally, the items on this scale were adapted to apply to the Volkswagen brand. This scale included items such as "Volkswagen
20 treats its customers honestly" and "Volkswagen is concerned with improving the general well being of society".
As the original scale aimed to measure different facets of corporate social responsibility a factor analysis was run. This analysis showed that the three items concerned with shareholders loaded on a different factor meaning they were not correlated to the items. It was decided to remove these items from the CSR scale.
The CSR scale in this research thus measured how Volkswagen was perceived to treat their customers, society and environment. The internal consistency for these items was high (α=0.829). In total seven items made up this scale. The internal consistency would not get any
higher if any items were removed from the scale. A factor analysis showed that three items of the CSR scale that were concerned with shareholders load onto the same factor which means that the items were correlated. The internal consistency for this shareholder scale was questionable with (α=0.604). As such, no further measures were conducted using this scale from
here on.
This gave three workable sets of items which can be transformed into scales that measured some aspect of the Volkswagen brand perception. "Trust perception in Volkswagen", "Volkswagen reputational perception" and "Volkswagen's CSR perception". All three scales were based on the mean scores of the items included in the scales. As such, the highest score a participant could obtain on a scale was a 7 and the lowest obtainable score was a 1. A 7 indicated very high levels of trust in Volkswagen or a very positive perception of the (CSR) reputation. A 1 on the other hand indicated a very low level of trust in Volkswagen or a very negative perception of the (CSR) reputation.
After these three reputation evaluating scales some demographic questions were asked to complete the survey.
Results
A number of one-way analysis of covariance's (ANCOVA's) were conducted for this study. In addition a number of ANOVA's were run if the ANCOVA proved inappropriate. The goal of these analyses was to test the hypotheses. For first three hypotheses the sample was split into two groups. Namely, the group people exposed to a news story with negative comments and the group of people exposed to a news story with positive comments. For the next three hypotheses
21 the sample was split into two groups as well. Namely, the group exposed to a news story by the NY times and the group exposed to a news story by Volkswagen. For the last three hypotheses the sample was split into four groups based on the four different stimuli.
H1: People exposed to the negative comments perceive Volkswagen as less trustworthy than people exposed to positive comments.
For the first ANCOVA the stimuli groups were grouped into a negative and a positive group. The negative group contains the NY times article combined with the negative Facebook comments and the Volkswagen article combined with the negative Facebook comments. The positive group on the other hand contains the NY times article combined with the positive Facebook comments and the Volkswagen article combined with the positive Facebook comments. This grouping will be referred to as "Tone of comments". The dependent variable was the perceived trust in the Volkswagen company and the covariate was the news media exposure factor. Before conducting the ANCOVA the homogeneity of regression assumption was tested to evaluate the interaction between the covariate and the independent variable. If a significant interaction occurs between the covariate and the independent variable it would suggest that differences on the dependant variable are due to the covariate. The analysis shows that the assumption of homogeneity of regression slopes is not broken. F(1,95)=.017, p(0,896) α > 0.05. As such the results from the ANCOVA can be considered meaningful.
A one-way ANCOVA was then conducted to determine if a statistically significant difference between the positive and negative toned groups on the perceived trust in Volkswagen after controlling for news media exposure exists. A Levene's test was run to see if the variances between the groups differs significantly. This test shows that the underlying assumption of homogeneity of variance for the one way ANCOVA has been met F(1, 97) = 1.245, p=0.267. The analysis shows that there is no significant effect of positive or negative stimulus group on perceived level of trust in Volkswagen. There was no significant effect for the positive or negative stimulus group on the perceived level of trust after controlling for news media exposure. F(1,96) = 3.586, p = 0.061, ηp2=0.036. However, the analysis shows that there is a near significant difference in perceived level of trust in Volkswagen, with people exposed to negative comments showing lower levels of trust (M =3.13, SD =1.02) than people exposed to positive comments (M =3.47, SD =.84 ). As there is no significant effect for the positive and negative stimulus group, or news media exposure F(1, 96) = 1.028, p = 0.279, ηp2= 0.012 on perceived levels of trust in
22 Volkswagen while controlling for news media exposure no support was found for this hypothesis. However, the data does show near significant results.
H2: People exposed to the negative comments perceive Volkswagen as less reputable than people exposed to positive comments.
The same two groups are being used as independent variable for the second analysis. For the second ANCOVA the dependent variable was the perceived reputation of Volkswagen and the covariate was the news media exposure factor. Before conducting the ANCOVA the homogeneity of regression assumption was tested to evaluate the interaction between the covariate and the independent variable. The analysis shows that the assumption of homogeneity of regression slopes is not broken. F(1,95)=0.00, p=0.982, α > 0.05. As such, the results from the ANCOVA can be considered meaningful.
A one-way ANCOVA was conducted to determine a statistically significant difference between the different stimulus groups on the perceived reputation of Volkswagen after controlling for news media exposure. A Levine's test was run to see if the variances between the groups differs significantly. This test shows that the underlying assumption of homogeneity of variance for the one way ANCOVA has been met F(1, 97) = 0.121, p=0.729.
The analysis shows that there is no significant effect of positive or negative stimulus on perceived reputation of Volkswagen. There was no significant effect of stimulus group on the level of perceived reputation after controlling for news media exposure. F(1,96) = 3.688, p = 0.058 ηp2=0.037. However, the analysis shows that there is a near significant difference in perceived reputation of Volkswagen, with people exposed to negative comments perceiving Volkswagen as less reputable (M =3.92, SD =1.31) than people exposed to positive comments (M =4.37, SD =1.30). As there is no significant effect for the positive and negative stimulus group, or news media exposure F(1, 96) = 2.612, p = 0.109, ηp2= 0.026 on the perceived reputation of Volkswagen while controlling for news media exposure no support is found for the second hypothesis. However, the data does show near significant results.
H3: People exposed to the negative comments perceive the CSR reputation of Volkswagen as less reputable than people exposed to positive comments.
For the third ANCOVA the same two groups were used as independent variable. The dependent variable was the perceived CSR reputation of Volkswagen and the covariate was the news media exposure factor. Before conducting the ANCOVA the homogeneity of regression
23 assumption was tested to evaluate the interaction between the covariate and the independent variable. The analysis shows that the assumption of homogeneity of regression slopes is not broken. F(1,95)=0.002, p=0.963, α > 0.05. As such the results from the ANCOVA can be considered meaningful.
A one-way ANCOVA was conducted to determine a statistically significant difference between the positive and negative stimulus groups on the perceived CSR reputation of Volkswagen after controlling for news media exposure. A Levene's test was run to see if the variances between the groups differs significantly. This test shows that the underlying assumption of homogeneity of variance for the one way ANCOVA has been met. F(1, 97) =0.009. p=0.925.
The analysis shows that there is a significant effect of positive or negative stimulus group on perceived CSR reputation of Volkswagen. There was no significant main effect for the positive or negative stimulus groups on the level of perceived CSR reputation after controlling for news media exposure F(1,96) = 2.122, p = 0.148, ηp2=0.022. For media exposure no main effect was found either F(1, 96) = .776, p = 0.38, ηp2= 0.008. As there is no significant effect for positive or negative stimulus groups on the perceived CSR reputation of Volkswagen while controlling for news media exposure no support is found for the third hypothesis.
H4: People exposed to the Volkswagen frame have more trust in Volkswagen than people exposed to the NY times frame.
To test the next set of hypotheses the Volkswagen stimuli and the NY Times stimuli were grouped together. This will provide insight into how people perceive crisis communication from different sources. For this ANCOVA the dependent variable was the perceived trust in the Volkswagen company and the covariate was the news media exposure factor. Before conducting the ANCOVA the homogeneity of regression assumption was tested to evaluate the interaction between the covariate and the independent variable. The analysis shows that the assumption of homogeneity of regression slopes is broken. F(1,95)=8.198, p(0,005) α < 0.05. The results from the ANCOVA cannot be considered meaningful. As such, an ANOVA has been conducted to test this hypothesis.
A one-way ANOVA was conducted to determine if a statistically significant difference between the perceived trust in Volkswagen exist for people exposed to the different sources. The one-way ANOVA shows that there was no statistically significant difference between groups F(1, 97) = 2.059, p = .155).
24 The analysis shows that there is no significant difference in perceived level of trust in Volkswagen for people exposed to a NY times stimuli and people exposed to a Volkswagen stimuli. There was no significant difference for people exposed to different crisis articles on the perceived level of trust in Volkswagen thus no support is found for this hypothesis.
H5: People exposed to the Volkswagen frame are more positive towards Volkswagen's reputation than people exposed to the NY times frame.
For the next hypothesis ANCOVA the dependent variable was the perceived reputation of Volkswagen and the covariate was the news media exposure factor. Before conducting the ANCOVA the homogeneity of regression assumption was tested to evaluate the interaction between the covariate and the independent variable. The analysis shows that the assumption of homogeneity of regression slopes is not broken. F(1, 95)=0.099, p=0,753, α > 0.05. As such the results from the ANCOVA can be considered meaningful.
A one-way ANCOVA was conducted to determine a statistically significant difference between the different news sources on the perceived reputation of Volkswagen after controlling for news media exposure. A Levene's test was run to see if the variances between the groups differs significantly. This test shows that the underlying assumption of homogeneity of variance for the one way ANCOVA has not been met F(1, 97) = 4.216, p=0.043.
The analysis shows that there is no significant effect of crisis communication frame on perceived reputation of Volkswagen. There was no significant effect of stimulus group on the level of perceived reputation after controlling for news media exposure. F(1,96) = .694, p = 0.407 ηp2=0.007. As there is no significant effect of stimulus group or news media exposure F(1, 96) = 1.777, p = 0.186, ηp2= 0.018 on the perceived reputation of Volkswagen while controlling for news media exposure no support is found for the fifth hypothesis.
H6: People exposed to the Volkswagen frame are more positive towards Volkswagen's CSR reputation than people exposed to the NY time frames.
For the following ANCOVA the dependent variable was the perceived CSR reputation of Volkswagen and the covariate was the news media exposure factor. Before conducting the ANCOVA the homogeneity of regression assumption was tested to evaluate the interaction between the covariate and the independent variable. The analysis shows that the assumption of homogeneity of regression slopes is broken. F(1,95)=3.663, p=0.041, α < 0.05. As such the results from the ANCOVA cannot be considered meaningful.
25 A one-way ANOVA was conducted to determine a statistically significant difference between the different frames of news on the perceived CSR reputation of Volkswagen. A Levene's test was run to see if the variances between the groups differs significantly. This test shows that the underlying assumption of homogeneity of variance for the one way ANCOVA has been met. F(1, 97) =0.473. p=0.493. As such, an ANOVA has been conducted to test this hypothesis. A one-way ANOVA was conducted to determine if a statistically significant difference between the perceived CSR reputation of Volkswagen exist for people exposed to the different news sources. The one-way ANOVA shows that there was a statistically significant difference between the groups F(1, 97) = 8.533, p = .004). The analysis shows that there is a significant difference in perceived level CSR reputation of Volkswagen for people exposed to a NY times frame (M =3.05, SD =.94) and people exposed to a Volkswagen frame (M =3.6 , SD =.93 ). Thus people exposed to a Volkswagen news frame perceive Volkswagen to have a higher CSR reputation than people exposed to a NY times news source. As such, data was found that supports this hypothesis.
For the following three ANCOVA's the independent variable, stimulus group, included four levels: NY times crisis communication article with positive comments, Volkswagen crisis communication article with positive comments, NY times crisis communication article with negative comments, Volkswagen crisis communication article with negative comments.
H7: People exposed to the Volkswagen news frame with positive comments will score the highest on the perceived trust scale and people exposed to the NY Times news frame with negative comments the lowest.
For the next ANCOVA the dependent variable was the perceived trust in the Volkswagen company and the covariate was the news media exposure factor. Before conducting the ANCOVA the homogeneity of regression assumption was tested to evaluate the interaction between the covariate and the independent variable. The analysis shows that the assumption of homogeneity of regression slopes is not broken. F(3,91)=2,286, p(0,084) α > 0.05. As such the results from the ANCOVA can be considered meaningful.
A one-way ANCOVA was then conducted to determine if a statistically significant difference between the different stimulus groups on the perceived trust in Volkswagen after controlling for news media exposure exists. A Levene's test was run to see if the variances between the groups differs significantly. This test shows that the underlying assumption of homogeneity of variance for the one way ANCOVA has been met F(3, 95) = 2.914, p=0.038.
26 The analysis shows that there is no significant effect of crisis communication stimulus group on perceived level of trust in Volkswagen. There was no significant effect of stimulus group on the perceived level of trust after controlling for news media exposure. F(3,94) = 2.265, p = 0.086, ηp2=0.067. As there is no significant effect of stimulus group, or news media exposure F(1, 94) = 1.151, p = 0.286, ηp2= 0.012 on perceived levels of trust in Volkswagen while controlling for news media exposure no support is found for the seventh hypothesis.
H8: People exposed to the Volkswagen news frame with positive comments will score the highest on the perceived reputational scale and people exposed to the NY Times news frame with negative comments the lowest.
For the following ANCOVA the dependent variable was the perceived reputation of Volkswagen and the covariate was the news media exposure factor. Before conducting the ANCOVA the homogeneity of regression assumption was tested to evaluate the interaction between the covariate and the independent variable. The analysis shows that the assumption of homogeneity of regression slopes is not broken. F(3,91)=0.003, p=1,000, α > 0.01. As such the results from the ANCOVA can be considered meaningful.
A one-way ANCOVA was conducted to determine a statistically significant difference between the different stimulus groups on the perceived reputation of Volkswagen after controlling for news media exposure. A Levene's test was run to see if the variances between the groups differs significantly. This test shows that the underlying assumption of homogeneity of variance for the one way ANCOVA has been met F(3, 95) = 1.025, p=0.385.
The analysis shows that there is no significant effect of crisis communication stimulus on perceived reputation of Volkswagen. There was no significant effect of stimulus group on the level of perceived reputation after controlling for news media exposure. F(3,94) = 1.586, p = 0.198 ηp2=0.048. As there is no significant effect of stimulus group or news media exposure F(1, 94) = 2.501, p = 0.117, ηp2= 0.026 on the perceived reputation of Volkswagen while controlling for news media exposure no support is found for this hypothesis.
H9: People exposed to the Volkswagen news frame with positive comments will score the highest on the CSR scale and people exposed to the NY Times news frame with negative comments the lowest.
For the final ANCOVA the dependent variable was the perceived CSR reputation of Volkswagen and the covariate was the news media exposure factor. Before conducting the ANCOVA the
