From a “breach of trust” to the evolution of a corporate crisis

From a “breach of trust” to the evolution

of a corporate crisis

Master’s Thesis

From a “breach of trust” to the evolution of a corporate crisis:

Capturing the organization’s crisis response strategies and the primary public’s reactions on the Facebook Cambridge Analytica scandal

Lea Steinmann 12088897

Master’s Program Communication Science Graduate School of Communication

University of Amsterdam

Supervisor: dhr. prof. dr. R. (Rens) Vliegenthart

Abstract

The purpose of this study is to extend the current conception of crisis communication research by analyzing both organizational responses and primary public’s reactions throughout the course of a corporate crisis. The organization’s crisis response strategies, the organizational engagement, the organizational stance, as well as the primary public’s displayed emotions, coping strategies, and attribution of crisis responsibility are investigated during the temporal development of the Facebook Cambridge Analytica scandal. A manual content analysis in the context of social media communication was conducted by examining Facebook posts (n = 21), and comments (n = 901) to capture direct public reactions on the organization’s crisis responses. The findings demonstrate a change of emotions, increasing levels of the conative coping strategy, and decreasing levels of the attribution of responsibility to actors involved as displayed by the primary public within the first six months after the initial crisis outbreak. Whereas the scapegoat strategy is primarily present towards the beginning and the end of the period analyzed, no other changes in regard to crisis response strategies, the level of

organizational engagement, or the organizational stance are evident. Lastly, the scapegoat and justification strategy are found to lead to more anger and less sympathy among Facebook users. Based on Integrated Crisis Mapping (ICM), the data breach scandal is generally related to a high level of organizational engagement and the public’s conative coping strategy.

Hence, the study contributes to the state of research in several ways: First, an integrated framework may focus on examining direct relationships between an organization’s crisis responses and public reactions during crises. Secondly, the evolving character of crises remains of crucial importance when investigating the latter in a systematic manner. Thirdly, public relations practitioners may monitor and assess primary public’s reactions on corporate crises to enhance the effectiveness of strategic crisis management.

Key words: crisis communication, situation-based crisis responses, integrated crisis mapping, crisis evolution, social media communication, content analysis

Introduction

On March 17, 2018, The New York Times and The Observer published high-profile stories featuring Christopher Wylie, a former employee at Cambridge Analytica, who claimed that the political consulting firm harvested data from over 50 million Facebook users without their consent (Cadwalladr & Graham-Harrison, 2018; Rosenberg, Confessore, & Cadwalladr, 2018). Wylierevealed that Cambridge Analytica developed a quiz app with Aleksandr Kogan, a researcher from Cambridge University, which illegally collected Facebook user data from test takers and their Facebook friends. In addition, Cambridge Analytica worked closely with the Trump campaign during the U.S. Presidential Election in 2016, and the Brexit referendum campaign group Leave.EU in the U.K., which may have contributed to the success of both campaigns (Bump, 2018; Scott, 2018). Worldwide, over 87 million were affected by the data breach, including more than 70 million Facebook users in the U.S., 1.1 million users in the U.K., and 600.000 users from Canada (Bowcott & Hern, 2018). As a result, Cambridge Analytica declared bankruptcy and Facebook was fined 500.000 pounds for a lack of transparency and the failure to protect its user data (Hern & Pegg, 2018). Therefore, data breaches remain a growing concern in society (Elhai & Hall, 2016) when it comes to

protecting the organization’s reputation, the government’s sovereignty, and their citizens, and may result in a decrease of monetary, brand, or consumer trust (Demirkan et al., 2015). Thus, the Facebook Cambridge Analytica scandal is a prime example of a data breach crisis, which entails millions of consumer accounts affected, and may severely impact a company’s

performance by a decline in stock prices, and reduced revenue growth rates (Kim, Johnson & Park, 2017; Neate, 2018).

When it comes to counteracting reputational threat resulting from a crisis, an

organization’s post-crisis communication plays a substantial role(Benoit, 1997; Coombs & Holladay, 2002). Hereby, situation-based crisis response strategies may be applied to repair corporate reputation, to shape the public’s perception of the crisis, or to prevent negative

behavioral intentions towards the organization (Coombs, 2007). Besides, time is a crucial element, since organizations face a high pressure of immediately reacting with concrete explanations to the public (Fishman, 1999). In this regard, the suspension of Cambridge Analytica’s services from Facebook was announced one day before the news reports were published (Grewal, 2018). Additionally, Facebook’s CEO Mark Zuckerberg (2018) released a statement on his Facebook account, referring to the data breach incident as a “breach of trust” (para. 8), and pointed to specific measures to address the crisis, such as technical and

procedural improvements. However, Zuckerberg’s reaction was criticized for being released “after days of delay” (Haggerty, 2018, para. 6), since the statement was published four days after the news broke.

Apart from solely considering organizational responses on corporate crises, Jin, Pang, and Cameron (2010) focused on mapping public reactions on the latter, examining crisis management from a “public-based, emotion-driven perspective” (p. 428). Hereby, an organization’s primary public is defined as being the most affected by the crisis in terms of cognition, emotion, and behavior, holds long-term interests towards the organization, and influences the firm’s reputation and operations (Jin et al., 2010). To that end, the bridging role of communication (Grunig, 2006) remains a crucial task for organizations, since organizations operate in an environment, where issues of interest for both the organization and its primary public are constantly being discussed (Mitchell, Agle & Wood, 1997). Therefore, Luoma-aho, Tirkkonen & Vos (2013) consider a more dialogical approach when it comes to interaction processes between an organization and its primary public, and further argue that for

organizations, “interaction with and among stakeholders moves onto new stages facilitated by technology and the social media” (p. 2). Yet, public reactions on corporate crises were

previously examined through news coverage, whereby events are filtered through journalists’ perceptions (Jin & Cameron, 2006; Jin et al., 2010), rather than assessing crises through social media communication, which allows capturing immediate public reactions on

organizations’ crisis responses (Porter, 2009; Rybalko & Seltzer, 2010). Besides, Coombs (2007), and Jin et al. (2010) have fallen short on examining the change of crisis response strategies and public reactions throughout the unfolding of corporate crises. However, the temporal development of a crisis due to its “rapidly evolving and escalating character” (Van der Meer & Verhoeven, 2013, p. 231) was previously emphasized in the field of crisis communication research(Van der Meer, Verhoeven, Beentjes & Vliegenthart, 2014)

In general, this study considers both the organization-centered approach by Coombs (2007), when it comes to detecting crisis response strategies, and the primary public-based approach by Jin et al. (2010), capturing public reactions on corporate crises. Furthermore, the public’s attribution of crisis responsibility (Kelley & Michela, 1980) is investigated to obtain a more detailed account of the primary public’s perception of a crisis. Thus, this paper aims to (a) analyze the crisis response strategies applied by Facebook addressing the data breach scandal, and (b) examines the level of organizational engagement, the organizational stance, as well as primary public’s displayed emotions, coping strategies, and the attribution of crisis responsibility as the crisis evolved in the context of social media communication. The (c) direct relationship between the presence of the organization’s crisis response strategies and emotions expressed by the primary public is further investigated, suggesting the definition of a new framework regarding an organization’s crisis management and its dependency on public reactions during crises (Jin et al., 2010). An integrated approach may further consider the understanding of corporate crises as an evolving process over time (Weick, 1988), which affects both the organization and its primary public when responding to a complex and rapidly developing crisis situation. Consequently, detecting appropriate crisis responses in regard to reactions by an organization’s primary public (Jin, Pang & Cameron, 2012) may enhance the current conception of crisis communication for researchers and public relations practitioners.

Theoretical Framework

Situation-Based Crisis Responses from an Organization-Centered Perspective

Coombs (2007) describes a corporate crisis as an unexpected event, which may disrupt organizational processes and pose financial and reputational threat. Specifically, a data breach refers to the loss of consumer information (Ayyagari, 2012), defined as “an incident in which an individual name plus a social security number, driver’s license number, medical record or financial record (…) is potentially put at risk because of exposure“ (ITR, 2014). The latter is due to outright data theft, intentional and unlawful sale of customer information, loss of equipment, or unintentional leaks of personal data (Acquisti, Friedman, & Telang, 2006).

When facing reputational threat due to a crisis, organizations may apply crisis response strategies based on the type of crisis and the locus of control (Coombs, 1998). The Situational Crisis Communication Theory (SCCT) (Coombs, 2007) conceptualizes how corporate reputation is affected by the initial crisis responsibility, the crisis history, the prior relational reputation, and the crisis severity. The strongest attribution of crisis responsibility and a severe reputational threat are present in preventable crises, characterized by an

intentional placement of people at risk, inappropriate or unlawful actions, and violation of laws and regulations (Coombs, 2007). The Facebook Cambridge Analytica scandal can be classified as a preventable crisis, since Facebook failed to protect personal information of its users from being obtained by Kogan and Cambridge Analytica (Lagone, 2018), and did not take sufficient action when first becoming aware of the data breach in 2015 (Browne, 2018).

In regard to organizations’ crisis responses, Coombs (2007) introduced denial (i.e. attack of the accuser, denial, scapegoat), diminish (i.e. excuse, justification), and rebuild crisis response strategies (i.e. compensation, apology), and suggested accommodative strategies for organizations facing preventable crises, including compensation and apology. In the event of data breach crises, organizations may primarily inform individuals affected, and provide advice on how to protect themselves in data breach situations (Bentley, Oostman & Shah,

2018). The study found that when organizations were clearly responsible for a data breach, they were more likely to signal empathy and acknowledge responsibility. Besides, Jenkins, Anandarajan, and D’Ovidio (2014) underlined apology and regret as the most effective strategies in data breach crises, whereas Kim et al. (2017) found that breached firms chose different types of strategies, including denial, ingratiation, and regret.

Furthermore, Weick (1988) considered the complex development of crises by stating: “To sort out a crisis as it unfolds often requires action which simultaneously generates the raw material that is used for sensemaking and affects the unfolding crisis itself” (p. 305). Thus, crises, due to low levels of probability, may trigger sensemaking processes and a constant evaluation of past behavior by the actors affected (Weick, 1988). The characterization of crises as complex situations for organizations and the primary public implies collective sensemaking to provide coherence of the situation, and to attribute crisis responsibility to reduce ambiguity and uncertainty (Van der Meer & Verhoeven, 2013; Van der Meer et al., 2014). Hereby, actions taken by the organization or the public may solve or worsen

immediate problems, whereby the latter predominantly prolongs a crisis (Shrivastava, 1987). While past studies acknowledged the evolving character of crises when analyzing the

alignment of framing processes by public relations practitioners, the media, and the public by defining different crisis phases (Van der Meer & Verhoeven, 2013; Van der Meer et al., 2014), research has fallen short on investigating organizations’ crisis response strategies throughout the course of a crisis (Coombs, 2007; Jenkins et al., 2014; Kim et al., 2017; Bentley et al., 2018). Consequently, this study examines the crisis response strategies applied by Facebook as the data breach scandal evolved within the first six months after the initial crisis outbreak, and poses the following question:

RQ1: How do the crisis response strategies applied by Facebook change as the Facebook Cambridge Analytica scandal evolves?

Examining Public Reactions on Corporate Crises through Integrated Crisis Mapping Jin et al. (2010) developed the Integrated Crisis Mapping (ICM) model, which focuses on capturing primary public’s emotions in times of corporate crises on two continua, namely the primary public’s coping strategy (conative to cognitive coping), and the level of

organizational engagement (low to high). The ICM model aims to offer practical insights for practitioners to understand public’s emotions experienced during crises to develop appropriate crisis responses and tools for effective crisis management (Jin et al., 2012). First, the

emotions anger, fright, anxiety, or sadness are identified, and mark the primary or secondary level emotions displayed by the primary public in crisis situations (Jin et al., 2010). However, the authors further questioned a concrete distinction between primary (i.e. immediately expressed emotion), and secondary level emotions (i.e. experienced in subsequent instances), since emotions experienced in crises are usually intertwined with one another. Moreover, the primary public’s coping strategy is either characterized as conative coping, such as the public aiming to alter a troubled relationship with the organization by taking concrete action, or cognitive coping, defined as the public’s tendency to change the interpretation of a crisis in terms of one’s own well-being (Jin et al., 2010). Thus, conative coping refers to the concept of problem-focused coping, whereas cognitive coping corresponds with emotion-focused coping (Folkman & Lazarus, 1988). The level of organizational engagement is indicated by the organization’s appraisal of crisis responsibility, and the relevance of the crisis to

organizational objectives (Jin et al., 2012). Additionally, the organizational stance refers to qualified-rhetoric-mixed, and action-based accommodation (Jin & Cameron, 2006), based on the contingency theory of accommodation in public relations (Cancel, Cameron, Sallot & Mitrook, 1997). An action-based stance characterizes the organization as willing to yield to the public’s propositions on future procedures, whereas the organization admits wrongdoing, expresses regret, or apologizes to the public in terms of the qualified-rhetoric-mixed stance (Jin et al., 2012).

Furthermore, Jin et al. (2010) examined five corporate crises situated in the first quadrant of the ICM model (high organizational engagement, conative coping). Hereby, the primary public’s displayed emotions predominantly implied anger, anxiety and sadness, and a more qualified-rhetoric-mixed stance by the organization was detected, which may or may not result in concrete action. Thus, Jin et al. (2012) suggested that further research may focus on analyzing a broader variety of emotions experienced during crises, as well as using a “more in-depth case analysis approach to examine how the individual crisis evolved” (p. 289 f.). To that end, the Facebook Cambridge Analytica scandal may be positioned in the first quadrant of the ICM model, since Facebook acknowledged crisis responsibility in front of the U.S. Congress, and was prompted to revise security tools (Levin, 2018) in terms of a high level of organizational engagement. Besides, users were advised to alter their security settings, or to stop using the social media platform (McGauley, 2018) in the sense of conative coping.

Moreover, the public’s interpretation of organizational behavior leading to a crisis plays a significant role, since individuals tend to search for the causes of negative and unexpected events (Weiner, 1985). The initial crisis responsibility, based on the Attribution Theory (Kelley & Michela, 1980), is one factor contributing to the reputational threat of a firm in crisis situations (Coombs, 2007).By attributing the crisis responsibility to the

organization, individuals may experience emotions, such as anger or sympathy, which in turn may impact behavioral intentions (Weiner, 2006). Kim et al. (2017) further proposed to investigate public’s perceptions on data breach crises, and the attribution of crisis

responsibility, which suggests examining the latter on the Facebook Cambridge Analytica scandal. Regarding the temporal unfolding of crises, Jin, Liu, and Austin (2014) noted that “publics could respond differently depending on whether it is the first time they are ever exposed to a given crisis or they have heard about the crisis and are in the process of deciding (…) how to react” (p. 90). By taking the limitations mentioned by past studies into account, the study aims to answer the following research questions:

RQ2a: How do the emotions displayed by the primary public change as the Facebook Cambridge Analytica scandal evolves?

RQ2b: How do the coping strategies (cognitive and conative coping) evident in the primary public change as the Facebook Cambridge Analytica scandal evolves?

RQ2c: How does the level of organizational engagement by Facebook change as the Facebook Cambridge Analytica scandal evolves?

RQ2d: How does the organizational stance (action-based and qualified-rhetoric-mixed) by Facebook towards the primary public change as the Facebook Cambridge Analytica scandal evolves?

RQ2e: How does the attribution of crisis responsibility evident in the primary public change as the Facebook Cambridge Analytica scandal evolves?

Moving Towards an Integrated Perspective

When examining corporate crises based on the ICM model, Jin et al. (2010) took crisis response strategies by organizations (Coombs, 2007) into account, however, the latter was exclusively analyzed in terms of the organizational stance, as well as the message attributes of the crisis response (i.e. defensive-accommodative, ambiguous-specific, emotional-factual). Moreover, Jin et al. (2014) found that when a crisis was internal, the public was more likely to accept accommodative crisis responses, and reported more attribution-dependent emotions, such as anger, contempt or disgust, compared to an external crisis origin. Whereas the study only analyzed the acceptance of crisis response strategies and public’s displayed emotions as outcomes, the direct relationship between organizations’ crisis responses and the publics-based emotion-driven perspective may be further investigated. This is due to the fact that organizations may aim at influencing public’s attitudes and behavior (Porter, 2009), and attempt to shape the public’s perception of the organization to elicit favorable public relations outcomes, such as positive reputation (Carroll & McCombs, 2003; Kim & Kiousis, 2012). For

instance, affective attributes of a corporate blog message were found to influence the public’s overall, cognitive and affective evaluation of organizations (Kim & Kiousis, 2012). Thus, the presence of a crisis response strategy may influence primary public’s displayed emotions when responding to a crisis. McDonald, Sparks, and Glendon (2010) found that excuse, justification, and denial evoked negative reactions, such as anger or negative word-of-mouth, whereas confession and no comment resulted in sympathy, loyalty and positive attitudes. However, the study only considered a small number of crisis response strategies (i.e. no comment, denial, excuse, justification, confession), and primary public’s displayed emotions (i.e. anger, sympathy), compared to Coombs’ (2007) operationalization of crisis response strategies, and the variety of emotions expressed during crises (Jin et al., 2010). Extending from the findings by McDonald et al. (2010), this paper aims to shed light on the following research question:

RQ3: Does the presence of a crisis response strategy applied by the organization relate to the presence of certain emotions displayed by the primary public?

Methodology

To answer the research questions, a manual quantitative content analysis was conducted to examine the crisis response strategies applied by Facebook, and to map the primary public’s reactions on the Facebook Cambridge Analytica scandal.

Sample and Coding Procedure

Past studies have predominantly relied on news coverage when examining public reactions on crises (Jin et al., 2010, 2012), however, news coverage may be subject to selection and interpretation processes by journalists (Bennett, 2016). Hence, a content analysis in the context of social media is performed, since social media facilitate interactive

communication and content exchange, and are embedded as an integral part in the organization’s communication strategy (Tinker & Fouse, 2009). In times of crises, the primary public may use social media to gather insider information (Austin, Liu & Jin, 2012), and to voice personal opinions by reacting to organizations’ crisis responses (González-Herrero & Smith, 2008).

To examine the crisis response strategies applied by the organization, Facebook posts published through the account of Facebook’s CEO Mark Zuckerberg are chosen as a unit of analysis. Whereas the organization itself exclusively communicates on its company pages targeted to the user’s location, e.g. FacebookNL, Mark Zuckerberg’s Facebook account counts over 119 million followers, and is displayed to all users regardless of individual’s location settings. The posts published between March 17, to September 16, 2018 were manually retrieved and analyzed. The relevance of the post in regard to the crisis was indicated if the post implied key words, such as “Cambridge Analytica”, “security settings”, or “privacy control” (see Appendix B). To gain a more comprehensive account of issues related to the data breach scandal, the posts were classified as directly, or indirectly referring to the crisis. Hereby, only the sections of the post containing a direct, or indirect reference were coded. If a post both directly and indirectly referred to the crisis, only the sections directly relating to the crisis were examined. A direct reference was given when a post included key words, such as “Cambridge Analytica”, “privacy settings”, and “General Data Protection Regulation”, or the event “F8”, in which Zuckerberg introduced steps taken by Facebook against the data breach incident (Guynn, 2018). When a post exclusively referred to election integrity and content regulation in terms of hate speech, or political advertising, an indirect reference to the crisis was present, since the data breach scandal was generally related to election interference (Bump, 2018; Scott, 2018).

The codebook was applied on the written section of the post, texts displayed on images, and videos embedded in posts, which were transcribed beforehand. Additionally,

external content linked to by several posts was coded, including the Facebook Transparency Report, Community Standards, a search engine for political content advertisement, an opinion piece published by The Washington Post, and two articles from Facebook Newsroom. As a result, 45 posts were analyzed, whereby 13 posts directly, and eight posts indirectly referred to the data breach scandal.

Examining public’s reactions on the crisis was achieved by analyzing the comments each Facebook post referring to the data breach scandal received by Facebook users. Hereby, the 70 “most relevant” comments, as marked by Facebook in terms of responses from verified pages or profiles, and comments with the most reactions or replies (“What does most

relevant”, 2018), were manually collected. The sample (n = 1470) implied comments published within 14 days after the release of the organization’s post, and comments directly responding to the post, whereas users’ replies to comments by other users were excluded. In regard to the final sample used for examining the research questions, 1323 comments were analyzed, whereas 901 comments referred to the crisis.

Facebook user reactions on posts addressing the data breach scandal in terms of likes, reactions, comments, and shares varied over time (Figure 1). Overall, user engagement was the highest when a post addressed the crisis for the first time, announced Zuckerberg’s testifying in front of the Congress, or broadcasted the conventions F8 and ViaTech via video stream. The highest levels of user reactions were present on posts published within the first two and a half months after the initial crisis outbreak, and steadily declined in regard to posts released between July 31, to September 13, 2018. However, posts published between May 25, to July 24, 2018 did not refer to the data breach incident.

Figure 1. User reactions on Mark Zuckerberg’s Facebook posts directly or indirectly addressing the Facebook Cambridge Analytica scandal

Measures

The organization’s crisis response strategies (Coombs, 2007) were examined based on the adaption by Jin et al. (2010), including attack the accuser, excuse, justification,

ingratiation, corrective action, and full apology. Besides, scapegoat was added, since the strategy is implied in Coombs’ (2007) typology of primary crisis response strategies, and

corrective action further comprised the compensation strategy.The variables were

dichotomously measured by indicating the presence of the crisis response strategy by either “Yes” (1), or “No” (0). Organizational engagement was measured on two seven-point Likcrt-like scales. The first scale relevance of the crisis to organizational goals was adapted from Jin et al. (2012), ranging from “The post implies that the crisis has nothing to do with the

organization’s goals, and is not likely at all to put the organization operation and reputation in danger” (1) to “The post implies that the crisis hits the organization’s goals and creates devastating damage” (7). The second scale organization’s appraisal of crisis responsibility ranged from “very low” (1) to “very high” (7). If the issues related to the variables were not addressed in the post, the code “99” was used (Jin et al., 2012). Furthermore, message

attributes of a post were examined using a seven-point Likert-like scale, including the

defense-accommodation continuum (“defensive” (1) to “accommodative” (7)), the ambiguity-specification continuum (“ambiguous” (1) to “specific“ (7)), and the emotion-cognition continuum (“emotional” (1) to “factual” (7)) (Jin et al., 2010). The organizational stance was measured on a seven-point Likert-like scale based on items from Jin and Cameron (2006), and implied the indication of “not evident” (1) to “very evident” (7) for the action-based stance, and the qualified-rhetoric-mixed stance. The action-based stance comprised several indicators (e.g. agree with the public on future action or procedures; agree to try the solutions suggested by the public), as well as the qualified-rhetoric-mixed stance (e.g. express regret or apologize to the public; change own position towards that of the public) (Jin et al., 2010).

The emotions displayed by the organization and the primary public were examined in terms of no display of emotion, anger, fright, anxiety, or sadness (Jin et al., 2010). Fright was operationalized in the sense of shock as a response to immediate or overwhelming danger, whereas anxiety was characterized as the public facing existential, ambiguous, or uncertain threat in terms of fear (Lazarus, 1993). Furthermore, sympathy (Jin, 2014), and contentment in the sense of optimism, satisfaction (Fredrickson, 1988; McDonald et al., 2010), hope or positive anticipation (Takahashi, 1999; Toikkanen & Muurinen, 1999), and other emotion(s) were added. The variables were dichotomously measured by indicating the presence of the emotion by either “Yes” (1), or “No” (0).Adapted from the operationalization by Jin et al. (2010), the primary public’s coping strategies were measured on a seven-point Likert-like scale, ranging from “not evident” (1), to “very evident” (7). Cognitive coping was measured between “The comment implies that the user is not willing to change its perception of the crisis, e.g. the crisis is what it is and there is no need to explain it further or change the angle of looking at it” (1) to “The comment implies that the user is extremely proactive in taking another look at the crisis and providing full explanation on the new perception” (7). Conative coping ranged from “The comment implies that the user is not willing to take any action

regarding the crisis” (1) to “The comment implies that the user is extremely proactive in taking actions with detailed plans against the crisis“ (7). The attribution of crisis

responsibility was examined in terms of Facebook, Cambridge Analytica, Aleksandr Kogan, or other actor(s). The items were dichotomously measured by indicating the mentioning of actor(s) as deemed responsible for the crisis by either “Yes” (1), or “No” (0).

Reliability

Intercoder reliability testing was conducted with the help of a second coder, who was familiar with crisis communication research. After performing a round of test coding, using five random organizational posts published between March 17, to September 16, 2018, and 70 random Facebook comments responding to the posts, the results were discussed, and detailed coding instructions and examples were provided. Secondly, approximately ten percent of the whole sample (Lombard, Snyder-Duch, & Bracken, 2002) was drawn, which comprised every ninth Facebook post (n = 5) published within the first six months after the initial crisis

outbreak. Out of the 70 most relevant comments each Facebook post referring to the corporate crisis (n = 21) received, every tenth Facebook comment was analyzed (n = 147). However, the comments included in the reliability sample were not a part of the final sample (n = 1323) used for examining the research questions.

In regard to Facebook comments, Krippendorff’s alpha ranged between .69 to 1, which is above the lowest conceivable limit, and tentative conclusions are acceptable, α ≥ .67 (Krippendorff, 2004). For Facebook posts, sufficient levels were achieved for continuous variables, whereby Krippendorff’s alpha ranged between .72 and .99 (see Appendix A). Due to the small number of posts analyzed, a consent between coders was considered for

dichotomous variables, whereby a full consent was achieved for most variables. The variables excuse, justification, no display of emotion, sadness, and contentment showed a consent of four out of five. Hereby, the reliability levels were regarded as sufficient.

Analysis

To conduct the analyses, the metric variable time was calculated to capture a linear time trend based on a daily level, whereby the values indicated the number of days within the first six months after the initial crisis outbreak. Day one marked the day the news about the data breach scandal broke. Therefore, the values ranged between “1” (March 17, 2018), to “184” (September 16, 2018). For instance, when a post was published on March 21, 2018, the value for the variable time was “5” in regard to the Facebook post and the comments by users responding to the post. In addition, the variables relevance of the crisis to organizational goals, and organization’s appraisal of crisis responsibility were used to calculate the mean variable for organizational engagement.

To examine how the organization’s crisis response strategies (RQ1), the level of organizational engagement (RQ2c), and the organizational stance (RQ2d) changed throughout the data breach scandal, descriptive statistics were considered, which is due to the small number of posts published within the examined period. Secondly, binary logistic regressions were performed to analyze how the primary public’s displayed emotions (RQ2a), and the attribution of crisis responsibility (RQ2e) changed as the crisis evolved. When investigating how the primary public’s coping strategies changed over time (RQ2b), two linear regression analyses were conducted. Lastly, several chi-square tests for association were performed between the crisis response strategies, which showed a varying degree of being present in organizational posts, in regard to all emotions to analyze whether the presence of a crisis response strategy in a post was related to emotions expressed by Facebook users (RQ3).

Furthermore, to examine whether the level of the primary public’s conative coping strategy significantly differed from the cognitive coping strategy, a paired-samples t-test was performed. The latter was further conducted in regard to detecting differences in the presence of the organization’s action-based stance, compared to the qualified-rhetoric-mixed stance.

Results Crisis Response Strategies

RQ1 asked how the organization’s crisis response strategies changed as the data breach scandal evolved. Hereby, the first post (P001) addressing the crisis implied scapegoat, justification, ingratiation, and corrective action as crisis response strategies. Throughout the temporal evolvement of the crisis, no other crisis response strategies were present in

Facebook posts (Table 1). Corrective action was implied in all posts directly, and indirectly referring to the corporate crisis, and was exclusively present in eight posts throughout the development of the crisis (P003, P009, P010, P012, P013, P017, P019, P033). The strategy exclusively referred to the introduction of new tools and the revision of current practices, whereas no indication of compensating victims was present. Furthermore, the combination of scapegoat, justification, and corrective action (P006, P007, P008, P037, P039, P045),

ingratiation and corrective action (P011, P033), justification, ingratiation and corrective action (P014), justification and corrective action (P022), and scapegoat and corrective action (P041) was implied in posts addressing the data breach. Whereby corrective action, the presentation of reasons for prior actions (P001, P006, P007, P008, P014, P022, P037, P039, P045), and the expression of gratitude towards the public’s understanding (P011, P011, P014, P033) were present in posts throughout the temporal development of the crisis, blaming an individual or group outside of the organization for the crisis was present only at the

beginning, and towards the end of the first six months after the initial crisis outbreak (P001, P006, P007, P008, P034, P037, P039, P041, P045).

Table 1

Descriptive statistics: The organization’s crisis response strategies within the first six months after the initial crisis outbreak

Crisis response strategy N M SD Crisis response strategy N M SD Attack the accuser 21 .00 .00 Justification 21 .43 .51

Denial 21 .00 .00 Ingratiation 21 .19 .40

Scapegoat 21 .43 .51 Corrective action 21 1.00 .00

Excuse 21 .00 .00 Full apology 21 .00 .00

Primary Public’s Displayed Emotions

RQ2a examined how the emotions displayed by the primary public changed throughout the crisis. A binomial logistic regression was performed to ascertain the linear effect of time on the likelihood that comments show no display of emotion (M = .36, SD = .48), anger (M = .20, SD = .40), fright (M = .00, SD = .06), anxiety (M = .02, SD = .14), sadness (M = .02, SD = .15), sympathy (M = .39, SD = .49), or contentment (M = .10, SD = .30). The logistic regression model regarding no display of emotion was statistically highly significant, χ2(1) = 13.08, p <. 001. However, the model only explained 2% (Nagelkerke R2) of the variance in users not displaying any emotion. 64.2% of cases were correctly classified. The predictor variable was statistically highly significant, p < .001 (Table 2). For one day further in time, the odds of comments showing no display of emotion increased by factor 1.01.

Additionally, the logistic regression model was statistically highly significant when examining the linear effect of time on the likelihood that comments show sadness, χ2(1) = 14.57, p < .001. 97.6% of cases were correctly classified. However, the model only explained 7.8% (Nagelkerke R2) of the variance in comments showing sadness over time. The predictor variable was statistically highly significant, p = .006, and for one day further in time, the odds

of comments showing sadness decreased by factor .97. In turn, no significant results were found regarding anger, χ2(1) = 3.83, p > .05, fright, χ2(1) = .21, p > .05, anxiety, χ2(1) = 1.50, p > .05, sympathy, χ2(1) = 2.03, p > .05, and contentment, χ2(1) = 2.59, p > .05, when it comes to changes in the presence of emotions displayed by the primary public as the crisis evolved.

Table 2

Binary logistic regressions: Primary public’s emotions as the crisis evolves

Emotion b* SEb Emotion b* SEB

No display of .005*** .001 Anxiety .005 .004

emotion Sadness -.029** .011

Anger -.003 .002 Sympathy -.002 .001

Fright .005 .010 Contentment -.004 .002

Note. All tests have one degree of freedom. *p < .05, **p < .01, ***p < .001. b* = unstandardized regression coefficient; SEb = Standard error of the coefficient.

Primary Public’s Coping Strategies

RQ2b examined how the coping strategies evident in the primary public changed as the data breach scandal evolved. A linear regression was calculated to predict the public’s cognitive coping strategy throughout the first six months after the initial crisis outbreak (Table 3). However, no significant regression equation was found, F(1, 899) = .002, p > .05, R2 = .00, indicating that the primary public’s coping strategy did not change as the crisis progressed.

Table 3

Linear regression: The primary public’s cognitive coping strategy associated with the temporal evolvement of the crisis

Variable b* SEB β t p 95% CI

(Constant) 3.370 .076 44.575 .000 [3.222, 3.519]

Time .00004 .001 .001 .044 .965 [-.002, .002]

Note. b* = unstandardized regression coefficient; SEb = Standard error of the coefficient; 𝛽 =

standardized coefficient.

Secondly, a linear regression was conducted to predict the public’s conative coping strategy within the first six months after the initial crisis outbreak (Table 4). A significant linear regression equation was found, F(1, 899) = 7.71, p = .006, R2 = .01. Hereby, only 0.8% of the primary public’s conative coping strategy can be predicted by the time. The model shows that time has a very weak association with the primary public’s conative coping strategy. During the first six months after the initial crisis outbreak, the primary public’s conative coping strategy as implied in the comments increased 0.002 times every day, in terms of the public altering the troubled relationship with the organization by taking concrete action.

Table 4

Linear regression: The primary public’s conative coping strategy associated with the temporal evolvement of the crisis

Variable b* SEB β t p 95% CI

(Constant) 3.988 .046 86.579 .000 [3.898, 4.078]

Time .002 .001 .092 2.776 .006 [.000, .003]

Note. b* = unstandardized regression coefficient; SEb = Standard error of the coefficient; 𝛽 =

Organizational Engagement

RQ2c investigated how the level of organizational engagement changed throughout the development of the crisis. When considering the values in regard to the relevance of the crisis to organizational goals, and the organization’s appraisal of crisis responsibility, no changes in terms of increasing or decreasing levels of both variables are evident. Hereby, the relevance of the crisis to organizational objectives ranged from rather high to very high, indicating that the organization predominantly perceived the crisis as affecting organizational operations and posing devastating damage. The organization’s appraisal of crisis

responsibility, referring to the degree to which the organization deemed itself responsible for the crisis event, ranged from rather low to very high, whereby the issue was not addressed in five organizational posts. Consequently, the level of organizational engagement remained high throughout the first six months after the initial crisis outbreak (Table 5).

Table 5

Descriptive statistics: The level of organizational engagement throughout the first six months after the initial crisis outbreak

Organizational engagement N M SD

Relevance of the crisis to organizational goals 21 6.62 .59 Organization’s appraisal of crisis responsibility 16 5.50 1.16

Level of organizational engagement 16 6.13 .67

Organizational Stance

RQ2d examined how the organizational stance changed throughout the crisis. When considering the values for the presence of the action-based and the qualified-rhetoric-mixed stance in regard to each Facebook post, no tendency in terms of increasing or decreasing levels of both organizational stances throughout the crisis are evident. The level of the

organization’s action-based stance remained high as the crisis evolved, and ranged between rather evident to very evident, implying that the organization was predominantly willing to yield to the public’s demands, or to agree with the public on future actions. The level of the qualified-rhetoric-mixed stance, indicating the organization’s willingness to express regret, to apologize to the public, to admit wrongdoing, or to change its position towards public

opinion, ranged between neither not evident nor evident to very evident. In general, the presence of both organizational stances was high throughout the development of the crisis (Table 6).

Table 6

Descriptive statistics: The organizational stance throughout the first six months after the initial crisis outbreak

Organizational stance N M SD

Action-based stance 21 6.29 .64

Qualified-rhetoric-mixed stance 21 5.86 .91

Attribution of Crisis Responsibility

RQ2e examined how the attribution of responsibility by the organization’s primary public changed as the crisis evolved. A binomial logistic regression was performed to ascertain the linear effect of time on the likelihood that the primary public attributed the responsibility of the crisis to Facebook (M = .18, SD = .38). The logistic regression model was statistically highly significant, χ2(1) = 31.99, p < .001. The model only explained 5.7% (Nagelkerke R2) of the variance in comments attributing the crisis responsibility to Facebook and correctly classified 82% of cases. The predictor variable was statistically highly

significant, p < .001 (Table 7). For one day further in time, the odds of Facebook users attributing the responsibility of the crisis to Facebook decreased by factor .99.

Besides, the logistic regression model was statistically highly significant regarding the linear effect of the primary public deeming Cambridge Analytica responsible for the data breach throughout the first six months after the initial crisis outbreak (M = .02, SD = .15),

χ2(1) = 47.22, p < .001. The model explained 26.6% (Nagelkerke R2

) of the variance in comments attributing the responsibility of the crisis to Cambridge Analytica and correctly classified 97.8% of cases. The predictor variable was statistically highly significant, p < .001, and for each day further in time, the odds of Facebook users deeming Cambridge Analytica responsible for the crisis decreased by factor .90, which is substantially the strongest effect in regard to all findings. Since the responsibility of the crisis was not attributed to Alexkandr Kogan (M = .00, SD = .00), as implied in the comments, no further statistical analysis was performed.

Table 7

Binary logistic regressions: Primary public’s attribution of responsibility as the crisis evolves

Attribution of responsibility b* SEB

Facebook -.011*** .002

Cambridge Analytica -.111*** .025

Aleksandr Kogan -- --

Note. All tests have one degree of freedom. *p < .05, **p < .01, ***p < .001. b* = unstandardized regression coefficient; SEb = Standard error of the coefficient.

Crisis Response Strategies and Primary Public’s Displayed Emotions

RQ3 examined whether the presence of a crisis response strategy was related to the presence of certain emotions displayed by the primary public. Several chi-square tests for association were conducted between the crisis response strategies scapegoat, justification, and ingratiation, showing varying degrees of being present in organizational posts, and the

primary public’s displayed emotions. All expected cell frequencies were greater than five. The results of the chi-square tests for association were not statistically significant for scapegoat, and justification regarding no display of emotion, fright, anxiety, sadness, and contentment. Thus, no significant associations between ingratiation and all emotions were found (Table 8).

However, the results showed a statistically significant association between scapegoat and anger, p = .009. The association was very small, and indicated that more anger was present among users when the organization deemed an actor outside of the organization responsible for the crisis (23.9%), compared to when no attribution of blame to the latter was implied in a post (16.9%). Moreover, a statistically significant association between scapegoat and sympathy was found, p = .001. The association was small, and showed that more

comments contained expressions of understanding towards the actors involved in the data breach scandal when the scapegoat strategy was not present in the organization’s crisis response (43.4%), compared to when the latter strategy was implied in a post (32.9%). Additionally, there was a statistically significant association between justification and anger, p = .013. The association was very small, and showed that more anger was expressed by users when the organization presented reasons for its actions prior to the crisis (23.7%), compared to when no justification of past behavior was given (17.0%). Furthermore, a statistically significant association between justification and sympathy was found, p = .003. The association was small, implying that more sympathy was expressed when the justification strategy was not present (43.1%), compared to when the organization provided reasons for past behavior (33.2%).

Table 8

Chi-square tests of association: Crisis response strategies and primary public’s displayed emotions

Crisis response strategy

Emotion

Scapegoat Justification Ingratiation χ2

phi χ2 phi χ2 phi

No display of emotion 2.74 .06 .75 .03 .74 -.30 Anger 6.87** .09 6.19* .08 .12 -.01 Fright 2.20 -.05 2.19 -.05 .46 .02 Anxiety 2.00 .05 .22 .02 .11 .01 Sadness 1.42 .04 1.44 .04 .00 .00 Sympathy 10.16** -.11 8.97** -.10 .30 .02 Contentment 2.18 -.05 1.51 -.04 1.93 .05

Note. All tests have one degree of freedom. *p < .05, **p < .01, ***p < .001.

Further Analyses

To examine the overall dominant coping strategy evident in the primary public in regard to the continuum of conative versus cognitive coping, as implied in the ICM model, a paired-samples t-test was conducted. The results were statistically highly significant, t(900) = -12.11, p < .001, and showed that the primary public predominantly engaged in conative coping (M = 4.08, SD = .91), compared cognitive coping (M = 3.37, SD = 1.49) throughout the crisis. Secondly, the results of a paired-samples t-test in terms of the organizational stance were statistically significant, t(20) = 2.63, p = .016, indicating that the organization

predominantly took an action-based stance towards the public (M = 6.29, SD = .64), compared to a qualified-rhetoric-mixed stance (M = 5.86, SD = .91) during the crisis.

Discussion

The examination of the organization’s crisis responses, and the primary public’s reactions in terms of the temporal evolvement of the Facebook Cambridge Analytica scandal yielded several significant results. First, the study found that the organization justified its actions prior to the crisis, expressed gratitude for the public’s support, or revised current practices throughout the crisis. A change in the application of crisis response strategies was only evident in terms of blaming a group or actor outside the organization for the crisis, which was present towards the beginning and the end within the first six months after the initial crisis outbreak (RQ1). Hereby, the blame was directed to Cambridge Analytica or Aleksandr Kogan when the organization directly referred to the data breach scandal, or to coordinated information campaigns from Iran or Russia, when election interference was mentioned. However, it should be noted that no crisis responses were published in the second third of the examined period, whereby posts addressed the organization’s charity work, the launch of new applications, and national holidays. In general, the results supported previous findings

regarding the use of multiple crisis response strategies by breached firms (Kim et al., 2017). Corrective action was present in every crisis response, such as a reference to Facebook’s new tool Clear History, which will allow users to change their security settings. This finding is in line with previous research stating that breached firms predominantly focus on providing advice on how to secure one’s data in the future (Bentley et al., 2018). Whereas presenting an apology may be regarded as one of the most effective strategies in data breach crises (Jenkins et al., 2014), the organization did not apologize for the crisis on its social media platform throughout the crisis.

Secondly, Facebook users became less emotional when responding to the crisis, and decreasingly signaled sadness as the crisis evolved (RQ2a). Although anger, anxiety, and sadness were found to be the key emotions experienced by the primary public in crises defined by high organizational engagement and the public’s conative coping strategy (Jin et

al., 2010, 2012), this study found that sympathy, no display of emotion, and anger were the most dominant emotions expressed by Facebook users. Besides, Coombs and Holladay (2005) found that crisis responsibility was positively associated with anger and schadenfreude, and negatively correlated with sympathy. This tendency corresponds with previous studies suggesting an association between the attribution of crisis responsibility and negative

emotions experienced by the public (Choi & Lin, 2009; Jin, 2010). Therefore, the presence of both sympathy and anger as displayed in Facebook comments may be due to the fact that not all users perceived Facebook as responsible for the crisis.

In terms of conative coping, Facebook users were increasingly willing to take action due to the data breach scandal throughout the development of the crisis (RQ2b). However, no change in regard to the public taking another look at the crisis, according to cognitive coping, was detected. In accordance with the first quadrant of the ICM model (Jin et al., 2010, 2012), the primary public was generally more likely to engage in conative coping, compared to cognitive coping. For instance, Facebook users predominantly changed their account’s security settings, or reported fake accounts to counteract election interference, rather than changing their interpretation of the crisis, such as minimizing the severity of the data breach scandal to foster their own well-being. Yet, cognitive coping may be regarded as the

antecedent to conative coping (Jin et al., 2012), implying that the internal processing of the crisis may have already taken place in the minds of Facebook users, leading to concrete action to cope with the crisis.

Besides, the study found that the level of organizational engagement did not change over time, but remained high as the crisis progressed (RQ2c). This may be due to the fact that the organization acknowledged crisis responsibility, faced an anti-social media campaign, and was prompted to develop security tools for data protection, and against election interference (Levin, 2018), some of which were implemented throughout the first six months after the initial crisis outbreak (Haselton, 2018). The previous findings suggest situating the Facebook

Cambridge Analytica scandal in the first quadrant of the ICM model (high organizational engagement, conative coping). However, Jin et al. (2012) mapped corporate crises connected to security issues in the fourth quadrant (low organizational engagement, conative coping). To that end, the ICM model may primarily be regarded as a simplification when it comes to examining crises by means of a 2 x 2 typology. Therefore, crisis clusters based on certain issues involved may not fit into one specific quadrant, which is due to varying degrees of the presence of coping strategies and organizational engagement during crises.

In addition, no changes were evident when analyzing the change of the organization’s action-based, and the qualified-rhetoric-mixed stance throughout the crisis (RQ2d). However, the presence of the action-based stance was generally higher compared to the qualified-rhetoric mixed stance, which is contrary to findings by Jin et al. (2010) in terms of crises positioned in the first quadrant of the ICM model. The authors suggested that a qualified-rhetoric-mixed stance “contains more rhetoric or posturing by the organization, and may or may not lead to action that supports the rhetoric” (Jin et al., 2010, p. 443). A lower degree of a qualified-rhetoric-mixed stance may be due to the fact that the organization did not publish an apology, or primarily expressed regret, which partly indicates the organizational stance. Besides, corrective action was implied as a crisis response in all posts, which may relate to higher levels of the organization’s action-based stance. However, Mark Zuckerberg publicly apologized when testifying in front of the U.S. Congress, taking on full responsibility and admitting that the data breach scandal was his fault (Timberg & Romm, 2018). This indicates that the organization applied different types of crisis response strategies, or showed varying levels of organizational stances dependent on the situational context, such as Congressional hearings, or social media communication.

Moreover, Facebook users decreasingly held Facebook, or Cambridge Analytica accountable for the data breach scandal as the crisis evolved (RQ2e). The findings may illustrate that the primary public was the most likely to discuss the actors responsible for the

crisis after the initial crisis outbreak. Since the attribution of crisis responsibility may impact the reputational threat of an organization (Coombs, 2007), Facebook may have faced

decreasing levels of reputational damage over time, corresponding with improved financial performance, or rising user numbers (Carmeli & Tishler, 2005). Yet, Facebook’s rising stock performance in the second quarter, and its drastic decline in the third quarter of 2018 (Nagel, 2018), stagnating user numbers in the U.S., and decreasing user numbers in Europe due to the General Data Protection Regulation (Dodds, 2018) contradict the assumption. This tendency underlines the fact that the attribution of crisis responsibility merely represents one factor contributing to the reputational threat of a firm, whereas the organization’s crisis history and prior reputation may be further considered (Coombs, 2007). Hence, the decreasing tendency of Facebook users directing the crisis responsibility to Cambridge Analytica throughout the crisis marked the strongest result of the study. This finding may be explained by the fact that the firm filed for bankruptcy shortly after the initial crisis outbreak (Watkins & Sutton, 2018), which may have led to Facebook users forgetting about the involvement of Cambridge

Analytica in the data breach scandal over time, or resulted in a decreasing interest in discussing the latter.

Lastly, the study demonstrated that blaming an actor outside the organization, and justifying actions prior to the crisis was related to more anger, and less sympathy expressed by Facebook users (RQ3). The results correspond with previous research suggesting that crisis responses, such as excuse, justification, or denial increased responsibility judgements, and lead to negative reactions among the public (McDonald et al., 2010). However,

McDonald et al. (2010) found that crisis causes and crisis responsibility were pivotal to determining public’s reactions on crises. Coombs and Holladay (2005) affirmed that crisis responsibility may be the most powerful predictor of public’s displayed emotions and future interactions with the organization. Whether the attribution of crisis responsibility towards the organization evoked emotions, such as anger (Coombs, 2007; Weiner, 2006) among the

public was not investigated. Consequently, this study found an impact of crisis response strategies on primary public’s displayed emotions, however, multiple factors, such as crisis causes and crisis responsibility, draw a more complex picture in regard to post crisis communication and its impact on public’s reactions throughout the course of a crisis.

Theoretical Implications

The results of this study suggest an integrated approach when it comes to examining corporate crises by considering the organization-centered perspective by Coombs (2007), and the primary public’s emotion-driven approach by Jin et al. (2010). Whereas the ICM model (Jin et al., 2012) previously mapped primary public’s emotions based on the level of

organizational engagement, additional constructs, such as crisis response strategies, the organizational stance, emotions displayed by the organization, and message attributes may be considered when it comes to investigating the effect of organizational behavior on primary public’s expressed emotions. Furthermore, positive emotions, such as sympathy or

contentment, may be added to the repertoire of emotions experienced during crises, since the latter were found to be present throughout the data breach scandal, and thus provide a more comprehensive account of public’s responses. The attribution of crisis responsibility may be further considered, not only to capture the primary public’s perspective in crises more precisely, but to analyze the construct regarding emotions experienced during crises, since attributing crisis responsibility was previously associated with a range of emotions (Choi & Lin, 2009; Coombs & Holladay, 2005; Jin, 2010). Consequently, the ICM model may not only map primary public’s displayed emotions in crises based on two continua, but rather investigative causal relationships in regard to various indicators of organizational behavior and public’s reactions on crises.

Secondly, the findings indicate that the evolving character of crises (Weick, 1988) plays a substantial role when it comes to changes in terms of emotions expressed, the conative

coping strategy, and the attribution of crisis responsibility, which were detected as the crisis progressed. Based on the understanding of crises as a dynamic and rapidly unfolding situation (Van der Meer & Verhoeven, 2013), an integrated approach may further consider the timely development of a crisis in association to both the organization-based, and the public-centered perspective in a dynamic, and at the same time, systematic manner.

Additionally, this study has shown that crisis communication may be investigated by taking social media communication into account. Since past studies have analyzed public reactions on crises through news coverage, which is filtered through the perception of journalists (Jin et al., 2010, 2012), this study allowed capturing direct responses by the organization and its primary public on a crisis, which may eventually serve as a sufficient method to test future theoretical assumptions.

Practical Implications

As crises may take unanticipated turns throughout time (Shrivastava, 1987),

communication professionals may monitor and assess the evolution of public’s perceptions of crises. Hereby, practitioners may strategically develop organizational responses to a current, or potential crisis in the future by examining public reactions on crisis responses as displayed in the context of news coverage, or social media communication. For instance, if a crisis response strategy evokes negative emotions, which may result in complaining and negative word of mouth, practitioners may alter crisis management practices to foster positive

emotions, such as sympathy, and to reinforce the public’s loyalty towards the firm (McDonald et al., 2010).

In general, crisis management in the context of social media communication may not only be useful to monitor public’s reactions on crisis responses, but social networks further remain a tool for the organization to communicate directly to actors affected by a crisis. Contrary to crisis news coverage, an organization’s web page or social media account remains

under its locus of control, which allows to ensure the inclusion of all statements and information relevant to the firm (Coombs & Holladay, 2009).

Lastly, crisis managers may strategically optimize social media tools due to the public’s increasing interest in using social media networks during crises (Jin et al., 2014). This may include addressing false information spread on social media, or emergency live chat sessions. To assess the public’s needs in crises (Bentley et al., 2018), user polls may capture public’s reactions, such as trust in the organization or changes in attitudes towards current products or services, to ultimately foster a dialogical relationship between an organization and its primary public.

Limitations and Future Research

The study faces a number of limitations and comprises potential for further research. Due to Facebook’s recent application programming interface restrictions (Schroepfer, 2018), Facebook posts and comments were manually collected. The amount of Facebook posts referring to the data breach scandal (n = 21) did not allow a statistical analysis in terms of the change of crisis response strategies, the level of organizational engagement and the

organizational stance throughout the crisis. In addition, a visual content analysis in regard to pictures or videos implied in posts was not conducted. Therefore, future research may consider accumulating posts for statistical analyses by defining different crisis phases (Van der Meer & Verhoeven, 2013; Van der Meer et al., 2014) across an extended period of time, and conduct visual content analyses to obtain a more detailed account of an organization’s crisis management and public responses in the context of social media communication.

Moreover, the analysis was only applied on one prime example of a corporate crisis, and the variables message attributes and the organization’s displayed emotions as implied in the codebook were not examined due to the scope of the research questions. Whereas the scale for the emotion-cognition continuum as a dimension of message attributes showed

sufficient reliability levels after a second round of test coding, the question remains whether facts and emotions implied in a message remain two opposite ends of a continuum. In fact, organizations were advised to incorporate both relevant information, and the expression of empathy towards actors affected when responding to a crisis (Reynolds, 2004). Thus, future research may investigate a number of crises to determine whether the integrated model as previously proposed is applicable to various crises, and all types of crisis clusters (Coombs, 2007). The organization’s displayed emotions and message attributes of the crisis response may be investigated in regard to primary public’s reactions, whereby the development of an alternative scale for the emotional-factual continuum is suggested. Besides, the emotions fright and anxiety relate to danger/threat, which may pose the difficulty of clearly

distinguishing between the latter when examining emotions experienced in crises. Hereby, Smith and Lazarus (1993) defined fright/anxiety as one concept, which may be considered in further analyses.

Since the study was conducted in terms of one social network only, and exclusively in the context of social media communication, further research may investigate how

organizations apply post-crisis communication across different types of social networks, media, or various situational contexts. This is due to the fact that a different kind of crisis response (Timberg & Romm, 2018) was apparent outside the organization’s crisis

communication on social media. Hereby, the extent to which the public displays reactions on crisis responses may be examined, such as the type of coping behavior, when being exposed to a certain type of source, medium, or post-crisis event (i.e. speech, court hearing).

Lastly, Facebook comments written in English were exclusively examined. Hereby, the sample only implied responses by users with the confidence and the ability to respond in English, whereby comments by users from specific countries were more likely to be

overrepresented. For instance, individuals from different countries may vary in the way of attributing crisis responsibility, or prioritizing crisis causes during crises (Papastamou et al.,

2017). Consequently, further research may investigate the involvement of audiences in regard to the cultural origin, and examine cultural differences when responding to corporate crises.

Conclusion

The strategic management of crisis responses, as well as maintaining communication processes with its primary public may serve as the key responsibility for organizations facing crises (Lukaszewski, 1997). In the event of the Facebook Cambridge Analytica scandal, the study has shown that the evolving character of crises (Weick, 1988) plays a substantial role when it comes to organizational crisis responses, public’s displayed emotions, the conative coping strategy, and the attribution of crisis responsibility throughout the crisis. Therefore, the traditional organization-centered approach by Coombs (2007) is becoming outdated, and an integrated approach alongside the public-based perspective by Jin et al. (2010) may capture the dynamic interplay between organizations’ and primary public’s reactions to a crisis. Not only did the content analysis in the context of social media communication allow the

detection of immediate responses by the primary public on the organization’s crisis responses, but the public’s social media empowerment (Castells, 2007) further contributes to the crucial role of social media communication during crises. The latter may further influence crisis news coverage, since media professionals use communication processes on social media as a source (Miel & Faris, 2008). Consequently, social media communication with actors affected by a crisis remains of vital importance for organizations, and the examination of public’s reactions to crisis communication may fundamentally contribute to the effectiveness of strategic crisis management in times of corporate crises.

References

Acquisti, A., Friedman, A., & Telang, R. (2006). Is there a cost to privacy breaches? An event study. ICIS 2006 Proceedings, 94, 1563-1580.

Austin, L., Liu, B. F., & 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.

Ayyagari, R. (2012). An exploratory analysis of data breaches from 2005-2011: Trends and insights. Journal of Information Privacy and Security, 8(2), 33-56.

doi:10.1080/15536548.2012.10845654.

Bennett, W. L. (2016). News: The politics of illusion (Tenth Edition). University of Chicago Press.

Benoit, W. L. (1997). Image repair discourse and crisis communication. Public relations review, 23(2), 177-186.

Bentley, J. M., Oostman, K. R., & Shah, S. F. A. (2018). We're sorry but it's not our fault: Organizational apologies in ambiguous crisis situations. Journal of Contingencies and Crisis Management, 26(1), 138-149. doi:10.1111/1468-5973.12169.

Bowcott, O., & Hern, A. (2018, April 18). Facebook and Cambridge Analytica face class action lawsuit. The Guardian. Retrieved from:

https://www.theguardian.com/news/2018/apr/10/cambridge-analytica-and-facebook-face-class-action-lawsuit.

Browne, R. (2018, April 26). Facebook ‘concealed the truth’ of what it knew about Cambridge Analytica, lawmaker claims. CNBC. Retrieved from:

https://www.cnbc.com/2018/04/26/facebook-concealed-truth-of-cambridge-analytica-scandal-uk-mp-says.html.