Expressing emotion in corporate crisis communication

Expressing Emotion in

Corporate Crisis Communication

Research Master Thesis by

Jeroen van der Putten

10004624

Supervised by

Dhr. Dr. Piet Verhoeven

University of Amsterdam, Graduate School of

Communication

Submitted at the Graduate School of

Communication

1 Abstract

This study aims to shed light on the effect that communicating emotion in a corporate crisis response directed at the general public has on corporate reputation, purchase intention and negative word-of-mouth (NWOM) and to what extent these effects are moderated by crisis type and mediated by the public’s feelings of anger and sympathy. To effectuate this, an online experiment with a 2 (communication response: emotional vs non-emotional) x 3 (crisis type: victim vs accident vs preventable) between subjects design was conducted. The experiment involved a mock-up news article of a corporation in crisis followed by several questions. In total, data from a convenience sample of 392 respondents were used to analyze the effects. A good distribution of age (M = 42.4, SD = 15.3), gender (49% men) and education was achieved. The results reveal that communicating emotion in corporate crisis situations, irrespectively of the type of crisis, helps reduce reputation damage (p < .001) and increase purchase intention (p < .001). Mediation analyses show that these effects can partially be explained because feelings of anger are reduced by

communicating emotion. However, this is only the case for corporate reputation (p < .001) and purchase intention (p < .001). No effects were found regarding NWOM nor for the mediating influence of sympathy. All in all, this study enlarges the body of literature on public relations by showing the value of communicating emotion in a crisis response and paves the way for further exploration of the effects of expressing emotion in times of crisis.

Introduction

Crisis situations are often characterized by high levels of emotion, both for the public as well as for the organization and its members. The latter however, has received little attention in the field of crisis communication when it comes to emotion. Most studies

concerning emotion in crisis situations have shown that when a crisis strikes, publics who are either directly or indirectly influenced by the crisis experience high emotional reactions (Jin, 2010, Jin, 2014). These emotions are a pivotal factor in how publics respond to a crisis situation (Jin & Pang, 2010). The negative emotion most commonly associated with a crisis situation is anger. This is because people tend to get angry when an unexpected and negative event strikes them (McDonald & Härtel, 2000). Anger in turn has shown to lead to several negative behavioral intentions such as negative purchase intentions (Coombs & Holladay, 2007; Jorgensen, 1997) and negative word-of-mouth (Coombs & Holladay, 2007),

2 which makes it an important ramification for an organization in crisis (Coombs, 2007a). However, more positive emotional reactions such as sympathy have also been shown to arise in a crisis situation (Coombs & Holladay, 2005; Kim & Niederdeppe, 2013). Public’s sympathy toward and organization is found to vary as function of perceived attribution of responsibility; stronger attribution of responsibility reduces feelings of sympathy toward an organization (Coombs & Holladay, 2005). Sympathy, in turn, may positively affect corporate reputation (Coombs, 2007b). Therefore, taking into account what emotional reactions the public experiences when faced with a crisis is important for any practitioner or scholar in the field of crisis communication. Although, important as this may be, this element of emotion only deals with one side of the spectrum: the public. The other side, that of the organization itself, is often ignored.

Reacting to a crisis situation in an appropriate way in order to limit reputational damage and negative behavioral intentions is indisputably of vital importance for any organization facing a crisis. However, a gap in the knowledge concerning how organizations should react to crisis situations remains. The largest body of literature concerning

organizational communication in times of crisis deals with ‘what’ information an

organization ought to communicate (Kim, Avery, & Lariscy, 2009). These studies are mostly based upon the Situational Crisis Communication Theory (SCCT) devised by Coombs (2007b). However, less attention has been paid to ‘how' crisis responses should be conveyed; what tone the response should have. An essential, but often overlooked, element of tone within a corporate crisis response is emotion (Read, 2007). When faced with a crisis, a corporation can choose to suppress the emotions present within the organization because they are deemed as inappropriate. But alternatively, a corporation can also choose to adhere to these emotions and respond by expressing regret and shame and therewith show how the

corporation and its members feel about the situation (Van der Meer, 2014). In that sense, emotions ‘are’ communication; they reveal information about the sender, of its feelings, motives and position relative to others (Ekman, 1993; Fridlund, 1994; Kutson, 1996). Recently, crisis communication studies have touched upon the role of an emotional tone in outgoing communication. Results show that emotional crisis responses can indeed diminish reputation damage after a crisis situation (Van der Meer, 2014). Not only knowing ‘what’ to communicate during a crisis situation, but also ‘how’, is thus of great importance. Especially

3 the interplay between emotional experiences perceived by the public and emotional

responses expressed by the organization is an interesting topic to clarify further.

Crisis communication professionals and scholars have been slow to recognize the importance of emotional communication aside to rational arguments often discussed in the SCCT theory. More attention should be paid to how emotional appeals can help strengthen organizational reputations against crisis situations (Read, 2007). Therefore, the main

question at hand is what the relation is between communicated emotion by the organization in crisis, emotional reactions of the public and negative outcomes such as reputational damage, negative word-of-mouth and reduced purchase intention. Because research documents that as perceptions of crisis responsibility intensify, negative emotions (e.g. anger) increase (Coombs & Holladay, 2005; Jorgensen, 1996) and positive emotions (e.g. sympathy) reduce (Cooms, 2007b) an emphasis will be placed upon different types of crisis (differing in attribution of responsibility) and the effects of emotional crisis communication responses. The research question, therefore, reads as follows:

RQ: What is the effect of the use of emotion (emotion vs. no-emotion) in corporate crisis communication aimed at the general public on corporate reputation and

behavioral intentions (i.e. word-of-mouth and purchase intention) and to what extent are these effects underlain by emotional reactions (i.e. anger and sympathy) and crisis type (victim vs. accidental vs. preventable)?

Theoretical Framework

Crisis situations; the trigger event

Although there are numerous articles and books written about corporate crises, there’s no accepted definition what a corporate crisis exactly entails (Coombs, 2014). The word crisis is most often used to describe difficult times and bad experiences. However, not all difficult times and bad experiences are necessarily a crisis (Ulmer, Sellnow, & Seeger, 2014). Hermann (1963) suggests that any crisis is characterized by three different dimensions. He argues that corporate crises (1) threaten high-priority values of a

corporation; (2) present a restricted amount of time in which a corporation can respond to the situation; and (3) are unexpected and, most often, unanticipated by the corporation. Based on these three dimensions, some researchers have identified a corporate crisis as: “a

4 specific, unexpected, and non-routine event or series of events that create uncertainty and threaten or are perceived to threaten an organization’s high-priority goals” (Ulmer, Sellnow, & Seeger, 2010: p. 8). Other researchers, such as Fediuk, Coombs and Botero (2010) focus less on the dimensions identified by Hermann (1963) and more on the notion that a crisis should be perceived from the point of view of the stakeholder. They follow the notion of Benoit (1997) who argues: if stakeholders believe an organization is in crisis, a crisis does exist and stakeholders will react to the organization as if it is in crisis. Fediuk, Coombs and Botero (2010) therefore define a crisis situation as: “the perception of an unpredictable event that threatens important expectancies of stakeholders and that can seriously impact the organization’s performance and generate negative outcomes” (p. 3).

As becomes clear, where some definitions such as that of Ulmer, Sellnow and Seeger (2010) take into account the dimensions identified by Hermann (1963), other researchers such as Fediuk, Coombs and Botero (2010) acknowledge the importance of the perception of stakeholders. Therefore, finding consensus regarding what a crisis situation entails, a joint understanding of both definitions is followed. This definition reads: a crisis situation is a specific, unexpected, and non-routine event or series of events that creates uncertainty and threaten important expectancies of stakeholders that can seriously impact the organization’s performance and generate negative outcomes. This definition incorporates the dimensions identified by Hermann (1963) as well as the importance of the perception of the stakeholder as argued by Benoit (1997).

Negative outcomes; reputational damage and negative behavioral intentions

After a crisis has struck, the organization in question has to face the possibility of experiencing serious negative outcomes. Thus, one of the primary objectives in crisis

management is to reduce these negative outcomes. Indeed, when stakeholders believe they are wronged in any way, the overall perception they have of an organization is damaged and it is likely that they will proceed to engage in certain negative behaviors. Therefore,

reputation damage and negative behavioral intentions arguably pose the biggest threat for any organization in crisis.

There is strong consensus among scholars that a strong corporate reputation is one of the most valuable intangible resource an organization can possess. A favorable reputation has been linked to numerous positive outcomes such as attracting new customers,

5 motivating employees, increasing job satisfaction, attracting top employee talent, generating positive media coverage and generating investment interest (Alsop, 2004; Davies, Chun, Da Silva, & Roper 2003; Dowling, 2002; Fombrun & Van Riel, 2004; Kim & Yang, 2013).

According to Van Riel and Fombrun (2007) “corporate reputations are overall assessments of organizations by their stakeholders” (p. 43). Corporate reputations in turn are formed based on the direct and indirect interactions of the stakeholder and the organization (Fombrun & Van Riel, 2004). Positive information about and interactions with an organization builds positive corporate reputation, whereas negative information about and interactions with an organization result in negative corporate reputation (Davies, et al., 2003).

Although related to corporate reputation, negative behavioral intentions pose a more tangible threat making it part of a different spectrum when it comes to negative outcomes as a result of a crisis situation. Crisis events are perceived by stakeholders as an injustice. Previous research in organizational justice suggests that when an injustice occurs,

stakeholders are motivated to engage in justice restoration activities such as revenge behavior as a way to restore justice (Bies & Tripp, 1996) and the greater the perceived injustice the stronger the motivation for revenge (Bies & Tripp, 1996; Tripp, Bies & Aquino, 2007). More specifically, researchers have documented that crises can have a negative effect on market share, sales of recalled products, stock prices and sales of other products by the company (Cleeren, Van Heerde, & Dekimpe, 2013; Dawar, 1998; Siomkos & Kurzbard, 1994). Indeed, purchase intention of stakeholders can be seriously hindered by a crisis situation.

Negative word-of-mouth is another behavioral intention that can result from a crisis situation. Wetzer, Zeelenberg and Pieters (2007) describe negative word-of-mouth as “[communication that] includes all negatively valenced, informal communication between private parties about goods and services and the evaluation thereof” (pp. 661-662). They stress that, as is the case with purchase intention, that negative word-of-mouth is depended on the perceived injustice that has been done upon stakeholders and on the severity of the crisis. Organizations value word-of-mouth because of the benefits associated with positive word-of-mouth and the harm inflicted by negative word-of-mouth (Laczniak, DeCarlo, & Ramaswami, 2001).

6

Emotional crisis communication; expressing shame and regret

To avoid negative outcomes as a result of a crisis situation, it is of pivotal importance for any organization to respond in a manner that is deemed appropriate. However, until now, crisis communication research has mostly focused on rational arguments a crisis response should have and less on the emotional tone with which a response can be conveyed (e.g. Coombs, 2007b). Exploring the value of expressing emotion in crisis communication is, above all, of importance because emotions ‘are’ communication; they reveal information about the sender, of its feelings (Ekman, 1993), of its motives and intentions (Fridlund, 1994) and of its position relative to others (Knutson, 1996). In

accordance with this notion, research shows that recipients use communicated emotion of the sender to determine its intention and adjust their behavior and thoughts

correspondingly (Lewis, 2000; Tiedens, 2001; Van Kleef, De Dreu, & Manstead, 2006). Specifically regarding corporate crisis situations, this makes it of great importance to determine how a crisis communication response with an emotional frame affects perceptions of stakeholder regarding the crisis situation and the corporation.

When it comes to expressing emotion in a crisis situation two types of emotions stand out: shame and regret. Shame results when a person (or in this case an organization) judges its actions as conflicting with its internal standards, norms, goals and those of

important others (Lewis, 1995). Regret results from a situation in which someone feels he or she made a mistake that could have been prevented would a different choice have been made (Van Dijk & Zeelenberg, 2002). Moreover, interpersonal regret is most likely to be perceived when someone is responsible for a bad experience for someone else (Zeelenberg, Van der Plight, & Manstead, 1998). Although some researchers (e.g. Wesseling, Kerkhof, & Van Dijk, 2007) suggest that shame and regret should be treated as different forms of emotion (even though they are associated with each other) when it comes to expressing emotion in a crisis response, recent studies (e.g. Van der Meer, 2014) have shown that neither cognitive nor affective responses from stakeholders to an emotional crisis communication differ between messages characterized by regret or shame.

The psychology behind the value of emotion in corporate crisis responses is threefold. First of all, the phenomenon metonymy is of importance. A metonymy is “a linguistic and cognitive operation in which we imply certain parts (e.g. buildings) directly to stand in for the larger idea or conceptualization (e.g. the corporation)” (Christensen &

7 Cornelissen, 2010, p. 6). This means that when a metonymy comes into play, people tend to see the parts and the whole as one and the same thing. This leads people to experience specific activities such as, for example, a corporate crisis response, as a defining

characteristic for the organization as a whole. According to Christensen and Cornelissen (2010) metonymies are quite pervasive in our way of thinking. Secondly, stakeholders, have a natural inclination to ascribe identity traits to an organization and treat it as (if it were) a person. Stakeholders of an organization effectively credit organizations with identity traits, just as they would an individual person (Cohen & Basu, 1987). And finally, emotions help us form and maintain social relationships. More specifically, emotions can reduce the distance between self and others, or between one’s own group and other groups (Fischer &

Manstead, 2008). Moreover, when regret and shame is expressed this is a ‘cue’ for the receiver that the person (or organization) in question possesses integrity and that this person feels bad and responsible regarding the given situation. This communication of ‘selfpunishment’ as Wesseling, Kerkhof and Van Dijk (2007) call it, might convince the receiver that further punishment is not necessary. The same may be true for organizations. In this sense, expressing emotion as a corporation can reduce negative outcomes as a result of a crisis situation since stakeholders will feel less inclined to distance themselves from the organization and punish them for what they have caused. Thus, how an organization

expresses itself is evaluated by stakeholders as if it were from an individual and treated as if that form of communication is characteristic for the entire organization. Furthermore, since expressing emotion has the ability to form and maintain social relationships, incorporating this element into a crisis response might diminish negative outcomes as a result of a

corporate crisis situation. Deducing from this thought, the following hypotheses are formed:

Hypothesis 1a: Communicating emotion in a crisis response after a corporate crisis situation will result in a more positive corporate reputation than not communicating emotion in a crisis response.

Hypothesis 1b: Communicating emotion in a crisis response after a corporate crisis situation will result in more positive purchase intentions than not communicating emotion in a crisis response.

8

Hypothesis 1c: Communicating emotion in a crisis response after a corporate crisis situation will result in more positive word-of-mouth than not communicating emotion in a crisis response.

Stakeholder reaction; the role of anger and sympathy

Crisis incidents do not only form unpleasant times for organizations, but also form important psychological events experienced by individuals that are struck by a corporate crisis. More often than not, these experiences are heavily emotion-laden (Coombs & Holladay, 2005). When an injustice is experienced, the incident can prime different emotional feelings. Most often these emotions are negative emotions such as anger

(Bembenek, Beike, & Schroeder, 2007). However, more positive emotions such as sympathy have also been identified (Cooms, 2007b). Positive and negative emotions have even been found to co-occur alongside each other during stressful situations (Folkman & Moskowitz, 2000). Knowledge of both positive and negative emotions is of great value in order to gain more insight on how to communicate with publics in a more reasonable and sensible way (Jin, 2009; Jin, 2010).

According to Iyer and Oldmeadow (2006), ‘‘anger is usually experienced when people blame specific agents (i.e., individuals, groups, or institutions) for a transgression or

injustice’’ (p. 637). Therefore, stakeholders tend to experience anger when faced with an offense from an organization against them or their well-being (Jin, 2009; Jin, 2010). When individuals feel negative emotions, such as in response to perceived violations, these

emotions increase the need to correct the wrong or engage in retaliatory behaviors (Barclay, Skarlicki & Pugh, 2005; Tripp, Bies, & Aquino 2007). The detrimental effect of anger is

confirmed by several studies that show a relationship between anger and negative behavior intentions (Coombs et al. 2007; Coombs & Holladay, 2005). The same may well be the case when it comes to corporate reputation as stakeholder anger can damage the organization– stakeholder relationship or even lead stakeholders to end the relationship (Jorgensen, 1996).

Sympathy occurs when there is awareness of others’ suffering, especially when the suffering is seen as undeserved (Iyer & Oldmeadow, 2006). However, witnessing and recognizing others’ suffering may not always elicit a direct sympathetic response. This is mainly so because sympathy involves sensitivity to and understanding of the feelings of the other (Gruen & Mendelsohn, 1986) as well as a certain distance from the situation (Iyer &

9 Oldmeadow, 2006). Given this knowledge, sympathy is most likely to be felt when indirectly exposed to a crisis. When sympathy is felt, it may reduce tendencies to engage in negative behaviors and reputational damage may be reduced (Coombs, 2007b).

Feelings of anger and sympathy may well be altered by a crisis response that is characterized by the expression of the organization’s sincere shame and regret concerning the situation. Moreover, since expressing emotion has the ability to form and maintain social relationships (Fischer & Manstead, 2008), expressing these emotions might elicit compassion with the receiver and therefore reduce negative emotions such as anger and increase

positive ones such as sympathy, resulting in less reputational damage and less negative behavior intentions. Following this thought, the following hypotheses are formed:

Hypothesis 2a: Anger and sympathy mediate the effect as proposed in H1a:

communicating emotion in a crisis response leads to less anger and more sympathy which subsequently leads to a better corporate reputation.

Hypothesis 2b: Anger and sympathy mediate the effect as proposed in H1b:

communicating emotion in a crisis response leads to less anger and more sympathy which subsequently leads to better purchase intentions.

Hypothesis 2c: Anger and sympathy mediate the effect as proposed in H1c:

communicating emotion in a crisis response leads to less anger and more sympathy which subsequently leads to better word-of-mouth.

Evaluation of the crisis; attribution of responsibility

The attribution of responsibility is a widely discussed and generally accepted element in any crisis situation. The main idea is that during a crisis situation people search for the cause of the event and an understanding of who is responsible (Coombs, 2007b). A theory that helps understand this, is the attribution theory. According to this theory, individuals make judgments on the cause of events that they experience by assessing locus of control, stability and controllability (Weiner, 1985). Locus of control refers to whether the cause of the event was internal or external to the actor. In the case of a corporate crisis, stakeholders will look at whether the cause of the crisis lies within the organization (e.g. wrong policy) or outside of the organization (e.g. a natural disaster). Stability refers to whether there is a pattern of crisis events or behavior. In this case people tend to look whether the corporation

10 in question has underwent similar crisis situation in the past. Controllability refers to

whether the actor could control the cause of the crisis or whether the cause is beyond the actor’s control.

Building on this theory, Coombs identified three clusters of crises that vary in the attribution of responsibility. Coombs (2007b) named the first crisis cluster the victim cluster. This cluster includes all crisis situations with the least responsibility for the organization. Examples of crisis include natural disasters, rumors, workplace violence, and external product tampering. The second cluster is titled the accident crisis cluster. This cluster includes primarily technical breakdowns (e.g., accidents, product recalls). This cluster

attributes moderate responsibility to the organization. The last cluster, the preventable crisis cluster, includes intentional human errors and organizational misdeeds. Crises in this cluster attribute the most responsibility to the organization.

Of notable importance is that the experienced emotion is related to the attribution of responsibility. Studies comparing anger and sympathy in crisis situations show that

uncontrollability of a crisis gives rise to sympathy and associated emotions such as compassion and judgments of personal responsibility generates anger (Weiner, 2006). So whereas anger emerges from responsibility judgments, sympathy is evoked from non-responsibility judgments.

Because experienced emotions become more negative depending on the attribution of responsibility, it may well be the case that emotional crisis communication is most

effective when used in a crisis of high organizational responsibility (i.e. preventable crises) in comparison to the other crisis situations (e.g. accident and victim cluster). For this reason the following hypotheses are devised:

Hypothesis 3a: Crisis type will moderate the effect of emotional crisis communication on anger: emotional crisis communication will negatively affect anger but this effect will be more pronounced for the preventable crisis situation, followed by the accident crisis situation and lastly the victim crisis situation.

Hypothesis 3b: Crisis type will moderate the effect of emotional crisis communication on sympathy: emotional crisis communication will positively affect sympathy but this effect will be more pronounced for the preventable crisis situation, followed by the accident crisis situation and lastly the victim crisis situation.

11 Resulting from the proposed hypotheses a conceptual model was devised (see figure 1). In short, this model predicts a negative effect of emotional crisis communication on negative word-of-mouth and a positive effect on purchase intention and corporate reputation (H1a, 1b and 1c). Furthermore, degree of sympathy and degree of anger are predicted to mediate the relation between emotional crisis communication and the dependent variables (H2a, 2b, 2c, 3a, 3b and 3c). More specifically, it is predicted that emotional crisis communication indirectly decreases negative word-of-mouth and increases purchase intention and corporate reputation by increasing feelings of sympathy and

reducing feelings of anger. Lastly, it is predicted that crisis type will moderate the relation between emotional crisis communication and degree of sympathy and anger (H4a and b). Emotional crisis communication will positively affect sympathy and negatively affect anger but this effect will be more pronounced for the preventable crisis situation, followed by the accident crisis situation and lastly the victim crisis situation.

Figure 1: Conceptual model

Method

Experimental design

To test the proposed model, an online experiment with a 2 (use of emotion:

emotional vs non-emotional) × 3 (type of crisis: victim vs accident vs preventable) factorial between-subjects design was conducted. This resulted in six conditions: (1) an emotional

12 response combined with a victim crisis type (N = 72), (2) an emotional response combined with an accident crisis type (N = 62), (3) an emotional response combined with a preventable crisis type (N = 56), (4) a non-emotional response combined with a victim crisis type (N = 69), (5) a emotional response combined with an accident crisis type (N = 63) and (6) a non-emotional response combined with a preventable crisis type (N = 70).

Sampling procedure & participants

Respondents for this study were approached based on convenience. Most of these respondents were contacted by means of social media, such as Facebook. For example, various Facebook communities were used to approach large groups of people. These Facebook communities include gardening, cooking, tools, motorcycles, cars, pets and reading communities. Personal social networks proved to be effective as well. However, a considerable number of respondents were also gathered by approaching people in public places, such as train stations (i.e. Haarlem Central Station and Amsterdam Central Station), or by going door-to-door (in and around Haarlem centre) with the request to participate in this research.

When approached, the respondents were told that the study was conducted in the context of a master thesis and that participation would take approximately ten minutes. The call for respondents included a link. This link, when clicked on, forwarded the respondent to the experiment. Here more elaborate information regarding the study and the rights of the respondents was provided. Participation in this study was in accordance with the ethical rules of ASCoR. Participation was completely anonyms, respondents were informed that they could terminate their participation at any time without consequences, the procedure of the study was made clear prior to the participation and a debriefing regarding the fictitious nature of the study was provided at the end of the experiment. After having given consent for participating in the study, the respondents could begin the experiment. All respondents were randomly assigned to one of the six different conditions. In total, 404 respondents participated in the study. However, because some respondents were excluded, which will be discussed later on, data from 392 respondents were used to test the proposed hypotheses.

Because this study targets the general public, demographics such as age, education and gender of the sample group should preferably be consistent with demographics of the general public. Furthermore, since this study was conducted in The Netherlands, the sample

13 of this study should therefore be comparable with the Dutch population. Data from the Central Bureau of Statistics (CBS) were used to compare the sample with the Dutch population. Especially age and gender showed to be very similar with that of the Dutch population. The sample group was, however, better educated than the average Dutch resident (see table 1).

Table 1 - Comparison of demographics of the sample group with the Dutch population.

Sample Dutch population

Demographics M SD M Age 42.4 15.3 41.0 N % % Male 192 49.0 49.5 Female 200 51.0 50.5 Education Elementary School 8 2.0 4.8

VMBO, MBO 1, AVO 20 5.1 17.4

MBO 2, MBO 3 29 7.4 14.2

MBO 4 50 12.8 19.4

HAVO, VWO 72 18.4 10.0

HBO, WO bachelor 151 38.5 22.0

WO masters, Doctor 62 15.8 12.2

Note. For the actual distribution of age, see appendix 1. Statistics about Dutch Residents: CBS (2015) were retrieved June 1, 2015 from http://www.cbs.nl/nl-NL/menu/cijfers/kerncijfers/default.htm.

Stimulus material

The stimulus material consisted of a mock-up news article of the Volkskrant (see appendix 2), one of the leading national newspapers in the Netherlands. The news article reported on a food scandal in which a certain type of bacteria (i.e. the E-coli bacteria) was found to be present in a large number of ready-to-serve pasta sauce jars. In order to avoid pre-existing attitudes towards the company that was responsible for the pasta sauce jars, a pasta sauce company named ‘La Cucina’ was invented. The topic of a food scandal was chosen because of the relatively often occurrence of these types of crises in the Netherlands in the past couple of years (Gussow & Kuiper, 2014). Also, a food scandal is likely to be a relevant crisis for a broad public making it ideal for studies in which a broad public is targeted.

The independent variable ‘emotional crisis communication’ was manipulated by developing two types of responses: a response in which emotion was shown by the company

14 in question and a response in which no emotion was shown. In the emotional response the company expressed their regret and shame regarding the situation. In the response without emotion the company expressed the seriousness of the situation and their intention to solve the problem. The independent variable ‘type of crisis’ was manipulated by developing three types of crises: a victim crisis, an accidental crisis and a preventable crisis. In the victim crisis the article emphasised the responsibility of a third party regarding the situation. More specifically, the articles with a victim crisis reported on another company named Gel Alpes that was responsible for the E-coli bacteria in the pasta sauce jars of La Cucina. In the accidental crisis the article reported on a production error that was the cause of the presence of the E-coli bacteria in the jars. More specifically, because of a faulting thermometer, the E-coli bacteria were able to survive the heating process. Lastly, in the preventable crisis the article emphasized the responsibility of La Cucina for the situation. In this case the article reported that La Cucina used contaminated meat in their pasta sauce knowing that this could result in pasta sauces that contained the E-coli bacteria.

Dependent variables

The dependent variable ‘negative word-of-mouth’ (NWOM) was measured using three questions on a seven-point Likert scale (Bougie, Pieters & Zeelenberg, 2003). Example questions include: “I intend to say negative things about this company to family” or “I intend to say negative things about this company to friends”. The answers ranged from very much disagree to very much agree. The items showed to form a very reliable scale (M =3.54, SD = 1.54, α = .98).

Purchase intention was also measured using three questions on a seven-point Likert scale (Souiden & Pons, 2009). This scale is specifically designed for crisis situations. Example questions include: “because of the incident, I will not buy products from this company” or “the likelihood of me buying products from this company is really low”. The answers ranged from very much disagree to very much agree. The items were found to form a reliable scale (M =3.08, SD = 1.37, α = .88).

Reputation was measured using five questions on a seven-point semantic differential scale. Because the company used in this study is fictitious, it was decided to follow the example of Claeys and Cauberghe (2014) and use items that measure the general reputation of a company. Actual corporate reputation incorporate several dimensions that go beyond

15 the scope of the stimuli material used in this study, which is why a scale that measures the general reputation of a company yields more valid results (Claeys, Cauberghe & Vyncke, 2010). The respondents were asked how they evaluated the company in general. The items included bad-good, negative-positive, unfavourable-favourable, etc. The scale was found to be very reliable (M = 3.61, SD = 1.34, α = .94).

Degree of sympathy was measured using four items on a seven-point Likert scale (Lee, 2004). Example questions include: “I think [the organization] should be punished” or “I feel sympathetic towards [the organization]”. The answers ranged from very much disagree to very much agree. The items showed to form a sufficiently reliable scale (M = 4.11, SD = 1.24, α = .70).

Degree of anger was measured using three items on a seven-point Likert scale (Coombs & Holladay, 2008; Kim & Niederdeppe, 2013). Example questions include: “I have feelings of anger toward [the organization]” or “I feel annoyed toward [the organization] for what happened”. The answers ranged from very much disagree to very much agree. The items showed to form a very reliable scale (M = 3.71, SD =1.46, α = .92).

Participation checks

To ensure that all the respondents paid sufficient attention to the news article they had to read, two questions were asked. First of all the respondents were asked whether they could remember the name of the pasta sauce company that was mentioned in the news article. They were given four answer options. The second question asked the respondent whether they could remember the situation the pasta sauce company was in. Again four answer options were given. In total, twelve persons were not able to answer the first and/or the second question correctly. Not answering these questions correctly clearly shows that these participants did not pay sufficient attention towards the news article. Because of that reason, they were excluded from the study entirely.

Manipulation check

Two variables were included to check whether the manipulations of the study had been successful. These variables are ‘perceived emotion’ and ‘perceived responsibility’. Firstly, in order to determine that the news article in which emotion was shown by the company was also perceived as such by the respondents, the following statement was presented to the respondents: “La Cucina responded in an emotional way on the situation.”

16 Respondents could indicate on a seven-point Likert scale to what extent they agreed with this statement. The results of an independent sample t-test with perceived emotion as dependent variable and emotional response as independent variable shows that the manipulation was successful. Respondents that were shown a news article with an

emotional response of the company indicated a much higher perceived emotion (M = 4.83,

SD = 1.67) than respondents that were shown an article without an emotional response by

the company (M = 3.14, SD = 1.61), t (402) = -10.33, p< .001, 95% CI [-2.01, -1.37].

The second variable, perceived responsibility, was also measured using one item on a seven-point Likert scale. All respondents were asked to what extent they agreed with the following statement: “considering the cause of the incident, La Cucina is responsible for the contaminated pasta sauce jars.” The results of a one-way ANOVA with perceived

responsibility as dependent variable and crisis type as independent variable shows that this manipulation was successful as well. The type of crisis has a significant effect on the

perceived responsibility of the company, F (2, 402) = 55,43, p < .001. According to the Bonferroni post hoc test, the company was perceived as the least responsible in the victim crisis scenario (p< 0.001, M= 3.86, SD= 1.84), followed by the accident crisis scenario (p < 0.001, M= 5.07, SD= 1.56) and finally the preventable crisis scenario (p < 0.001, M = 5.89, SD = 1.36). The complete questionnaire can be found in appendix 3.

Data screening

Before turning to the results section, it is important to pay some attention to the distribution of the data to check whether certain assumptions concerning regression analyses are met. As Osborne, Christensen, and Gunter (2001) observe, few articles report having tested assumptions of the statistical tests they rely on for drawing their conclusions. This creates a situation where there is an abundance of rich literature, but the validity of many of these results, conclusions and assertions need to be questioned, as we have no idea whether the assumptions underlying the statistical tests were met. Therefore the data of this study was checked for normally distributed variables, extreme values (outliers), homoscedasticity and linearity.

First of all, normally distributed variables. Non-normally distributed variables (highly skewed or kurtotic variables, or variables with substantial outliers) can distort relationships and significance tests (Van Peet, Van den Wittenboer & Hox, 2009).For that reason, several

17 histograms of the continuous variables in this study’s proposed model (i.e. anger, sympathy, reputation, purchase intention and NWOM) were examined along with their skewness and kurtosis values (see table 2 for the kurtosis and skewness values; see appendix 4 for the histograms). Most of these variables have histograms that show sufficiently normally

distributed data with good skewness and kurtosis (not lower than -1 and not higher than 1). However, an exception should be made for NWOM. This variable has an oddly shaped data distribution. Especially the kurtosis value confirms this since the kurtosis value of NWOM is -1.14. However, because the value is just slightly outside the margin, no adjustments

regarding this variable were made.

Table 2 - Skewness and kurtosis values for the variables: anger, sympathy, corporate reputation, purchase intention and negative word-of-mouth.

Variable Skewness SE Kurtosis SE

Anger -.90 .12 -.82 .24

Sympathy .04 .12 -.57 .24

Corporate reputation .22 .12 -.37 .24

Purchase intention .62 .12 -.03 .24

Negative word-of-mouth .01 .12 -1.14 .24

Next up are outliers. To assess whether there are outliers in the data, several boxplots were produced and analyzed (see appendix 5). Fortunately, the boxplots did not show any signs of outliers that needed to be addressed. Because the histograms, overall, showed sufficient normally distributed variables, no further outlier tests were performed.

Moving on to homoscedasticity. When homoscedasticity is violated, also called heteroscedasticity, the variance of errors differs at different values of the independent variable. Slight heteroscedasticity may have little effect on significance tests, but greater heteroscedasticity can lead to serious distortion of findings (Tabachnick & Fidell, 1996). In order to determine whether the data shows homoscedasticity, several histograms and scatterplots of the regression standardized residuals were examined for every dependent variable (see appendix 6). The graphs clearly indicate homoscedasticity since the histogram of the residuals of all dependent variables are normally distributed and the scatterplots shows no real difference in variance of errors between different values of the independent variables. However, an exception should be made for NWOM, since the histogram of the standardized residuals of that variable, again, is oddly shaped.

18 Lastly, linearity was assessed. Regression analyses can only accurately estimate the relationship between dependent and independent variables if the relationships are linear in nature (Van Peet, Van den Wittenboer, & Hox, 2009). To test whether the relation between the independent variables and the dependent variables in this study are linear in nature a scatterplot matrix of all numerical variables was produced and analyzed (see appendix 7). Fortunately, the relations depicted in the matrix do not show any non-linear relations (i.e. U-shaped or inverted U-U-shaped lines) which means that linearity can be assumed.

Results

Correlations

To test whether any of the demographics should be held into account in the analyses a series of correlations between all demographic variables and all dependent variables were calculated. The results of these correlations show that education correlates negatively with purchase intention (r = -.14, p = .004) and that gender correlates positively with anger (r = .11, p = .034). As these variables correlate significantly with some of the dependent variables (regardless of the condition), it should be tested whether these variables are evenly

distributed across the groups (i.e. the non-emotional versus emotional response groups and the crisis type groups). Age did not correlate with any of the dependent variables and is for that reason excluded from further analyses.

Distribution of demographic variables

To test the distribution of education across the groups an independent t-tests and a one-way ANOVA were conducted. The independent t-test with education as dependent variable and emotional response as independent variable shows an insignificant result (t (390) = -.91, p = .365, 95% CI [-.16, .48]). This means that it can be assumed that education is evenly distributed between the emotional versus non-emotional groups. The one-way ANOVA with education as dependent variable and crisis type as independent variable also shows an insignificant result (F (2,289) = 1.503, p = .224). So it can also be assumed that education is evenly distributed across the crisis type groups.

Moving on to the distribution of gender across the groups. To assess the distribution of gender, a chi-square test was performed. Since this test shows an insignificant result (χ² (1, N = 392) = .67, p = .412), it can be assumed that the distribution of gender is equal

19 between the emotional versus non-emotional groups. A chi-square test was also performed to assess the distribution of gender across crisis type groups. Again this test shows an insignificant result (χ² (2, N = 392) = 3.77, p = .152) meaning that the distribution of gender across crisis type groups is equal.

Despite the fact that gender and education correlate with some of the dependent variables it has been decided not to include these variables as covariates because the above presented tests suggest that these variables are equally distributed across the groups. Including them as covariates would therefore be redundant. Gender and education have for that reason been excluded from further analyses.

Effectiveness of an emotional crisis communication

To assess the effectiveness of an emotional crisis communication on corporate reputation, purchase intention and NWOM (hypotheses 1a, b and c), three one-way ANOVA’s were conducted. Beginning with the effect of an emotional crisis communication on corporate reputation (hypothesis 1a). The results of the one-way ANOVA with emotional crisis communication as independent variable and corporate reputation as dependent variable show a significant, weak positive effect (F (1, 390) = 25.51, p < .001, η2 = .06). Respondents who were exposed to the emotional crisis communication reported higher values of corporate reputation (M = 3.96, SD = 1.25) than respondents who were exposed to the non-emotional crisis communication (M = 3.31, SD = 1.34). The one-way ANOVA with emotional crisis communication as independent variable and purchase intention as

dependent variable also shows a significant, weak positive effect (F (1, 390) = 32.75, p < .001, η2 = .08). Respondents who were exposed to the emotional crisis communication reported higher purchase intention (M = 3.47, SD = 1.52) than respondents who were exposed to the non-emotional crisis communication (M = 2.71, SD = 1.11). Based on these two ANOVA’s, hypothesis 1a and 1b are supported. The last ANOVA with emotional crisis communication as independent variable and NWOM as dependent variable shows no significant result (F (1, 390) = .88, p < .348). Respondents exposed to either an emotional crisis communication or a non-emotional crisis communication reported similar values for NWOM. Based on these results, hypothesis 1c is rejected.

20

Mediating influence of anger and sympathy

By means of a series of mediation analyses using the PROCESS macro developed by Preacher and Hayes (2008) hypotheses 2a, b and c were tested. The PROCESS macro offers, among others, a statistical test for mediation, using bootstrapping. The significance of the mediation effects are calculated by means of the confidence intervals (Preacher & Hayes, 2008). According to Preacher and Hayes (2008) bootstrap intervals are preferred over the traditional Sobel test because bootstrapping respects the non-normality of the sampling distribution of the indirect effect. If the intervals include a zero, there is no mediation. However, because the data in this study shows, overall, normally distributed data, the Sobel test was performed as well to make sure that both the confidence intervals as well as the Sobel test indicate mediation. The mediation analyses were tested using a bootstrapping (5000 resamples) method with confidence estimates that are bias-corrected (Preacher & Hayes, 2004).

To assess the mediating influence of anger and sympathy, regression analyses were analyzed as suggested by Baron and Kenny (1986). This method includes looking at the regressions (1) for the direct effect of the independent variable on each mediating variable (i.e. the a-path), (2) for the direct effect of each mediating variable on each dependent variable (i.e. the b-path), (3) for the direct effect of the independent variable on each dependent variable (i.e. the c-path) and (4) for the effect of the independent variable on each mediation variables, when controlled with each mediating variable (i.e. the c’-path). If the results show that there is a difference between c and c’ such that c’ is no longer or less significant, a mediation effect is present.

First of all, the direct effect of emotional crisis communication on anger and

sympathy (a-paths). The regression output of the mediation analyses show that emotional crisis communication has a significant effect on anger (b = -.59, t (390) = -4.10, p < .001, 95% CI [-.88, -.31]), but an insignificant effect on sympathy (b = .17, t (390) = -4.10, p = .161, 95% CI [-.06, .42]).

Moving on to the direct effects of anger and sympathy on corporate reputation, purchase intention and NWOM (b-paths). Starting with corporate reputation, the regression output shows that both anger (b = -.46, t (389) = -11.45, p < .001, 95% CI [-.53, -.38]) and sympathy (b = .65, t (389) = 15.49, p < .001, 95% CI [.56, .73]) have a significant effect on corporate reputation. The same goes for purchase intention as dependent variable. Both

21 anger (b = -.41, t (389) = -9.84, p < .001, 95% CI [-.51, -.33]) and sympathy (b = .51, t (389) = 10.43, p < .001, 95% CI [.40, .59]) show to have a significant effect on purchase intention. Lastly, both anger (b = .62, t (389) = 13.84, p < .001, 95% CI [.53, .71]) as well as sympathy (b = -.81, t (389) = -16.92, p < .001, 95% CI [-.90, -.72] also show to have a significant effect on NWOM.

Then the c-paths. Although these paths have also been discussed for hypotheses 1a, b and c, they will be repeated for better understanding of the mediation effects. Emotional crisis communication shows to have a significant effect on corporate reputation (b = .66, t (390) = 5.05, p < .001, 95% CI [.40, .92]) and purchase intention (b = .76, t (390) = 5.72, p < .001, 95% CI [.50, 1.02]), but not on NWOM (b = -.14, t (390) = -.94, p = .351, 95% CI [-.45, .16]).

The lasts paths to be analysed are the c’-paths. The effects in these paths resemble the effect of emotional crisis communication on corporate reputation, purchase intention and NWOM when controlled with anger and sympathy. The results show that, when controlled with anger, emotional crisis communication still has a significant effect on corporate reputation (b = .39, t (389) = 3.38, p = .008, 95% CI [.16, .62]) and purchase

intention (b = .52, t (389) = 4.25, p < .001, 95% CI [.28, .76]) and still an insignificant effect on NWOM (b = -.04, t (389) = -.36, p = .721, 95% CI [-.28 .21]). When controlled with sympathy, the effect of emotional crisis communication on corporate reputation (b = .54, t (389) = 5.29,

p < .001, 95% CI [.34, .75]) and purchase intention (b = .68, t (389) = 5.70, p < .001, 95% CI

[.34, .81]) remain significant and the effect of emotional crisis communication on NWOM (b

= -.01, t (389) = -.02, p = .982, 95% CI [.44, .91]) remains insignificant.

The final step to determine whether mediation effects are present is to assess whether the difference between the c and c’ paths is significant. Both the bootstrap

confidence intervals for the indirect effect of anger on the relation between emotional crisis communication and corporate reputation (b = .27, 95% CI [.14, .42]) as well as the Sobel test (Sobel’s Z = 3.85, p < .001) indicate that anger has a significant mediation effect. However, because emotional crisis communication still has a significant effect on corporate reputation when controlled with anger (c’-path), the mediation is partial. Sympathy shows to have no mediation effect on the relation between emotional crisis communication and corporate reputation. Both the confidence intervals (b = .11, 95% CI [-.05, .28]) as well as the Sobel test

22 (Sobel’s Z = 1.41, p = .159) show insignificant results. These results indicate that hypothesis 2a is partially supported.

Next up is to determine whether anger and sympathy mediate the relation between emotional crisis communication and purchase intention. The bootstrap intervals for the indirect effect of anger on the relation between emotional crisis communication and purchase intention (b = .24, 95% CI [.12, .39]) and the Sobel test (Sobel’s Z = 3.77, p < .001) indicate that anger mediates the effect between emotional crisis communication and purchase intention. However, because emotional crisis communication still has a significant effect on purchase intention when controlled with anger (c’-path), the mediation is partial. Sympathy shows to have no mediation effect (b = .09, 95% CI [-.03, .22]; Sobel’s Z = 1.41, p = .162). These results indicate that hypothesis 2b is partially supported.

Finally, the mediation effect of anger and sympathy on the relation between

emotional crisis communication and NWOM is assessed. The results show that neither anger (b = -.10, 95% CI [-.29, .08]; Sobel’s Z = -1.06, p = .287) nor sympathy (b = -.14, 95% CI [-.34, .06]; Sobel’s Z = -1.41, p = .158) mediate the effect between emotional crisis communication and NWOM. These results indicate that hypothesis 2c is rejected.

Moderating influence of crisis type

By means of two moderation analyses using the PROCESS macro, hypotheses 3a and b were assessed. Just like mediation analyses, PROCESS allows for moderation analyses using bootstrapping. 5000 bootstrapping samples were used. By looking that the change in R2 and the pairwise comparisons, the presence moderation can be determined (Hayes, 2013).

The results of the analyses show that crisis type does not have an interaction effect on the relation between emotional crisis communication and anger. Inserting crisis type as a moderator does not improve the fit of the model (ΔR2 = .005, F (2,286) = 1.13, p = .324). The ΔR2 indicates that allowing the relationship between emotional crisis communication and anger to differ between the three crisis types results in an additional .5 percent of variance explained relative to when this relationship is fixed to be the same across the crisis types. This increase is not significant, indicating that there is no moderation. The pairwise comparisons of the individual interactions verify this presumption. There is no difference between victim crisis types and accidental crisis types (b = .44, t (386) = 1.31, p = .191, 95% CI [.22, 1.12]), no difference between victim crisis types and preventable crisis types (b = -.40, t (386) = -1.26, p = .207, 95% CI [-.89, .09]) and no difference between accident crisis

23 types and preventable crisis types (b = .02, t (386) = .05, p = .962, 95% CI [-.67, .71]). Based on these results, hypothesis 3a is rejected.

The second moderation analysis (with sympathy as dependent variable) shows insignificant results as well. Allowing crisis type to moderate the relation between emotional crisis communication and sympathy does not improve the fit of the model (ΔR2 = .001, F (2,286) = .24, p = .789). Having crisis type as moderator only accounts for .1 percent extra explained variance, indicating that there is no moderation. The pairwise comparisons confirm this presumption. There is no difference between victim crisis types and accidental crisis types (b = .14, t (386) = .49, p = .626, 95% CI [-.43, .71]), no difference between victim crisis types and preventable crisis types (b = .19, t (386) = .66, p = .511, 95% CI [-.38, .76]) and no difference between accident crisis types and preventable crisis types (b = .05, t (386) = -.16, p = .871, 95% CI [-.64, .54]). Based on these results, hypothesis 3b is rejected.

Additional analyses

Considering the fact that the proposed mediators in the conceptual model of this study do not or just partially mediate the effects between emotional crisis communication and corporate reputation, purchase intention and NWOM, it makes sense to test whether crisis type moderates the relation between emotional crisis communication and corporate reputation, purchase intention and NWOM. To test this, three moderation analyses were performed (5000 bootstrap samples). The results show that crisis type does not moderate the relation between emotional crisis communication and corporate reputation (ΔR2 = .002,

F (2,286) = .43, p = .652), nor does it moderate the relation between emotional crisis

communication and purchase intention (ΔR2 = .008, F (2,286) = 1.63, p = .196) and it also does not moderate the relation between emotional crisis communication and NWOM (ΔR2 < .001, F (2,286) = .04, p = .959). The pairwise comparisons for these tests can be found in appendix 8.

Discussion

The present study sought to determine the value of communicating emotion as a corporation in times of crisis. Incorporating the element of emotion into a crisis response was argued to diminish negative outcomes as a result of a corporate crisis situation (Claeys & Cauberghe, 2014; Christensen and Cornelissen, 2010; Van der Meer, 2014; Wesseling,

24 Kerkhof & Van Dijk, 2007). Moreover, expressing these emotions was thought to elicit

compassion with the receiver and therefore reduce negative emotions such as anger and increase positive ones such as sympathy (Fischer & Manstead, 2008), resulting in less reputational damage and less negative behavior intentions. Lastly, it was argued that that emotional crisis communication is most effective when used in a crisis of high organizational responsibility in comparison to the other crisis situations (Coombs, 2007b; Weiner, 1985).

To grasp the effects of expressing emotion during a corporate crisis situation, an online experiment was conducted. The results confirm the value of an emotional crisis response. Expressing emotion during a crisis situation proves to reduce reputation damage (hypothesis 1a) en improve purchase intentions (hypothesis 1b). No significant effects were found for negative word-of-mouth (hypothesis 1c). Furthermore, anger was shown to partially mediate the relation between emotional crisis communication and corporate reputation, but no indirect effect of sympathy was found (hypothesis 2a). Anger also partially mediates the effect between emotional crisis communication and purchase

intention, but no such effect was found for sympathy (hypothesis 2b). Also no mediation of neither anger nor sympathy on the relation between emotional crisis communication and negative word-of-mouth were found (hypothesis 2c). Lastly, crisis type did not have a moderation effect on the relation between emotional crisis communication and anger (hypothesis 3a) nor sympathy (hypothesis 3b).

Overall, these results adhere to the idea that expressing emotion during a crisis situation can result in less reputation damage and better behavioral intentions.

Communicating emotion has the ability to form and maintain social relationships (Fischer & Manstead, 2008) and therewith reduce negative outcomes. By communicating emotion, an organization shows itself to be human and empathetic. This in turn helps to solidify

relationships with stakeholders (Kelleher, 2009). Especially diminishing negative cognitive processes such as feelings of anger have shown to underlie the effect of expressing emotion. However, this study has also shown that communicating emotion cannot avert the loss of sympathy toward the organization as a result of the crisis. Arguably, increasing positive emotions such as sympathy is much less likely than reducing negative ones such as anger. This may be the case since feelings of sympathy only occurs when there is awareness of others’ (in this case the corporation’s) suffering (Iyer & Oldmeadow, 2006). In a crisis

25 situation, the public is much more likely to contemplate the offense that is made against them and their well-being by the corporation than to think of the corporation’s suffering. Consequently, negative emotions may play a more important role in the effects of

expressing emotion in a crisis response than positive emotions. Another point of interest is the fact that the effect of expressing emotion did not differ depending on the type of crisis. Despite the fact that experienced emotions become more negative depending on the attribution of responsibility (Weiner, 2006), expressing regret and shame results in a stable decrease of negative emotions such as anger. Considering the SCCT theory of Coombs (2007b) this is remarkable. This theory is built on the presumption that based on the type of crisis more or less responsibility should be taken for the incident. When a corporation is responsible for a crisis situation the SCCT theory teaches us that the corporation should take the blame and offer apologies, whereas when the corporation is not or less responsible a more defensive reaction in the form of denial is more appropriate. This seems not to be the case regarding expressing emotion. It seems as though responding to a crisis situation with emotion, irrespectively of the type of crisis, strengthens relations with the public and therewith solidifies corporate reputation and lessens negative behavioral intentions. This means that when a corporation is faced with a crisis situation, the public does not

necessarily take into account the responsibility of the corporation with regard to the crisis situation when evaluating the corporation. So, merely communication emotion in a crisis situation helps bolster corporate reputation and reduce negative outcomes.

The results of this study add to existing literature in several ways. This study is one of the few to focus attention on ‘how’ crisis responses should be conveyed, instead of ‘what’ a crisis response should entail. Moreover, it helps unearth the value of expressing emotion in a corporate crisis situation. Although the effect of communicating emotion in other fields such as marketing (Cotte & Ritchie, 2005) and political psychology (Ridcout & Searles) has been researched. This study helps to shed light on the effect of communicating emotion in a corporate crisis situation. Lastly, this study adds to existing literature by focusing on different crisis types. Where previous literature (Van der Meer, 2014) have tried to link the types of crisis responses identified by Coombs (2007b) with emotion, this study has helped unravel the value of communicating emotion in different kinds of crises.

As is the case with most research done, the findings must be interpreted with several limitations in mind. Furthermore, some of these limitations offer opportunities for future

26 research. First of all, this study used only one (rather specific) product/organization, namely a pasta sauce company. However, considering the fact that a lot of people come into contact with these kinds of products/companies (directly or indirectly), the choice for this company is justified. Future research, however, should consider different types of products and organizations to verify that the results found in this study are applicable to different types of organizations. Secondly, the present studies used a reputation scale that measured the overall evaluation of the corporation. Although this measurement does provide a good indication as to how the corporation is perceived by respondents, it does not capture the multidimensional character of a corporate reputation. However, considering that a fictitious organization was used, respondents would have had a hard time rating the reputation of a fictitious organization using a multidimensional construct such as that of Fombrun, Gardberg and Sever (2013). Thirdly, the crisis response of the company in question was communicated by means of a news article. This is not a limitation in the sense that it is unrealistic, as most people gather news indirectly through news agencies and therewith shape their (i.e. the publics) agenda (McCombs, 1978). But, future research should focus on different kinds of media. It would be interesting to see whether more direct means of communication such as corporate websites or social media such as Facebook and Twitter result in different effects. Fourthly, this study focused on only two types of perceived emotions (i.e. anger and

sympathy). Especially because negative cognitive processes such as anger have shown to underlie the effect of expressing emotion, future research may focus on more kinds of negative emotions the public can perceive during a corporate crisis. Other prominent emotions during a crisis such as fear, despair or disappointment (Jin, Pang, & Cameron, 2010) may be diminished as well by expressing emotion. Lastly, although this study confirms the positive effect of expressing emotion by written word, no attention has been paid to the non-verbal part of emotion such as gestures, facial expression and tone of voice. Non-verbal signals are arguably even more important in communication than verbal information

(Mehrabian, 1981). Therefore, future studies should research whether more rich media, with which non-verbal signals can be captured and communicated, add to the results found in this study.

All in all, the results of the present research provides both theoretical as well as practical implications. From a theoretical point of view, the current study provides new insights in the field of PR and crisis communication. More specifically, by incorporating the

27 element of communicating emotion, this study enlarges the body of literature on PR and crisis communication. Not only should emotion be studied from the publics’ but also from the corporation’s point of view. Moreover, the research provides insight in the effects of expressing communication in different crisis types and discusses the underlying affective processes. Furthermore, the current study is of great interest to PR managers and executives because it provides practical suggestions for the implementation of emotion in crisis

communication. When faced with times of turmoil, both the organization and its

stakeholders are faced with stress resulting in various negative emotions. In these situations corporations ought not to ignore the emotional side of crises. They should show they care instead of responding in a coldly and distant manner. By expressing emotions of regret and shame the corporation shows its human side and therewith reduce negative outcomes as a result of a crisis situation.

Literature

Alsop, R. J. (2004). The 18 immutable laws of corporate reputation. Dow Jones & Company. Avery, E. J., Lariscy, R. W., Kim, S., & Hocke, T. (2010). A quantitative review of crisis

communication research in public relations from 1991 to 2009. Public Relations

Review, 36(2), 190-192.

Barclay, L. J., Skarlicki, D. P., & Pugh, S. D. (2005). Exploring the role of emotions in injustice perceptions and retaliation. Journal of Applied Psychology, 90(4), 629.

Baron, R. M., & Kenny, D. A. (1986). The moderator–mediator variable distinction in social psychological research: Conceptual, strategic, and statistical considerations. Journal

of personality and social psychology, 51(6), 1173.

Benoit, W. L. (1997). Image repair discourse and crisis communication. Public relations

review, 23(2), 177-186.

Bembenek, L. J., Beike, D. R. & Schroeder, D. A. (2007). Justice violations, emotional reactions, and justice-seeking responses. In D. De Cremer (Ed.), Advances in the

psychology of justice and affect (pp. 16-36). Charlotte, NC: Information Age.

Bies, R. J., & Tripp, T. M. (1996). Beyond distrust: “Getting even” and the need for revenge. In R. M. Kramer & T. Tyler (Eds.), Trust and organizations (pp. 246–260). Thousand Oaks, CA: Sag

28 Bougie, R., Pieters, R., & Zeelenberg, M. (2003). Angry customers don't come back, they get

back: The experience and behavioral implications of anger and dissatisfaction in services. Journal of the Academy of Marketing Science, 31(4), 377-393.

Christensen, L. T., & Cornelissen, J. (2010). Bridging corporate and organizational communication: Review, development and a look to the future. Management

Communication Quarterly, 25(3), 383-414.

Claeys, A. S., & Cauberghe, V. (2014). What makes crisis response strategies work? The impact of crisis involvement and message framing. Journal of Business

Research, 67(2), 182-189.

Claeys, A. S., Cauberghe, V., & Vyncke, P. (2010). Restoring reputations in times of crisis: An experimental study of the Situational Crisis Communication Theory and the

moderating effects of locus of control. Public Relations Review, 36(3), 256-262. Coombs, W. T. (2004). West Pharmaceutical's explosion: structuring crisis discourse

knowledge. Public Relations Review, 30(4), 467-473.

Coombs, W. T., & Holladay, S. J. (2009). Further explorations of post-crisis communication: Effects of media and response strategies on perceptions and intentions. Public

Relations Review, 35(1), 1-6.

Coombs, W. T. (2007b). Protecting organization reputations during a crisis: The development and application of situational crisis communication theory. Corporate reputation

review, 10(3), 163-176.

Coombs, W. T. (2014). Ongoing crisis communication: Planning, managing, and responding. Sage Publications.

Coombs, W. T., & Holladay, S. J. (1996). Communication and attributions in a crisis: An experimental study in crisis communication. Journal of public relations research, 8(4), 279-295.

Coombs, W. T., & Holladay, S. J. (2005). An exploratory study of stakeholder emotions: Affect and crises. Research on emotion in organizations, 1, 263-280.

Coombs, W. T., & Holladay, S. J. (2007). The negative communication dynamic: Exploring the impact of stakeholder affect on behavioral intentions. Journal of Communication

Management, 11(4), 300-312.

Coombs, W. T., & Holladay, S. J. (2008). Comparing apology to equivalent crisis response strategies: Clarifying apology's role and value in crisis communication. Public