Non-Profit Organizations in Times of Crisis
The Effects of Emotional Crisis Response Strategies and Crisis Clusters on theReputation of a Charitable Organization
Lonneke van de Ven 11308869
Master thesis
Corporate Communication
Graduate School of Communication University of Amsterdam
26th of June 2017
Abstract
During a crisis, non-profit organizations (NPOs) are highly vulnerable. Especially for NPOs such as charities, a damaged reputation can have an enormous impact, since these
organizations are largely dependent on monetary donations, investments in time and tangible gifts. For this reason, proper crisis communication is highly important for charitable
organizations.
For the profit sector, previous studies show positive effects on post-crisis evaluations and reputation when using emotions in an organizational crisis response. However, the effects of emotional crisis response strategies have not been examined for NPOs. Therefore, this study aims to examine the effects of communicated emotions in crisis response strategies and crisis cluster on the reputation of a charitable organization. An online experiment with a 3 (crisis response strategy: anger vs. sadness vs. non-emotional) x 2 (crisis cluster: victim vs. preventable) experimental design was conducted to investigate the hypotheses.
Findings show that crisis response strategy has no significant influence on the
dependent variables of this study: willingness to forgive, reputation and willingness to donate. However, it appears that the relationship between willingness to forgive and willingness to donate is partially mediated by reputation. The results thus expand our knowledge of how people’s willingness to donate arises in times of crisis and show that it is important for charitable organizations to create a sense of forgiveness amongst their stakeholders, since forgiveness positively stimulates donating behavior.
Keywords: crisis communication, non-profit organizations, charities, crisis cluster, emotions, crisis response, reputation.
Introduction
In times of crisis, non-profit organizations (NPOs) are highly vulnerable (Sisco, 2012). A crisis can be defined as “a sudden and unexpected event that threatens to disrupt an
organization’s operations and poses both a financial and reputational threat” (Coombs, 2007, p. 164). Due to a crisis, people can think badly of an organization which can result in a damaged reputation (Coombs, 2007). Especially for NPOs such as charities, a damaged reputation can have serious consequences, since this type of organization highly depends on monetary donations, investments in time and tangible gifts (Beldad, Snip, & Van Hoof, 2014). Moreover, charities would not be able to help their beneficiaries when they do not receive sufficient monetary donations from generous parties (Beldad et al., 2014).
Meisenbach (2006) examined a crisis situation of the American Red Cross (ARC), a well-established charitable organization in the United States. The ARC is founded to help victims of disasters and provide relief to them. In order to facilitate this, they collect monetary donations. After the attacks on 9/11 they established a special Liberty Fund which would support the ARC to reduce human suffering. However, a few weeks later, it appeared that the Liberty Fund was used to avoid the chance of a new attack by investing in extra audit services such as information systems instead of providing direct help to victims. Donators and the public felt that they had been misled about how their donations were going to be used. This resulted in negative news coverage and angry donators who stopped donating money.
During a crisis, NPOs such as the ARC in the above mentioned case disappoint the public. In this way, they risk not only their reputation, but possibly their very existence, since NPOs are dependent on their publics (Sisco, 2012). Therefore, proper post-crisis
communication plays a crucial role in situations like this. It can be used to repair the
reputation or even prevent reputational damage (Coombs & Holladay, 2005). In order to avoid a damaged reputation and the corresponding negative consequences in times of crisis, it is useful for NPOs to know what proper crisis communication is and thus how to react on a crisis. However, research about protecting the reputation of NPOs in times of crisis is scarce.
Plenty of studies about crisis communication focused on the role of emotions in post-crisis communication (Choi & Lin, 2007; Claeys & Cauberghe, 2015; Grappi, Romani, & Bagozzi, 2015; Iyer & Oldmeadow, 2006; Jin, Pang, & Cameron, 2007; Kim & Cameron, 2011; Van der Meer & Verhoeven, 2014). Not only because crises can be described as highly emotional events for both the public and the organization itself (Coombs & Holladay, 2005),
but also because emotions can have a framing function in which they can stimulate how people interpret and respond to negative events (Nabi, 2003). An organization can therefore decide to communicate emotions in its post-crisis response.
For the profit sector, previous research shows positive effects on post-crisis evaluations and reputation when using emotions in a crisis response (Choi & Lin, 2007; Claeys et al., 2013; Van der Meer & Verhoeven, 2014). However, nothing is known about the extent to which and the way in which such communicated emotions affect the reputation of NPOs. This is relevant to investigate because NPOs are perceived as warm and kind (Aaker, Vohs, & Mogilner, 2010) and organizations who do not communicate emotions might be perceived as cold and uncommitted (Van der Meer & Verhoeven, 2014). It can therefore be expected that emotional crisis response strategies are more suitable for NPOs compared to rational crisis responses. However, this has not yet been proved.
To investigate the role of emotions in post-crisis communication for NPOs, this study focuses on how communicated emotions (i.e. anger and sadness) in crisis response strategies may affect the reputation of a charitable organization. To improve understanding of this relationship, willingness to forgive is identified as a mediator in this study since forgiveness has the potential to overcome negative thoughts (Claeys, Cauberghe, & Leysen, 2013) which might positively stimulate an organization’s reputation. Furthermore, because Coombs’ Situational Crisis Communication Theory (SCCT; Coombs, 2007) shows that the reputation of an organization in times of crisis depends on attributions of crisis responsibility, the use of emotional crisis response strategies will be investigated in relation with two crisis clusters (i.e. victim and preventable cluster) in which different attributions of responsibility are
emphasized. This leads to the following research question:
RQ: What are the effects of using an emotional crisis response strategy on the reputation of a charitable organization and to what extent is this influenced by the crisis cluster?
The current study contributes to the literature in two ways. First, to my knowledge, this is the first (experimental) research studying the effects of emotional crisis response strategies and crisis clusters on the reputation of NPOs (i.e. charitable organizations). Second, it extends the existing crisis literature with the potential influence of willingness to forgive. The results will provide practical implications regarding post-crisis communication for charitable
Theoretical Framework
Emotions in Crisis Communication
Crisis communication can be used to protect reputational assets in times of crisis (Coombs & Holladay, 2005). SCCT (Coombs, 2007) is an important framework in the literature about crisis communication. It identifies how key facets of a particular crisis situation influence stakeholders’ attributions about the crisis and the organizational reputation. As a result, the framework provides an understanding how to minimize a damaged reputation based on post-crisis communication.
Post-crisis communication literature, such as SCCT, has mainly focused on how an organization can effectively communicate with its key publics in order to diminish
reputational threat as much as possible. Much attention has been paid to the rational
arguments that organizational crisis responses should include. For instance, Coombs (2007) distinguished different response strategies such as denial, diminish and rebuild and
investigated which response strategy suits a certain crisis type best. However, the use of emotions in organizational crisis responses is often omitted. For charitable organizations, there even is no literature about the effects of emotional crisis response strategies.
The aim of this research is to fill this gap in literature, since emotions can change the way people think, feel and behave (Roseman, Wiest, & Swartz, 1994). This can be explained by means of the Appraisal-Tendency Framework (ATF; Han, Lerner, & Keltner, 2007), which addresses how and why specific emotions affect consumer decision-making. More
specifically, the ATF assumes that specific emotions evoke specific cognitive and
motivational processes, which account for the effects of each emotion upon judgment and decision-making. Later on, both the SCCT and ATF will be discussed more in depth as a general framework for this study.
The Effect of Communicating Emotions on Reputation
In times of crisis, emotional reactions of the public are inevitable (Coombs & Holladay, 2005). As a result, a series of researchers (Coombs, 2007; Grappi et al., 2015; Iyer & Oldmeadow, 2006; Jin et al., 2007; Van der Meer & Verhoeven, 2014) has investigated the type of emotions that people experience after or during a crisis (i.e. emotions as a dependent variable). However, the role of emotions as independent variable may not be neglected since the formulation of messages about corporate crises can guide how the public reacts on a crisis
(Kim & Cameron, 2011). More specific, emotions can have a framing function: “repeated pairing of certain emotions with particular ideas or events shapes the way in which one interprets and responds to those events” (Nabi, 2003, p. 227).
To my knowledge, there are only a few studies that focused on the effects of emotional crisis response strategies on an organization’s reputation (i.e. emotions as independent
variable). Reputation, in this study, is defined as a stakeholder’s overall evaluation of an organization over time (Abratt & Kleyn, 2012). Moreover, these studies are accomplished among companies in the profit sector.
Choi and Lin (2007) for example, conducted an experiment wherein they compared organizational crisis responses with a rational and emotional appeal. Their results demonstrate that messages with an emotional appeal (i.e. emotions in general, not specific ones) result in more positive crisis evaluations than when a rational appeal was used. In line with this research, Claeys et al. (2013) investigated the role of expressing signs of sadness in specific. They show that expressing sadness results in a better post-crisis reputation than a rational message in case of an organizational self-disclosure. Finally, Van der Meer and Verhoeven (2014) also investigated the role of emotions in corporate crisis responses and distinguished between two specific emotions: they examined which type of emotions, shame vs. regret vs. no emotions, suits the rebuild and diminish crisis response strategy of Coombs (2007) best. Results reveal that corporate reputation is positively affected by a crisis response that contains signs of shame or regret compared to communicating no emotions. The authors clarify this result by stating that a lack of emotions may indicate lower levels of organizational
involvement, through which the organization may be perceived as cold and uncommitted. In contrast to the use of emotional crisis response strategies for profit organizations, nothing is known about using emotional crisis response strategies with regards to NPOs. However, based on a study which states that NPOs are perceived as warm (Aaker et al., 2010), it could be argued that non-emotional crisis response strategies will not suit these type of organizations. Feelings of warmth generally include perceptions of generosity, kindness, trustworthiness, sincerity, helpfulness, honesty and thoughtfulness (Aaker et al., 2010). With regards to organizations, this means that consumers believe that employees and consumers are treated fairly by the organization and that they have the feeling they can trust the organization (Aaker et al., 2010). It can therefore be expected that non-emotional crisis responses, which might enhance feelings of coldness and low levels of commitment (Van der Meer &
Verhoeven, 2014), do not positively affect a charitable organization’s reputation in times of crisis. Therefore, this study assumes that the public’s perception of a charitable crisis will be
positively influenced by communicating emotions in general (not specific ones) in a crisis response. This leads to the first hypothesis:
H1: The resulting reputation of a charitable organization communicating emotions in their crisis response strategy is more positive than the resulting reputation of a
charitable organization which does not communicate emotions in their crisis response strategy.
Besides the studies of Claeys et al. (2013) and Van der Meer and Verhoeven (2014), little is known about which communicated emotions in crisis response strategies affect corporate reputations. Van der Meer and Verhoeven (2014) did not find any differences between communicating signs of shame and regret and recommend future research to focus on other emotions in crisis responses to get a more complete image about how communicating
emotions can help organizations in crisis with restoring their reputation. Therefore, this study focuses on two negative emotions, anger and sadness, since negative emotions are common and seen as natural organizational reactions during a crisis (Tiedens, Ellsworth, & Mesquita, 2000). Moreover, anger and sadness appear to have different appraisals of control and responsibility which as a result might influence how people process organizational messages and how they evaluate the organization (Kim & Cameron, 2011).
The Effect of Communicating Anger and Sadness on Reputation
Anger and sadness are both described as negative emotions (Tiedens et al., 2000), but with different definitions and associations. Literature about the characteristics of basic emotions states that when experiencing sadness, there is a failure to attain or maintain a goal (Stein & Trabasso, 1992). Additionally, anger is often experienced when an agent causes a loss of a goal (Stein & Trabasso, 1992). Both emotions are considered as powerful emotional states which affect thoughts and behavior (Darwin, 1965).
Due to the fact that anger and sadness have different associations, communicating these emotions does not have the same influence on a public. Madera and Smith (2009) examined for example how expressions of anger and sadness influence the evaluation of leaders during a crisis. Results of this research show that expressing sadness leads to a more favourable leader evaluation than expressing anger. The authors explain these results by the fact that anger might be viewed as too defensive or hostile which negatively affected the public. Moreover, they argue that expressing sadness demonstrates signs of sympathy and
concern (Shaver, Schwartz, Kirson, & O'Connor, 1987), which was therefore seen as more appropriate and did not evoke negative affect in others.
In line with the study of Madera and Smith (2009), Gallois (1993) states that sadness is associated with warmth, concern and affiliation. This is also supported by other scholars who demonstrate that sadness besides warmth and affiliation is a sign of remorse and
sympathy (Shaver et al., 1987; Tiedens, 2001). In contrast to the positive feelings associated with sadness, communicating anger is often perceived as being dominant, competent and importantly: less warm, friendly and nice (Tiedens, 2001).
Because NPOs are seen as warm, which include perceptions of kindness, honesty, sincerity, helpfulness, trustworthiness and thoughtfulness (Aaker et al., 2010), it can be
expected that communicating sadness has a more positive effect on a charitable organization’s reputation than communicating anger. Hence, Hypothesis 2 is posed as follows:
H2: The resulting reputation of a charitable organization communicating signs of sadness in their crisis response strategy is more positive than the resulting reputation of an organization communicating signs of anger in their crisis response strategy.
The Mediating Effect of Willingness to Forgive
Forgiveness is the ideal outcome for a transgressor who wishes to repair a damaged relationship (Davis & Gold, 2011). In times of crisis, it is therefore important for
organizations that stakeholders will forgive them so that the organization can start to rebuild their reputation. In this study, willingness to forgive is defined as “the public’s effort to reduce negative thinking, overcome unpleasant emotions and restore their damaged relationship with an organization due to a crisis that affected both the organization and the public” (Moon & Rhee, 2012, p. 680).
Research shows that people are more likely to forgive others for wrongdoing when they experience sincerity (Claeys et al., 2013) and empathy (Davis & Gold, 2011). Feelings of empathy result in greater care for the transgressor and the affected relationship as a result of which people are more likely to forgive (Davis & Gold, 2011). Furthermore, when an organization offers a response that is considered sincere, stakeholders may be more likely to forgive the organization for its wrongdoing and importantly: evaluate the event and the organization less negatively (Claeys et al., 2013).
The amount of empathy and sincerity can be influenced by the communicated
the state of the person experiencing them (Parkinson, Fischer, & Manstead, 2005). Wubben, De Cremer, and Van Dijk (2009) investigated for example how communicating anger and disappointment guides reciprocal cooperative behavior. It appears that communicating disappointment or anger is more likely to result in forgiveness and therefore establish a mutually beneficial relationship. However, the effect was stronger for communicating disappointment than anger. Furthermore, Hareli and Eisikovits (2006) examined how
communicating emotions of guilt and shame affects forgiveness. Their findings demonstrate that expressions of guilt as well as shame strengthen people’s willingness to forgive.
Blaney, Benoit, and Brazeal (2002) emphasize the importance of sincerity and forgiveness with regards to crisis communication in specific. They suggest that stakeholders are more likely to forgive an organization for a crisis when an organization offers a response that is considered sincere. In line with these scholars, several studies state that emotional message framing has a positive impact on an organization’s perceived sincerity and the public’s willingness to forgive (Benoit & Brinson, 1999; Kauffman, 2008; Legg, 2009). Although there is no literature about the specific effects of anger and sadness on willingness to forgive, based on the above mentioned studies, it is expected that communicating emotions positively influences this concept. Therefore, Hypothesis 3a can be posed as follows:
H3a: The public is more likely to forgive a charitable organization when they communicate emotions in their crisis response strategy rather than when they do not communicate emotions in their crisis response strategy.
As mentioned earlier, sadness is associated with feelings of warmth, concern and affiliation (Gallois, 1993), which corresponds to the fact that NPO’s are seen as warm. Hence, the following is expected:
H3b: The positive effect of communicating emotions on willingness to forgive will be more pronounced if a charitable organization communicates signs of sadness in their crisis response strategy compared to communicating signs of anger.
Psychological research shows that expressions of forgiveness convey a willingness to maintain a positive relationship (Wallace, Exline, & Baumeister, 2008). When this willingness to forgive is applied to crisis communication, stakeholders might evaluate the crisis and the organization less negatively (Claeys et al., 2013). It can therefore be expected
that the relationship between an organization’s crisis response strategy and reputation is mediated by stakeholder’s willingness to forgive. This leads to the following hypothesis:
H4: The relationship between crisis response strategy and reputation is positively mediated by willingness to forgive.
The Moderating Effect of Need for Affect on Willingness to Forgive
In general, it is expected that communicating emotions in a crisis response strategy will positively influence both the public’s willingness to forgive and the resulting reputation of a charitable organization. However, due to individual differences, some persons could
appreciate affective messages more, while others prefer cognitive messages. To illustrate this, it appears that persuasive messages stimulate a greater attitude change when the message fits the recipient’s desired level of affect and cognition (Maio & Haddock, 2007). This can also be applicable when it comes to post-crisis communication.
Need for affect (Maio & Esses, 2001) is an important construct in understanding emotion-related processes and should therefore be taken into account in this study. The term is described as follows: “the general motivation of people to approach or avoid situations and activities that are emotion inducing for themselves and others” (Maio & Esses, 2001, p. 585). A distinction is made between being low vs. high in the need for affect. Research shows that individuals who are high in the need for affect are more likely to use affective information in forming attitudes (Maio & Esses, 2001).
Haddock, Maio, Arnold, and Huskinson (2008) investigated the role of need for affect and need for cognition in affect- and cognition-based persuasive messages. They found that an affective message evokes more positive attitudes among people high in the need for affect and low in the need for cognition. In line with these findings, Fabrigar and Petty (1999) also demonstrated that an affect-based message was more effective in convincing a person high in the need for affect compared to a person low in the need for affect. The findings of these studies all demonstrate that the need for affect influences the impact of affect-based messages. Because this study is also about affect-based messages, the need for affect is important to take into account.
Although there is no literature about the need for affect in relation to willingness to forgive, based on the fact that messages stimulate a greater attitude change when they fit the recipient’s desired level of affect (Maio & Haddock, 2007), the following moderation effect is expected:
H5: The positive effect of emotional crisis response strategies on willingness to forgive is more pronounced for individuals high in the need for affect compared to individuals low in the need for affect.
The Moderating Effect of Crisis Clusters on Willingness to Forgive
Coombs’ SCCT (2007) is built on a highly influential theory, namely the Attribution Theory (Weiner, 1985). The Attribution Theory states that people make attributions based on the cause of negative and unexpected events. According to Weiner (1985), attributions can be classified along three causal dimensions. The first is locus of control, which focuses on the question if an event is internally or externally caused. The second dimension, stability, is about the cause and if this will change over time or not. Controllability is the last dimension and questions if the cause can be controlled or not. The more responsible a person, in this case a charitable organization, is perceived for a negative event (i.e. crisis), the more negative emotions will be experienced according to this theory.
Coombs (2007) elaborates on the Attribution Theory by identifying three crisis clusters in which each has a different perceived responsibility: the victim, accidental and preventable cluster. The victim cluster has very weak attributions of crisis responsibility, since the cause of the crisis was outside the control of the organization, for example a natural disaster or an external agent who causes damage to an organization. In this cluster, the organization is considered as a victim of the event. The accidental cluster has minimal
attributions of crisis responsibility and is considered as unintentional or uncontrollable by the organization (e.g. technical errors). The third cluster, which is called the preventable cluster, is considered as purposeful. Therefore, this cluster has very strong attributions of crisis responsibility. Examples are human-error accidents and organizational misdeed. All three crisis clusters are incorporated in SCCT and influence the amount of damage a crisis could cause to an organization’s reputation.
This study does not include all three of Coombs’ crisis clusters for a 3x3 experiment, but focuses only on the victim and preventable cluster. This is because previous researches show difficulties with finding significant results on an organization’s reputation when facing an accidental crisis (Claeys, Cauberghe, & Vyncke, 2010; Sisco, 2012).
It is assumed that the victim and preventable crisis cluster differ in attributions of perceived responsibility (Coombs, 2007). Previous research (Coombs & Holladay, 1996; Coombs & Holladay, 2002) has shown that these attributions of crisis responsibility affect the way how stakeholders perceive a crisis. Additionally, research by Moon and Rhee (2012)
shows that the public is more likely to forgive an organization when a crisis is perceived as external and uncontrollable (i.e. victim cluster) than when it is perceived as internal and controllable (i.e. preventable cluster). The authors explain these results by stating that internal and controllable crises give the public the feeling that the organization should have taken more responsibility in order to avoid the crisis situation. Due to these feelings, the public may not easily forgive the organization. Based on these findings, the following moderation effect is expected:
H6: The public is more likely to forgive a charitable organization when the crisis is perceived as a victim crisis rather than a preventable crisis.
The Relationship between Communicating Emotions, Crisis Clusters, Willingness to Forgive and Reputation
As mentioned earlier, the current study aims to investigate the effects of communicating anger and sadness in relation with crisis clusters on the reputation of a charitable organization. This can be done by drawing on the ATF (Han et al., 2007), a framework that distinguishes
between six cognitive dimensions that affect the underlying appraisal patterns of different emotions: certainty, pleasantness, attentional activity, control, anticipated effort and responsibility. The effects of anger and sadness will be explained by means of control and responsibility, since these two dimensions play an important role with regards to an organization’s reputation during a crisis (Coombs, 2007).
Previous researches consider the emotions of anger and sadness both as negative in valence, but different with respect to the appraisal of control and responsibility (Dunn & Schweitzer, 2003; Ellsworth & Smith; 1988; Han et al., 2007; Iyer & Oldmeadow; 2006; Keltner, Ellsworth, & Edwards, 1993). Literature about the feelings associated with anger show that anger is characterized by high other-person control (Dunn & Schweitzer, 2003; Ellsworth & Smith; 1988; Han et al., 2007; Iyer & Oldmeadow; 2006; Keltner et al., 1993). This means that people feel angry when they perceive some other person(s) to be the cause of their misfortune. Additionally, Roseman et al. (1994) state that anger is usually experienced when people blame specific agents for a transgression or injustice, for example during the 9/11 attacks. The same scholars have also studied people’s feelings associated with sadness. These studies show that people feel sad when they perceive impersonal circumstances beyond human control to be the cause of their misfortune (Ellsworth & Smith, 1988; Han et al., 2007;
Keltner et al., 1993). In other words, in this case an organization could not have prevented the cause.
Both emotions can be applied to the SCCT crisis clusters by Coombs (2007). More specifically, anger as well as sadness can be linked to the victim cluster, which is associated with weak attributions of crisis responsibility. Natural disasters, rumor, workplace violence, product tampering and malevolence are all examples of crises that belong to the victim cluster. When experiencing anger, one can think of malevolence (i.e. a person causes damage to an organization), while a natural disaster is an example of a crisis when the organization experiences feelings of sadness according to the above mentioned literature.
In order to know what the effect is of communicating these emotions in post-crisis communication, Claeys and Cauberghe (2015) conducted an experiment in which they investigated an organization’s perceived crisis responsibility. Findings reveal that expressing feelings of sadness rather than anger stimulates the idea that an organization is taking their responsibility for a negative event. Because the preventable cluster is concerned with high levels of perceived responsibility (Coombs, 2007), it is expected that communicating sadness will suit this crisis cluster better than communicating anger. Hence, Hypothesis 7a can be posed as follows:
H7a: Communicating sadness in a crisis response strategy results in a higher
willingness to forgive, and thus a more positive reputation, in the preventable cluster compared to the victim cluster.
Since the victim cluster is not associated with feelings of perceived responsibility (Coombs, 2007), it is expected that anger is the most appropriate emotion to communicate in a crisis response strategy for this type of crisis. This leads to the following hypothesis:
H7b: Communicating anger in a crisis response strategy results in a higher willingness to forgive, and thus a more positive reputation, in the victim cluster compared to the preventable cluster.
The Relationship between Reputation and Willingness to Donate
If a crisis negatively affects the reputation of a charity, but does not impact behavioral intentions, there would be no or less reason to worry about the effects of crises for charitable organizations. For this reason, willingness to donate is an important variable to take into
account. As mentioned earlier, charitable organizations are largely dependent on regular monetary donations (Beldad et al., 2014). It is therefore relevant to understand why people want to donate and what triggers them to keep doing this.
Beldad et al. (2014) examined the effects of organizational reputation for charitable organizations and show that organizational reputation plays an important role in influencing both the intention to donate to a charitable organization and to continue donating. Meijer (2009) shows the same effect for organizational reputation: people would be more likely to donate to charitable organizations with a strong and positive reputation, which also positively influences people’s repeating intention to donate. Maintaining their positive reputation is thus of extreme importance for charitable organizations in order to encourage regular donations.
However, in times of crisis these are one of the most difficult aspects to manage (Coombs, 2007). The negative publicity where a company is faced with is deleterious for an organization’s reputation and increases people’s perceptions of the risks of donating (Beldad et al., 2014). Moreover, a charitable crisis often results in a disappointed public, a loss of public confidence and a negative reputation (Sisco, 2012). When a charitable organization faces a crisis, it is therefore expected that the damaged reputation leads to a lower willingness to donate among (potential) donators.
By drawing on previous literature (Beldad et al., 2014; Meijer, 2009) that describes the reputation of a charitable organization as an important predictor of people’s willingness to donate, this study suggests a mediation of the relationship between crisis response strategy and willingness to donate by the reputation of the charitable organization. This leads to the final hypothesis:
H8: The relationship between crisis response strategy and willingness to donate is positively mediated by the reputation of the charitable organization.
Based on the proposed hypotheses, a conceptual model has been developed (Figure 1). In short, the model predicts both a positive direct and a positive indirect effect of emotional crisis response strategy on reputation (H1, H2). Since willingness to forgive is identified as a mediator between emotional crisis response strategy and reputation (H4), there is also a positive direct effect predicted from emotional crisis response strategy on willingness to forgive (H3a, H3b). Furthermore, it is predicted that crisis type and need for affect moderate the relation between crisis response strategy and willingness to forgive (H5, H6, H7a, H7b).
Finally, it is expected that reputation positively mediates the relation between crisis response strategy and willingness to donate (H8).
Figure 1. Conceptual Model
Method
Design
In order to answer the research question and test the hypotheses, an online experiment with a 3 (crisis response strategy: anger vs. sadness vs. non-emotional) x 2 (crisis cluster: victim vs. preventable) factorial between-subjects design was conducted. This resulted in six conditions. Each participant was randomly exposed to one of these conditions.
Procedure
To accomplish the experiment, the online survey software from Qualtrics was used.
Respondents were approached via convenience sampling, which means that the survey was distributed to friends, colleagues, classmates and relatives of the researcher by means of social media, such as Facebook and LinkedIn, and via e-mail. Respondents were therefore free to choose the time and location for filling in the survey.
The questionnaire, which can be found in Appendix A, started with a short introduction wherein the ethical rules of ASCoR and an explanation of the study were provided. It was also mentioned that the survey took approximately five minutes and that the
data were going to be processed anonymously. After reading the introduction, respondents had to give their permission for participation.
In the first part of the survey, the respondents’ demographics were asked. The second part consisted of some general questions with regards to emotions (i.e. need for affect), followed by some questions to measure the respondents’ general attitude towards charitable organizations and their willingness to donate. These questions were considered as control questions. Next, participants were exposed to a news article for a minimum of 30 seconds about which they had to answer twelve questions. These questions included the dependent measures: willingness to forgive, reputation and willingness to donate. After reading the news article, respondents were not able to go back and read it again. In the final part, participants had to answer two manipulation checks. When the respondents were finished, they were thanked for their participation, which was followed by a short debriefing regarding the fictitious nature of the study. Also, each participant had the ability to ask questions about the study.
Respondents
205 respondents agreed to participate in this study and completed the questionnaire. 26.3% of the sample were male and 73.7% were female. The age of the respondents ranged from 18 to 68 years, with an average age of 27.43 (SD = 11.06). Furthermore, 79.2% of the respondents was highly educated, which means they completed a HBO or University study. With regards to charitable organizations, 56.1% had a positive attitude about charitable organizations in general and 88.3% said they had donated to a charity once or more.
Randomization Check
Since this study consisted of six conditions, a randomization check was applied to test the distribution of the demographic variables in each condition. The included demographic variables in this randomization check were gender, age and education level.
In order to check if participants’ gender was comparable over the six conditions, a Chi-square Test was conducted. Results showed that participants’ gender was not
significantly different when comparing the six conditions with each other, χ2 = 6.462, p = .264.
Furthermore, with a One-way ANOVA, it was checked if participants’ age and
education level were evenly distributed over the same six conditions. The means for age (F(5, 199) = .28, p = .927) as well as education (F(5, 199) = .10, p = .991) did not significantly
differ among the six conditions. Therefore, it can be stated that all the tested demographics were equally divided over the groups.
Stimulus Material
As previously mentioned, this study employed a 3 x 2 experimental design which resulted in six different conditions. Based on these conditions, six organizational crisis responses were developed about a fictitious foodbank in Amsterdam, called Food4You. A story about a foodbank was consciously chosen due to the clear goal and activities of foodbanks, with which most people are familiar. Furthermore, a fictitious charitable organization was used in order to prevent prior relationships and history between the organization and the participant. The independent variable crisis cluster was manipulated by changing the cause of the crisis as done by Sisco (2012). In the victim cluster, the charitable organization was described as a victim of the crisis because an external agent caused damage to the organization. In contrast, in the preventable cluster, it became clear that laws and regulations were violated by the management of the charitable organization. Moreover, this could have been avoided with a better money administration.
The independent variable crisis response strategy was manipulated by adding angry and sad statements in the organizational crisis responses as Van der Meer & Verhoeven (2014) did in their study. Dependent on the condition, the organization expressed feelings of anger, sadness or communicated no emotions. Examples of these statements are: “We feel very sad/anger that this could happen”, “I could only scream” (anger), “Tears welled in my eyes” (sadness). In the non-emotional condition, these sentences were left out. The full stimulus material is provided in Appendix B.
Instrumentation
The constructs in this study were all measured with multiple items on a 7-point Likert scale ranging from ‘fully disagree’ (0) to ‘fully agree’ (7). Table 1 includes the bivariate
correlations and descriptive statistics of the main concepts and Table 2 in Appendix C shows all the items inclusive factor loadings.
Table 1. Bivariate correlations and descriptive statistics of main concepts.
Variable M (SD) 1 2 3 4 5
1. Willingness to forgive 4.90 (1.13) 1
2. Reputation 5.31 (0.89) .65**
3. Willingness to donate 3.32 (1.36) .36** .25** 4. Need for Affect 5.15 (0.87) .04 .08 -.01
5. General willingness to donate 4.84 (1.33) .08 .18** .28** .19** 6. General attitude 4.98 (0.80) .19** .30** .16* .15* .50**
Note: N = 205. *Correlation is significant at p <.05 (2-tailed). **Correlation is significant at p <.01 (2-tailed).
Measurement Dependent Variables
The first construct, willingness to forgive, was measured with four questions derived from Xie and Peng (2009). Example questions were: “Given the company’s response, I would forgive the foodbank” and “I would disapprove of the foodbank”. An Exploratory factor analysis indicated that the scale was unidimensional by showing that there was one component meeting the Eigenvalue-criterion (bigger than 1: Eigenvalue component is 2.597), explaining 64.92% of the variance. The factor loadings for this construct ranged from 0.71 to 0.87 and the 4-item scale also proved reliable as indicated by a Cronbach’s Alpha of .82.
Reputation was assessed using five items based on Coombs (1998). Example items were: “The foodbank of Amsterdam is concerned with the well-being of its public” and “The foodbank of Amsterdam is basically dishonest”. An Exploratory factor analysis indicated that the scale was unidimensional by showing that there was one component meeting the
Eigenvalue-criterion (bigger than 1: Eigenvalue component is 2.776), explaining 55.51% of the variance. The factor loadings for this construct ranged from 0.53 to 0.81 and the 5-item scale also proved reliable as indicated by a Cronbach’s Alpha of .77.
The third and last construct, willingness to donate, was measured using three questions based on a study by Coyle and Thorson (2001). Questions which were included are: “I am likely to donate to the foodbank of Amsterdam” and “There is a small chance that I will donate to the foodbank of Amsterdam”. An Exploratory factor analysis indicated that the scale was unidimensional by showing that there was one component meeting the Eigenvalue-criterion (bigger than 1: Eigenvalue component is 2.324), explaining 77.46% of the variance. The factor loadings for this construct ranged from 0.86 to 0.88 and the 3-item scale also proved reliable as indicated by a Cronbach’s Alpha of .85.
Measurement Moderating Variable
Need for affect is considered as a moderator in this study and was measured using six questions derived from Maio and Esses (2001). Example questions are: “I am a very emotional person” and “Emotions help people get along in life”. An Exploratory factor analysis indicated that the scale was not unidimensional by showing that there were two components meeting the Eigenvalue-criterion (bigger than 1: Eigenvalue components are 1.167 and 2.068). The scale also did not prove reliable with a Cronbach’s Alpha of .55. Due to this low reliability, two items were deleted. A second Exploratory factor analysis indicated that the scale was now unidimensional by showing that there was one component meeting the Eigenvalue-criterion (bigger than 1: Eigenvalue component is 1.988), explaining 49.69% of the variance. The factor loadings ranged from .60 to .78 and Cronbach’s Alpha was improved to .65, which was sufficiently reliable.
Measurement Control Questions
The attitude towards charitable organizations in general and the willingness to donate to charitable organizations in general can influence the results of this study. Therefore, both constructs were included as control variables in the survey.
Participants general attitude towards charitable organizations was measured with five items which were based on research by Webb, Green, and Brashear (2000). Examples are “Charities have good causes” and “I have a positive image on charities”. An Exploratory factor analysis indicated that the scale was unidimensional by showing that there was one component meeting the Eigenvalue-criterion (bigger than 1: Eigenvalue component is 2.295), explaining 45.90% of the variance. The factor loadings for this construct ranged from 0.58 to 0.81 and the 5-item scale also proved reliable as indicated by a Cronbach’s Alpha of .69.
Willingness to donate to charitable organizations in general was measured with the same three questions as the dependent variable willingness to donate (Coyle & Thorson, 2001). An Exploratory factor analysis indicated that the scale was unidimensional by showing that there was one component meeting the Eigenvalue-criterion (bigger than 1: Eigenvalue component is 2.427), explaining 80.90% of the variance. The factor loadings for this construct ranged from 0.78 to 0.84 and the 3-item scale also proved reliable as indicated by a
Cronbach’s Alpha of .88.
To check if the time that participants took to read the news article about the charitable crisis influenced the results, reading time was considered as a third control variable.
Pre-test
To ensure that the survey and the stimuli were clear and comprehendible, a pre-test was conducted. The pre-test consisted of the measurement scales of the dependent variables and two manipulation checks in order to check if the manipulations of the two factors were successfully implemented and respondents have perceived them as they were intended.
The pre-test was divided in four conditions: (1) the anger response strategy combined with a victim crisis cluster, (2) the anger response strategy combined with a preventable crisis cluster, (3) the sad response strategy combined with a victim crisis cluster, and (4) the sad response strategy combined with a preventable crisis cluster. The non-emotional crisis
response strategies were not taken into account. For every condition, five respondents filled in the questionnaire. This means that a total sample of 20 respondents took part in the pre-test.
The first manipulation check examined if the emotional crisis response strategy (anger vs. sadness) was also perceived as such by the respondents. The question was therefore posed as follows: “Which emotions do you believe the charitable organization communicated in their crisis response?”. This question was formulated as a multiple choice question with the following answer possibilities: anger, shame, sadness and no emotions. Results showed that for the anger conditions (condition 1 and 2) as well as the sad conditions (condition 3 and 4) only one out of ten respondents did not provide the right answer. It can therefore be stated that the manipulation of the emotional crisis response strategy was well implemented.
The second manipulation check was formulated for the factor crisis cluster and measured the perceived responsibility of the charitable organization with the following question: “How much do you believe the charitable organization was responsible for the crisis?”. This question was answered on a 7-point Likert-scale ranging from ‘not responsible at all’ (0) to ‘fully responsible’ (7). Respondents who were exposed to the news article with the preventable crisis cluster indicated a higher perceived responsibility (M = 4.90, SD = 1.45) than respondents who were exposed to the news article with the victim crisis cluster (M = 2.10, SD = 0.88), t(18) = -5.23, p <.001, 95% CI [-3.93, -1.68]. This means that the manipulation was perceived as it was intended.
In order to increase the chance to find differences between the conditions in the main study, the foodbank as named in the news article has been changed from Antwerp to
Amsterdam. This is because of the idea that respondents of this study will feel more connected to Amsterdam instead of Antwerp.
Manipulation Checks
Although the manipulation checks were included in the pre-test, they were examined one more time since the pre-test was conducted among 20 respondents and only focused on the emotional conditions. With regards to the first manipulation check, which examined if the crisis response strategy (anger vs. sadness vs. non-emotional) was perceived as intended by the respondents, results showed that for the anger conditions (condition 1 and 2) 80.3% of the respondents indicated they had read a news article with anger emotions. In the sad conditions (condition 3 and 4), 84.4% gave the right answer and perceived the news article as it was intended. For the non-emotional conditions (condition 5 and 6), results showed that only 32.9% indicated that there were no emotions used in the news article. This means that the non-emotional news article was not perceived as non-emotional. In the Conclusion & Discussion section, there will be more elaboration on this finding.
With regards to the manipulation check about the perceived responsibility,
respondents who were exposed to the news article with the preventable crisis cluster indicated again a higher perceived responsibility (M = 5.39, SD = 1.39) than respondents who were exposed to the news article with the victim crisis cluster (M = 2.95, SD = 1.40), t(203) = -12.51, p = .040, 95% CI [-2.83, -2.06]. This means that the manipulation was still perceived as it was intended.
Analysis
In order to test Hypothesis 1 until 3, four Univariate ANOVAs were conducted. Because of the moderation effects in Hypothesis 5, 6 and 7 and the mediation effects in Hypothesis 4 and 8, these hypotheses were tested with PROCESS, a modelling tool for SPSS that integrates many of the functions of existing statistical tools for mediation and moderation analysis (Hayes, 2012). The advantage of using PROCESS in this study is that it combines the different steps of a regression analysis into one simple-to-use procedure.
To test the mediation analyses, PROCESS model 4 (simple mediation model) was used. The two moderated relationships were tested with PROCESS model 2, since this model allows to test a relationship with two moderators between the independent variable and the mediating variable. All tested models controlled for any influences of gender, age, education, reading time and general attitude towards charitable organizations. The models which
included the dependent variable willingness to donate also controlled for respondent’s general willingness to donate.
Results of the analysis will be discussed per step concerning the conceptual model of this study. First the direct effects of using emotions in crisis response strategies on the reputation of a charitable organization and the public’s willingness to forgive will be
examined. Afterwards the results of the mediation and moderation effects will be discussed.
Results
Direct Effects of Crisis Response Strategy on Reputation and Willingness to Forgive The first hypothesis (H1) expects that the resulting reputation of a charitable organization communicating emotions in their crisis response strategy is more positive than the resulting reputation of a charitable organization which does not communicate emotions in their crisis response strategy. To test this hypothesis, a Univariate ANOVA was conducted with crisis response strategy (emotional vs. non-emotional) as independent variable and reputation as dependent variable. The results show no significant effect, F(1, 197) = .016, p = .901, which means that there is no significant difference between the perceived reputation of respondents who were exposed to the emotional crisis response strategies (M = 5.32, SD = .96) and respondents who were exposed to the non-emotional crisis response strategies (M = 5.29, SD = .74). As a result, H1 needs to be rejected. With regards to the control variables, the only significant result in this model is for the control variable general attitude towards charitable organizations, F(1, 197) = 11.890, p = .001, η2 = .00, meaning that people who have a generally positive attitude towards charitable organizations score higher on reputation (M = 5.51, SD = .73) than people who have a negative attitude towards charitable organizations in general (M = 5.05, SD = 1.00). The results of the other (non-significant) control variables can be found in Appendix D (Table 3).
The second hypothesis (H2) expects that the resulting reputation of a charitable organization is more positive when communicating signs of sadness in their crisis response strategy instead of signs of anger. To test this hypothesis, a second Univariate ANOVA was conducted wherein crisis response strategy (anger vs. sad) was the independent variable and reputation the dependent variable. Results show a marginal significant effect, F(1, 127) = 2.841, p = .094. However, due to this significance level, H2 is rejected. This means that there is no significant difference between the perceived reputation of respondents who were
exposed to the anger crisis response strategies (M = 5.18, SD = .93) and respondents who were exposed to the sad crisis response strategies (M = 5.48, SD = .98). Concerning the
control variables, again only general attitude towards charitable organizations shows a significant result, F(1, 127) = 7.818, p = .003, η2 = .11 (Appendix D, Table 4).
The third Univariate ANOVA with crisis response strategy (emotional vs. non-emotional) as independent variable and willingness to forgive as dependent variable was conducted to examine the effects of emotional crisis response strategies on willingness to forgive (H3a). The results of this test show no significant effect of communicating emotions on willingness to forgive, F(1, 197) = .002, p = .966. This means that there is no significant difference between the willingness to forgive of respondents who were exposed to the
emotional crisis response strategies (M = 4.91, SD = 1.15) and respondents who were exposed to the non-emotional crisis response strategies (M = 4.88, SD = 1.08). As a result, Hypothesis 3a is rejected. However, for this Univariate ANOVA not only the control variable general attitude towards charitable organizations is significant F(1, 197) = 6.976, p = .009, η2 = .05, but also sex, F(1, 197) = 4.866, p = .029, η2 = .02 (Appendix D, Table 3). It appears that women (M = 5.00, SD = 1.03) are more likely to forgive a charitable organization than men (M = 4.61, SD = 1.32).
To assess the possible differences of the two emotions in specific on willingness to forgive, a fourth Univariate ANOVA is conducted with crisis response strategy (anger vs. sad) as independent variable and willingness to forgive as dependent variable. Again, no
significant effect is found, F(1, 127) = 1.803, p = .182, which means that there is no
significant difference between the perceived reputation of respondents who were exposed to the anger crisis response strategies (M = 4.79, SD = 1.10) and respondents who were exposed to the sad crisis response strategies (M = 5.04, SD = 1.20). Hence, also H3b is rejected. Concerning the control variables, the same to control variables, namely general attitude towards charitable organizations, F(1, 127) = 9.613, p = .002, η2 = .71, and sex, F(1, 127) = 13.606, p <.001, η2 = .09, show significant results (Appendix D, Table 4).
Mediating Effect of Willingness to Forgive
To assess the mediating influence of willingness to forgive in the relation between crisis response strategy and reputation (H4), PROCESS model number 4 was used (Hayes, 2012). The model with crisis response strategy as independent variable, willingness to forgive as mediating variable and reputation as dependent variable appeared to be significant F(7, 197) = 6.28, p = .002, R2 = .106. The output shows that crisis response strategy has no significant effect on willingness to forgive (b = .05, t(197) = .55, p = .585, 95% CI [-.13, .24]). This is in line with the results of the earlier mentioned Univariate ANOVAs. For the relationship
between willingness to forgive and reputation, the output does show a positive significant effect (b = .49, t(196) = 11.21, p <.001, 95% CI [.40, .57]). Furthermore, crisis response strategy shows to have an insignificant direct effect on reputation (b = .06, t(197) = .76, p = .446, 95% CI [-.09, .20]), which is also consistent with findings of the Univariate ANOVA. The final path examines the effect of crisis response strategy on reputation when controlled for willingness to forgive. The output shows that, when controlled for willingness to forgive, crisis response strategy still has no significant effect on reputation (b = .03, t(196) = .54, p = .591, 95% CI [-.08, .15]).
Because there is only one significant relationship in this mediation output (between willingness to forgive and reputation), it can be stated that the relationship between crisis response strategy and reputation is not mediated by willingness to forgive. This is confirmed by the Sobel test, which is also insignificant (Sobel’s Z = .54, p = .587). As a result,
Hypothesis 4 is rejected.
Regarding the control variables in this model, general attitude towards charitable organizations appears to be significant in the output with reputation as outcome variable (b = .198, t = 3.29, p = .001) as well as willingness to forgive as outcome variable (b = .315, t = 3.50, p <.001). In Appendix E, both the results of the other (non-significant) control variables (Table 5) and the direct and indirect effects can be found (Table 6).
Moderating Effects of Need for Affect and Crisis Cluster
By means of a moderation analysis using PROCESS model 2, the effects of need for affect (H5) and crisis cluster (H6) on the relation between crisis response strategy and willingness to forgive were investigated. First, the results show that there is no interaction between crisis response strategy (emotional vs. non-emotional) and need for affect (low vs. high), (b = .22,
t(201) = .66, p = .513, 95% CI [-.44, .89]). This means that individuals high in the need for
affect who are exposed to an emotional crisis response strategy (M = 4.95, SD = 1.06) are not more willing to forgive a charitable organization compared to individuals high in the need for affect who are exposed to a non-emotional crisis response strategy (M = 4.86, SD = 1.26). As a result, Hypothesis 5 is rejected.
Hypothesis 6 expects that the public is more likely to forgive a charitable organization when the crisis is perceived as a victim crisis rather than a preventable crisis. The PROCESS output shows that crisis cluster has a significant influence on willingness to forgive (b = -.91,
t(201) = -3.72, p = <.001, 95% CI [-1.19, -.63]), indicating that people who were exposed to
than people who were exposed to the preventable crisis (M = 4.45, SD = 1.18). Based on this result, Hypothesis 6 is confirmed.
Hypothesis 7a and 7b are about the interaction effect of crisis response strategy (anger vs. sadness) and crisis cluster (victim vs. preventable) on willingness to forgive and
reputation. The results show that there is no interaction for the model with willingness to forgive as outcome variable, (b = .12, t(131) = .32, p = .794, 95% CI [-.61, .85]). This means that communicating sadness does not result in a higher willingness to forgive in the
preventable cluster (M = 4.65, SD = 1.34) compared to the victim cluster (M = 5.47, SD = .86), and that communication anger does not result in a higher willingness to forgive in the victim cluster (M = 5.27, SD = .85) compared to the preventable cluster (M = 4.57, SD = 1.12).
With regards to the model with reputation as outcome variable, again no interaction effect was found (b = -.24, t(131) = -.74, p = .463, 95% CI [-.88, .40]). This means that
communicating sadness does not result in a more positive reputation in the preventable cluster (M = 5.22, SD = 1.04) compared to the victim cluster (M = 5.75, SD = .85), and that
communication anger does not result in a more positive reputation in the victim cluster (M = 5.32, SD = .97) compared to the preventable cluster (M = 5.04, SD = .88). Hence, both Hypothesis 7a and 7b are rejected.
Effects of Reputation on Willingness to Donate
H8 expects that the relationship between crisis response strategy and willingness to donate is positively mediated by the reputation of a charitable organization. PROCESS model 4 with crisis response strategy as independent variable, reputation as mediating variable and willingness to donate as dependent variable appears to be significant F(7, 197) = 2.98, p = .005, R2 = .096. With regards to the direct effect of crisis response strategy on reputation, the output shows no significant effect (b = .06, t(197) = .76, p = .446, 95% CI [-.09, .20]). Moving on to the effect of reputation on willingness to donate. The output shows that reputation has a significant effect on willingness to donate, (b = .31, t(197) = 2.93, p = .004, 95% CI [.10, .52]). Furthermore, crisis response strategy has no direct significant effect on willingness to donate (b = -.05, t(197) = -.46, p =.644, 95% CI [-.27, .17]). The final path that needs to be analysed examines the effect of crisis response strategy on willingness to donate when controlled for reputation. The output shows that, when controlled for reputation, crisis response strategy still has no significant effect on willingness to donate (b = .07, t(196) = -.63, p = .530, 95% CI [-.28, .15]).
Because only one significant relationship is found in this output (between reputation and willingness to donate), the relationship between crisis response strategy and willingness to donate is not mediated by reputation. This is confirmed by the Sobel test, which is also insignificant (Sobel’s Z = .70, p = .483). Hence, the final hypothesis (H8) is rejected.
Regarding the control variables in this model, general willingness to donate to charitable organizations has a positive significant influence in the output with willingness to donate as outcome variable (b = .281, t = 3.62, p <.001), and general attitude towards charitable organizations has a positive significant influence in the output with reputation as outcome variable (b = .166, t = 2.38, p = .018). In Appendix E, both the results of the other (non-significant) control variables (Table 7) and the direct and indirect effects can be found (Table 8).
As crisis response strategy has no influence on reputation as well as willingness to forgive, it was also checked if the relationship between willingness to forgive and willingness to donate is positively mediated by reputation (in this model crisis response strategy is not included). PROCESS model 4 with willingness to forgive as independent variable, reputation as mediating variable and willingness to donate as dependent variable appears to be
significant F(7, 197) = 6.316, p <.001, R2 = .183. With regards to the effect of willingness to forgive on reputation, the results show that willingness to forgive has a positive significant effect on reputation (b = .49, t(197) = 11.26, p <.001, 95% CI [.40, .57]). Moving on to the effect of reputation on willingness to donate. The output shows that reputation has a significant effect on willingness to donate, (b = .21, t(195) = 3.51, p = .002, 95% CI [.10, .52]). Furthermore, willingness to forgive also shows to have a direct significant effect on willingness to donate (b = .37, t(197) = 4.62, p <.001, 95% CI [.21, .53]). The final path examines the effect of willingness to forgive on willingness to donate when controlled for reputation. The output shows that, when controlled for reputation, willingness to forgive still has a significant effect on willingness to donate (b = .36, t(196) = 3.51, p <.001, 95% CI [.16, .56])
Because all the relationships in this model are significant, a mediation effect is
present. This is confirmed by the Sobel test which shows that the difference between the c and c’-path is significant (Sobel’s Z = 2.52, p = .012). However, because willingness to forgive has a significant effect on willingness to donate when controlled for reputation, the mediation is partial. Regarding the control variables in this model, general willingness to donate to charitable organizations has a positive significant influence in the output with willingness to donate as outcome variable (b = .281, t = 3.62, p <.001), and general attitude towards
charitable organizations has a positive significant influence in the output with reputation as outcome variable (b = .166, t = 2.38, p = .018). In Appendix E, both the results of the other (non-significant) control variables (Table 9) and the direct and indirect effects can be found (Table 10). Although these results do not belong to one of the hypotheses, it is an interesting finding which partially confirms the conceptual model of this study.
Conclusion and Discussion
NPOs such as charities are highly vulnerable in times of crisis (Sisco, 2012). Especially for this type of organization, a damaged reputation can have an enormous impact, since these organizations are largely dependent on monetary donations (Beldad et al., 2014). Previous studies show positive effects on post-crisis evaluations and reputation when using emotions in a crisis response (Choi & Lin, 2007; Claeys et al., 2013; Van der Meer & Verhoeven, 2014). However, because the effects of emotional crisis response strategies have not been examined for NPOs, the current study has been conducted to examine whether using an emotional crisis response strategy affects the reputation of a charitable organization and to what extent this is influenced by the crisis cluster. To improve understanding of this relationship, willingness to forgive is identified as a mediator. In order to accomplish this research, an online experiment with a 3 (crisis response strategy: anger vs. sadness vs. non-emotional) x 2 (crisis cluster: victim vs. preventable) factorial between-subjects design was conducted. In contrast with the expectations, communicating emotions in an organizational crisis response does not influence the reputation of a charitable organization as well as willingness to forgive and willingness to donate. Below, the findings will be discussed in more detail.
The Insignificant Effects of Emotional Crisis Response Strategies
Results of this study show some remarkable findings. First and most important, the emotional crisis response strategy appears to have no influence on both willingness to forgive and reputation: there was no difference in willingness to forgive and reputation when comparing the emotional and non-emotional crisis response strategy as well as the sad and anger crisis response strategy. As this is the first study combining the effects of emotional crisis response strategies with crisis clusters, it does not contradict previous studies. However, the results do contradict the expectations of this research.
The findings might be explained by the influence of respondent’s attributed
crisis responsibility to the preventable crisis cluster and less to the victim situation, which is in line with the SCCT (Coombs, 2007). However, it can occur that the attributed
responsibility had too much influence as a result of which the emotional appeals were not newsworthy enough. Consequently, respondents might have answered the questions in the survey only with the attributed responsibility in mind due to which emotions did not affect willingness to forgive, reputation and willingness to donate. Kim and Cameron (2011) advise practitioners who attempt to use emotional appeals in crisis response strategies to ensure that the public recognizes the organization’s emotional response as a newsworthy point. In this study, this could have been done by omitting the effects of crisis cluster, because in this way the findings cannot be influenced by attributed responsibility as a result of which respondents might be focusing more on the communicated emotions. Further research is thus required to examine if emotional crisis response strategies indeed have no influence on a charitable organization’s reputation.
A second reason which can explain the insignificant effects of emotional vs. non-emotional crisis response strategies, is the fact that the non-non-emotional crisis response
strategies were not experienced as non-emotional by the respondents. This appears from the manipulation check which shows that only 32.9% of the respondents indicated that there were no emotions used in the news article. If the non-emotional crisis response strategy is not perceived as intended, it is difficult to find differences between the emotional and non-emotional conditions.
As previous studies show clear differences between the effects of emotional and non-emotional crisis responses on reputation (Choi & Lin, 2007; Claeys et al., 2013; Van der Meer & Verhoeven, 2014) and willingness to forgive (Benoit & Brinson, 1999; Kauffman, 2008; Legg, 2009), the problem might be linked to the stimulus material and/or manipulation check used in this study. It is therefore recommended for future research to include the
non-emotional stimulus material in the pre-test too. In this way, the researcher knows in an earlier stage if the non-emotional crisis response strategies are perceived as intended. Furthermore, in this study, the manipulation check existed of a multiple choice question with four answers (anger, sadness, shame, no emotions). To be sure if the non-emotional crisis response strategies were indeed perceived as non-emotional, it is better to use more comprehensive multiple-item measures in future research instead of one multiple choice question with four answers. A question on a 7-point semantic differential scale can for example be added to improve the manipulation check (Kim & Cameron, 2011).
The Moderated Effects
Need for affect was considered as a moderator in this study, since messages stimulate a greater attitude change when they fit the recipient’s desired level of affect (Maio & Esses, 2001). However, this finding from previous literature is not proved in the current study. Respondents high in the need for affect were not more willing to forgive when they were exposed to the emotional crisis response strategy instead of the non-emotional crisis response strategy.
A possible explanation can be that the stimulus material in this study was too short for respondents to feel involved in the crisis situation. Hoffman (1984) namely states that feelings of empathy are important when examining the need for affect. Compared to previous research on the need for affect wherein narratives (Appel & Richter, 2010), movies (Maio & Esses, 2001), or several short text fragments instead of one (Haddock et al., 2008) are used as stimulus material, in the current study one short news article is used wherein only three emotional sentences were included. This can impede feelings of empathy as a result of which the need for affect does not show the intended results.
Furthermore, the insignificant results of need for affect can also be linked to the discussion in the previous section. As stated in the Theoretical Framework, individuals who are high in the need for affect are more likely to use affective information in forming attitudes (Maio & Esses, 2001). However, if respondents indeed did not recognize the emotional response as newsworthy because of the influential attributed responsibility, respondents might not experience a sense of affect. As a result, it is difficult to find differences on the concepts reputation and willingness to forgive between respondents who are low and high in the need for affect.
The second moderation effect was expected for the factor crisis cluster, which is based on different attributions of responsibility (Coombs, 2007). This factor is taken into account, since Coombs’ SCCT (2007) shows that the reputation of an organization in times of crisis depends on attributions of crisis responsibility. Findings demonstrate a direct effect of crisis cluster on willingness to forgive, which means that the public is more likely to forgive a charitable organization when the crisis is perceived as a victim rather than a preventable crisis. This result is in line with research by Moon and Rhee (2012) which shows that people are more likely to forgive an organization when the crisis is perceived as external and
uncontrollable (i.e. victim cluster) than when it is perceived as internal and controllable (i.e. preventable cluster).
