Is Being Funny a Useful Policy? How Local Governments’Humorous Crisis Response Strategies and Crisis Responsibilities Inf luence Trust, Emotions, and Behavioral Intentions

Janna H?mpke · Stefan R?seler ·Meinald T. Thielsch

Abstract This study is the f irst to investigate how a local government’s humorously framed response strategy on social media to a low-severity crisis inf luences people’s trust in the local government and their crisis-related behavioral intentions, specif ically when considering the government’s responsibility for the crisis. Based on the situational crisis communication theory, we examined the mediating role of experienced positive or negative aff ect on people’s responses to a local government’s crisis communication strategy. Further, we exploratorily examined the predictive power and moderating role of demographics, sense of humor, disposition to trust, and the respective crisis scenarios. A total of 517 people participated in an online experiment in which they were confronted with three randomly presented f ictive crisis scenarios where the local government’s crisis responsibility (high versus low) and the framing of their crisis response strategy (in form of humorous versus rational Twitter posts) were systematically varied between subjects.First, the results mostly corroborate earlier f indings about the degree of crisis responsibility (that is, when a government’s crisis responsibility is high, people have less trust and behavioral intentions) and about the mediating role of experienced aff ect. Second, we found that humorously framed strategies negatively inf luence trust and positive aff ect (but not behavioral intentions). In contrast to earlier f indings,the crisis responsibility × framing interaction was not signif icant. Altogether, the results advise against using humor in crisis communications on social media, even in lowseverity crisis. Exploratory analyses indicate that further investigations should focus on specif ic crisis characteristics and potential moderators.

Keywords Aff ect · Crisis communication · Crisis responsibility · Humor · Situational crisis communication theory · Social media

1 Introduction

In crisis situations, social media off ers local governments a way to communicate directly and quickly with the public and provide up-to-date information. Such crisis communication allows local governments to stay in contact with the public during a crisis, gain their trust, and encourage them to take appropriate actions (Bakker et al. 2018). Prior research indicates that crisis managers might eff ectively use humor in social media communications during low-severity crises,such as misconducts or scandals (Avidar 2012), as humorous messages during such crises generally reach more people and are more popular (Choudhary et al. 2012; Fraustino and Ma 2015 ). Even local governments and public authorities have increasingly begun using less formal and even humorous language on social media (Fraustino and Ma 2015; Rasmussen 2017). For instance, the NSW Australian Police is famous for using humor to communicate public safety messages (NSW Police Force n.d.).

However, research and practical examples indicate that humorous crisis response strategies do not always result in positive responses (Honisch and Más Manchón 2020).For instance, the brand Tropicana missed the mark when responding to the COVID-19 pandemic in a humorous way(Tropicana 2020). For local governments, such misuse can have severe consequences: Inappropriate humor may threaten citizens’ trust in the local government and discourage them from taking appropriate actions (Bitterly 2022). Hence, it is necessary to understand the conditions under which humorous crisis communication on social media might be useful. We contribute to this topic by examining how a local government’s humorous versus rational communications, depending on the government’s responsibility for the crisis, inf luence people’s responses, and we examine the mediating role of experienced positive and negative aff ect on this response. To our knowledge, this is the f irst study to investigate humorous crisis communication among public sector organizations.

2 Theoretical Background

The following section f irst provides brief background information on crises and crisis communication in general. Then relevant literature regarding the role of attributed crisis responsibility, emotions, and humor in crisis communication on social media is described to later derive our study hypotheses from.

2.1 Crises, Crisis Communication, and Situational Crisis Communication Theory (SCCT)

A crisis can be described as “an event that is an unpredictable, major threat that can have a negative eff ect on the organization, industry, or stakeholders if handled improperly” (Coombs 1999, p. 2). Furthermore, it is characterized by “ambiguity of cause, eff ect, and means of resolution”(Pearson and Clair 1998, p. 60). Managing a crisis requires an organization to interact with the stakeholders to mitigate negative consequences, and crisis communication is the “verbal, visual, and/or written interaction between the organization and its stakeholders (often through the media)prior to, during and after a negative occurrence” (Fearn-Banks 2002, p. 480). These communication processes aim to provide information and support to stakeholders or the public, but they also help the organization manage and repair its reputation and legitimacy (Sturges 1994). Coombs’ ( 2007)situational crisis communication theory (SCCT) offers empirically based guidelines of how organizations can minimize reputation damage. According to SCCT, central factors aff ecting stakeholders’ emotional, attitudinal, and behavioral responses to a crisis are the degree to which the organization is responsible for the crisis (crisis responsibility) and the organization’s crisis response strategy, which should match its crisis responsibility (Coombs 2007; Ma and Zhan 2016).

2.2 Attributed Crisis Responsibility Has Negative Consequences

According to Coombs ( 1998, p. 180), crisis responsibility can be def ined as “the degree to which stakeholders blame the organization for a crisis event.”. When the organization is seen as a victim of the crisis, it is minimally held accountable (e.g., in the case of natural hazard-related disasters or rumors). However, a preventable crisis, such as a humanerror accident, leads to strong attributions of crisis responsibility (Lee 2004). When the attributed crisis responsibility is high, several negative consequences follow: (1)Reduced trust and reputation.Several studies have found a negative eff ect of crisis responsibility on trust and organizational reputation (Coombs and Holladay 2007; Kim and Niederdeppe 2013; Ma and Zhan 2016) in private corporations and in public organizations. For instance, Bakker et al. ( 2018)revealed that high attribution of crisis responsibility leads to reduced trust in the local government, whereas trust in the local government scored higher when the government was not made accountable for the crisis. (2)Damaging behavior.When the attributed crisis responsibility is high, stakeholders are more willing to engage in damaging behavior,for example boycotting an organization (Coombs and Holladay 2007; Grappi and Romani 2015) or spreading negative word-of-mouth, which further damages the organization’s reputation (McDonald et al. 2010). (3)Potential decrease in supportive behavior.Studies show that the level of crisis responsibility does not inf luence stakeholders’ willingness to follow the local government’s advice (Bakker et al. 2018) or to share the organization’s crisis communication with others(secondary crisis communication, Utz et al. 2013). However,since high crisis responsibility has a negative eff ect on trust,it is conceivable that high crisis responsibility lowers people’s willingness to behave in ways that support the organization. (4)Negative emotions.The attribution of high crisis responsibility generally leads people to have more negative aff ect, especially more feelings of anger toward an organization (McDonald et al. 2010; Kim and Niederdeppe 2013; Utz et al. 2013), and it decreases people’s positive aff ect, such as joy or sympathy for an organization (Coombs and Holladay 2005; McDonald et al. 2010).

2.3 Emotions Partially Mediate the Eff ect of Crisis Responsibility

Emotions play an important role in explaining the eff ectiveness of crisis communication (Coombs and Holladay 2005),since emotions function as a type of information processing system that helps people f igure out how to view and respond to a specif ic situation (Loewenstein et al. 2001).Several studies suggest that emotions partially mediate between attributed crisis responsibility and people’s trust in an organization and behavioral intentions. On the one hand,having a negative aff ect damages the relationship between a stakeholder and an organization by reducing organizational trust and reputation (Choi and Lin 2009; Wang and Wanjek 2018); it also leads to more negative word-of-mouth comments and intentions to boycott an organization (Utz et al.2013). On the other hand, positive aff ect, such as sympathy,improves one’s attitudinal and behavioral responses toward an organization (Kim and Niederdeppe 2013; Grappi and Romani 2015).

2.4 Does Humorous Framing of Crisis Response Strategies Reduce Negative Emotions and Elicit Positive Emotions?

Several authors have suggested that using humor in crisis response strategies can decrease negative aff ect and increase positive aff ect (Vigs? 2013; Kim et al. 2016). The use of humor is expected to enhance stakeholders’ psychological coping mechanisms in crisis situations (Fredrickson et al.2003) because humor can work as a buff er, lowering stress levels and the intensity of and the focus on negative emotions, such as fear and anger. Instead, it elicits feelings of happiness or cheerfulness (Fredrickson et al. 2003; Martin 2007). These positive emotions then broaden people’s thinking, leading to a more f lexible and open-minded interpretation of a situation (Gulas and Weinberger 2006). Along that line of thinking, three decades of advertising research and even studies on ethically sensitive and threatening topics,such as climate change (Skurka et al. 2018), indeed reveal that humor inf luences emotional and cognitive processes,positively enhancing stakeholders’ attitudinal and behavioral reactions toward an organization (Eisend 2009, 2011). It leads to decreased counter-arguing and increased attention,liking of the source, purchase intentions, and generally to a more positive attitude toward a brand (Nabi et al. 2007;Eisend 2009).

According to these f indings, humor might be useful in crisis communication. Kim et al. ( 2016) found that humorous crisis communication strategies on social media, such as a self-mocking and mocking-the-accuser strategy, cause recipients to have a more positive attitude toward an organization than when rational, non-humorous strategies are used.Furthermore, several case studies have verif ied the positive impact of humor in crisis communication on the relationship between an organization and its stakeholders during or after a crisis (Vigs? 2013). In line with SCCT (Coombs 2007), humorously framed crisis responses should match the organization’s crisis responsibility. Xiao et al. ( 2018)revealed that using humorously framed crisis responses on social media is less appropriate when the organization is responsible for a crisis, whereas in the case of a rumor (that is, low crisis responsibility) it is more eff ective.

Nevertheless, humor may reduce source credibility(Eisend 2009), which could reduce its positive eff ect on stakeholders’ trust in an organization (Mayer et al. 1995).Furthermore, humor can trivialize the perceived seriousness of a topic (that is, a crisis), making stakeholders less willing to perform the supportive behaviors proposed during a crisis(Fraustino and Ma 2015). Hence, humor is inappropriate in serious crises, as it ref lects a lack of concern and empathy for the circumstances and decreases the perceived sincerity of an organization (Vigs? 2013; Xiao et al. 2018). However, it might be an eff ective tool in low-severity crises or in a paracrisis, def ined as publicly visible crisis threats and accusations against an organization (Coombs and Holladay 2012). Especially on social media, where informal language is preferred, a humorously framed crisis response to a lowseverity crisis could result in greater trust and contribute to the acceptance of the crisis response (Nabi et al. 2007;Kelleher 2009).

2.5 The Present Study

Humorous crisis communication on social media might be useful in low-severity crises. By reducing negative aff ect and eliciting positive aff ect, humorous crisis communication might have positive eff ects on people’s trust in an organization, their view of the organization’s reputation, and their behavioral intentions. In this study, using an experimental research design, we investigated how people respond to humor in a local government’s crisis communication (presented as tweets, a typical social media format) depending on the government’s crisis responsibility. We chose trust as the central outcome variable, as the value of information given by governments is strongly reduced if they are not trusted (Steelman and McCaff rey 2013). This, in turn, can signif icantly inf luence citizens’ behavior in crises (Rubin et al. 2009).

Considering the damaging effect of attributed crisis responsibility for organizations (Coombs 2007), we postulate that in low-severity crises, high (versus low) crisis responsibility leads to less trust in the local government(H1a), more negative aff ect (H1b), and less positive aff ect(H1c). With regard to the inf luential role of emotions in explaining the eff ectiveness of crisis communication (Choi and Lin 2009), we expect that the eff ect of crisis responsibility on trust is mediated by the experienced positive or negative aff ect (H1d).

Considering that humor positively inf luences people’s emotions and stakeholders’ relationships with organizations(Eisend 2009, 2011; Kim et al. 2016), our second hypotheses state that humorously (versus rationally) framed crisis communication leads to more trust in the local government(H2a), less negative aff ect (H2b), and more positive aff ect(H2c). Moreover, based on f indings by Eisend ( 2009, 2011),we expect that the eff ect of humorously framed crisis communication on trust is mediated by experienced positive and negative aff ect (H2d). Another question is whether the inf luence of humorously framed crisis communication on trust in the local government and on positive and negative aff ect is dependent upon whether the local government is responsiblefor the low-severity crisis (RQ1), since, to be most eff ective,crisis response strategies should match the organization’s crisis responsibility (Coombs 2007).

Fig. 1 Research design illustrating the hypotheses H1a?H1d and H2a?H2d and the research question RQ1

Fig. 2 Research design illustrating the research questions RQ2?RQ5

Considering the inf luence of crisis responsibility and the framing of crisis responses on stakeholders’ diff erent types of actions (Coombs 2007; Eisend 2009, 2011), we assume that crisis responsibility and the humorous framing of crisis response strategies inf luence people’s behavioral intentions toward an organization, but clear evidence for this is lacking, especially for behavioral intentions indicating that crisis communication was successful, such as people’s willingness to follow advice (Bakker et al. 2018) or secondary crisis communication (Utz et al. 2013). For this reason, we decided to examine exploratorily whether key actions–namely, people’s willingness to follow the local government’s advice,their willingness to seek information from the local government’s communication channels in crisis situations, and their secondary crisis communication–are inf luenced by the local government’s crisis responsibility (RQ2) and the framing of the crisis response strategy (RQ3). In addition, based on the conceptual model of SCCT, we wondered whether experienced positive or negative aff ect partially explain the eff ect that crisis responsibility (RQ4) and framing of the crisis response strategy (RQ5) have on people’s behavioral intentions. Our research design is visualized in Fig. 1 for H1, H2, and RQ1 and in Fig. 2 for RQ2?RQ5, respectively.

3 Methodology

This section presents the methodology used in this study,including information on the pre-studies conducted to test study materials, the participants and their recruitment, the study design and procedure, and the materials and measures.Additional information on these contents can be found in the online supplement (including instructions and measures).1https:// doi. org/ 10. 5281/ zenodo. 69831 91

3.1 Pre-Studies

Two pre-studies were conducted to select the stimulus materials for the main study. In pre-study I, 10 participants (100%female) between 19 and 56 years old (Mage=25.70,SDage=10.75) were presented with nine diff erent f ictive scenarios(randomly ordered) describing crises in local governments’areas of responsibility. Participants read two versions of each scenario, one where the local government was responsible for the crisis and one where it was not. After reading each scenario, participants rated the attributed crisis responsibility of the local government on a seven-point scale (a translated and adapted version of Lee 2004 ). A series oft-tests was carried out to check whether the local government was indeed considered responsible for the crisis in the respective scenarios.

In pre-study II, 13 participants (92% female) between 20 and 58 years old (Mage= 27.62,SDage= 12.88) saw 32 f ictive Twitter tweets posted by local governments in response to one of nine f ictive crisis situations. Each post was preceded by a brief description of the background of the f ictive crisis to which the post was related. Information on crisis responsibility was not given to avoid possible interaction eff ects between the government’s responsibility and the framing of the crisis response strategies (Xiao et al. 2018).Half of the tweets were humorously framed, while the other half were rationally formulated. Using an adapted version of Nabi et al.’s ( 2007) humor questionnaire, participants rated the perceived humor of each tweet on a seven-point scale.By using a series oft-tests or Wilcoxon signed-rank tests(if the assumption of normal distribution was violated) we compared whether humorous tweets were indeed perceived as funnier than rational ones. The three tweets perceived as the funniest, their rational equivalents and their respective scenarios were included in the main study.

3.2 Participants

The a priori power analysis indicated that a sample size of 225 would be suffi cient to detect signif icant eff ects in the planned parallel mediation models with a power of 0.80, an alpha of 0.05 and a medium eff ect size ofr= |0.30|, which could be roughly expected following existing f indings (Choi and Lin 2009; Eisend 2009; Kim and Niederdeppe 2013).Participants were recruited via two German online panels–PsyWeb2https:// psyweb. uni- muens ter. deand Fire Feedback3https:// f ire. uni- muens ter. de–and diff erent social media platforms ensuring participants’ familiarity with online communication platforms, such as Twitter. Participants had to be at least 18 years old and speak German as their mother tongue or second language. Participation was voluntary,anonymous, and incentivized by the possibility of receiving a research report after data collection. Of 936 participants who started the questionnaire, 542 completed the full survey.Data from 25 participants were excluded from the analysis because they did not speak German as their mother tongue or second language (3), they did not allow the use of their data (17), or they did not answer the questionnaires seriously (5), as identif ied by an unrealistic short overall survey response time under seven minutes and a wrong answer in the attention check (cf. Meade and Craig 2012). Thus, the f inal sample consisted of 517 participants (272 male, 236 female, 4 diverse, 5 not specif ied). Ages ranged between 18 and 74 years old (M= 43.67,SD= 15.97). Participants’highest completed educational levels were compulsory basic secondary schooling (4.4% of participants), a general certif icate of secondary education (13.0%), a general or specialized university entrance qualif ication (30.4%), a degree from a university or college of higher education (47.8%), or a doctorate (3.3%); the remainder (1.2%) had an unspecif ied school-leaving qualif ication. Of the participants, 9.1%pursued a career in local government, 5.4% in crisis management, 34.8% in f ields of f ire safety, rescue services, or disaster control, and 3.9% in the police force or similar security services; 57.8% worked in none of these areas.4Multiple choice was allowed. Some participants worked in more than one job sector.The f inal dataset and data from excluded participants did not diff er signif icantly in central characteristics, except for sense of humor,Z= 1,660.5,p= 0.048.

3.3 Study Design and Procedure

This study was preregistered on the open science platform As Predicted5https:// aspre dicted. org/ f8yy5. pdfand approved by the ethics board of the University of Münster’s Faculty 7, Psychology & Sports Science (ID 2021-41-MT). Data were collected on the online platform EFS Survey (Questback GmbH, 2017) between 25 May 2021 and 18 June 2021. The completion of the study took about 13 min on average (median = 12.40;M= 13.35 min,SD= 5.69). This study used a 2 (crisis responsibility:high versus low) × 2 (crisis response strategy: humorous versus rational) between-subjects design, resulting in four experimental conditions.6The four experimental groups consisted of 131, 131, 119, and 136 participants, respectively. They did not diff er signif icantly in central characteristics, except for education, p = 0.016, and sense of humor,F(3, 513) = 3.91, p = 0.009, η p2 = 0.02.All instructions, manipulations,and questionnaires were written in German. Participants were informed that the purpose of the 15-min study was to f ind out more about the eff ectiveness of local governments’crisis communication. No information on humorous crisis communication was given at the beginning.

3.4 Materials and Measures

The following section provides detailed information on the study materials and measures. Examples for the materials and scales used are given.

3.4.1 Crisis Scenarios and TweetsFictional crisis scenarios were used to avoid interference from local governments’ crisis histories and negative precrisis reputations that could have inf luenced the eff ectiveness of crisis communication (Coombs 2007). All crisis scenarios started with a short introduction where participants were asked to imagine that they lived in a f ictional city;then they were given a description of a low-severity crisis having a certain importance for the city’s residents. It was decided to choose scenarios that can basically occur anywhere and aff ect anyone so that the participants could easily relate to the situations described. The scenarios were about a road salt shortage after an intense snowfall, a dysfunctional speed camera, and dysfunctional ticket machines in the local transport system. At the end of each scenario, information on the crisis’ cause was given. The local government could be either held responsible or not responsible for the crisis.

Fig. 3 Examples of two humorous tweets and their corresponding rational tweets

In the respective tweets, the local government informed the public about the situation either in a rational or in a humorous way. All tweets used a subtype of the rebuilding response strategy (e.g., apology) (see Coombs 2007), and the same subtype was used in both versions of the tweet. No further details were given in the tweets concerning the cause of the crisis. Figure 3 illustrates examples of the tweets. We used a visualization based on Twitter as exemplary social media platform as Twitter is often linked to offi cial German news sites and can be accessed easily without a user account.

3.4.2 MeasuresThe scales we used to measure participants’ positive and negative aff ect, trust in the local government, and behavioral intentions, including willingness to follow advice,secondary crisis communication, and willingness to seek information, are depicted in Table 1. To check whether the manipulation worked, we measured the crisis responsibility attributed to the local government and the perceived humor of the tweets. Furthermore, we checked attention by including one item within the questionnaire measuring behavioral intentions; this item asked participants to select the specif ic answer “Agree” for quality control reasons (Meade and Craig 2012). We decided to measure two control variables–namely, participants’ propensity to trust and sense of humor–because individual diff erences in these two dispositions could have inf luenced the eff ectiveness of the study manipulation. More precisely, disposition to trust can signif icantly inf luence the perceived trustworthiness of others (Mayer et al. 1995; Beierlein et al. 2014), whereas interindividual diff erences in senses of humor can explain why people diff er in their sensitivity to humorous cues,in how much they like other humorous people and social interactions, and in expressions of mirthfulness and laughter(Svebak 2010).

4 Results

In this part, the results of the manipulation checks, the hypotheses testing, and exploratory analyses are presented.Detailed information on descriptive data and additional statistical analyses can be found in the online supplement.7https:// doi. org/ 10. 5281/ zenodo. 69831 91

4.1 Manipulation Checks: Attributed Crisis Responsibility and Perceived Humor

Two ANOVAs suggested that the experimental manipulations worked very well. Local governments were held more responsible for the crisis in the high-responsibility crisis scenarios (M= 5.61,SD= 0.94) than in the low-responsibility ones (M= 3.44,SD= 1.26),F(1, 513) = 495.90,p< 0.001,ηp2= 0.49, and humorously framed crisis response strategies were perceived as funnier (M= 4.25,SD= 1.70) than the rational ones (M= 2.23,SD= 1.50),F(1, 513) = 356.25,p< 0.001, η p 2 = 0.41.

Table 1 Measures of aff ect, trust, behavioral intentions, the manipulation check, and control variables

4.2 H1, H2, and RQ1: The Eff ect of Crisis Responsibility and Humorous Framing on Trust,Negative Aff ect, and Positive Aff ect

Hypotheses about the main effect of crisis responsibility(H1a?H1c) and humorous framing (H2a?H2c) and their interaction eff ect (RQ1) on trust, positive aff ect, and negative aff ect were tested with a MANOVA. Here, Pillai’s trace indicated signif icant eff ects of crisis responsibility,V= 0.04,F(3,511) = 6.76,p< 0.001, ηp2= 0.04 (1 – β = 1.00), humorous framing,V= 0.02,F(3, 511) = 3.81,p= 0.010, ηp2= 0.02(1 – β = 0.95), and a crisis responsibility × humorous framing interaction eff ect,V= 0.02,F(3, 511) = 2.97,p= 0.032, η p 2= 0.02 (1 – β = 0.95).

The follow-up ANOVA for trust yielded a signif icant result for crisis responsibility,F(1, 513) = 9.57,p= 0.002, ηp2=0.02 (1 – β = 0.90), suggesting that trust in the government was higher when the crisis responsibility was low (M= 4.78,SD= 0.95) than when it was high (M= 4.51,SD= 1.02). Furthermore, a signif icant main eff ect of framing was found,F(1,513) = 5.67,p= 0.018, ηp2= 0.01 (1 – β = 0.63), suggesting that trust in the local government was higher with the rational framing (M= 4.75,SD= 1.02) than the humorous framing(M= 4.53,SD= 0.96). Yet, there was no signif icant crisis responsibility × framing interaction eff ect on trust in the local government,F(1, 513) = 3.55,p= 0.060, η p 2 = 0.01.

The follow-up ANOVA for negative aff ect showed a signif icant eff ect for crisis responsibility,F(1, 513) = 12.14,p< 0.001, ηp2= 0.02 (1 – β = 0.90), suggesting that negative aff ect was higher when the crisis responsibility was high(M= 1.63,SD= 0.46) than when it was low (M= 1.49;SD= 0.44). Neither the main eff ect of framing,F(1, 513)= 0.53,p= 0.467, ηp2< 0.01, nor the interaction eff ect on negative aff ect,F(1, 513) =1.80,p= 0.180, ηp2< 0.01, were signif icant.

The follow-up ANOVA for positive aff ect revealed that crisis responsibility did not signif icantly inf luence positive aff ect,F(1, 513) = 3.41,p= 0.065, η p 2 = 0.01. However, it indicated a signif icant eff ect for framing,F(1, 513) = 5.96,p= 0.015, ηp2= 0.01 (1 – β = 0.63), suggesting that positive aff ect was lower with the humorous framing (M= 2.17;SD= 0.83) than with the rational framing (M= 2.35;SD=0.82). The interaction eff ect on positive aff ect was not signif icant either,F(1, 513) = 0.35,p=0.553, ηp2< 0.01. All eff ects found in the MANOVA and its follow-up ANOVAs are small according to Cohen ( 1988), who def ined η 2 < 0.06 as small eff ects.

4.3 H1d and H2d: The Mediating Role of Experienced Positive or Negative Aff ect on Trust

Fig. 4 Negative aff ect mediates the eff ect of crisis responsibility on trust. Representation of the unstandardized direct eff ects and the total eff ect (in brackets). a 0 = low crisis responsibility, 1 = high crisis responsibility. ** p < 0.01. *** p < 0.001.

Fig. 5 Positive aff ect mediates the eff ect of framing on trust. Representation of the unstandardized direct eff ects and the total eff ect (in brackets). a 0 = rational framing, 1 = humorous framing. * p < 0.05.*** p < 0.001.

Due to missing effects of crisis responsibility on positive aff ect and humorous framing on negative aff ect, we subsequently conducted simple instead of parallel mediation analyses. The f irst mediation analysis tested whether negative aff ect mediates the relationship between crisis responsibility and trust (H1d). We found a signif icant indirect eff ect of crisis responsibility mediated by negative aff ect,ab= ?0.12, 95% CI [?0.20, ?0.05],p< 0.001 (1 – β =0.93). High crisis responsibility led to higher negative aff ect,resulting in lower trust. The other path coeffi cients are presented in Fig. 4.

The second mediation analysis investigated the mediating role of positive aff ect on the relationship between framing and trust (H2d). It revealed a signif icant indirect eff ect of framing mediated by positive aff ect,ab= ?0.05, 95% CI[?0.10, ?0.01],p= 0.002 (1 – β = 0.70). Participants seeing the rational framings experienced more positive aff ect, and,in turn, participants who experienced more positive aff ect gave higher ratings of trust. The other path coeffi cients are presented in Fig. 5.

4.4 RQ2 and RQ3: The Eff ect of Crisis Responsibility and Framing on Behavioral Intentions

The main eff ect of crisis responsibility (RQ2) and humorous framing (RQ3) on variables measuring behavioral intentions were tested with a MANOVA. Here, Pillai’s trace indicated a signif icant eff ect of crisis responsibility,V= 0.04,F(6,508) = 3.39,p= 0.003, ηp2= 0.04, but not for framing,V= 0.02,F(6, 508) = 1.43,p= 0.201, ηp2= 0.02, nor for the crisis responsibility × framing interaction eff ect,V= 0.01,F(6, 508) = 0.74,p= 0.619, ηp2= 0.01.

Table 2 Mediation eff ect of crisis responsibility or framing on behavioral intentions by negative or positive aff ect, respectively

When the local government’s crisis responsibility was high, participants showed less willingness to follow the local government’s advice (high:M= 4.90,SD= 1.04; low:M= 5.15,SD= 1.04), they were less willing to follow local government’s advice as a matter of principle (high:M=5.58,SD= 1.35; low:M= 5.89,SD= 1.25), they were less willing to ask family and friends to follow advice (high:M=4.60,SD= 1.24; low:M= 4.83,SD= 1.21), and they were less willing to seek information on the local government’s website (high:M= 5.21,SD= 1.23; low:M= 5.55,SD=1.14). Moreover, the willingness to ask family and friends to follow advice was lower when a humorous (M= 4.59,SD= 1.26) rather than a rational framing was used (M= 4.83,SD= 1.19). All signif icant eff ects found in the MANOVA and its follow-up ANOVAs are small according to Cohen’s( 1988) conventions.

4.5 RQ4 and RQ5: The Mediating Role of Positive and Negative Aff ect on Behavioral Intentions

We tested whether negative or positive aff ect mediates the relationship between crisis responsibility or humorous framing and behavioral intentions (RQ4 and RQ5). Similar to the mediation analyses above, simple instead of parallel mediation analyses were conducted. Mediation analyses were only conducted for the behavioral intention variables that were signif icantly inf luenced by crisis responsibility or framing. All results of the mediation analyses can be found in Table 2; they show that aff ect fully mediated the eff ect of crisis responsibility on willingness to follow advice and secondary crisis communication but not on willingness to seek further information. Aff ect partially mediated the eff ect of framing on secondary crisis communication.

Table 3 Results of the hierarchical regression analysis predicting trust

4.6 Exploratory Analyses

In four exploratory analyses, we examined (1) further predictors of trust; (2) the moderating role of participants’demographics and characteristics; (3) the moderating eff ect of the crisis scenarios themselves; and (4) participants’ liking of the tweets depending on both experimental factors.Additionally, four exploratory analyses were examined.8In the interest of parsimony, only relevant signif icant eff ects are reported for the exploratory analyses. Complete results and eff ect sizes are in the online supplement ( https:// doi. org/ 10. 5281/ zenodo.69831 91).The results of the f irst exploratory analysis can be found in Table 3; these results showed that both control variables and some demographic variables contributed to the predication of trust. The second and third exploratory analyses indicated that some characteristics (e.g., disposition to trust) and the scenarios themselves moderated the eff ect of crisis responsibility and framing on trust. The forth exploratory analysis yielded a signif icant eff ect of framing,F(1, 513) = 8.39,p=0.004, ηp2= 0.02, and indicated that participants preferred rational framing (M= 4.42,SD= 1.23) rather than humorous framing (M= 4.07,SD= 1.44).

5 Discussion

Inspired by recent propositions that local governments and public authorities may want to use less formal crisis communication strategies on social media, this study off ers f irst insights into the eff ectiveness of humorous crisis communication on social media by public sector organizations in low-severity crises. It further highlights once again the importance of crisis responsibility in crisis communication.

5.1 Attributed Crisis Responsibility Evokes Negative Aff ect and Reduces Trust

According to our f irst hypothesis, high attributed crisis responsibility predicted a reduction in trust (H1a) and a rise in negative aff ect (H1b). This is in line with earlier f indings(Coombs 2007; Choi and Lin 2009; Bakker et al. 2018). We also replicated the mediating role of experienced negative aff ect in explaining the decrease in trust (H1d; see Choi and Lin 2009; Kim and Niederdeppe 2013). However, a local government’s high crisis responsibility did not reduce or increase positive aff ect (H1c). This stands in contrast to the f indings by Coombs and Holladay ( 2005), McDonald et al.( 2010), and Kim and Niederdeppe ( 2013). Consequently,the mediating eff ect of experienced positive aff ect on the relationship between attributed crisis responsibility and trust found by past researchers (Kim and Niederdeppe 2013;Grappi and Romani 2015) did not occur in the present study(H1d).

The lack of correlation between crisis responsibility and positive aff ect might have occurred because the study’s opposing eff ects neutralized each other: For example, while positive emotions directed to the organization, such as sympathy, are negatively related with crisis responsibility, the positive emotions associated with one’s own situation in a crisis, such as relief, are partially positively correlated with crisis responsibility (McDonald et al. 2010).

5.2 Humorous Crisis Communication Does Not Have a Benef icial Eff ect

Contrary to our second hypothesis, we did not f ind a positive eff ect of humor on trust (H2a) or on experienced emotions(H2b and H2c), such that emotions played no mediating role(H2d), even though the manipulation check showed that humorously framed crisis responses were indeed perceived as humorous. This pattern of results contradicts earlier studies in humor (Martin 2007) and advertising research (Eisend 2009, 2011), making clear that f indings on humorous communication from these research areas cannot be transferred to a public sector’s online crisis communication. Our results also disagree with f indings from private sector online crisis communication (Kim et al. 2016), but they are in line with more recent research by Honisch and Más Manchón ( 2020)and Xiao et al. ( 2018), who advised against using humorously framed crisis response strategies on social media, as they resulted in lower levels of organizational reputation compared with more rational ones.

The independence of perceived humor and emotional reactions found in our study can be explained by the benign violation theory of humor (Warren and McGraw 2015).Humor can violate a person’s well-being, normative beliefs,or identity, and the degree to which the violation is acceptable explains the extent to which negative emotional reactions are evoked (Warren and McGraw 2015). Especially in a crisis, humor can be perceived as inacceptable (Liu et al.2013), thus evoking less positive and more negative emotions. Several participants indeed commented at the end of the study that they considered humor inappropriate in a crisis. The exploratory analysis also conf irmed that participants liked the humorous tweet less than the rational one. However, it is interesting to recognize that only positive aff ect,not negative aff ect, was signif icantly inf luenced by the framing in this study. The overall low level of negative aff ect in the present sample might explain this.

Furthermore, several moderators might elucidate the results. For instance, the exploratory analysis revealed that individual diff erences in disposition to trust moderated the eff ect of humorous framing on trust. Additional potential moderators are participants’ prior image of local governments (Kim et al. 2016), which may have inf luenced perceived message credibility (Nabi et al. 2007), the social media culture itself (Kelleher 2009; Kim et al. 2016), and the quality of participants’ prior relationship with real local governments (Jahng and Hong 2017). Moreover, the signif icant scenario × framing interaction eff ect on trust in our exploratory analysis indicates that the eff ectiveness of humorously framed crisis response strategies depended on the specif ic context’s characteristics. One important contextual factor might be the individuals’ level of involvement with the crisis, which could aff ect the eff ectiveness of humorous communication (Yoon and Tinkham 2013).

Interestingly, we did not f ind a signif icant interaction eff ect of crisis responsibility and humorously framed crisis response strategies on trust in local government, positive aff ect, or negative aff ect. This contrasts with some basic assumptions of SCCT (Coombs 2007) and f indings by Xiao et al. ( 2018), which assume that crisis response strategies should match the organization’s crisis responsibility to protect the organizational image. However, our f indings are in line with research by Honisch and Más Manchón ( 2020),who found that humorous crisis communication on social media seems to be ineff ective regardless of an organization’s level of attributed crisis responsibility.

5.3 Behavioral Intentions

In exploratory analyses, we found that behavioral intentions were predicted by attributed crisis responsibility but not by humor. In line with earlier research (McDonald et al. 2010;Grappi and Romani 2015), high attributed crisis responsibility led to a decrease in supportive behavior. Humorously framed crisis communication, however, did not inf luence behavioral intentions. This ref lects recent ambiguous f indings in humor research (see, e.g., Liu et al. 2013; Honisch and Más Manchón 2020) and illustrates that public authorities should use humor with caution.

5.4 Limitations and Future Research

The experimental design is a key strength of our study.However, it imposes some limitations in terms of variables included and ecological validity, which, in turn, off er opportunities for future research. A f irst limitation is that we only tested the humorous framing of rebuilding response strategies using affi liative humor and rhyme. In terms of future research, it would be useful to extend current f indings by examining the humorous framing of diff erent crisis response strategies (Coombs 2007) and using diff erent humor types(Martin et al. 2003). As we only used a visualization based on Twitter as exemplary social media platform, it might also be interesting to examine whether there are any platformspecif ic eff ects. Further, we only investigated experienced positive and negative aff ect as mediating variables. As specif ic emotions associated with crisis responsibility inf luence attitudes and behaviors toward an organization diff erently,future studies should investigate the eff ects of aff ect beyond their valence (Raghunathan et al. 2006). Additionally, the results of the exploratory analyses revealed the need to further examine other, potentially opposing mediating variables, such as cognitive mechanisms (e.g., perceived severity), which might inf luence participants’ evaluation of their relationship with the local government as well as their decision to act (Moyer-Gusé et al. 2011).

It is unclear to what extent intentions, as measured in our study, would translate into actual behavior in real-life scenarios. According to the literature, behavioral tendencies do not always lead to actual behavior (Baumeister et al. 2007).Thus, in addition to intention itself, it may be important here to consider in parallel other key variables related to individuals’ ability and conf idence to act: For example, in a recent experiment on laypersons behavior in a low-severity crisis situation (the confrontation with an incipient f ire in a work setting), self-effi cacy beliefs were a signif icantly stronger predictor of behavior than intentions (Thielsch et al. 2021).However, some research in crisis communication indicates that behavioral intentions can be important drivers for actual behavior in crisis situations (Weyrich et al. 2020). Hence,future research should analyze the impact of humorous crisis communication in diff erent crisis types on both short- and long-term crisis-related behavior in f ield experiments while at the same time considering other key variables such as self-effi cacy.

We tested our experimental material in two pre-studies.Yet, the exploratory analysis revealed that the results were not independent of the specif ic scenario presented. On the one hand, this f inding can be considered a study limitation,in that we failed to create low-severity crisis scenarios with the same level of perceived crisis severity. On the other hand,these f indings off er promising avenues for future research.As already discussed, we see that contextual and individual factors (such as personal involvement) could strengthen or reduce the eff ects found.

Finally, one might criticize our use of SCCT (Coombs 2007) as a theoretical framework for investigating communication during a crisis, although results from this study clearly support its postulated structure. Coombs ( 2007) himself described SCCT as a post-crisis communication theory designed to repair an organization’s reputation after having addressed the physical and psychological threats of crisis victims. However, Olsson ( 2014, p. 117) opposed this by claiming that dividing crisis responses into ones that provide instructing information and ones that address reputational concerns is “somewhat artif icial,” as an organization’s crisis communication and actions during the whole course of a crisis inf luence the relationship with the stakeholders. Scholars (e.g., Liu et al. 2012) have started to establish models specifying which crisis response strategies should be used in a specif ic phase of a crisis, but empirical research on critical response time in crisis scenarios is limited. It might be interesting to explore whether humorous online crisis communication works better after a crisis rather than during the incident and how knowledge on a local government’s crisis responsibility aff ects its eff ectiveness at that specif ic point in time.

5.5 Practical Implications

Our core message for local governments’ crisis communication managers is as follows: If your local government is responsible for a crisis, do not use humorous framing in crisis communication on social media, even in low-severity crises. For humor to have a positive eff ect, crisis communication managers would have to accurately consider many factors, such as the specif ic characteristics of the crisis (e.g.,crisis severity), the local government (e.g., prior image), and the target group (e.g., involvement). Additionally, the style of humor (Martin et al. 2003) should be chosen wisely. If crisis communication managers choose poorly, this can cause a loss in trust, induce negative emotions, and encourage the circulation of rumors (Park et al. 2012), which further damage citizens’ relationships with the local government.Because local governments are required to communicate quickly about crises via social media, crisis communication experts recommend preparing adequate crisis responses for diff erent incidences in advance (Claeys and Opgenhaffen 2016). As such, it is questionable whether integrating humorous elements in formal online crisis communications might be eff ective, as humor is often found in the surprising and funny aspects of current circumstances, which cannot be known beforehand. Therefore, it might be helpful to use humorous elements only in post-crisis communication when the relationship between citizens and the local government has to be restored or improved.

6 Conclusion

In summary, this study conducted the f irst experiment on the eff ectiveness of humor in local governments’ crisis communication via social media, its conditions, and working mechanisms. Our results mainly argue against the use of humor in online crisis communication because it negatively inf luences trust and positive aff ect. However, if the application is well planned and both the target group and framework conditions are suitable, humorous crisis communication on social media might be appropriate under specif ic circumstances (e.g., for post-crisis communication, low crisis involvement of the target group) in low-severity crises. We hope that the current study will motivate further investigation of possible moderating and mediating variables, such as quality of emotions and perceived crisis severity. Furthermore, the present research contributes to a growing body of evidence for the SCCT (Coombs 2007), suggesting that an organization’s level of crisis responsibility is one of the most important factors in crisis communication, as high levels of crisis responsibility negatively inf luence aff ect, trust, and behavioral intentions. If a public organization has caused a crisis, humorous communication cannot compensate for the perceived negative impact of this responsibility.

AcknowledgmentsWe thank Gerrit Hirschfeld and Simon Wilms for their statistical support as well as Marie Kersting for her creative input in developing the study materials. Further, we thank Celeste Brennecka for her very helpful comments on earlier versions of this manuscript.

Open AccessThis article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source,provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http:// creat iveco mmons. org/ licen ses/ by/4. 0/.