«13TH INTERNATIONAL PUBLIC RELATIONS RESEARCH CONFERENCE “Ethical Issues for Public Relations Practice in a Multicultural World” Holiday Inn ...»
Positive emotions. Although negative emotions have been the primary focus of crisis research (e.g., Choi & Lin, 2009; Coombs, Fediuk, & Holladay, 2007; Jin et al., 2007; Kim & Yang, 2009), positive emotions have received far less attention in the literature. The role of positive emotions in a crisis has been largely neglected since they are believed to be less intense and less enduring compared to negative emotions (Folkman et al., 1997). However, growing number of researchers emphasize that people, even in the most difficult circumstances, experience positive emotions which play an important role mitigating negative impact of crises (Folkman & Moskowitz, 2000).
Fredrickson et al. (2003) found that positive emotions indeed co-occurred alongside negative emotions. Gratitude, interest, and love were the three most frequent positive emotions drawn by the September 11 tragedy (Fredrickson et al., 2003). As described above, many people experienced a feeling of alertness after the first Mattel toy recall. This emotion was found to be a significant predictor of the company’s perceived reputation (Choi & Lin, 2009).
Positive emotions not only provide more pleasant subjective experiences than negative emotions, but they also help reduce an exclusive focus on negative emotions (Fredrickson et al., 2003). Fredrickson and Joiner (2002) suggested that positive emotions work as a “breather” by undoing physiological arousal and enhancing broadminded coping. In contrast to negative emotions, which narrow people’s attention to specific action tendencies (e.g., attack), positive emotions tends to broaden attention, thinking and behavioral responses (Fredrickson, 1998, 2000). Positive emotions can aid an organization by allowing stakeholders to be more flexible in interpreting a crisis situation and more open-minded in processing relevant information (Reed & Aspinwall, 1998; Trope & Pomerantz, 1998).
429 Our review of previous research on emotional responses to crisis situations leads to our first two research questions. First, building on studies demonstrating both negative and positive
emotional responses to crises (e.g., Choi & Lin, 2009; Fredrickson et al., 2003), we ask:
RQ1: Which discrete emotions will students experience in response to an H1N1 influenza outbreak on a college campus?
Second, based on Kim and Yang’s (2009) proposition and the revised model of SCCT (Choi & Lin, 2009), we ask:
RQ2: Do positive and negative emotions mediate the relationship between crisis responsibility and both relational trust and intention to seek crisis-related information?
A Model of Emotion-Based Interpretation of a Crisis Based on the literature review, this study proposes the model of emotion-based interpretation processes of a crisis to explain the linkages among the areas of both positive and negative emotions, judgment of crisis responsibility, relational trust, and intention to seek information. Figure 1 presents the hypothesized direct paths in the proposed model. Crisis responsibility and relational trust are at the center of the dynamic. Crisis responsibility shapes both positive and negative emotional reactions, and has direct and/or indirect influence on relational trust and intention to seek crisis-related information. Relational trust is posited as a mediator between emotional responses and information seeking intention.
RQ3: Will the proposed model achieve reasonable fit in explaining stakeholders’ interpretation of an H1N1 influenza crisis?
Method An online survey was conducted at a large-sized private university in upstate New York to collect data used in this study. In the fall 2009 semester, cases of H1N1 influenza increased rapidly across the campus and even caused the death of a student, producing widespread media coverage and campus concern. At the time of the survey being conducted, more than 500 students had been suspected of being infected by the H1N1 influenza virus and the story was reported in the local and national news media. The university and its health center established a website to provide relevant information about flu prevention and shut down several campus social events.
Participants The sampling frame consisted of 25% of total undergraduate students (n=3,495) randomly selected from a list of student email addresses provided by the university registar’s office. To make sure that the sample has representation from various student groups, participants were randomly chosen in proportion to the share of each school year group (i.e., 25% Freshmen, 25% Juniors, 25% Sophomores, 25% Seniors). Each sampled email address was sent an email invitation on October 23, 2009, which linked to a survey questionnaire on the Web. Participants were asked to complete a questionnaire that included items about demographics, their perception of crisis responsibility and relational trust toward a college health center, and intention to engage in crisis-related information seeking. A reminder email was sent on October 29, and the survey was closed on November 2, 2009. Among a total of 429 questionnaires that were initiated (12.2%), 327 participants completed the entire questionnaire (a 9.4% response rate).
Respondents consisted of 70% women (n=231) and 30% (n=100) men. Age of respondent varied from 17 to 40, with an average age of 20. Of the respondents, 23% (n=77) were Freshmen, 26% (n=88) were Juniors, 21% (n=70) were Sophomores, and 29% (n=98) were 430 Seniors. Two out of three (n = 225; 68%) were White, 20% (n = 67) were Asian/AsianAmerican, and 11 % (n = 36) were Black, Hispanic or multiple races. The study was approved by the university’s Institutional Review Board (IRB).
Measurement Instrument Four theoretical constructs were measured to test research questions and the proposed model. One variable was considered exogenous (i.e., crisis responsibility) while others were treated as endogenous variables (i.e., negative and positive emotions, relational trust, and intention to seek crisis-related information) in the proposed model.
Positive and negative emotions. Modified Izard’s (1977) Differential Emotions Scale (DES) was used to assess discrete emotions that students experienced during the influenza pandemic on a campus. Four discrete negative emotions and four positive emotions were measured: (a) negative: anger, fear, sad, and anxious; (b) positive: grateful, interested, hopeful, and love. These emotions were selected because they were found to be most frequently experienced emotions in crisis cases (e.g., Fredrickson et al., 2003; Jin et al., 2007, 2008).
Referencing Fredrickson et al. (2003), we also added an item to measure sympathy. On a 5-point scale from 1 never to 5 most of the time, participants were asked to “think back to the beginning of this semester when cases of H1N1 influenza (“the swine flu”) were increasing rapidly across campus” and report on how often they had felt each of 9 different emotions (e.g., “I have felt grateful, appreciative or thankful”).
We examined the discriminant validity of the two groups of emotions (i.e., negative and positive emotions) using principal component analysis with Varimax rotation. Based on the Kaiser’s rule (Tabachnick & Fidell, 2001), any component with an eigenvalue greater than or equal to 1 was extracted. Two components were extracted for the nine discrete emotions. Each of the extracted components clearly corresponds to the theoretical categorization of emotions: The first factor is positive emotion (eigenvalue=2.97; 32.9% variance explained) and the second is negative emotion (eigenvalue=2.57; 28.5% variance explained). While sympathy was correlated with the composite score of negative emotions in a previous study (Fredrickson et al., 2003), our analyses revealed that sympathy is more closely related to positive emotion in the context of an influenza outbreak case (See Table 1).
Accordingly, subscales for positive and negative emotions were created. The Positive Emotions subscale is a composite of 5 positive emotions (including sympathy), with coefficient.82. The Negative Emotions subscale is a composite of four negative emotions, with coefficient.78.
Crisis responsibility. On a scale from 1 to 7, where 1 means not at all responsible and 7 means totally responsible, crisis responsibility was measured with the following three items: how much responsibility should (a) the university health center, (b) other organizations on campus (for example, Campus Life, Administration, or Facilities), and (c) students themselves bear for reducing transmission of H1N1 influenza (“the swine flu”) among students? Items (a) and (b) were used to measure organizational attribution and (c) was to assess personal attribution towards the issue. Cronbach’s alpha of the item (a) and (b) was.82 (n=330; 2 items).
Relational trust. Referencing previous studies (e.g., Yang, 2007; Yang & Lim, 2009), six 5-point scale items were used to measure three dimensions of relational trust: competence, dependability, and integrity. The measurement items were slightly modified to make them more relevant to the organization being studied (i.e., a college health center). Two items were used for 431 each dimension (e.g., competence, “I feel very confident about [the health center’s] expertise in the health needs of college students”; dependability, “[The health center] can be relied upon to provide the health information and resources I need”; and integrity, “Whenever [the health center] makes an important decision, I know it will be concerned about students like me”).
Cronbach’s alpha of the items were.92 (n=328; 6 items).
Intentions to seek crisis-related information. To measure intentions to seek information, three items were developed modifying items from Grunig and Hung’s (2002) questionnaire.
While original item includes information searching only on the websites of organizations, this study added two social media platforms (i.e., Facebook and Twitter) considering participants’ media usage tendency. The three items are (a) I would join a Facebook group of [The health center] that provides information about the flu on campus, (b) I would follow [The health center] on Twitter to learn about the flu on campus, and (c) I am interested in learning more information about influenza on [The health center’s] website. Cronbach’s alpha of the items was.64 (n=327;
Results Dominant Emotions Experienced During an Influenza Pandemic: Test of Research Question 1 The first research question was posed to identify dominant emotions felt during an influenza outbreak on a college campus. The most frequently experienced emotion was the feeling of “interest” (M=3.22, SD=1.06) which was grouped with “alert” and “curious,” followed by “sympathy or compassion” (M=3.14, SD=1.0), and “anxious, worried or concerned” (M=2.80, SD=1.15). Table 1 presents the result of descriptive statistics for each emotion.
As for the latent variables of positive and negative emotions, “scared, fearful or afraid” (B=.82, p.001) and “anxious, worried or concerned” (B=.84, p.001) were the strongest indicators of negative emotions, while “grateful, appreciative or thankful” (B=.80, p.001) was the strongest indicator of positive emotion followed by “love, closeness or trust” (B=.79, p.001), “hopeful, optimistic or encouraged” (B=.75, p.001) (See Table 3).
Structural Equation Modeling: Test of Hypotheses and Research Question 2 & 3 To empirically test theoretically derived paths in the model, structural equation modeling (AMOS 6.0) was used. Parameters were estimated by maximum likelihood method and a twostep process of latent path modeling was followed. Imposing all factors in the proposed model to covary, an initial confirmatory factor analysis (CFA) was conducted. The initial measurement model fitted satisfactorily, we did not revise the model.
A model can be retained as a valid model when (a) the value of χ2/df is less than 3, (b) the value of CFI is equal to or greater than.90, and (c) the value of RMSEA is equal to or less than.08 (Byrne, 1994, 2001; Kline,1998). Following this guideline, the proposed CFA model is a valid model indicating good measurement reliability and validity of the variables: χ2/df = 2.57, CFI =.92, and RMSEA=.06.
All latent variables were significantly correlated with each other except for the correlation between relational trust and crisis responsibility. Other latent variables were correlated at least at.05 level ranging from r = -.112 between negative emotion and relational trust, as the lowest correlation, and r =.415 between positive and negative emotions, as the highest correlation. Table 2 reports correlation and descriptive statistic results.
Regarding RQ3, structural relations were constructed among four variables based on research hypotheses proposed in this study (See Figure 1). The proposed structural equation model yielded the following data-model fits: χ2/df = 2.81, CFI =.904, and RMSEA =.065.
432 Thus, we concluded that the proposed model was a valid model that explained patterns of association between variables involved in the interpretation process of a crisis.
Direct effects in the proposed model. Among nine paths that explain the mediators (i.e., positive and negative emotions and relational trust) and the dependent variable (i.e., intention to seek crisis-related information), six significant paths (p.05) were found. Crisis responsibility was a significant predictor for both negative (H1a, β =.37, B =.43, S.E. =.06) and positive emotions (H1b, β =.25, B =.29, S.E. =.06, both at p.001). Only positive emotion was a significant predictor for relational trust (H2b, β =.17, B =.19, S.E. =.06, p=003). Negative emotion was found to predict information seeking intention (H3a, β =.14, B =.17, S.E. =.06, p=.03). Contrary to our expectation, crisis responsibility was not a significant predictor for relational trust (H4, β = -.04, B =-.05, S.E. =.06, p=ns). Both relational trust (H5, β =.39, B =.41, S.E. =.07, p.001) and crisis responsibility (H6, β =.19, B =.26, S.E. =.06, p =.002) were significant predictors for intention to seek crisis-related information (See Table 4).