ABSTRACT
In the context of climate change communication, this study explores the process through which exposure to media messages about a risk leads to recommended behavioral intentions. We propose a model of this process based on the Extended Parallel Process Model (EPPM) and the Risk Perception Attitude (RPA) framework. Our model analyzes how risk perception, negative emotion, and efficacy beliefs mediate and moderate the effects of media messages on people’s intention to engage in pro-environmental behaviors. A national survey among 1,000 adults in South Korea was analyzed, and the fitting of PROCESS Models 4 and 15 yielded four main findings. First, media exposure was directly and positively related to risk perception, negative emotion, and pro-environmental behavioral intention. However, the significant relation between media exposure and behavioral intention was partly conditional upon efficacy beliefs. Second, risk perception and negative emotion were also significantly related to behavioral intention conditional upon efficacy beliefs. Third, efficacy beliefs significantly moderated the relation between risk perception and behavioral intention, but not between negative emotion and behavioral intention. Fourth, efficacy beliefs served as a moderator for the indirect effect of media exposure on behavioral intention via risk perception and negative emotion.
Disclosure statement
No potential conflict of interest was reported by the authors.
Notes
1. We originally used four perceived severity items that included “The impact of climate change is very worrying” and four negative emotion items that included “worry.” However, a reviewer raised an important question about potential multicollinearity and threat to construct validity. To address these issues, we performed the following additional statistical procedures: (a) multicollinearity diagnostics in a multiple regression model using SPSS; (b) Exploratory Factor Analysis (EFA) with varimax rotation for the two variables – risk perception and negative emotion; and (c) Confirmatory Factor Analysis (CFA) for the two latent variables (risk perception and negative emotion) and the original nine observed indicators using LISREL 8.8. Results are as follows. First, we detected no multicollinearity problems (tolerance = .73, VIF = 1.37) based on the conventional guidelines (see Thompson et al., Citation2017). Second, EFA clearly indicated two-factor solutions (first factor among five risk perception items = 43.68% of total variance explained; second factor among four negative emotion items = 38.52% of total variance explained). However, our CFA results showed that the “worry” item was highly correlated to both latent variables of risk perception and negative emotion, thereby failing to meet several cutoff criteria for goodness-of-fit indexes (Hu & Bentler, Citation1999). After removing the item (“worry”) in accordance with a modification index and theoretical reasoning, the goodness-of-fit indexes met the cutoff criteria, χ2 (12) = 69.89, p < .001, RMSEA = .07, NNFI = .99, SRMR = .03, GFI = .99, CFI = .99 (for traditional cutoffs, see Hu & Bentler, Citation1999; Kline, Citation2016). In addition, results showed that all standardized factor loadings were greater than .7 and therefore met the criteria for securing convergent validity (Hair et al., Citation2009). Also, the correlation between the two latent factors was .64, meeting the criteria of .75 or lower for not ruling out discriminant validity issues (Voorhees et al., Citation2016). Based on these tests, we conclude that no evidence was found to threaten the construct validity of our measurement. We would like to thank the reviewer for noticing this issue.