Mobilizing the Public in Saving the Bonneville Salt Flats: Understanding Blame as a Psychological Construct

ABSTRACT

This study seeks to explicate blame as a psychological construct. Situated in the environmental problem of the decline of the Bonneville Salt Flats (BSF), the study examines the structure of blame and how it may mediate message effect on the public’s issue engagement to save the BSF. With data collected from a Web-based experiment on a college student sample, blame was shown to be best modeled as a latent construct consisting of both the cognitive component of responsibility judgment and the affective component of anger. Blame was stronger when the scientific research findings were communicated in certain vs. uncertain language and was in turn a predictor of behavioral intentions to protect the BSF, including punitive attitudes against the perceived wrongdoers, information seeking and sharing activities, and civic participatory behaviors. Our study offers a psychological account of blame as a useful explanation of the effect of environmental communication on issue engagement.

KEYWORDS:

• Blame
• responsibility
• anger
• scientific uncertainty
• public engagement

Acknowledgements

The authors wish to thank Ryan Kor-Sins (PhD candidate, University of Utah) for her assistance with the project.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Notes

1 Although in everyday language we can also say “We blame the rain for the delay,” we only consider blame as arising from events caused by human agents.

2 In the blame literature, responsibility mostly refers to moral accountability, as opposed to legal responsibility (see Shaver, 2012).

3 Project funded by National Science Foundation: Adaptation, Mitigation, and Biophysical Feedbacks in the Changing Bonneville Salt Flats: NSF Coupled Natural Human Systems, Principal Investigator: Brenda Bowen, University of Utah, 2016–2020.

4 The experimental study involved other two factors: positive images (yes vs. no), and inclusion of negative images (yes vs. no). These two factors were included for questions not addressed in this paper and were shown to have no effects on the variables examined in this study. We analyzed the effects of the three factors (all the 3-way and 2-way interactions and main effects) on all the endogenous variables used in the SEM analyses reported in this study: perceived certainty, anger, responsibility attribution, information activities, civic participation, and prohibitive attitudes. On each variable, we conducted a MANOVA to examine the multivariate effects, followed by univariate analyses. From these analyses, the only significant effects were the main effect of certainty vs. uncertainty manipulation. None of interaction effects or the other main efforts were significant. We therefore proceeded by only analyzing the language certainty as the experimental factor.

Given the design, the required sample size was estimated via G*Power for a power of .80 at two effect sizes, f = .15 and f = .20, picked between the small (f = .10) and medium (f = .25) effect sizes. The estimated required sample size was n = 199 for f = .20 and n = 351 for f = .15.

5 To add degrees of freedom to the model for model comparison, certainty manipulation was put as an exogenous variable leading to either two latent factors consisting of two items each, or one latent factor with four items. The two-factor model showed a poor fit: χ²(4) = 84.13, p < .001, RMSEA = .255, CFI= .780, SRMR= .167. One factor model was a better fit: χ²(5) = 33.42, p < .001, RMSEA = .136, CFI = .922, SRMR = .050. Including correlated error between the two items of scientific evidence being conclusive or established further improved the fit of the one-factor model: χ²(5) = .59, p = .96, RMSEA = .000, CFI = 1.00, SRMR = .007.

6 To reduce demand characteristics, these three items were embedded in a randomized list with other emotions such as sad, surprise, happy, depressed. To ascertain that the item “disgusted” belonged to the same scale as “angry” and “outraged,” we ran confirmatory factor analyses (CFA). Since there were only three items in this theorized scale (hence a just-identified model), we introduced two additional emotional measures, “sad” and “depressed,” to construct models for comparison. One measurement model had two factors: anger (“angry,” “outraged,” “disgusted”) vs. sadness (“sad,” “depressed”). The competing model contained three factors: anger (“angry,” “outraged”), disgust (“disgusted”), and sadness (“sad,” “depressed”). As the two models were not nested, the BIC difference was used to assess model comparison. The two-factor model (BIC = –18.39) was shown to be a superior model than the three-factor model (where “disgusted” was its own factor, BIC = –7.28). This BIC difference (11.11) is regarded as strong evidence in favor of the two-factor model. Hence, the CFA results demonstrated that “disgusted,” “angry,” and “outraged” formed one factor.

7 The competing models in Figure 2 were tested as well though not reported here. The BIC value for the integrated model with blame as a latent construct was –20.13. The BIC difference was 17.54 against the Dual Process model, and above 80 for the two serial causal models. Again, the evidence strongly favored the intertwined model of blame.

8 This rather high coefficient bespeaks the potential lack of discriminant validity; the operationalization of punitive attitudes needs to be refined in future research.

Additional information

Funding

This work was supported by the National Science Foundation [grant number 10040652].