Stigmatizing ‘evildoers’: how beliefs about evil and public stigma explain criminal justice policy preferences

ABSTRACT

Characterizations of offenders as ‘evil’ appear in popular entertainment, political rhetoric, and scholarship and may be rooted in widely endorsed cultural narratives. Integrating psychological and criminological literatures, we argue that endorsement of good-and-evil myths may be associated with criminal justice policy support via stigmatizing attitudes about those who engage in crime. We partially tested these hypotheses in a preliminary study using an online convenience sample (recruited using Amazon MTurk in 2015), then tested the full theoretical model using more comprehensive measures in a survey of American adults conducted via Qualtrics Panel in 2021 (N = 1,162). We found that both belief in evil (i.e. belief in evil forces) and belief in redemptive violence (i.e. belief in a clash between good and evil) were positively associated with punitive policy support, and stigmatizing attitudes partially mediated the effects of both. While belief in redemptive violence was negatively associated with rehabilitative policy support and stigmatizing attitudes mediated that relationship, belief in evil was not associated with support for rehabilitation. We consider the implications of our findings for mitigating the harms of public stigma and punitive criminal justice policy.

KEYWORDS:

• Beliefs about evil
• public stigma
• policy preferences
• punishment
• rehabilitation

Open Scholarship

This article has earned the Center for Open Science badge for Open Data. The data are openly accessible at https://osf.io/bk9jf/?view_only=fc7f4cd974104fec96082a6f9a7dd765.

Disclosure statement

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

Notes

1 It is important to note that while many religions invoke concepts of good and evil, beliefs about good and evil are not specific to a particular religion, nor are they exclusive to religious people (e.g., Bastian et al., 2015). As well, beliefs in evil and redemptive violence predict policy views independently of religious fundamentalism (Campbell & Vollhardt, 2014). Thus, endorsement of good-and-evil narratives should not be considered synonymous with endorsement of religious beliefs emphasizing the concept of evil.

2 As an anonymous reviewer insightfully pointed out, there are parallels between belief in redemptive violence and feelings of retribution, as both belief in redemptive violence and retribution as a punishment goal may justify the use of violence as a response to perceived immorality. In the current study, we take the view that belief in redemptive violence reflects endorsement of a belief system, while feelings of retribution more likely reflect punitive responses to specific transgressions; indeed, research suggests that punishment severity preferences track feelings of retribution closely (Carlsmith et al., 2002). However, this parallel raises interesting questions about the potential relationships between beliefs in redemptive violence and preferences for retribution in punishment (e.g., whether belief in redemptive violence might elicit preferences for retribution in specific cases), and we urge researchers to explore such questions in future work.

3 To the extent that good-and-evil narratives parallel the ‘crime master narrative’ in their emphasis on the existence of evil agents, the current study may be viewed as a partial test of Haney’s (2020) framework, particularly with regard to the dispositional attribution dimension of stigma.

4 Another study (Hickert et al., 2024) was also published using the 2021 Qualtrics data. However, this study did not incorporate beliefs about evil or use the novel stigma scale developed in Shi et al. (2022).

5 Neither study was pre-registered. Data from the preliminary study sample (MTurk) and main sample (Qualtrics Panel) are available at https://osf.io/bk9jf/?view_only=fc7f4cd974104fec96082a6f9a7dd765.

6 Recent findings show that online survey respondents became less attentive during the pandemic, possibly due to the COVID-19 interruptions (Peyton et al., 2022; Ternovski & Orr, 2022).

7 In our sample, failure to pass the attention check was significantly associated with political party affiliation, race/ethnicity, age, income, and region; descriptive statistics for the full sample are provided in Appendix B.

8 Consistent with research validating these scales (Campbell & Vollhardt, 2014), exploratory factor analysis including the items for both measures indicated that belief in belief in evil and belief in redemptive violence demonstrated discriminant validity (i.e., loaded on distinct factors). While the redemptive violence items loaded on a single factor (all loadings > .44), the belief in evil items were split between two correlated (r = .55) factors (all loadings > .53). Given that the two items that loaded on a second factor had a different wording structure than the other three items (i.e., which all read ‘There are __ in this world that are evil’), we interpreted this multidimensionality as likely reflecting a methodological artifact and retained a single measure.

9 As noted above, data used for the current study were also used to develop and validate this scale.

10 A commonly cited minimum appropriate level of alpha is 0.7. While the reliabilities for our punitive policy support measure falls below this benchmark, it is important to note that alpha levels are lower for scales with few items, as is the case for this measure (Cortina, 1993). More generally, methodological work recommends against applying hard-and-fast rules to reliabilities, particularly when values approach the 0.7 benchmark (Cho & Kim, 2015). While we retain this measure for its theoretical utility, we note that lower alpha values may reduce the predictive precision of the measures, and we urge readers to interpret the results accordingly.

11 We used the user-written ‘sgmediation2’ command in Stata (see https://www.trentonmize.com/software/sgmediation2). This command provides equivalent results to the PROCESS macro available for SAS, SPSS, and R (Hayes, 2018).

12 Percent reduction was calculated as follows and reflects the percent reduction in the regression coefficient following the introduction of the mediator: 100*(total effect – direct effect)/total effect. For example, for the belief in evil coefficients in Models 2 and 3, the percent reduction was calculated as: 100*(.25−.19)/.25 = 24%.

13 In the sensitivity analysis using the full sample (Appendix B), belief in evil was significantly and positively associated with support for rehabilitation (b = .09, p = .002); however, we urge readers to interpret this unexpected effect with caution given that it appears to be driven by less attentive respondents.

Additional information

Funding

The current study is funded using internal funds from Rutgers University—Newark, Bridgewater State University, and the University at Albany, SUNY.