Investigating the role of group-based morality in extreme behavioral expressions of prejudice

The study addresses why certain perceived threats lead to acts of hate such as hate speech, hate group activity, and violence—collectively termed extreme behavioral expressions of prejudice (EBEPs). Prior work links EBEPs to perceived intergroup threats and ideological dispositions (e.g., right-wing authoritarianism, social dominance). The authors propose the moralized threat hypothesis: EBEPs are often motivated by beliefs that an outgroup has committed moral violations, with susceptibility shaped by individuals’ moral values. Using Moral Foundations Theory, they focus on binding values (loyalty, authority, purity) tied to group preservation and hypothesize these values predict EBEP justification, particularly for targets marginalized by right-wing ideologies.

Research has documented rising hate crimes and hate group activity in the U.S. and Europe. Intergroup threat theory links realistic and symbolic threats to prejudice, mediated by authoritarianism and social dominance. Fiske and Rai argue many violent behaviors are morally motivated. Moral Foundations Theory (MFT) distinguishes individualizing (care, fairness) and binding (loyalty, authority, purity) values; conservatives generally endorse binding values more than liberals. Prior geographical and criminological studies relate structural covariates to hate group distribution. This work integrates these strands, proposing that moralization of perceived threats—especially binding values violations—underpins justifications for EBEPs.

Five studies were conducted. Study 1 (geospatial analysis): County-level moral values were estimated from 2012–2018 YourMorals.org data (N=106,465) via Multilevel Regression and Synthetic Poststratification (MrsP) with spatial smoothing to derive county scores for individualizing and binding values. County-level hate group counts (2012–2017) were obtained from the SPLC and modeled as rates per 10,000 inhabitants using negative binomial regression with Bayesian Spatial Filtering, adjusting for poverty, education, racial composition, 2016 Democratic vote share, and rural/urban status; sensitivity analyses included state fixed effects using GLMs. Validation of MrsP used county conservatism estimates compared to 2016 Republican vote share. Studies 2–3 (observational surveys): Participants completed the Moral Foundations Questionnaire (MFQ) to measure binding and individualizing values, rated perceived moral wrongness (PMW) of outgroup behaviors (Muslims “spreading Islamic values” in Study 2; Mexican immigrants “taking jobs” in Study 3), and judged justification for four EBEPs (Facebook hate speech, flyer hate speech, verbal assault, physical assault). Analyses used hierarchical Bayesian cumulative logistic regression with varying intercepts for participants and items and varying slopes for conditions. Mediation by PMW was assessed via Bayesian posterior simulations. Samples: Study 2 N=511 (Qualtrics panel, stratified), Study 3 N≈324 (post-exclusions) from MTurk. Study 4 (experiment): MTurk participants (analyzed N=294) were randomized to high vs. low moral threat vignettes about a fictional outgroup (“Sandirians”) framed as harming vs. helping the local economy. PMW and EBEP justification were measured as in earlier studies; models mirrored hierarchical ordered logistic regression. PMW’s mediating role in the effect of condition on EBEP justification was tested via posterior simulation. Study 5 (experiment with moderated mediation): Qualtrics national panel participants (N=1026 post-QC) were randomized to control, binding-violation (purity/convention violations via graphic sexual ritual), or individualizing-violation (harm/care violation via eating pets) vignettes about a fictional outgroup (“People of the Earth”), designed to elicit moral violations without direct material threat. Measured PMW, EBEP justification, religiosity, ideology, and MFQ. Bayesian regressions tested moderated mediation: PMW mediating condition effects on EBEP justification, moderated by binding or individualizing values (treatment susceptibility). Effects were summarized as Average Causal Mediation Effects (ACME) and Average Direct Effects (ADE), evaluated at low (−1 SD) and high (+1 SD) levels of the moderator. Across studies, ideological covariates were included in sensitivity analyses.

• Study 1: County-level binding values predicted higher hate group rates; per 1 SD increase in binding, odds ratio=1.32 (posterior SD=0.14, 95% HPDI [1.05, 1.61]) after adjusting for covariates; individualizing values showed no such effect. Model predictions aligned with observations (RMSE=0.15). With state fixed effects, binding association attenuated (~40%) and became indistinguishable from null due to reduced power and low within-state variance.
• Study 2 (Muslims): Binding values strongly associated with perceiving EBEPs as justified (b=2.29, SD=0.35, 95% CI [1.67, 2.96]); individualizing values negatively associated (b=−1.70, SD=0.32, 95% CI [−2.32, −1.10]). PMW positively associated with EBEP justification (b=1.72, SD=0.51, 95% CI [0.75, 2.77]) and statistically mediated the binding→EBEP link. Effects robust to adjustment for political ideology.
• Study 3 (Mexican immigrants): Binding values positively associated with EBEP justification (b=1.60, SD=0.27, 95% CI [1.12, 2.14]; OR≈4.95 per SD). Individualizing values negatively associated (b=−1.15, SD=0.41, 95% CI [−1.88, −0.35]; OR≈0.31). PMW positively associated with EBEP justification (b=1.63, SD=0.46, 95% CI [0.70, 2.49]; OR≈5.10) and partially mediated binding effects; adjusting for PMW reduced binding and individualizing effects.
• Study 4 (causal test): High-threat condition increased EBEP justification (b=1.44, SD=0.68, 95% CI [0.16, 2.80]; OR=4.21 [1.18, 16.45]). PMW strongly associated with EBEP justification (b=2.21, SD=0.42, 95% CI [1.48, 2.94]; OR≈9.20). Including PMW reduced the condition effect to nonsignificant (b=0.28, SD=0.63, 95% CI [−0.86, 1.57]), and mediation analyses supported PMW as mediator.
• Study 5 (moderated mediation): Binding-violation condition showed strong moderated mediation by binding values: ACME_high (binding +1 SD)=0.53 (SD=0.21, 95% CI [0.13, 0.87]); ACME_low=0.01 (SD=0.01, 95% CI [0.0007, 0.026]); ADEs near zero. For individualizing-violation, binding values showed smaller ACME_high=0.25 (SD=0.15, 95% CI [0.04, 0.57]) vs. ACME_low=0.05 (SD=0.06, 95% CI [0.004, 0.16]) with overlapping CIs. Individualizing values did not moderate ACME in the individualizing-violation domain (ACME_low=0.08, SD=0.10, 95% CI [0.004, 0.34]; ACME_high=0.09, SD=0.07, 95% CI [0.009, 0.27]). Overall supports that only binding values function as a treatment susceptibility factor, strongest for binding-domain violations.

Findings support the moralized threat hypothesis: people are more likely to justify EBEPs when they perceive an outgroup’s behavior as morally wrong, and group-preserving (binding) moral values amplify this effect. Geospatially, counties with higher binding values have higher hate group rates, though state-level fixed-effects analyses were underpowered and yielded null associations, suggesting potential unmeasured confounds or power limitations. Observational and experimental studies demonstrate that PMW mediates the link between outgroup behavior and EBEP justification, even without material threats, and that binding values uniquely increase susceptibility when binding-domain violations are perceived. The results underscore the role of moralization and group-based moral values in extreme prejudice, aligning with theories of morally motivated aggression and highlighting implications in digital environments that can amplify moral outrage.

This work integrates geospatial modeling and experimental evidence to show that group-focused moral values (binding) and perceived moral wrongdoing are key factors in justifying extreme behavioral expressions of prejudice. Binding values predict regional hate group prevalence and, at the individual level, strengthen the pathway from perceived outgroup moral violations to EBEP justification—most strongly for binding-domain violations. The findings suggest interventions should consider moralization processes and binding values when addressing EBEP risk. Future research should clarify discrepancies between spatial smoothing and fixed-effects estimates, examine generalizability to contexts where groups coalesce around individualizing values (fairness, harm), test cross-cultural applicability, and explore dynamics in online ecosystems that may catalyze moral outrage and EBEPs.

• SPLC hate-group data may be incomplete or biased due to volunteer-based collection; underreporting of hate crimes is common.
• County-level moral values estimated from non-representative YourMorals.org data, adjusted via MrsP; residual biases may remain.
• State fixed-effects models attenuated the binding–hate group association and were underpowered due to low within-state variance and measurement shrinkage.
• Vignette-based experiments use fictional scenarios; ecological validity and generalizability to real-world behavior are limited.
• Moral threat manipulations, especially for individualizing violations, may not be perfectly domain-specific and could partially engage other moral domains.
• Findings may be most applicable to groups organized around loyalty, authority, and purity; generality to fairness/harm-centric groups is uncertain.
• Self-report measures of justification may not translate directly to behavior; causality for long-term outcomes was not tested.