Authoritarianism, Threat, and Prejudice: An Analysis of Mediation and Moderation

Abstract

Right-wing authoritarianism is a central construct in individual differences approaches to prejudice. Its power to predict prejudice is often attributed to perceived threat. However, the exact moderating and mediating processes involved are little understood. In two studies (Ns = 53, 84), exposure to threatening versus nonthreatening information about an ethnic out-group had reliable indirect effects on prejudice in authoritarians, but not in nonauthoritarians, largely because authoritarians were more likely to perceive actual threat when they interpreted the information received to represent a threatening argument. Additionally, in Study 2, authoritarians reacted more strongly with negative emotions when they perceived actual threat.

ACKNOWLEDGEMENTS

Study 2 was supported by a post-doctoral stipend of the Deutsche Forschungsgemeinschaft (GRK 622; International Research Training Group “Conflict and Cooperation between Social Groups”) to the first author. We thank Frank Asbrock, Natascha de Hoog, Philipp Jugert, Thomas Kessler, and Jost Stellmacher for helpful comments on earlier versions of this paper and Bettina Doering and Christina Schilling for collecting data for Study 2.

Notes

¹Although there are some differences between both conceptualizations of authoritarianism (see Feldman, 2003), high correlations between their respective measures (rs = .70, .71; Cohrs, 2005; Feldman, 2003) show that they are very similar empirically. The measures also bear substantial item overlap.

²One additional case was removed from the data set because it was a clear outlier with a studentized deleted residual of 4.21 regarding the prediction of perceived threat.

³Text length was 514 (threat) and 413 words (control), respectively. The texts were perceived as equal in plausibility and conclusiveness/argument strength in a pretest with 40 students. The full texts (in German) are available from the authors. Reading times ranged from 73 to 273 seconds and were longer in the threat (M = 183) than in the control condition (M = 132), t(51) = 3.32, p = .002, η² = .18, but there were no influences of RWA or RWA × condition. We thank Barbara Moschner for data collection in the pretest.

Note. Exposure to threatening information represents the experimental condition, coded +1 for threat and −1 for control.

°p < .08. ∗p < .05. ∗∗p < .01 (two-tailed).

⁴Perceived actual threat, anti-intimacy, and denial of positive emotions were also regressed on RWA, experimental condition, and their interaction. Perceived actual threat was predicted by RWA, b = .49, p < .001, and experimental condition, b = .31, p = .009; the interaction effect was positive but nonsignificant, b = .11, p = .35 (R ² = .37). Anti-intimacy was predicted by RWA, b = .35, p = .011, but not significantly by experimental condition, b = .19, p = .16, or the interaction, b = .07, p = .60 (R ² = .17). Denial of positive emotions was not predicted significantly by RWA, b = .17, p = .20, or experimental condition, b = .20, p = .13, but by their interaction, b = .28, p = .039 (R ² = .16). Thus, there were no significant interaction effects of RWA and experimental condition on perceived actual threat and anti-intimacy even though the indirect effects of experimental condition on these variables were significant at high-but not at low-RWA levels and there was a significant interaction effect of RWA and perceived argument threat on perceived actual threat. It seemed not to be the presence of contextual threat (exposure to threatening vs. nonthreatening information) per se that produced effects but rather its subjective interpretation (perceiving the information this way), reflecting that the power to detect an effect can sometimes be higher when mediation is taken into account (see Shrout & Bolger, 2002). Still, simple slope analyses showed that experimental condition affected perceived actual threat, b = .42, p = .013, and denial of positive emotions, b = .48, p = .013, at high levels of RWA but not at low levels of RWA: perceived actual threat: b = .20, p = .22; denial of positive emotions: b = − .08, p = .67.

⁵Five additional participants were excluded from the analyses because they did not give us permission to use their data, showed an undue response pattern (i.e., ticking the middle category for every emotions item), distrusted the cover story or did not fill out the emotions items.

Note. Exposure to threatening information represents the experimental condition, coded +1 for threat and −1 for control.

∗p < .05. ∗∗p < .01 (two-tailed).

⁶Perceived actual threat, positive emotions, and negative emotions were also regressed on RWA, experimental condition, and their interaction. Perceived actual threat was predicted by RWA, b = .46, p < .001; the effects of experimental condition, b = .12, p = .23, and the interaction, b = .14, p = .16, were positive but nonsignificant (R ² = .22). Positive emotions were predicted by RWA, b = −.24, p = .030, but not by experimental condition, b = .01, p = .93, or the interaction, b = .11, p = .33 (R ² = .08). Negative emotions was predicted by RWA, b = .35, p = .001, and experimental condition, b = .21, p = .045; the interaction effect was positive but nonsignificant, b = .15, p = .15 (R ² = .16). Thus, analogous to Study 1, there were no significant interaction effects of RWA and experimental condition on these variables even though the indirect effects of experimental condition were significant at high- but not at low-RWA levels and there was a significant interaction effect of RWA and perceived argument threat on perceived actual threat. Still, simple slope analyses showed that experimental condition affected perceived actual threat, b = .26, p = .071, and negative emotions, b = .36, p = .017, at high levels of RWA but not at low levels of RWA: perceived actual threat: b = −.02, p = .88; negative emotions: b = .06, p = .69.