The influence of threat and right-wing authoritarianism on the selection of online (dis)information—a conceptual replication and extension of Lavine et al. (2005)

Abstract

Over the decades, communication research has investigated the situational and personal conditions under which people particularly prefer attitude-consistent over attitude-inconsistent content (confirmation bias). In a central study, Lavine et al. (2005) [Lavine, H., Lodge, M., & Freitas, K. (2005). Authoritarianism, threat, and motivated reasoning. Political Psychology, 26(2), 219–244.] examined how right-wing authoritarianism (RWA) and threat cause bias when processing political information. Their laboratory experiment suggested that right-wing authoritarians prefer attitude-consistent information in the presence of a threat. Given new crisis environments accompanied by various threats, we re-examined this interaction effect and conceptually replicated Lavine et al.'s central hypothesis in a contemporary media environment. In an online experiment (N = 1,118), we focused on selective exposure to verified news and disinformation and tracked participants’ selection unobtrusively. Contrary to expectations, the interaction between different threats and RWA did not increase selective exposure to attitude-consistent (dis)information. The results challenge the hypothesis’ underlying framework and make it necessary to consider new ways of advancing the theoretical model.

How people select information is a central topic in communication research. The selective exposure approach (Knobloch-Westerwick et al., 2015), which assumes that people prefer information consistent with their pre-existing attitudes (confirmation bias), has been robustly supported in numerous studies considering personal and situational conditions (threat in particular; Fischer et al., 2011). In this context, a central study by Lavine et al. (2005) investigated the interaction effect of threat and right-wing authoritarianism (RWA) on selective exposure behavior to political information. Consistent with the threat-RWA activation hypothesis (Feldman & Stenner, 1997), they argued that threat activates otherwise dormant RWA values (authoritarian aggression, submission, and conventionalism), thereby increasing their influence on other political judgments and preferences. In a laboratory experiment, they asked 92 participants to read one of three articles that contained either pro-, con-, or balanced information about capital punishment after a writing task reminding them of their death threatened half of the sample. Lavine et al. (2005) showed that in the presence of the threat, high (but not low) right-wing authoritarians preferred political information consistent with their pre-existing attitudes. Thus, the effect of RWA on selective exposure to information was fully dependent on threat.

Beyond testing the interaction of RWA and threat within the confines of a controlled experimental setting, their study addressed a pertinent gap in research at this time. In addition to other studies that established a positive relationship between RWA, threats and a range of attitudes and policies such as prejudice, intolerance toward marginalized groups, and endorsement of authoritarian measures (e.g., Cohrs & Ibler, 2009; Hartman et al., 2021), Lavine et al. (2005) were the first to show the influence of threat on the cognitive strategies employed by right-wing authoritarians when processing new political information. Given the extensive research conducted in communication science on selective exposure (see Hart et al., 2009), the finding that threat leads to biased rather than open-minded information processing in right-wing authoritarians was particularly noteworthy and contributed to understanding which situational and personal conditions (and the interaction of both) can intensify the confirmation bias. However, although recent societal crises (e.g., the coronavirus disease 2019 (COVID-19) pandemic, climate change, and refugee crisis) created increasingly suitable conditions for the threat-RWA interaction effect to unfold (increased threat level, the prevalence of right-wing authoritarian attitudes, increase in (dis)information), the frequently cited result was not revisited in subsequent years. Therefore, to advance the underlying theoretical framework in a contemporary information environment, the primary objective of this study was to conceptually replicate the core hypothesis put forth by Lavine et al. (2005) under more rigorous circumstances and examine the existence (and direction) of the interaction effect between threat and RWA on selective exposure (see goal #1; Anderson & Maxwell, 2016). We deliberately opted for a conceptual replication that allowed us to evaluate the theoretical robustness of the threat-RWA activation dynamic.¹ At the same time, this gave us the opportunity to make substantial changes to the measurements and procedures that we considered more appropriate for retesting the critical predictions of Lavine et al. (2005), accepting that the approach may limit full comparability between both studies. We hypothesized that people show a stronger confirmation bias in selecting information when threat is high than when threat is low and that RWA strengthens this effect (just as in Lavine et al., 2005), but made significant adjustments to our experimental conditions, considering the current global crisis climate and utilizing a more natural news selection environment with different news options, measures, stimuli, and statistical analysis, as detailed in the Method section. The full preregistered hypotheses and a complete study overview are available online at the Open Science Framework (OSF) repository (https://osf.io/z82re/?view_only=d54676f684ba405d95937bd0106ae539).

Method

Data and participants

Data were collected between August 5 and 17, 2021, from an online access panel with quotas set for gender, age, and education, representative of German Internet users aged 18–75 years (for additional information on the participants see the expanded methodological write-up in the Supplementary Appendix A). Data files and materials are available at OSF (see link above).

Procedure and variables of selective exposure, threat, and RWA

Instead of conducting the experiment in a laboratory setting (N = 92), we implemented an online approach to test Lavine et al.’s (2005) central hypothesis in a contemporary information environment with a larger sample size (N = 1118).² Participants browsed a mock news site for a limited time (four minutes) and clicked on headlines to read full articles (see Figure 1; English versions of the headlines are available in Supplementary Appendix C). A software-based tool (Unkel, 2021) tracked their browsing history, click rate, and time spent reading the articles, replacing hand-coding used by Lavine et al. (2005). We varied the stance of the articles (pro versus con) and the news type (verified news versus disinformation) as within-subject factors. As our study centered around Lavine et al.’s (2005) theoretical rational (threat-RWA activation on selective exposure behavior), we refrained to add assumptions on participants’ balanced news selection, and, therefore, balanced news options. Instead, we provided disinformation to (a) simulate a realistic information environment in times of crisis in which disinformation spread alongside verified news and, (b) examine whether the general susceptibility of right-wing authoritarians to disinformation (e.g., Frischlich et al., 2021; Klebba & Winter, 2024) is increased under threat (see the commentary on the influence of right-leaning factors in the Supplementary Appendix D). To further reflect a crisis information environment, we chose COVID-19 vaccine-related news and disinformation as a more suitable topic than capital punishment in Lavine et al. (2005). As the independent variable, we elicited an existential threat as in Lavine et al. (2005) and a social-normative threat according to Stenner’s (2005) theory of threat-RWA activation to better reflect the state of the research debate (for further elaboration, refer to the additional commentary in the Supplementary Appendix D on the modification of manipulating different types of threats). We varied the induction of threat as a between-subjects factor, using news articles instead of the original death-thought manipulation due to ethical concerns and the recommendation of a local ethics committee (see Supplementary Appendix C for stimulus material). Participants were randomly assigned to three conditions before the selection task: existential threat (article on the high risk of developing cancer in the course of a lifetime; n = 376), social-normative threat (article on the influx of African refugees to Europe as a result of the COVID-19 pandemic; n = 368), or a control group (article about trips to a tourist destination in Germany; n = 374). We conducted a pilot study to identify the most threatening articles for each threat types, and manipulation checks confirmed that participants in the threat induction groups felt significantly more threatened than the control group (see manipulation check and correlations in Supplementary Appendix A). Instead of Altemeyer’s (1996) RWA scale, to rate RWA in its subdimensions, we used the KSA-3 short scale (Beierlein et al., 2014) as a well-tested German scale, including three items measuring authoritarian aggression, authoritarian submission, and conventionalism, on a five-point scale (M = 3.02, SD = .82, α = .86) (see an expanded description of key measures in the Supplementary Appendix A).

Results

Data analysis plan

We used moderated regression models rather than a mixed-effects ANOVA as in Lavine et al. (2005).³ Following current empirical studies (e.g., Winter et al., 2016; Wojcieszak et al., 2020), we recorded which articles were selected for further reading. These clicks were categorized into attitude-consistent or attitude-inconsistent selection of (dis)information based on participants’ prior attitudes toward COVID-19 vaccination (e.g., attitude-consistent if a proponent/opponent read a full pro/con article on COVID-19 vaccination; attitude-inconsistent if a proponent/opponent read a full con/pro article on COVID-19 vaccination). Then, we calculated a ratio score by subtracting the attitude-consistent news selection variable from the attitude-inconsistent news selection variable for both verified news and disinformation, representing the extent of the confirmation bias. Thus, the higher the value, the higher the participants’ selective exposure behavior (see an expanded description in the methodological write-up in Supplementary Appendix A).

Testing the existence (and direction) of the interaction effect

We conducted moderated regression analysis with the predictors of threat (threat of cancer or African COVID-19 refugees), RWA, and the dependent variable of the confirmation bias (ratio score, selection rates) for verified news and disinformation. The moderated regression analyses found no significant interaction effects for RWA and the threat of cancer (ß = −.001, p = .988) or African COVID-19 refugees (ß = −.047, p = .516) in predicting participants’ selective exposure behavior (preference for attitude-consistent over attitude-inconsistent information) to verified news on COVID-19 vaccination. However, a significant interaction effect was found for RWA and the threat of African COVID-19 refugees on participants’ selective exposure behavior to disinformation on COVID-19 vaccination. Unlike predicted, participants with high RWA scores threatened by incoming African COVID-19 refugees were more likely to select attitude-inconsistent over attitude-consistent disinformation on COVID-19 vaccination (ß = −.175, p = .025, see Figure 2 and a discussion in the Supplementary Appendix D).

Additional analysis

A repeated-measures ANOVA showed that participants selected significantly more attitude-consistent verified information (M = .64, SD = .75) than attitude-inconsistent verified information (M = .58, SD = .76; F (_{1, 940}) = 5.102; p = .024, η_p² = .005), and were more likely to select attitude-consistent disinformation (M = .67, SD = .77) than attitude-inconsistent disinformation (M = .56, SD = .73; F (_{1, 940}) = 18.274; p < .001, η_p² = .019) on COVID-19 vaccination. In contrast, no significant main effect of threat was found on participants’ selective exposure to verified news (F (_{2, 940}) = 1.794; p = .167, η_p² = .004) and disinformation (F (_{2, 940}) = 1.465; p = .232, η_p² = .003), but RWA predicted a decrease in participants’ time spent on verified news versus disinformation (ß = −.07; p = .036). All hypotheses’ tests and exploratory analyses are reported and discussed in Supplementary Appendix B.

Discussion

The lack of support for Lavine et al.’s (2005) central hypothesis that right-wing authoritarians are more likely to select attitude-consistent (versus attitude-inconsistent) information under threat raises issues regarding statistical power, experimental variation, and theoretical strength of the hypothesis.

Discussion of the methodological changes

While we acknowledge the limitations of comparability due to deviations from the original Lavine et al. (2005) design, we contend that several of the changes are superior options for rigorously testing the eighteen-year-old finding. Our decision to significantly augment the sample size vis-à-vis Lavine et al. (2005)—which is underpowered for testing mixed effects⁴—should have facilitated detection of the effect and helped to avoid inflated effect sizes and constitution of type-1-errors in smaller studies (Button et al., 2013). The omission of a balanced news option, and instead including disinformation to better simulate a crisis news scenario, might have influenced participants’ news selection in direct comparison to Lavine et al. (2005) but leaves the theoretical rationale of the threat-RWA activation hypothesis on selective exposure behavior, the formation of the dependent variable and statistical analysis untouched (in both Lavine et al. (2005) and our testing). Due to the ongoing debate surrounding which type of threat activates RWA and produces right-wing authoritarian outcomes, we opted to switch the type and topic of threat induction. We utilized existential and social-normative threats in the form of news articles to examine which apprehension provokes RWA more and mitigate ethical concerns by not threatening participants with reminding them of their death. The manipulation checks confirmed the effectiveness of the induction, and the different RWA measure displayed strong reliability, supporting an appropriate evaluation of the construct. However, we concur that a news-based threat induction could be less impactful, and that differences in the setup of the study (larger sample size, online setting, different measurements, adjusted news selection scenario) may partly explain diverging results. Despite these limitations and reasonable disagreement among readers about the degree of methodological changes that are “allowed” and valuable, we believe the current study to be a conceptual replication of Lavine et al. (2005) (and designed as such) that contributes to moving the literature on the threat-RWA activation hypothesis forward, and should be of interest to the research community.

Discussion of the results

Since we could not reproduce the direction and presence of the interaction effect on selective exposure behavior with a larger sample, which provides a higher probability of detecting the effect and more precise data analysis, it is necessary to question the underlying framework of the hypothesis. Our results showed that RWA responded exclusively to our social-normative threat induction (the threat of African COVID-19 refugees), but selective exposure was not a significant finding. Although other studies showed that RWA interacts with existential or social-normative threats, recent interactions were more likely to elicit right-authoritarian behaviors (e.g., nationalism and anti-immigrant sentiments, Hartman et al., 2021). Our results, therefore, question the significance of the interplay between RWA and threat for the preference of attitude-consistent over attitude-inconsistent (dis)information. Instead, they showed a general confirmation bias and a susceptibility of highly right-wing authoritarian individuals to disinformation independent of threat perception. Thus, this study (with a larger sample and a more appropriate data analysis) suggests that the interaction between threat and RWA plays a minor role in crisis information environments than we previously expected (see Supplementary Appendix D for further discussion). To further investigate the threat-RWA-activation hypothesis, it is crucial to address the conflicting results by solidifying the categorization of threat types and systematizing the outcomes. As we incorporated disinformation into the selection task, more detailed replications of existing studies and outcome variations may be fruitful in field surveys and experimental settings. The overarching goal should be to develop a more advanced model of the threat-RWA-activation hypothesis, capable of verifying existing contexts and looking at them more sophisticatedly with new crucial moderators and possible boundary conditions.

Supplementary material

Supplementary material is available online at Human Communication Research online.

Data availability

Data are publicly available online. Data files and survey materials are available at OSF (https://osf.io/z82re/?view_only=d54676f684ba405d95937bd0106ae539).

Conflicts of interest

The authors have no external funding or conflicts of interest to disclose.

Footnotes

References