Previous research

Perceptions of security threats are a function of the probability of a threatening event occurring combined with the severity of its consequences (HM Government, 2010, p. 37; Sjoberg, Reference Sjoberg1999). Thus, a security threat with a high probability of occurrence may still be regarded as less serious than a security threat with a low probability of occurrence if the consequences of the former are regarded as far less serious than the consequences of the latter. Perceptions of probabilities and consequences, in turn, are driven by contextual changes, including changes in an issue’s salience and framing (Haider-Markel & Vieux, Reference Haider-Markel and Vieux2008), and individual-level factors such as education, media habits, and authoritarian predisposition. Examples of contextual change, such as terrorist acts like 9/11, have the capacity to shift public opinion as a result of the shock of death and destruction and the shifts in media attention that follow (Berrebi & Klor, Reference Berrebi and Klor2008; Croft & Moore, Reference Croft and Moore2010; Davis & Silver, Reference Davis and Silver2004).Footnote ² These influences, in turn, are moderated by the individual-level attributes through which information is filtered, interpreted, and experienced, with authoritarianism being the key moderator according to contemporary research (e.g., Hetherington & Weiler, Reference Hetherington and Weiler2009, Reference Hetherington and Weiler2018; Huddy et al., Reference Huddy, Feldman, Taber and Lahav2005; Norris & Inglehart, Reference Norris and Inglehart2019; Stevens & Vaughan-Williams, Reference Stevens and Vaughan-Williams2016).

While infectious disease may also be regarded as a security threat (Stevens & Vaughan-Williams, Reference Stevens and Vaughan-Williams2016)—indeed, there is “an entire scientific literature on how infectious diseases promote authoritarianism” (Kealey, Reference Kealey2021)—the focus of much of that literature is on its relationships with authoritarian attitudes and practices at the aggregate rather than at the individual level, with countries that have a high incidence of infectious disease being more authoritarian. Parasite- or pathogen-stress theory (Thornhill & Fincher, Reference Thornhill and Fincher2014) finds a link between authoritarian governance and infectious disease over space and time (Pericas, Reference Pericas2020) that may be due both to evolutionary psychology and to the need for conformity in cultures to limit its spread (Tybur et al., Reference Tybur, Inbar, Aarøe, Barclay, Barlow, Barra, Becker, Borovoi, Choi, Choi, Consedine, Conway, Conway, Conway, Adoric, Demirci, Fernández, Ferreira, Ishii and Žeželj2016), and mediated by the behavioral immune system (Helzer & Pizarro, Reference Helzer and Pizarro2011; Murray & Schaller, Reference Murray and Schaller2012; Murray et al., Reference Murray, Schaller and Suedfeld2013). Zmigrod et al. (Reference Zmigrod, Ebert, Götz and Rentfrow2021) confirm a relationship between authoritarian attitudes and infectious disease in the 47 countries they examine, but they make an additional connection to voting behavior, with data from the United States showing a path from prevalence of infectious disease through authoritarian attitudes to voting for Donald Trump. Mixing individual-level survey data with county-level data on mortality rates, Hinckley (Reference Hinckley2021) also finds a relationship between existential threat and positive views of Donald Trump. However, other studies question the causal explanations in pathogen-stress theory (Bromham et al., Reference Bromham, Hua, Cardillo, Schneemann and Greenhill2018; Pollet, Reference Pollet2014) or suggest that the relationships are conditional on other societal-level factors such as economic wealth (Kusano & Kemmelmeier, Reference Kusano and Kemmelmeier2018).

More importantly for this article, most of the research on authoritarianism and infectious disease leaves questions about the micro-level processes open. We know little about how changes in context, such as increased threat of infectious disease over time, are related to individual-level perceptions of threat or their interaction with individual-level authoritarianism. Thus, they, too, cannot solve the puzzle of why the COVID-19 pandemic did not increase authoritarian support for more authoritarian measures.

How might contextual changes in security threats such as terrorism or infectious diseases combine with authoritarianism?Footnote ³ Stenner’s (Reference Stenner2005) answer is that authoritarians are highly sensitive to changes in context that have implications for threats to social norms or societal division: she describes authoritarians as “relentlessly sociotropic boundary maintainers” (p. 32). Indeed, authoritarianism has been identified as contributing to emergent cleavages in British public opinion, the focus of this article, because of perceptions of threats to social norms from immigration, which are then reflected in right-wing vote preferences in elections (Chrisp & Pearce, Reference Chrisp and Pearce2019; Fox & Pearce, Reference Fox and Pearce2018; Golec de Zavala et al., Reference Golec de Zavala, Guerra and Simao2017; Hetherington & Weiler, Reference Hetherington and Weiler2018; Kaufman, Reference Kaufman2016; Norris & Inglehart, Reference Norris and Inglehart2019; Peitz et al., Reference Peitz, Dhont and Seyd2018). Tillman (Reference Tillman2021) argues that as a result of immigration, the predisposition toward authoritarianism increasingly divides other publics in Western Europe, and Hetherington and Weiler (Reference Hetherington and Weiler2009, Reference Hetherington and Weiler2018) that it divides Republicans from Democrats in the United States.

However, little of this research has examined (1) the claim that authoritarians are especially sensitive to changes in context that have implications for social norms; (2) the corollary that perceptions of other threats over time—those without implications for social norms—are more stable for authoritarians; or (3) the kinds of contextual change, beyond shocks such as terrorist attacks, that may affect authoritarians. This reflects two problems. First, there is a tendency toward a narrow focus on security issues related to conflict, terrorism, and prejudice. This tells us neither about other threats to welfare, such as those from infectious diseases, nor about the “everyday threats,” such as crime that individuals experience, and whether there are differences in responses to these types of threats as opposed to the threats to social norms to which Stenner (Reference Stenner2005) refers. Second, we possess knowledge of perceptions of various different security threats in a series of changing contexts, such as after a terrorist attack, and their relationship with authoritarian predispositions (e.g., Stevens & Banducci, Reference Stevens and Banducci2022), but we know less about the same security threats in different contexts, such as the threat of terrorism when there is little discernible threat, and their relationship with authoritarian predispositions. We also know about the relationship between authoritarian attitudes and practices and the threat from infectious diseases, but neither how the threat from infectious diseases is actually perceived nor its relationship with authoritarian predispositions.

Theory and hypotheses

If perceptions of threat are a function of perceptions of the probability of an event occurring and the likely severity of its consequences, perceptions of threat will increase given (a) perceptions of an increase in the likelihood that an event will occur (e.g., another terrorist attack after 9/11) or (b) perceptions of an increase in the severity of a threat’s consequences (e.g., environmental degradation or fallout from an economic depression). Perceptions of the probability and severity of a threat will be affected not only by the real world but also by signals such as media attention to an issue (regardless of whether those signals reflect real-world change): as Ridout et al. (Reference Ridout, Grosse and Appleton2008) put it, “individuals adjust their perception of reality to fit the image of the world around them” (p. 576), rather than vice versa. Media agenda-setting research tells us that the amount of media attention to an issue is taken as a signal of its importance. With regard to security threats, such changes in attention alone may be taken to signal changes in the probability and/or consequences of threats, but particularly if the framing is negative: both point to a context of increased threat.

The question then becomes how authoritarian predispositions interact with such changes in context. Previous research offers four answers.

1. Authoritarians are impervious to contextual change. Authoritarians are in a permanently elevated state of threat, which, in turn, implies that they are relatively unresponsive to changes in security threats. It is libertarians who are more likely to respond to increased attention to an issue with perceptions of elevated threat (Hetherington & Weiler, Reference Hetherington and Weiler2009) because perceptions of high levels of threat are built into authoritarians’ worldviews.

2. Authoritarians are responsive to any contextual change. Altemeyer (Reference Altemeyer1996) argues that “authoritarians stand about ten steps closer to the panic button than the rest of the population” (p. 100) rather than being in a permanently elevated state of threat. This implies that in contexts of increased threat, authoritarian perceptions of threat should increase regardless of the issue.

3. Authoritarians are responsive to changes in security threats with implications for “normative threat” but impervious to contextual changes to security threats that do not have implications for normative threat. Stenner (Reference Stenner2005) suggests a more dynamic relationship between perceptions of threats and authoritarianism. She argues that while authoritarians are more likely to see the world as dangerous and “are inclined towards this peculiar fear of a dangerous world under any conditions” (Stenner, Reference Stenner2005, p. 69; italics in original), they are particularly attentive to “normative threat,” which can be defined as “threats to unity and consensus, or ‘oneness and sameness’” (Stenner & Haidt, Reference Stenner, Haidt and Sunstein2018, p. 180). Moreover, in stable contexts, Stenner suggests “somewhat wishful thinking” among authoritarians in which perceptions of normative threat are lower than for libertarians. But, at the same time, authoritarians have a “hypersensitivity” to changes with implications for normative threat (Stenner, Reference Stenner2005, pp. 69–70). The combination of “a dangerous world” with respect to perceptions of threat in general and wishful thinking and hypersensitivity to normative threat suggests that authoritarians are in a permanently elevated state of threat on most issues, but not those issues that have implications for normative threat. Therefore, we should expect greater responsiveness to changes in security threats with implications for normative threat among authoritarians; for security issues without implications for normative threat, authoritarians should have consistently elevated perceptions of threat that are impervious to contextual change.

4. Authoritarians are responsive to changes in security threats with implications for normative threat. They are also responsive to changes in security threats that do not have implications for normative threat but may respond to increased attention to them with decreased perceptions of those threats. A fourth set of research findings suggests that while Stenner (Reference Stenner2005) may be correct about normative threat, the notion that authoritarians simply perceive “a dangerous world,” implying a lack of sensitivity to other threats, also needs refining. Kahan et al. (Reference Kahan, Braman, Gastil, Slovic and Mertz2007) contend that “individuals selectively credit and dismiss asserted dangers in a manner supportive of their cultural identities” (p. 465, emphasis added; see also Hibbing, Reference Hibbing2020), implying that authoritarians should discount threats to difference and diversity (i.e., to individual autonomy). In addition, Duckitt and Sibley (Reference Duckitt and Sibley2010) argue that authoritarians are relatively unperturbed by security threats linked to inequality or scarcity because they tend not to impinge on authoritarians’ most fundamental concerns about group conformity and social norms (p. 1869). This also suggests that authoritarians may diminish threats that affect subgroups of society, such as hate crimes, rather than the whole of society. Finally, Choma et al. (Reference Choma, Hanoch, Gummerum and Hodson2013) argue that another salient characteristic of threats such as the environment is that they have diffuse rather than personal risks. It is not clear, however, that the threats that do seem to exercise high authoritarians, such as immigration, are perceived as personal (Stevens & Vaughan-Williams, Reference Stevens and Vaughan-Williams2016).

These perspectives imply that while there may be some consensus that an issue like the environment is a growing security threat, to the extent that environmentalism is associated with resource scarcity, authoritarians will be more likely to downplay the threat relative to libertarians. Even an existential threat like the COVID-19 health pandemic, while likely to be identified as a threat by the vast majority of people given its prominence, may be viewed as less of a threat by authoritarians if the emphasis is on its unequal economic effects or on mortality among out-groups that is unlikely to threaten social norms. We discuss this in more detail later. Similarly, signals of change in threat from crimes that affect all groups (e.g., burglary) may be viewed differently by authoritarians than threats from crimes that affect subgroups (e.g., hate crimes and crimes against women).

In sum, this fourth perspective suggests that for some issues beyond normative threat, rather than perceiving a dangerous world or being in a permanently elevated state of threat, authoritarians may respond to changes in context signaling increased threat from an issue by minimizing that threat. Authoritarians will respond to signals with implications for heightened normative threat with increased perceptions of threat but diminish other types of threat, such as those with diffuse risks or implications for inequality or scarcity.

A further important variable that may moderate perceptions of threats for authoritarians is whether they are sociotropic or personal. The corollary of Stenner’s (Reference Stenner2005) description of authoritarians as relentlessly sociotropic boundary maintainers is her claim that they are relatively unconcerned with personal threats. This is because personal threats have implications for the exercise of individual autonomy rather than the social norms and sociotropic boundary maintenance that authoritarians care most about (see also Feldman, Reference Feldman2013). However, Asbrock and Fritsche (Reference Asbrock and Fritsche2013) argue that perceptions of the personal more than the national threat of terrorism provoke authoritarian expression and suggest that “authoritarian responses may operate as a group-level coping strategy for a threat to the personal self” (p. 35). But their argument is based on two experiments rather than over-time analysis. More promisingly, Stevens and Banducci (Reference Stevens and Banducci2022) use survey data before and after a terrorist attack to claim that increases in perceptions of personal threat from terrorism increase perceptions of normative threat for authoritarians. But their claims are limited to a single issue over a short space of time.

In sum, we have gleaned three different categories of threat from the literature on threat and authoritarianism:

1. 1. Threats to social norms and the in-group, such as threats from immigration or perceptions of weak border control

2. 2. Threats with implications for inequality and resources, such as threats from the economy or infectious diseases such as COVID-19

3. 3. Threats from crime, with an additional distinction between those that affect everyone (e.g., burglary) and those that affect subgroups (e.g., hate crimes and crimes against women), as well as between threats that are national versus personal.

Threats from infectious diseases could have implications for social norms. Indeed, with COVID-19, there was some initial focus on potential threats to social norms such as food hoarding or from a severe economic depression. There was also evidence of growing racial antipathy in countries like the United Kingdom and the United States, manifesting in increased hate crimes (Borkowska & Laurence, Reference Borkowska and Laurence2021; Strassle et al., Reference Strassle, Stewart, Quintero, Bonilla, Alhomsi, Santana-Ufret, Maldonado, Forde and Napoles2022), perhaps fed by policies that closed borders to out-groups. However, over time, the emphasis shifted to other subjects, including the drag on resources affecting the health service and the economy and the unequal death rates by race (i.e., to resources and inequality). In addition, Ollerenshaw (Reference Ollerenshaw2022) and Geana et al. (Reference Geana, Rabb and Sloman2021) argue that right-wing opinion leaders in the United States, such as Donald Trump, diminished the threat from COVID-19—the overall message was “don’t worry about this” (Geana et al., Reference Geana, Rabb and Sloman2021, p. 6)—and that the perceptions of their supporters, who tend to be more authoritarian, were influenced by this messaging (Geana et al., Reference Geana, Rabb and Sloman2021; Ollerenshaw, Reference Ollerenshaw2022; see also Hibbing, Reference Hibbing2022). Similarly, in the United Kingdom, the right-wing prime minister, Boris Johnson, initially diminished the threat of COVID-19. While he then introduced measures such as lockdowns, these were largely attributed to “following the science” (Andreouli & Brice, Reference Andreouli and Brice2022), and the government subsequently transitioned to initiatives such as “eat out to help out” (a government-subsidized scheme in which people could get up to a 50% reduction on food and drinks at restaurants, cafés, and pubs) that appear to have contributed to a second wave of infections and deaths toward the end of 2020 (Morales, Reference Morales, Ring, Hutton and Paton2020; Parker & Payne, Reference Parker and Payne2021; Lintern, Reference Lintern2021). Finally, Pretus and Villarroya (Reference Pretus and Vilarroya2022) point out that the absence of a clearly defined “enemy” with respect to COVID-19 and other such catastrophes and natural disasters may reduce the influence of group threat.

Thus, perceptions of threats from COVID-19 to social norms or from out-groups were likely lessened in the United Kingdom (as in the United States) by right-wing elites’ efforts to minimize the overall threat of the disease. Indeed, our analysis here confirms that COVID-19 was viewed by the public as a threat more like the environment or problems of resource scarcity than terrorism or immigration.

Following from our categorization of threats to social norms, threats with implications for inequality and resources, and threats from crime, we test the following hypotheses:

H1 implies a positive interaction between authoritarianism and signals of increased threat for issues pertaining to out-groups, whereas H2 implies a negative interaction between authoritarianism and signals of increased threat pertaining to inequality, resource scarcity, or crimes against subgroups in society. H1 implies an interaction between authoritarianism and out-group threat that is consistent with the third and fourth possibilities suggested by the literature and inconsistent with the first (which implies a negative interaction) and second (which implies main effects of authoritarianism and context but no interaction). H2 implies an interaction between authoritarianism and threats pertaining to inequality, resources, and crimes against subgroups in society that is only consistent with the fourth.

Although H1 and H2 do not distinguish between perceptions of national and personal security threats, there is additional disagreement about whether authoritarians are concerned mainly by national (sociotropic) security threats (Feldman, Reference Feldman2013) or whether perceptions of personal security threats are paramount (Asbrock & Fritsche, Reference Asbrock and Fritsche2013). Therefore, we test two additional, mutually exclusive hypotheses:

What if there is less media attention to an issue over time—that is, a signal of decreasing threat? All else being equal, in this context, we would expect perceptions of threat to decrease for both authoritarians and libertarians. But for issues with implications for normative threat, Stenner (Reference Stenner2005) would lead us to expect such “normative reassurance” to lower threat levels for authoritarians more than libertarians.

We summarize the combinations of expectations around different kinds of security threats and changes in their prominence in Table 1: 0 represents no moderating influence of authoritarian predispositions, + a positive moderating effect, and – a negative moderating effect. We also indicate whether there is an expected main effect of authoritarian predispositions.

Table 1. Expectations of interactive relationships between authoritarian predispositions and attention to threats

Research design

The ideal test of our hypotheses would examine within-person changes in perceptions of security threats and their context with longitudinal panel data. However, longitudinal panels such as Understanding Society do not ask about security threats. The British Election Study (BES) regularly asks “most important issue” questions, but these do not necessarily elicit answers related to threat, and the BES does not regularly ask about perceptions of security threats.Footnote ⁴ To test our hypotheses without panel data, we thus draw inferences from two cross-sectional surveys that are unique in asking samples of (English)Footnote ⁵ respondents identical questions about perceptions of 10 security threats at two points in time: the June 2012 Perceptions of Security in an Age of Austerity online survey conducted by ICM for Stevens and Vaughan-Williams (Reference Stevens and Vaughan-Williams2016)Footnote ⁶ and an online survey conducted for the authors by Opinion Research Business (ORB) in July 2020.

The Perceptions of Security in an Age of Austerity survey asked whether 22 issues were a serious security threat to “(1) the world, (2) UK, (3) community in which you live, (4) you and your family at the moment.” The issues ranged from terrorism and the threat of Russia and China to burglary and online fraud. We replicated these questions in an online survey administered by ORB in July 2020, with four differences. First, we did not ask about community-level threats because Stevens and Vaughan-Williams (Reference Stevens and Vaughan-Williams2016) found minimal differences with perceptions of personal threats. Second, we randomly split the sample and asked different subsamples of respondents about their perceptions of threats to the United Kingdom or to “you and your family” to guard against possible effects of respondent fatigue—in the ICM survey, each respondent was asked four times about 22 issues. Third, because of additional concerns about respondent fatigue, we asked about a subset of the original 22 issues, dropping those such as “the far right” and “Islamophobia” that had been identified as threats by few respondents in 2012. And fourth, our sample in 2020 was from England rather than Great Britain. The analysis that follows excludes from the 2012 data respondents who resided in Scotland or Wales. In addition, given that our hypotheses include claims about perceptions of minority out-groups, we exclude non-White respondents,Footnote ⁷ providing a sample size of 1,414 for the 2012 survey and 2,926 in total for the 2020 survey, but with the latter sample split between respondents asked about national or personal threats rather than both.

The surveys captured authoritarian predispositions using Feldman’s (Reference Feldman and Stenner1997) four-question child-rearing values measure, in which respondents are given a choice of attributes and asked which is preferable in children—independence/respect for elders, obedience/self-reliance, curiosity/good manners, and considerate/well behaved. The premise of the scale is that child-rearing values reflect fundamental orientations toward conformity or autonomy—they are not affected, for example, by actual child-rearing practices (Hetherington & Weiler Reference Hetherington and Weiler2009, p. 50; Stenner, Reference Stenner2005, p. 24). The scale has been used widely to capture authoritarian predispositions, both in other research (e.g., Arikan, Reference Arikan2023; Hetherington & Weiler, Reference Hetherington and Weiler2009; Stenner, Reference Stenner2005; Zmigrod, Reference Zmigrod, Ebert, Götz and Rentfrow2021) and in major election studies in the United States, Britain, France, New Zealand, and Switzerland. Respondents with the highest authoritarian predispositions choose respect for elders, obedience, good manners, and well behaved. Respondents with lower levels of authoritarian predisposition choose one to three of these values, while libertarians choose independence, self-reliance, curiosity, and considerate. Both surveys also include the control variables for mortality salience, television news consumption, sex, age and education employed in Stevens and Vaughan-Williams’s (Reference Stevens and Vaughan-Williams2016) analysis.Footnote ⁸

For the comparison of the two surveys to be valid, we checked that the samples were similar on the observed demographics—and they were on age, sex, and education (see Table A1 in the Appendix). We also examined the mean levels of mortality salience and authoritarianism in the two surveys, for completeness rather than because we would expect them to be the same; there are differences of one-quarter to one-third of a standard deviation, with mean levels of authoritarianism slightly higher in 2012 and mean levels of mortality salience slightly lower.

The analysis that follows focuses on 10 issues that appear in both surveys and capture the kinds of security threats outlined in H1 and H2: immigration; terrorism; weak border control; health pandemics, such as avian flu (2012) or COVID-19 (2020); environmental issues, such as global warming or greenhouse gas effects; resource scarcity; economic depression, financial crisis, and unemployment; burglary; crimes against women; and racial or religious hate crimes.

Table 2 presents summary data on the proportions of the samples identifying each issue as a security threat. It shows that at the national level, perceptions of the top five issues of terrorism, the economy, immigration, weak border control, and racial or religious hate crimes as threats are quite stable across the two surveys, while the new issues of health pandemic, environmental issues, and racial and religious hate crimes loom larger in the later survey. At the same time, Table 2 indicates increases in perceptions of almost all issues as personal threats compared to 2012—only burglary shows a small decrease. While the economy and terrorism are among the top-ranked personal threats in both years, some of the increases in perceptions of threats from other issues are large, particularly a health pandemic and environmental issues. Reasons for the increase in the personal threat from terrorism could be related to the number of relatively small-scale attacks that were not a feature of Islamic terrorism in the United Kingdom up to 2012, such as those at Westminster, the Manchester Arena, and London’s Borough Market in 2017, as opposed to the larger-scale terrorism and loss of life of 9/11 and 7/7.Footnote ⁹ As we show later, media attention to the personal threat of terrorism was much greater in 2020 than in 2012.

Table 2. Perceptions of threats in 2012 and 2020

Note: Figures for 2012 and 2020 are percentages.

It is worth noting some of the differences between the threat presented by avian flu in 2012 and COVID-19 in 2020 that are reported in Table 2, as well as that while the survey question gave avian flu as an example of a health pandemic in 2012, there had been an outbreak of swine flu in 2009. According to the Office for National Statistics (2022), COVID-19 was the leading cause of death in England and Wales in 2020, with roughly 75,000 deaths within 28 days of a COVID-19 test by the end of the year, of which about 40,000 had occurred by the time the ORB survey was in the field. In October 2020, the reproduction rate (R0) of the coronavirus, or the average number of additional people infected from a single infection case, was estimated to be 1.40 to 6.49, compared to 1.30 to 1.71 for other flu viruses, with a case fatality rate (CFR) of 1.40% to 3.67% (Bai & Tao, Reference Bai and Tao2021). Infection rates for the swine flu in 2009 were high, with an estimated 11% to 21% of the world’s population contracting the virus, but the CFR was much lower at 0.1% to 0.5%. The estimated global number of deaths varies widely, but even the highest of more than 1 million is a fraction of the deaths from COVID-19 (Roser, Reference Roser2020). In contrast with the high case rate of swine flu, in the 20 years from 2003 to 2023, there were 878 cases of avian flu reported globally, of which 458 were fatal (World Health Organization, 2023)—it had a higher CFR, estimated at 20% to 40%, but person-to-person transmission was limited to nonexistent (FutureLearn, 2020).

Table 3 presents the factor analysis of the 10 issues as national and personal threats. Given that the data are categorical—an issue is identified as a threat or not—the factor analyses are of the polychoric correlation matrices for national and personal threats. Three factors have eigenvalues greater than 1 in each factor analysis. Table 3 highlights the factors (with orthogonal varimax rotation) on which the issues have the highest factor loading. The issues in each factor are the same for national and personal threats; they reflect threats from out-groups, such as those from immigration, terrorism, and weak border control; threats from crime, such as burglary and hate crimes; and threats to health and resources, such as those from health pandemics, the environment, and the economy. Thus, the factor analysis confirms that these 10 issues cover the threat types described in our hypotheses and that health pandemics load with issues such as the environment rather than threats from out-groups.Footnote ¹⁰

Table 3. Exploratory factor analysis of the 10 security threats

We capture changing national and personal security threat contexts on these 10 issues by comparing newspaper attention to them in the year prior to the survey, using Nexis. We conducted a keyword and thematic search of the 10 issues in national newspaper articles (see the Appendix for a description of the keywords and themes), then used Nexis’s filters to focus only on “negative news” stories (i.e., stories that are threatening).Footnote ¹¹ Nexis also allows separation of “personal news” stories—that is, stories that are more relevant to perceptions of individual-level threat—from all other stories (which we term “national-level” stories as shorthand for national and international stories on these topics). Table 4 shows the analysis of the 10 issues for national and personal news stories. A context of increased threat for an issue is signified by a substantial increase in the number of such newspaper stories.Footnote ¹² By this criterion, 2020 was a context of increased threat from health pandemics, environmental issues, resource scarcity, immigration, and crimes against women and of unchanged or reduced threat from the other five issues.

Table 4. Coverage in national newspapers one year before June 2012 and July 2020 surveys

Note: Figures represent the number of stories resulting from theme and keyword searches in Nexis in the 12 months before the first day of each survey; see the Appendix for details

Large increases or decreases in attention to threats from a national frame also tend to be accompanied by large decreases or increases in attention to threats from a personal frame. They move together in direction and size of change: exceptions are crimes against women, for which coverage with a more personal frame increased more than coverage with a national frame, although both increased, and terrorism, for which coverage with a national frame decreased by 12%, while more personal stories pertaining to terrorism increased by 12%. Overall, among the three categories of threats outlined in Table 1, we have examples of issues that grew or declined in prominence for all three categories, allowing for satisfactory tests of our hypotheses.

Analysis of perceptions of security threats

To test the hypotheses, we pool the 2012 and 2020 surveys. We estimate two sets of models for perceptions of national and personal security threats: (1) initial examinations of the probability of identifying each of the 10 issues as a threat as a function of authoritarian predispositions, a dummy variable for 2020, and the interaction between the two; (2) the probability of identifying the three security threat types in Table 1—out-groups, inequality/resources, and crime—treating each individual’s response on each issue as a separate observation by stacking the data and including dummy variables for two of three threat types, and an interaction with the 2020 dummy variable.

We control for the same influences as Stevens and Vaughan-Williams (Reference Stevens and Vaughan-Williams2016) with the exception of religion and religiosity: mortality salience, television news consumption, sex, age, and education. We also interact these variables with the dummy variable for survey year to account for any changes in their relationships with threats.Footnote ¹³

Table 5 shows the estimates for the key independent variables for the hypotheses, indicating whether the sign and significance of the interactions are consistent with H1 and H2 given the changes in media attention shown in Table 4. The results are consistent with the hypotheses for 7 of the 10 issues (3 of 5 for which media attention increased, and 4 of 5 for which media attention stayed the same or decreased). All three issues pertaining to out-group threat accord with expectations. Immigration, for which there was increased threat in the context of the 2020 survey according to our analysis, was disproportionately more likely to be seen as a threat by authoritarians than in 2012. In contrast, the issues of terrorism and border control, for which there was little or no change in threat according to the analysis, show only main effects of authoritarian predispositions—that is, authoritarians are more likely to identify them as a threat, but there is no difference between 2012 and 2020.

Table 5. Perceptions of threat, authoritarianism and media coverage in 2012 and 2020—individual logit models

* p < .05.

Notes: Estimates are from logit models. All models control for mortality salience, media consumption, sex, education and age, along with their interactions with the survey year.

The estimates for the threats pertaining to inequality and resource scarcity confirm the hypotheses for two of the four issues—the economy and the environment. Both show authoritarians less likely to identify them as a threat than libertarians. For the environment, there is an additional, and expected, negative impact for authoritarians in 2020, when the analysis in Table 4 indicates a context of increased threat compared to 2012. Of the other two issues, perceptions of the threat of a health pandemic increased enormously in 2020: the sign on the interaction with authoritarianism is negative and close to statistical significance, in line with the expectation that authoritarians will be more resistant to conceding that COVID-19 is a national threat.

The three issues pertaining to crime also conform to expectations for burglary and racial and religious hate crimes: for the former, which affects everybody, authoritarians are more likely to identify it as a threat—and with little change in media attention in 2020, that relationship does not shift. At the same time, however, authoritarians are less likely to identify crimes against women as a threat—and the relationship is somewhat stronger in 2020, when the issue had garnered somewhat more media attention. Perceptions of hate crimes show a similar relationship, but this does not conform to expectations because the interaction term shows a stronger relationship with authoritarianism in 2020, even though the issue had receded a little in prominence.

With respect to personal threats, we tested competing hypotheses regarding whether changed national (H3a) or personal (H3b) threat contexts should have stronger relationships with authoritarianism. Table 5 suggests patterns that are generally similar. There are few interactions with different signs, but they are somewhat weaker for personal threat; there is one statistically significant interaction compared to four for national-level threat, indicating some support for H3a.

In the second set of tests, we stack the data, treating perceptions of threat for each of the 10 issues as a separate observation, meaning that our observations increase 10-fold (we cluster the standard errors to account for the nonindependence of observations for the same respondent). We analyze perceptions of threat as a function of authoritarian predispositions, the threat type—out-group, inequality/resources, or crime—and the interactions between threat type, authoritarian predispositions, and year.

Combining the issues into threat types also necessitates that we reconsider the implications of media coverage for increased or decreased threat. Using the net change in the percentage of stories with a national frame in Table 4 within the three different threat types to test H1 and H2 indicates little change in out-group threat—a net increase of 6%—and a large increase in threat from issues pertaining to inequality/resources, leading us to expect a negative interaction between authoritarian predispositions and the year dummy. The context of crimes against subgroups is mixed, with reports of crimes against women up in 2020 but of racial and religious hate crimes down by a larger percentage. On this basis, we would expect negative main effects of authoritarianism that are not moderated by year—because these are issues that are of lesser concern to authoritarians, and decreased coverage should not change that. Media coverage of the personal threat of the three threat types in Table 4 shows some variation from the context of national threat. Steep increases in stories with a more personal frame on terrorism and on crimes against women signal increases in out-group threat and crimes against subgroups rather than decreasing or staying the same as at the national level.

The estimates for national and personal threat are shown in the Appendix in Table A3.Footnote ¹⁴ We focus here on the relationships as represented in Figure 1, which shows simulations of perceptions of national and personal threat for the three threat types in 2012 and 2020, conditional on authoritarian predispositions, based on the model estimates.Footnote ¹⁵ The relationships are in the expected direction for perceptions of national threat for two of the three threat types. First, authoritarians are more likely to identify out-group threats. This is true in both 2012 and 2020. With the context of out-group threats changing little in terms of media attention with a national frame, there is also no change in the slope of the relationship. Similarly, for inequality/resource scarcity, Figure 1 shows the negative relationship suggested by H2b and in previous research by Duckitt and Sibley (Reference Duckitt and Sibley2010), in which authoritarians are less likely to identify such issues as threats. While the increased threat posed by COVID-19 in 2020 raised threat levels on these issues for all respondents, there is the expected negative relationship between authoritarian predispositions and the 2020 context—that is, authoritarians’ perceptions of these threats do not increase as much as libertarians; indeed, the marginal effect of authoritarianism in 2020 is significantly more negative than in 2012, in line with expectations. For crimes against subgroups, however, there is a statistically significant and unexpected negative interaction with year. This is not because decreased coverage of crime unexpectedly lowers perceptions of threat for authoritarians, however. Authoritarian perceptions of the threat change little, as expected, but libertarian perceptions of the threat from crime against subgroups are considerably higher in 2020 than in 2012.

Figure 1. Relationships between authoritarian predispositions, media attention, and identifying issues as threats.

Turning to H3a and H3b, the lower half of Figure 1 depicts the relationships between personal threat context, authoritarian predispositions, and perceptions of personal threat. If change in personal threat context has effects on perceptions like those outlined in Table 1, we would expect a positive interaction between authoritarianism and out-group threat and negative interactions with inequality/resources and crimes against subgroups. We see these relationships for out-group and inequality/resources threat but, again, unexpected results for crimes against subgroups in that there is not a negative interaction with authoritarianism.