In recent years, North America and Europe have seen a wave of high-profile terrorist attacks with a significant number of casualties. This has raised concerns about the consequences of an increased perceived threat from terrorism for the civil liberties fundamental to liberal democracy. Indeed, prior research has found that terrorist attacks induce a greater willingness among citizens in liberal democracies to sacrifice civil liberties pertaining to privacy and personal freedom—for example, increasing surveillance of the general public—in order to increase security and prevent future terrorist attacks (Silver and Davis, Reference Silver and Davis2004; Sniderman et al., Reference Sniderman, Petersen, Slothuus, Stubager and Petrov2019).
Observing such illiberal reactions raises the question of why they come about and, relatedly, if this inclination is a universal feature of individuals, or rather confined to certain segments of the population. To this end, an important line of research within political psychology has theorized and empirically examined how deep-seated predispositions can be activated by threats from terrorism to alter and structure attitudinal responses including support for civil liberties. One manifestation of this research program has looked at the role of authoritarianism (Feldman and Stenner, Reference Feldman and Stenner1997; Hetherington and Suhay, Reference Hetherington and Suhay2011).
In this paper, we focus on how exposure to terrorism interacts with another fundamental psychological predisposition, namely ethnocentrism; a “deep” individual difference—partly hereditary, partly socialized (Kinder and Kam, Reference Kinder and Kam2010, 33–34)—spearheaded in the realm of political psychology by Kam and Kinder (Reference Kam and Kinder2007, Reference Kam and Kinder2012) and Kinder and Kam (Reference Kinder and Kam2010). While the literature offers different more or less elaborate conceptualizations of ethnocentrism, the essence revolves around individual differences in group-based, social-hierarchical “us against them” thinking (Kinder and Kam, Reference Kinder and Kam2010, 7–8, 31–32; Bizumic and Duckitt, Reference Bizumic and Duckitt2012; Bizumic, Reference Bizumic2019). More specifically, the core aspects of ethnocentrism are a tendency to differentiate categorically between in-groups and out-groups, and to maintain the “purity” and cohesion of the in-group by derogating out-groups (Kinder and Kam, Reference Kinder and Kam2010, 7–8, 31–32; Bizumic and Duckitt, Reference Bizumic and Duckitt2012).
Ethnocentrism has been found to influence a wide set of political attitudes, but it is especially consequential when activated (primed) by information or events that link a given policy issue to salient social groups, and therefore lend itself more strongly to group-based thinking (the activation hypothesis) (Kam and Kinder, Reference Kam and Kinder2007, Reference Kam and Kinder2012). Terrorism is an example of a focusing event, which—through intensive media coverage—highlights group conflict and, thus, can activate ethnocentrism to explain a range of attitudes, even toward phenomena that are only subtlety connected to such group-based thinking. Kam and Kinder (Reference Kam and Kinder2007) and Kinder and Kam (Reference Kinder and Kam2010) substantiate this claim by showing that the 9/11 attacks activated ethnocentrism to explain a wide range of attitudes, including support for war on terrorism, in the United States.
Kam and Kinder's studies of the terrorism-induced activation of ethnocentrism are foundational, but the scope of their finding is unknown. First, their study exclusively focuses on the United States; an ethnically and socially highly diverse country characterized by idiosyncratic group relations. As Kinder and Kam (Reference Kinder and Kam2010, 229) highlight themselves, this raises the question of whether their finding regarding the importance of ethnocentrism travels to other contexts characterized by different group constellations. Second, the 9/11 attacks are unique on many dimensions—they marked the beginning of a “new age of terrorism” and are unprecedented in magnitude. This in turn prompts the question of whether only very significant attacks—the 9/11 attacks arguably being a “most likely” case—have the leverage to activate ethnocentrism. Third, Kam and Kinder examine a wide range of policy and group attitudes and evaluations, but it is unknown whether the activation hypothesis extends to attitudes toward fundamental civil liberties. Beyond sentiments toward different groups and policies within a polity, it is important to know whether activated ethnocentrism also shapes support for the very principles that set the boundaries of liberal society. Further, such attitudes toward civil liberties are arguably more conceptually distant from ethnocentrism than, for example, group evaluations, and the ethnocentrism-civil liberties link therefore speaks to the reach of the activation hypothesis. In short, the scope of the terrorism-induced activation of ethnocentrism hypothesis beyond a restricted set of attitudes in the American context is an open question.
In this paper, we contribute with new empirical evidence on this question. Specifically, we study the activation of ethnocentrism vis-à-vis attitudes toward civil liberties among citizens in Germany—a large European country, markedly different from the US in terms of significant group categories and relations. We do this by using a cascade of terrorist attacks in 2016, coinciding with the collection of the German General Social Survey (GGSS) (Bauernschuster et al., Reference Bauernschuster, Diekmann, Hadjar, Kurz, Rosar, Wagner and Westle2017). Because all of the attacks revolved around conflicts between natives and immigrants, we expect them to activate ethnocentric predispositions among German citizens, specifically in relation to the fundamental tradeoff between civil liberties and security concerns prompted by the threat of terrorism. We find strong and consistent support for the activation of ethnocentrism in Germany; after the terrorist attacks, ethnocentric dispositions were much more predictive of native Germans' attitudes toward civil liberties, and the increase in restrictive responses is concentrated among the more ethnocentric individuals.
1. Research design, data, and measures
Parallel to related previous studies (e.g., Finseraas and Listhaug, Reference Finseraas and Listhaug2013; Legewie, Reference Legewie2013), we leverage a quasi-experimental interrupted time-series research design. However, we depart from common usage of the quasi-experimental set-up in that we are not primarily interested in comparing a given attitude before or after the attack (we do this in a secondary analysis), but instead focus on whether ethnocentrism more strongly predicts support for civil liberties (i.e., is activated) in the immediate aftermath of the attacks. To this end, we utilize that the German General Social Survey (GGSS) (response rate: 34.6 percent; see Appendix A for further information), collected between April 6 and September 18, 2016, coincided with several terrorist attacks in France and Germany in July 2016. More precisely, the Nice truck attack on July 14 (86 deaths, 434 injured); the Würzburg train attack on July 18 (five injured); the Munich shootings on July 22 (nine deaths; 36 injured); the Ansbach suicide bombing on July 24 (15 injured); and the Normandy Church attack on July 26 (one death, one injured).
The terrorist attacks took place in a climate in which native/citizen-immigrant conflict was already significant in Germany following recent highly publicized terrorist attacks (in Paris in January and November 2015, and in Brussels and Orlando earlier in 2016) and the influx of more than 1 million refugees from predominantly Muslim majority countries in 2015 (BAMF, 2016, 5, 20, 25). As such, the German public—German citizens to be precise—was plausibly already subject to ethnocentric political thinking vis-à-vis civil liberties and other attitudes.Footnote 1 In effect, our test of the activation hypothesis is a conservative one; we analyze if ethnocentric dispositions can be triggered even further in the face of geographically proximate terrorist attacks.
The interrupted time-series design allows us to examine the association between ethnocentrism and support for civil liberties in a brief time window around the terrorist attacks, which—contingent on the as-if randomness in the timing of the interview (Muñoz et al., Reference Muñoz, Falcó-Gimeno and Hernández2020)—are expected to induce quasi-experimental variation in the conditions prone to activate ethnocentrism (i.e., more fertile conditions after the attacks). Because a cascade of attacks occurred within 12 days of the Nice attacks, we effectively study the cumulative activation effect of several terrorist attacks. To minimize the risk of contamination of the terrorist attack with other temporally coinciding influences, we restricted our sample to a relatively narrow time window around the attacks (but expand this in auxiliary analyses, see Appendix Q). More specifically, the pre-attack period (i.e., the control group) was restricted to 20 days prior to the Nice attack (June 24 to July 13, n = 238) and the post-attack period (i.e., the treatment group, n = 59) to the first 7 days after the last attack (the Normandy attack) (July 27 to August 2). As we show in Appendix K, the groups interviewed before and after the attacks were generally similar on various demographic characteristics, thus indicating their comparability.
Inspired by previous work (Kleinpenning and Hagendoorn, Reference Kleinpenning and Hagendoorn1993; Bizumic et al., Reference Bizumic, Duckitt, Popadic, Dru and Krauss2009; Bizumic and Duckitt, Reference Bizumic and Duckitt2012; Rapp Reference Rapp2016), we measure ethnocentrism by an additive index (α = 0.85)—scaled so 0 reflects the least and 1 the most ethnocentric predispositions—based on ten survey questions tapping various aspects of in-group-centeredness vis-à-vis the native-foreigner distinction, e.g., “It is better for a country if all people belong to a common culture” (see Appendices E and F). Although varying in generality, the items ostensibly tap into the same underlying construct as indicated by the scales' high reliability. In Appendices G and N, we discuss an alternative measure of ethnocentrism, and show that the main results are generally robust to using this measure.
Underlying our analysis is the assumption that ethnocentrism is a stable predisposition that is not in itself affected by the terrorist attacks. In Appendix J, we evaluate this assumption empirically and show that ethnocentrism is indeed highly stable throughout the survey period and, correspondingly, we do not find significant differences in means between our pre- and post-attack windows. It is thus not ethnocentrism per se, but its relationship with support for civil liberties that may potentially change after the terrorist attacks.
The dependent variable, support for security concerns (relative to civil liberties), is measured by combining three survey items regarding governmental authorities' right to ignore civil liberties upon a suspected terrorist act (e.g., “to detain people for as long as they want without putting them on trial?”) into an additive index (α = 0.81) (see Appendix I).
2. Results
In line with previous studies, we observe a significantly greater willingness to sacrifice civil liberties for security in the immediate aftermath of the cascade of terrorist attacks in 2016 as indicated by the positive coefficient on the post-attacks (“treatment”) dummy in Model 2 (Table 1). Ethnocentrism is also positively associated, although not significantly, with a preference for safety over liberty as evidenced by Model 1 and 2. However, most important for our purposes, we find support for the activation hypothesis in observing that ethnocentrism much more strongly predicts willingness to compromise civil liberties after the terrorist attacks as indicated by the positive and significant interaction term (0.68) in Model 3. After the attacks, a change from the 5th to the 95th percentile in ethnocentrism corresponds to a dramatic increase in support for security concerns of 0.44 on the 0–1 scale. Further, perhaps surprisingly, ethnocentrism appears to be unrelated to the security-civil liberty trade-off before the terrorist attacks as is evident from the main term in Model 3. This strongly supports the idea of ethnocentrism being a latent predisposition that can be dormant at times, where group distinctions are less salient, and then suddenly activated to structure attitudes—even more distant and principled ones—by focusing events that bring such group thinking to the fore. In Appendices L–T we report results reproducing this finding in various robustness checks.
Activation of ethnocentrism vis-à-vis security concerns versus civil liberties
Note: OLS estimates with robust standard errors in parenthesis. Stars indicate level of significance. **p < 0.01; * p < 0.05.
The models include sociodemographic controls (sex, age, age2, education, employment status) and account for geographical clustering and oversampling respondents from the former GDR through weighting (see Appendices A and E).
Figure 1 elaborates on this finding by showing differences in predicted levels of support for security (based on Model 3, Table 1) before and after the terrorist attacks for different levels of ethnocentrism (5th, 50th, and 95th percentile). Correspondingly, Figure L1 in Appendix L shows the marginal effects of the attacks and associated confidence intervals for different levels of ethnocentrism. Whereas there is no difference across levels of ethnocentrism in the predicted means before the terrorist attacks, we see a marked polarization after the attacks. For highly ethnocentric respondents (95th percentile), we observe a marked and statistically significant change in predicted means of 0.37 (from 0.53 to 0.90) after the terrorist attacks, whereas there is no significant change for the less (5th percentile) ethnocentric individuals. A more moderate and significant shift (a predicted 0.11-point change) is observed among those with the median level of ethnocentrism. More generally, as Figure L1 shows, we observe statistically significant shifts (increases) in support for security concerns after the terrorist attacks for approximately the upper two-thirds of the ethnocentrism scale. In short, when ethnocentric dispositions are activated by terrorism, individuals, who harbor more ethnocentric dispositions, tend to downgrade the importance of civil liberties over security concerns, ostensibly because they come to view this trade-off through the lens of group hierarchies due to associations of out-groups with terrorism.
Predicted support for security concerns for different levels of ethnocentrism.
Note: Estimates are predicted means based on Model 3 (Table 1) for the pre-attacks and post-attacks groups on three levels of ethnocentrism (the 5th percentile, the 50th percentile, and the 95th percentile)
We can evaluate the specificity of the activation of ethnocentrism by looking at citizens' attitudes toward a broader range of government actions relating to the security versus civil liberties trade-off, but not specifically implicating terrorism. If group associations are the “activating mechanism”, as we have argued, we would only expect ethnocentrism to be activated by security concerns with reference to terrorism, but not those without such reference, which consequently do not directly implicate out-groups. In Model 4–6, we examine the activation hypothesis vis-à-vis an index (α = 0.78) composed of four trade-offs regarding civil liberties versus general security concerns (e.g., surveillance in public spaces; see Appendices E and I). For this alternative “non-group implicated” measure of the security versus civil liberties trade-off, we find no evidence for ethnocentrism being activated by the terrorist attacks; the interaction term between ethnocentrism and the post-attack dummy (Model 6) is essentially zero and not statistically significant. This indicates that ethnocentrism is activated by the specific reference to terrorism in the outcomes studied, which in turn evokes group-based thinking in the wake of terrorism.
Also related to the specificity of the activation of ethnocentrism, we examined to which extent the observed activation effect overlaps with the activation of authoritarianism—another prominent predisposition, which can also be activated by perceived threat (Feldman and Stenner, Reference Feldman and Stenner1997; Hetherington and Suhay, Reference Hetherington and Suhay2011). We assess this by including a commonly used measure of authoritarianism (relating to conformity in children, see Appendices E and H) and interacted it with the post-attack dummy. We see indications of an activation of authoritarianism (a positive interaction term in Model 9), although this is not statistically significant. This generally squares with the findings of a modest activation effect for authoritarianism reported in a recent multi-country study (Claassen and McLaren, Reference Claassen and McLaren2021). More important for our purposes, when we simultaneously examine the activation of ethnocentrism and authoritarianism (Model 10), we see that the activation effect for ethnocentrism remains significant and, if anything, is even stronger. Thus, ethnocentrism is not only conceptually distinct from authoritarianism; it also operates in a distinctive way vis-à-vis activation by terrorism. In Appendix O, we scrutinize the role of ideology (left-right self-placement) in a similar manner and find inconsistent indications of its post-terrorism activation, but no evidence that it confounds the activation of ethnocentrism.
Lastly, we examine when, during the cascade of terrorist attacks, the activation of ethnocentrism occurred and for how long it lasted. Figure 2 maps the coefficient of the interaction term (in 32 reestimations of the main model (Model 3), where the seven-day time window defining the post-attacks group is moved by one day at a time) in the first five weeks after the first attack (from July 15 to August 21) (for a parallel approach, see Legewie Reference Legewie2013; see Appendix P for details). From the figure it is evident that ethnocentrism is activated immediately after the first attack (the Nice attack), and that the activation effect gradually increases in strength as more attacks took place, before waning (the interaction term becoming insignificant and close to zero) around two weeks after the last attack (the Normandy attack).Footnote 2, Footnote 3 Interestingly, the pattern of a gradually stronger activation effect as more attacks occur fits with the logic of priming (Krosnick and Kinder, Reference Krosnick and Kinder1990); as the salience of terrorism (see Appendix C substantiating this in terms of media coverage) heightens after several terrorist attacks, people increasingly come to rely on their ethnocentric predispositions when forming attitudes toward civil liberties. Comparatively speaking, the persistence of the activation effect is roughly similar to the most persistent “direct” effects of terrorism on immigration attitudes observed by Legewie (Reference Legewie2013), and, more generally, it matches the relatively short-lived attitudinal effects of terrorism before the return to a baseline (“perturbation effects”) that Sniderman et al. (Reference Sniderman, Petersen, Slothuus, Stubager and Petrov2019) observe in their review. When primed by terrorism, ethnocentrism continues to feature prominently in public opinion formation for a while. While the effect is relatively short-lived, an interesting question for future work is whether more recent experiences of an activation of ethnocentrism sensitize citizens to its subsequent rekindling by future attacks.
The activation of ethnocentrism in the first five weeks after the Nice terrorist attack.
Note: Dots indicate the interaction terms based on 32 consecutive seven-day “rolling window” calculations based on Model 3 (Table 1) estimated by OLS regression with robust standard errors. In the first calculation, the post-attacks group consists of the respondents interviewed on the first day immediately after the Nice attack and during the following six days (i.e. 15 July to 21 July). In the second model, the treatment group consists of the respondents interviewed from the 16 July to 22 July and so on. The pre-attacks group remains the same in all calculations. Vertical lines indicate days of terrorist attacks. Solid dots indicate level of significance (p < 0.05).
3. Concluding remarks
The 9/11 attacks brought ethnocentrism to the fore as a core structuring factor of American public opinion (Kam and Kinder, Reference Kam and Kinder2007, Reference Kam and Kinder2012; Kinder and Kam, Reference Kinder and Kam2010). Yet, little is known about the role of ethnocentrism outside of the US, including whether terrorism serves the same activating function in other contexts with different in- and out-group constellations, nor whether the activation extends toward more principled attitudes regarding respect for civil liberties. In this paper, we have demonstrated, using a quasi-experimental design, that ethnocentrism operates in parallel ways in Germany. In the summer of 2016, a cascade of nearby terrorist attacks involving immigrants (a primary out-group) activated ethnocentrism among German citizens to powerfully predict (lower) support for civil liberties relative to security concerns after its influence had been absent just a month before.
Theoretically, finding that the activation of ethnocentrism extends to a context with different in- and out-group dynamics is consistent with an understanding of ethnocentrism as a universal feature of humans' (political) psychology. Obviously, more comparative work must be conducted to more firmly establish this claim, but our findings indicate that such efforts are worthwhile. One pertinent methodological challenge to studying the role of ethnocentrism systematically across contexts is to develop cross-culturally equivalent indicators of ethnocentrism. Further, understanding the conditions beyond terrorism under which ethnocentrism is activated—and toward which attitudinal objects—is another important theoretical and substantive question to pursue in future work.
Supplementary material
The supplementary material for this article can be found at https://doi.org/10.1017/psrm.2022.5
To obtain replication material for this article, please visit https://doi.org/10.7910/DVN/RRLCNP
Acknowledgments
The authors wish to thank Kim Sønderskov, UIrich Kohler, and Fabian Class for advice and helpful feedback on the paper. They also thank participants at the WZB kolloquium (2019) and the ISPP conference (2020) for inputs. Dinesen acknowledges support from Independent Research Fund Denmark (Grant number: 9038-00123B).
