Terrorism and Immigration Policy Preferences

ABSTRACT What is the causal impact of terrorism on immigration policy preferences? Under what circumstances and due to which psychological micro-mechanisms does this impact materialize? To answer these questions, we provide evidence from pre-registered and well-powered experiments for Germany and the United Kingdom. We find that anti-immigration responses to terrorism follow an emotional proximity rationale: terrorism leads to more restrictive migration policy preferences only among individuals with high levels of perceived insecurity, especially when terrorism occurs in their own country. Policy preferences are not affected by terrorism abroad or by information cues on the objectively low probability of being victimized.


Introduction
In recent years, social scientists have extensively studied the effect of terrorism on social trust, political attitudes and political behavior (e.g. Bali 2007;Berrebi and Klor 2008;Getmansky and Zeitzoff 2014;Geys and Qari 2017;Balcells and Torrats-Espinosa 2018;Castanho 2018;Breton and Eady 2022; for further discussion, see Sandler 2019, 2022). Within this research field, several studies have investigated the impact of terrorist attacks on attitudes towards immigrants and migration (e.g. Legewie 2013; Nussio, Bove, and Steele 2019; for an overview see Helbling and Meierrieks 2022;Gaibulloev and Sandler 2022). This literature tends to suggest that terrorism is associated with less favorable attitudes towards immigration, e.g. due to increased out-group hostility and a growth in political conservatism.
In our study, we are interested in the effect of terrorism on individual preferences about immigration policies. The nexus between terrorism and immigration policies has so far received only very limited attention (e.g. Bandyopadhyay and Sandler 2014;Finseraas, Jakobsson, and Kotsadam 2011;Finseraas and Listhaug 2013;Solheim 2021). Focusing on policy measures and preferences allows us to go beyond research on the attitudinal consequences of terrorism (e.g. in the form of increased xenophobia) and instead study the policy effects of terrorism associated with such attitudinal changes. For instance, terrorism may lead to stronger feelings of uncertainty and loss of control; this, in turn, may induce changing preferences about immigration policies, where more restrictive immigration policies might be seen as a means to regain control (Lerner and Keltner 2001;Lerner et al. 2003).
The literature on terrorism and attitudes towards migrants has largely relied on natural experiments, exploiting the plausibly random effect of terrorist attacks on respondents of surveys that were in the field during an attack. By doing so, these studies aim at a high level of internal validity, while also reaching a reasonable degree of external validity. A shortcoming of this literature, however, is that it has paid little attention to the mechanisms connecting terrorism with preferences, focusing on the question of whether rather than why terrorism should affect public opinion and political preferences. This is a consequence of natural experiments being event-specific (Muñoz, Falcó-Gimeno, and Hernández 2020). It is thus not possible to differentiate between different kinds of terrorist attacks; also, often survey questions have not been developed to examine micromechanisms and the underlying psychological linkages between the terrorist threat and individual preferences. An additional problem is that potential moderators will also be affected by the events surrounding these natural experiments, as -by design -it is not possible to avoid post-treatment effects (Muñoz, Falcó-Gimeno, and Hernández 2020: 194).
For this reason, we complement existing experimental studies on the role of terrorism in attitudes towards migrants and migration policy preferences with survey experimental evidence that allows us to use specific treatments and thus maximize internal validity. In contrast to natural experiments, survey experiments make it possible to completely isolate treatments and test different theoretical mechanisms of interest. For instance, real-world terrorist incidences might be directly followed by collateral reactions such as specific media coverage that frame the event in a certain way and thereby affect people's reactions (Solheim 2021). Similarly, terrorist attacks may occur in the context of other events such as elections or commemorations, which might change the meaning and perception of the terrorist attack (Muñoz, Falcó-Gimeno, and Hernández 2020: 190). Our survey experiment approach allows us to avoid such issues.
For our experiment, we examine people's attitudes towards immigration policies after they have read a vignette that includes information on terrorist attacks. We vary the saliency of terrorism by referring to the increase in the frequency and lethality of Islamist terrorism in either the respondent's own country or outside Europe since the 1990s. Moreover, people are randomly informed about the low probability of being killed in a terrorist attack. Providing such information is comparable to studies that investigate the psychological effect of news reports about terrorism (Matthes, Schmuck, and von Sikorski 2019;Makkonen et al. 2020). This research design guarantees a certain degree of external validity even if we cannot measure the effects of real-world attacks. While the geographical distance of the attacks and the probability of being affected by terrorist attacks constitute rational considerations that might moderate the effects of terrorism, we also argue that people's reactions depend on emotional considerations. In particular, we expect stronger effects on policy preferences among fearful people as they are anticipated to be more sensitive to violence and have a higher likelihood to avoid risks (e.g. Skitka et al. 2006;Huddy and Feldman 2011;Igarashi 2021).
To generalize our findings, we conduct our survey experiment among representative samples in Germany and the United Kingdom, where several major terrorist incidences occurred in the 2010s and where various natural experiments have already been conducted (e.g. Larsen, Cutts, and Goodwin 2020;Nägel and Lutter 2020;Nussio 2020). Our findings suggest that the causal effects of terrorism on immigration policy preferences are only observable among respondents whose preexisting levels of perceived insecurity are high; however, this sub-group accounts for a nonnegligible portion of our two nationally representative samples. We argue that anti-immigrant reactions to terrorism follow an emotional proximity rationale: respondents favor more restrictive migration policies due to terrorism when terrorism feels threatening (as it attaches to latent insecurity) and proximal (as it occurs in one's country); by contrast, policy preferences are not affected by our primes on terrorism abroad or by rational cues on the objectively low probability of becoming a victim.

Theoretical Arguments
Terrorism is commonly defined as the 'premeditated use or threat to use violence by individuals or subnational groups against noncombatants in order to obtain a political or social objective through the intimidation of a large audience beyond that of the immediate victims' (Enders, Sandler, and Gaibulloev 2011: 321). 1 A number of studies find that terrorism makes citizens more hostile towards immigrants (e.g. Boomgaarden and De Vreese 2007;Jungkunz, Helbling, and Schwemmer 2019;Böhmelt, Bove, and Nussio 2020;Ferrín, Mancuso, and Cappiali 2020; for contrary evidence, see Castanho 2018). This increased hostility may have two causes. First, if a terrorist attack can be linked to foreign actors (e.g. because a foreign terrorist group has carried out the attack), this may lead to increased threat perception which, in turn, increases hostile attitudes towards sections of the population that can be associated with the attack, e.g. because such sections have a similar ethnic or religious background to the attacking foreign actor. 2 Second, we may also expect that the intimidation caused by a terrorist attack produces feelings of uncertainty and loss of control (Lerner and Keltner 2001). To regain control and decrease the risk of further attacks, more restrictive immigration policies may be favored. For instance, as shown by Brader, Valentino, and Suhay (2008) and Iyer et al. (2014), the threat of terrorism leads to increasing support of conservative policies (which may include more restrictive immigration measures) to avoid further attacks. Whether or not closed borders and other more restrictive immigration policies actually reduce terrorism risks remains uncertain (Choi 2018;Dreher, Gassebner, and Schaudt 2020). However, such policies may nevertheless be favored in the aftermath of terrorism because they provide people with the impression that they are in control of the situation, making them feel more secure. 3 For these reasons, it is plausible to expect that in the context of terrorism people prefer increased border controls and a reduction of immigration flows: H1: Terrorist attacks lead to increased support for more restrictive immigration policies.
However, it is possible that the hypothesized response to terrorism (in the form of increased support for immigration restrictiveness) is not uniform but heterogeneous. Below, we develop three arguments and hypotheses that speak to this idea.
First, we may expect increasing support for restrictive immigration policies due to terrorism to be especially pronounced among people who are more fearful. Following other migration scholars, fearfulness refers to perceived insecurity with respect to crime and violence (e.g. Sniderman, Hagendoorn, and Prior 2004). Research shows that fearful people gravitate towards more conservative, right-wing and even extremist positions (e.g. Huddy and Feldman 2011;Jost et al. 2017). Despite the importance of fearfulness for attitudes and behavior (Brader, Valentino, and Suhay 2008;Jost et al. 2017), the direct and moderating impact of this trait on policy preferences in the context of migration in general, and in the context of terrorist threats in particular, is unknown. We argue that fearful people are more susceptible to intimidation and perceive higher levels of risk (Eysenck 1992;Lerner and Keltner 2001). Moreover, they should generally be more sensitive to violence (Igarashi 2021) and have a higher likelihood to show risk-avoiding behavior (Skitka et al. 2006;Huddy and Feldman 2011). In contrast to anger, anxiety makes people avoid out-group members (Mackie, Devos, and Smith 2000). For this reason, fearful people are anticipated to be rather opposed to military actions but favor preventive policy measures (e.g. Marcus, Neuman, and MacKuen 2000;Skitka et al. 2006). Accordingly, fearful people should be more supportive of more restrictive immigration policies, given that they feel more threatened by terrorist attacks and consider closed borders as a risk avoidance measure:

H2: The effects of terrorist attacks on increased support for more restrictive immigration policies is stronger among fearful individuals.
Besides emotional reasons, the effect of terrorism on immigration policy preferences could also be moderated by a geographical proximity rationale. It is reasonable to expect that attacks occurring in one's own country increase the intimidation effect, as the threat appears more realistic and the personal vulnerability more imminent (e.g. Huddy et al. 2005: 595). Our research design allows us to differentiate between the impact of terrorism abroad and at home on immigration policy preferences. In line with the geographical proximity rationale, we study whether people might be more eager to gain control over immigration flows and prefer more restrictive regulations when attacks occur in their own country: H3: The effects of terrorist attacks on increased support for more restrictive immigration policies is stronger if terrorism occurs in one's own country.
Finally, while some people are more easily affected by terrorists' intimidations for emotional reasons, others may rationally calculate the risks of being affected by terrorist attacks. Indeed, the probability of being killed in a terrorist attack is generally extremely low. For instance, it is much more likely to be killed in a road accident (Wilson and Thomson 2005). We therefore expect individuals to demand more restrictive migration policies especially when they are unaware of the extremely low probability of being affected by terrorism. Fear has detrimental effects on cognitive functioning (Eysenck 1992) and, consequently, a threat might appear bigger than it is. Under such circumstances, we might observe the phenomenon of probability neglect, meaning that individuals focus on the bad consequences of an attack and ignore the fact that the odds of being affected by such attacks are extremely low (Sunstein 2003). Accordingly, the support of measures for protection and control might be potentially excessive under such circumstances. 4 However, if people are aware of the low probability of being affected by such attacks, they might be less intimidated and thus less concerned about the regulation of immigration flows:

H4: The effects of terrorist attacks on increased support for more restrictive immigration policies is mitigated if individuals know that the probability of being affected is low.
The test of H2 to H4 will allow us to unpack some of the conditions (predisposition to fear, proximity to an attack and information/education about terrorism) that determine when terrorism can shape policy preferences. For instance, evidence in favor of H2 but not H4 would imply that connections between terrorism and immigration preferences are rooted in subjective feelings of threat proximity. As another example, support for H4 would imply that the connection between terrorism and immigration is simply based on misperceptions of terrorism threat than can be rationally countervailed with objective information.

Data
We conducted two pre-registered online surveys in Germany (N = 1,283) and the United Kingdom (N = 1,409) between 15 and 27 November 2019. We selected these two countries as they have witnessed several major terrorist incidences over the last decade, making the vignettes we used in our survey more realistic. Moreover, a comparison of these two countries allows us to generalize our findings to some extent. 5 Our sample includes individuals living in Germany or the United Kingdom between the ages of 18 and 69 who were sampled according to their gender, age and education. 6 The survey was conducted by Respondi; this company maintains an ISO-certified online access survey with around 100,000 potential respondents per country. 7 Our empirical analysis consists of a factorial survey experiment (Auspurg and Hinz 2015). We rely on a between-subject design in which each respondent answers a number of questions after reading a specific vignette. As we study the respondents' immediate reactions to different vignettes, our research design does not allow us to draw conclusions on long-term effects. Rather, our setting is comparable to studies investigating policy attitudes in the immediate aftermath of a terrorist attack (e.g. Finseraas, Jakobsson, and Kotsadam 2011;Finseraas and Listhaug 2013;Solheim 2021). In contrast to these natural experiments, however, we are able to unpack the causal effect of different theoretical micro-mechanisms connecting the threat of terrorism with immigration preferences.
To test the effect of terrorism on migration policy preferences, we use responses to the following question as the dependent variable: 'To what extent do you think Germany/the United Kingdom should allow immigrants to come and live here?' 8 Response categories are 'allow many', 'allow some', 'allow a few' and 'allow none'.
To test our hypotheses concerning the impact of terrorism on immigration policy preferences (especially H1 and H3), we randomly vary the saliency of terrorism by referring to the increase in the frequency and lethality of Islamist terrorism in either the respondent's own country or outside Europe since the 1990s; there is no reference to terrorism for the control group. 9 These statements prime participants to think of Islamist terrorism when answering the question about immigration. We refer to Islamist terrorism because this type of terrorism is responsible for most large-scale terrorist attacks that have affected Germany and the United Kingdom during the last decade and usually has a transnational dimension (e.g. as perpetrators have a migrant background), allowing for plausible links to migration.
To study the role of probability neglect (H4), respondents receiving the terrorism treatment were further randomly assigned to a group that was informed about the low probability of being killed by terrorism compared to being killed in a road accident or to a group that did not receive such information. 10 Furthermore, we argue that fearfulness could moderate the effect of terrorism on policy preferences (H2). Fearfulness refers to respondents' perceived insecurity with respect to crime and violence and has already been used in research on attitudes towards immigrants (e.g. Sniderman, Hagendoorn, and Prior 2004). This moderator variable is measured by answers to the following two questions: 'How safe do you feel when walking alone in your neighborhood after dark?' (answer categories are 'totally safe', 'partly safe', 'partly unsafe', 'totally unsafe') and 'To what extent do you agree with the following statement: 'I fear increasing violence and vandalism in my neighborhood?' (answer categories are 'fully agree', 'partly agree', 'partly disagree', 'fully disagree'). We use a summative score of both items, since they show high levels of scale reliability (Cronbach's α = 0.69). In contrast to natural experiments, we are able to ask these questions before the experiments to avoid post-treatment effects (see Muñoz, Falcó-Gimeno, and Hernández 2020: 194). People's fearfulness is thus not affected by the terrorism vignettes.
Our experiment is very well-powered according to standard calculations. As it is conventional to do, we assume a 95% level of statistical significance and 80% statistical power. We then calculate the standardized difference that we would expect to find between our control and each of our treatment groups concerning differences between the means and standard deviations of our outcome in the population (Jones, Carley, and Harrison 2003). We conservatively assume a 5% difference in the means of our outcomes between our treatment and control group (equivalent to a 0.2-point difference on a 4-point scale) and a standard deviation of 0.8 in our outcome. According to these assumptions, we would need around 176 individuals per treatment group for a well-powered analysis. The N of our treatment conditions range from 729 to 985 in the various terrorism saliency treatments and from 244 to 738 in the probability neglect treatment. It is thus unlikely that any null results are due to an underpowered design.

Empirical Results
We report our results assessing the effect of terrorism on migration policy preferences by means of OLS analyses in Table 1 . 11 Contrary to H1, Model 1 indicates that information on terrorism does not have a statistically significant effect on immigration policy preferences. Similarly, the results do not support H3: terrorism does not affect policy preferences regardless of whether it occurs at home or in other parts of the world. 12 By contrast, Model 2 provides support for H2: terrorism leads to greater demand for restrictive migration policies among individuals who are fearful. 13 As a robustness check, Model 3 in Table 1 replicates the same interaction effect while controlling for gender, age and education, since those demographic variables could drive both fearfulness and immigration policy preferences. 14 Indeed, men, individuals with lower education and older people prefer more restrictive immigration policies, as already shown by Finseraas, Jakobsson, and Kotsadam (2011). However, there is little evidence that these control variables affect our main findings, which speaks to the soundness of our experimental design. Figure 1 depicts the marginal effects associated with the interaction effect from Model 2. When primed on terrorism at home, policy preferences are significantly more hostile towards migrants among highly fearful respondents. By contrast, when primed on terrorism abroad, the sign of the interaction is the same, but the marginal effects are negligible in both magnitude and statistical significance. This suggests a two-fold proximity rationale linking security threats to migration. First, terrorism only creates demand for more restrictive policies when occurring at home. Second, terrorism only matters to a subset of the population for whom personal security threats are more salient. However, this subset (reporting values of 3 or above on our 4-points fearfulness scale) is not negligible, accounting for 31.4 percent of respondents in Germany and approximately 25 percent in the United Kingdom, respectively. Finally, Models 4 and 5 of Table 1 explore whether exogenously manipulated information on the actual probability to be a terrorism victim affect individual migration policy responses. 15 Contrary to H4, we do not find that information about the low probability of being a victim affects policy preferences, regardless of whether it concerns terrorism at home or abroad. Table A6 in the Appendix shows that the effect of the probability treatment is also not conditional on levels of perceived individual insecurity. This suggests that the terrorism threat effect among fearful people (Model 2, Table 1) is driven by an emotional reaction to terrorism that feels proximal (as it occurs in one's country) and not by rational calculations of terrorism as an objective threat.

Conclusions
We present results from a pre-registered and well-powered factorial survey experiment for Germany and the United Kingdom, testing whether terrorism induces stronger preferences for restrictive migration policies. Our study yields three main findings. First, information on terrorism does not shape migration policy preferences on average. This is in line with earlier studies that have shown that attitudes towards immigrants remain relatively stable, even in the light of economic and political shocks (Hopkins, Sides, and Citrin 2018;Kustov, Laaker, and Reller 2021). Among the studies that examined the effects of terrorist attacks via natural experiments, there are several contributions that also find no effects on attitudes towards immigrants (Castanho 2018;Larsen, Cutts, and Goodwin 2020;Giani 2021) or rather short-lived effects in only a small number of countries (Legewie 2013).
Second, however, terrorist attacks perpetrated in one's home country significantly and substantially increase anti-immigrant policy preferences among respondents with high levels of fearfulness, which accounts for a substantial share of the total population (31.4 percent and 25 percent in Germany and the United Kingdom, respectively). This effect is not observed when terrorism occurs in countries outside Europe. This highlights the importance of fearfulness and perceived insecurity as a psychological trait affecting political behavioral outcomes, speaking to, amongst others, Skitka et al. (2006), Huddy and Feldman (2011) and Igarashi (2021). Nevertheless, these factors have been so far largely ignored in research on the relationship between terrorism and immigration and migration policy preferences, respectively. Third, information on the objectively low probability to become a victim of terrorism has no significant effects on preferences. This suggests that anti-immigrant responses to terrorism, on average, follow an emotional proximity rationale and are generally not conditional on objective threats: that is, terrorism primarily produces a response when it feels close (i.e. when terrorism happens at home and meets a receptive audience that is sufficiently fearful with respect to this threat).
In contrast to the literature that conducted natural experiments, our factorial survey experiment allows us to investigate the mechanisms behind the effects of terrorism on policy preferences, circumventing the common problems of natural experiments being event-specific and not allowing to investigate moderators due to post-treatment effects (Muñoz, Falcó-Gimeno, and Hernández 2020). This is not to say that that there are no weaknesses associated with factorial survey designs; for instance, they always constitute an artificial setting and regardless of how realistic the vignettes are, such references to terrorism can hardly be as strong as a real attack. Consequently, we believe that the survey experiment approaches can complement -rather than replace -natural experiments and thus provide a more complete picture regarding the effect of terrorism on migration policy preferences. Indeed, there are ample research opportunities to pursue further factorial survey experiments, e.g. by considering the effect of terrorism on migration policy preferences in less developed countries (in which average policy preferences with respect to migration may be markedly different compared to developed economies) or in countries that are more regularly affected by terrorism than the countries we consider in this study (meaning that in the former countries' average levels of fearfulness with respect to violence may be noticeably different from those in Germany and the United Kingdom).
The political implications of our findings concerning the psychological micro-mechanisms that relate terrorism to immigration policy preferences are nuanced. The lack of a generalized effect of terrorism on preferences suggests that the increased politicization and securitization of migration since the turn of the 21 st century may not necessarily be due to salient terrorist attacks. At the same time, however, our findings suggest that there is a rather sizeable part of the population (i.e. fearful individuals) that is susceptible to associate terrorist threats with anti-immigrant policy preferences, offering, in turn, opportunities for political mobilization for political entrepreneurs with antiimmigration agendas. Indeed, those mobilization opportunities appear to be rather stable, given that the link between terrorism and anti-immigration policy preferences is rooted in feelings of geographical (i.e. by terrorism in one's home country rather than abroad) and emotional proximity (i.e. by pre-existing levels of insecurity), while being largely immune to objective information on the low probability of becoming a victim. This implies that in order to counter anti-immigration sentiment due to terrorism, a purely rational response to terrorism (i.e. by informing citizens of the low probability of terrorist events) may not be a sufficient response but that policy measures that adequately address the (rather irrational) emotional response to terrorism also need to be considered.

Notes
1. For a further discussion of this definition as well as an excellent general introduction to the literature on terrorism studies, see Gaibulloev and Sandler (2019). 2. This especially concerns transnational terrorism, i.e. terrorism that affects more than one country (Enders, Sandler, and Gaibulloev 2011). For instance, terrorist perpetrators may have migrated from other countries or may be supported by foreign actors. 3. Such feelings, in turn, may incentivize the actual implementation of these policies by governments that want to reap associated electoral rewards. For empirical tests of this proposition, see Helbling and Meierrieks (2020) and Choi (2021). 4. For example, such fear effects that induce excessive economic responses may also explain why terrorism can generate comparatively large and adverse macroeconomic effects (Naor 2015; see also Gaibulloev and Sandler 2019). 5. Due to financial constraints, we had to limit our analysis to two countries. While other countries such as France or Spain would also constitute relevant cases, we have no theoretical reasons to expect different effects for these countries.
6. Tables A1 and A2 in the Appendix report the summary statistics. Table A3 demonstrates that the in-sample distribution of education levels corresponds to their respective country-specific distribution, speaking to the sample's representativeness. 7. For more information on the Respondi access survey, see https://www.respondi.com.

Similar questions have been used in international surveys such as the European Social Survey and the
International Social Survey Program. We keep the wording of this question as broad as possible because immigration policies rarely target migrant groups from specific parts of the world. 9. The vignettes for saliency of terrorism are as follows: 'Since the 1990s, Islamist terrorist attacks [in Germany/in the United Kingdom/outside Europe] have become more frequent and have killed a larger number of people than before.' 10. The vignettes for probability neglect are as follows: 'Researchers have shown that the probability of being killed in an Islamist terrorist attack is 390 times smaller than being killed in a road accident.  Table A7, we find that terrorism also does not affect policy preferences related to integration or surveillance policies. 13. Table A7 in the Appendix replicates Table 1 using alternative measures of security values developed by Schwartz (2012) and included, among others, in the European Social Survey. Here, survey respondents were asked to indicate how similar they feel towards a person who is described as follows: 'It is important to him/her to live in secure surroundings. S/he avoids anything that might endanger his/her safety.' and 'It is important to him/her that the government ensures his/her safety against all threats. S/he wants the state to be strong so it can defend its citizens.' Answer categories are 'Very much like me', 'Like me', 'Somewhat like me', 'A little like me', 'Not like me', 'Not like me at all'. We use a summative score of both items, since they show high levels of scale reliability (Cronbach's α = 0.72). Using this alternative measure of security values, we again find support for H2. 14. In Tables A4 and A5 in the Appendix, we reproduce this finding for the German and United Kingdom subsample, respectively. While we find that there is support for H2 for the United Kingdom sub-sample, the results for the German sub-sample appear to be somewhat less clear-cut, as estimated interaction effects are marginally insignificant for one specification (p = 0.109). At the same time, the joint significance test for the interaction term and the two constituent components is highly significant (F = 19.36, p < 0.01). Potentially, then, the marginally insignificant interaction effect may point to a multicollinearity issue. In any event, while the effect associated with the interaction effect for Germany are somewhat smaller, statistically speaking (assessed by means of a seemingly unrelated estimation) it is not different from the effect size of the interaction effect for the UK. 15. This analysis is run with a reduced sample as the probability information was only manipulated among the respondents that received one of two the terrorism vignettes.   Notes: Own survey experiment. OLS models predicting 4-point scale measuring surveillance law preference (from 'fully disagree with more severe surveillance laws' to 'fully agree' with such laws) in models (1) to (4). OLS models predicting 4-point scale measuring integration policy preferences (from 'fully disagree that there should be new laws requiring immigrants to integrate better' to 'fully agree' with such laws) in models (5) to (8). OLS model predicting 4-point scale measuring immigration policy preferences (from 'allow many' to 'allow nobody' to enter the country) in model (9), which uses a measure of fearfulness using an alternative Schwartz (2012) security values measure (Cronbach's α = 0.72). Respondents have been asked to indicate how similar they feel towards a person who is described as follows: 'It is important to him/her to live in secure surroundings. S/he avoids anything that might endanger his/her safety.' and 'It is important to him/her that the government ensures his/her safety against all threats. S/he wants the state to be strong so it can defend its citizens.' Answer categories are 'Very much like me', 'Like me', 'Somewhat like me', 'A little like me', 'Not like me', 'Not like me at all'). Standard errors in parentheses. ***p ≤ 0.001, **p ≤ 0.05, *p ≤ 0.1