Does context matter? European-specific risk factors for radicalization


 In this study we sought to identify whether risk and protective factors for radicalization can be classed as ‘universal’ factors or whether they have heterogeneous cross-regional effects. Specifically, we sought to identify whether there were factors which displayed significantly different effects in European contexts compared to other democratic countries. We conduct a confirmatory meta-analysis based on a recent Campbell Collaboration systematic review and meta-analysis (Wolfowicz, Litmanovitz, Weisburd and Hasisi, 2021). Studies were classified as being from either EU or non-EU countries and moderator analysis was used to identify between-region heterogeneity. The analysis was possible for 23 factors pertaining to radical attitudes, 13 pertaining to radical intentions and 4 for radical behaviours. For radical attitudes, the estimates for European studies were significantly larger for Gender, Socio-economic status, and Parental involvement, whereas the estimates for Religiosity, Institutional trust, Integration, and Moral neutralizations were significantly smaller compared to other democratic countries in other regions. For radical intentions, the estimates for Self-esteem was significantly larger for European studies. For radical behaviours, the estimate for Unemployment was significantly larger for European studies than for democratic countries in other regions. Overall, most risk and protective factors for radicalization appear to have ‘universal’ effects across democratic countries, but there are some factors that may be more relevant for targeting by counter-radicalization in certain contexts. Although European counter-radicalization has often focused on factors such as integration and institutional trust, these factors have relatively small relationships with radicalization, and these relationships are even smaller in the European context compared to democratic countries in other regions. The findings suggest that mitigation strategies, and interventions providing employment opportunities in particular, may be well suited to the European context if the goal is to develop locally-oriented approaches to counter-radicalization.


Introduction
The ways in which democratic countries seek to combat violent extremism have changed considerably over the last two decades. Since the mid 2000's, these countries have increasingly adopted a prevention approach that seeks to tackle the radicalization that could potentially lead to violent extremism. It is in this context that we have witnessed a proliferation of counter and de-radicalization policies and initiatives. operating at primary, secondary and tertiary levels. Primary level interventions seek to reduce the likelihood of radicalization among the general population or wide sections of specific sub-populations. At the secondary level, interventions focus on individuals considered to be at risk for radicalization, or who have been classified as already radicalized, with the goal of preventing their progression to violent extremism. Tertiary level interventions seek to reduce the likelihood that extremist offenders will re-offend. In all cases, interventions generally seek to counter radicalization by targeting underlying risk and protective factors (Hardy, 2020;Silva & Deflem, 2020).
To a large degree, the development of counter-radicalization programs has adopted the same type of risk-factor paradigm that has become popularized over the last few decades for combatting a range of criminal and criminal-analogous outcomes (Farrington, 2000). Research on risk and protective factors has sought to identify and establish which factors have ›universal‹ relationships with such outcomes across contexts, with a view to identifying which types of interventions may be transferable. In order to accomplish these objectives, comparative multi-region and multi-country studies are needed, although they are unfortunately all too infrequent (Farrington, 2015).
In the development of counter radicalization interventions, and the selection of which factors ought to be targeted, there is a debate concerning the degree to which re-gions or countries should draw on the experiences of each other, or develop more locally relevant approaches. Differences in the approaches of democratic countries are often more of a reflection of political, social, and cultural norms than actual evidence pointing to the specific local relevance of targeted risk factors (Hardy, 2019;Hasisi, Perry and Wolfowicz, 2019). Common norms underpin current approaches to counter-radicalization in the European Union (EU). While there is certainly a degree of heterogeneity between EU member states and even their counter-radicalization approaches, all of them are tied to the EU's counter-radicalization strategy, which defines radicalization as »the phenomenon of people embracing opinions, views and ideas which could [sic] lead to acts of terrorism« (European Commission, 2005. In line with this definition, the European approach to combatting radicalization heavily focuses on prevention of the development of these ›risky‹ cognitions in the context of primary level interventions (Barlett et al., 2010;Silva & Deflem, 2020). And while different countries may give greater weight to certain factors over others, factors such as social inclusion, acculturation and identity are common across the European Union (Bakker, 2017;Bartlett et al., 2010;Korn, 2016;Proctor, 2019).
In the absence of cross-national and cross-regional comparative research, criminologists have often turned to confirmatory meta-analysis, which makes use of moderator analysis in order to examine specific hypotheses, such as whether risk and protective factor relationships differ between regions Lösel 2018). This study employs such an approach and sought to identify which risk and protective factors for radicalization may have unique effects in European contexts. Our data are drawn from a recent systematic review and meta-analysis of risk and protective factors for radicalization that we conducted for the Campbell Collaboration (Wolfowicz, Weisburd, Hasisi and Litmanovitz, 2021). The review included some 127 studies and 206 unique samples from which over 1300 effect sizes were extracted and synthesized across 101 risk and protective factors for both cognitive and behavioral outcomes of radicalization in OECD countries. In addition to identifying the relative magnitude of the estimates for the different factors, the review also identified study location. In the current study we analyze European and non-European located studies, employing both meta-regression and moderator analysis to identify cross-regional differences between risk factors (e.g. Pinquart & Kauser, 2018).
The original review found that the most salient factors of radicalization were, generally speaking, those associated with central criminological theories, specifically Social Learning and Social Control theories. While both of these ›general‹ theories hold that risk and protective factors associated with them ought to have universal effects, they also account for population heterogeneity, positing that differences in social structures may lead to cross-regional differences (Gottfredson & Hirschi, 1990;Akers & Jensen, 2006). We use these theories as a guide for contextualizing regional differences in the effects and discuss the possible implications of such differences for counter-radicalization interventions. 1 2 Cross-regional differences in the effects of risk factors The literature on risk factors for radicalization is highly similar to the broader literature on criminological risk factors in many ways. One area in which these literatures are similar is in the relative lack of cross-regional research (Wolfowicz et al., 2021). While some studies have compared democratic and non-democratic countries (e.g. Zhirkov, Verkuyten and Weesie, 2014), few have compared more similar countries. However, the cross-national research that does exist, and the relatively limited number of factors included in such studies, suggests that some differences may exist in the relationship between different risk factors and radicalization based on region (e.g. Berger, 2016; Gøtzsche-Astrup, 2021). Regional differences in the effects of risk factors have previously been understood by criminological perspectives to be primarily the result of different social structures and variance in opportunities (Akers, 1998;Gottfredson & Hirschi, 1990). One theory that has elaborated extensively on this relationship is Social Learning Theory, which holds that social learning related factors are variably moderated and mediated by a range of social structure factors (Akers, 1998;Verrill, 2008). According to Akers (1998), the general idea is that social structure encompasses: «the general culture and structure of society and the particular communities, groups, and other contexts of social interaction provide learning environments in which the norms define what is approved and disapproved, behavioral models are present, and the reactions of other people (for example, in applying social sanctions) and the existence of other stimuli attach different reinforcing or punishing consequences to individuals' behavior.« (p. 321).
Social structure variables can be grouped in different categories, both at the individual and macro levels, useful for understanding and explaining differences in risk factor effects across contexts. At the macro level, the most important factors are those associated with »differential social organization«, which refers to differences in cultural, social, and demographic characteristics of societies. Important factors related to »differential location in the social structure«, may include the relative position of different groupings, such as ethnic groups, within the social structure. Relatedly are factors associated with »social disorganization«, theoretically defined structural variables such as anomie, group-based inequality, and societal integration (Akers, 1998).

The systematic review and meta analysis
We draw in our analyses from our Campbell field-wide review of risk and protective factors for radicalization in democratic countries (Wolfowicz et al., 2021). The review's primary objectives were to investigate what the risk and protective factors for radicalization are, and to identify the relative magnitude of their relationships with radicalization outcomes. Given the diversity of the literature, especially with respect to definitions of radicalization, the review pre-determined its inclusion and exclusion criteria in a systematic review protocol (Wolfowicz, Litmanovitz, Weisburd and Hasisi, 2020). These criteria are summarized in the following sections.

Outcomes
In line with contemporary approaches towards (counter) radicalization, the review sought to include studies that assessed both cognitive and behavioral outcomes of radicalization. While there are a wide range of proxies that are used in the literature to assess »radicalization«, the review adopted the criteria of the outcomes from McCauley and Moskalenko's (2017) two-pyramid model (TPM) of radicalization to guide the inclusion and exclusion of studies. Unlike other models, the TPM is not a process model nor does it posit any specific direction or relationship between cognitive and behavioural radicalization. In fact, it is more of a typological model than it is a theoretical model and is therefore useful for the purposes of meta-analysis which seek to combine studies examining comparable outcomes (Wolfowicz et al., 2021). In line with the TPM, the review sought to include studies whose outcomes met the criteria set forth by the TPM for radical attitudes and intentions (cognitions), as well as behaviors. 1. Radical attitudes: Justification/support for radical behaviours carried out in the name of a cause. 2. Radical intentions: Willingness/intentions towards engagement in radical behaviours in the name of a cause. 3. Radical behaviours: Actual involvement in violent radical behaviours in the name of a cause, including terrorism.
In line with the TPM, radical behaviors included both sub-terroristic forms of radical violence, as well as terrorism. The review included studies employing both validated and non-validated instruments to measure cognitive radicalization, as well as measures of outcomes derived from self-reports, family reports, administrative reports (e.g. government or law-enforcement), practitioner/clinical reports, and databases whose data was derived from open-sources.

Types of studies
The review was limited to quantitative studies of an observational nature, namely cross-sectional, longitudinal, and case-control studies. For a study to have been included it must have employed a design that enabled the calculation of an effect size, either by including a direct comparison or control group, or a single sample in which there was variation on the dependent variable. The review excludes all factors that were not measured at the individual level, for example: -Meso-level factors: Community level deprivation, population density etc. -Macro level factors: GDP, GINI etc.

Risk and protective factors
-Time-series factors: The occurrence of specific events (e.g. terror attacks), or rates of social phenomena (e.g. crime rates).
Additionally, the review did not include experimentally manipulated factors as combining these with non-manipulated versions of the same factors would lead to methodological inconsistencies.

Context
As with other systematic reviews of risk factors, the Campbell review limited itself to specific types of countries/regions (Higginson et al., 2018;Litmanovitz & Montgomery 2016;Murray et al. 2018;Shenderovich et al. 2016). The review cross-referenced the Organisation for Economic Cooperation and Development (OECD) countries with the Democracy Index's list of democratic countries. The result of this approach led to an inclusion of all OECD countries with the exception of Turkey, which the Democracy Index ranks a »hybrid regime« rather than a democratic system. As such, studies were eligible for inclusion when their samples included participants from the following countries: Australia, Austria, Belgium, Canada, Chile, Columbia, Czech Republic, Denmark Estonia, Finland, France, Germany Greece Hungary, Iceland, Ireland, Israel, Italy, Japan, Korea, Latvia, Lithuania, Luxemburg, Mexico, Netherlands, New Zealand, Norway, Poland, Portugal, Slovakia, Slovenia, Spain, Sweden, Switzerland, UK, and USA. For the purposes of the current study, each included study was coded as being either »European« or »Non-European«. Although there is significant heterogeneity within these two groupings, there is also a significant degree of commonality. In particular, previous research has found that the countries that make up these groupings tend to cluster with each other in terms of patterns of deviant outcomes (Eisner, 2002;Junger-Tas, 2010). Moreover, previous meta-analytic research has found that these regional levels of aggregation are often more informative than using continuous measures pertaining to socio-cultural or socio-economic metrics (Fraguas et al., 2021;Pinquart & Kauser, 2018;Gaffney, Farrington and Ttofi, 2019).  In the current study we examine which factors show evidence of differences in their relationship with radicalization outcomes (attitudes, intentions and behaviours) between European and non-European studies. We first identified which of the factors examined in Wolfowicz et al (2021) included at least two European and two non-European studies. We subsequently carried out moderator analysis to identify statistically significant between region heterogeneity (p<.05).

Included studies
The samples in the included studies pertained to a relatively limited number of local contexts. In the EU, while a few samples included participants from multiple countries (N=10), the largest number of samples were from Denmark (N=17), Spain (N=15), and the UK (N=17), followed by the Netherlands (N=12), Germany (N=11), Belgium (N=9), and France (N=8). A small number of samples were from Sweden (N=5), Norway (N=4), Switzerland (N=4), and Poland (N=2), while a single sample represented the countries of Austria, Greece, Italy and Hungary. Overall, the countries represented in these samples can at the least be said to be representative of western Europe. For non-European democratic countries, the majority of the samples pertained to the U. S. (N=52). However, there were also a number of samples from Canada (N=12) and Australia (N=4), as well as two samples from Israel and South Korea. Another 12 samples included participants from multiple countries, while others included those from both Canada and the U. S. (N=3). Overall, these samples cover a list of countries that can at least be said to be representative of the Anglosphere of non-European democracies.

Results
As noted earlier the Campbell systematic review and meta-analysis identified over 100 risk and protective factors for the outcomes of radical attitudes (29 protective and 71 risk factors), intentions (8 protective and 37 risk factors) and behaviors (7 protective and 26 risk factors). With respect to the relationship between outcomes, the review identified and analyzed radical attitudes as a risk factor for both radical intentions and behaviors. However, the current analysis was limited to those factors for which a minimum of two European and two non-European effect sizes were present. As such, the current analysis was limited to 26 factors pertaining to radical attitudes, 13 factors pertaining to radical intentions, and 4 factors pertaining to radical behaviors. These factors are highlighted in Table 1 where they appear in rank-order according to the magnitude of the estimates. 2 Moderator analysis identified that between-region heterogeneity was absent for the majority of factors. For the outcomes of radical attitudes, statistically significant between-region heterogeneity (p<.05) was found for 6 factors, whereas for both the outcomes of radical intentions and behaviours it was only found for one factor for each. The results of the moderator analysis can be found below in Table 2 and are arranged alphabetically.
With respect to the statistically significant factors for radical attitudes, SOCIO-ECONOMIC STATUS had a very small but statistically significant estimate for EU studies but a non-significant estimate for non-EU countries. For INSTITUTIONAL TRUST, whilst estimates for both EU and non-EU studies were statistically significant, the size of the estimates differed substantially, with a small estimate for the EU studies, and a more robust, moderate estimate for non-EU studies. With respect to PARENTAL INVOLVE-MENT, although the estimates for both EU and non-EU studies were small, the estimate for the EU studies was significantly larger. Similar findings were made with respect to GENDER, with estimates being exceptionally small, although larger for EU studies. With respect to both INTEGRATION and MORAL NEUTRALIZATIONS, the esti-  (7) Effect sizes are r correlations with 95 % confidence intervals. Q is Cochran's Q statistics for heterogeneity with p-value from Chi2 test. I2 and T2=heterogeneity statistics. N=Combined sample size, K=Number of effect sizes. ***<.001, **<.01, *<.05. 2 For factors for which such analysis was not possible, please see the original review (Wolfowicz et al., 2020a). mates for EU studies were both significantly smaller than the more moderate estimates for non-EU studies. With respect to the other outcomes, for radical intentions the estimate for SELF-ESTEEM was significantly larger for EU-based studies than non-EU studies. For the outcome of radical behaviors, the estimate for UNEMPLOYMENT was significantly larger for EU studies, with a borderline moderate sized estimate being more than double the size of the small estimate for the non-EU studies. The two EU studies included one study from the Netherlands, and another from the UK, both of which compared terrorism offenders with general population samples (Table 2).

Discussion
The current study conducted a confirmatory meta-analysis in order to identify which risk and protective factors for radicalization may have universal effects, and which have unique effects in Europe. The original systematic review and meta-analysis included some 100 risk and protective factors for outcomes of both cognitive and behavioral radicalizations. For most of these factors, included studies were exclusively or primarily from the European region, thereby precluding their inclusion in our cross-regional analysis. But based on the factors that were included in our analysis, it was identified that between-region heterogeneity exists for only a small number of factors. This means, that similar to the broader literature on the generalizability of risk and protective factors for criminological outcomes, most factors may have essentially universal effects . For radical attitudes these include factors with relatively smaller effects, such as perceptions and experiences of discrimination, and factors with larger effects, such as anomia, authoritarianism and political extremism. These are all factors which may be, and are already being targeted by counter-radicalization initiatives in democratic countries (Wolfowicz et al., 2021).
As the objective of our study was to identify which factors may be more relevant in the European context, we focus our discussion on these factors only. We exclude from our discussion those socio-demographic variables for which differences were found, since these differences are likely to simply reflect differences in the distributions of these characteristics in the different populations. In focussing on only the more substantive factors, we discuss; 1) Parental involvement, 2) Self-esteem, and 3) Unemployment as it relates to radical behaviours.

Parental involvement and radical attitudes
Although not objectively large, relative to other protective factors found in the review, there is a modest relationship between parental involvement and radical attitudes. Parental bonds have repeatedly been found to be a key protective factor for a range of criminal and criminal analogous outcomes (Flanagan, Auty and Farrington, 2019). According to social control perspectives there are a number of elements to parental bonds, the most important of which are attachment and involvement. In addition, poor parental relationships are believed to be central to the development of low self-control (Gottfredson & Hirschi, 1990), a key universal factor for both criminal and radical outcomes (Wolfowicz et al., 2021). Changing social norms may serve to explain why the strength of the relationship between parental involvement and radicalization is stronger in Europe than other democratic countries. For example, a recent study of delinquency in the Netherlands found that over the last 15 years there has been an apparent improvement in parental bonds (and parental involvement in particular), and the increased presence of this important protective factor is partially responsible for a sustained decrease in youth criminal outcomes (van der Laan et al., 2021). Similar findings have been made in other countries, such as Sweden, leading scholars to suggest that cultural shifts have taken place in many parts of Europe that have led to this increase. Perhaps most relevant to our findings, this increase has been found to corresponded with a decrease in positive attitudes towards crime and offending behaviors (Svennson & Oberwittler, 2021).  In applying these findings to counter-radicalization, already across Europe, policy-makers have recognized parents and family as key figures in combatting radicalization. However, current approaches that deal with parents and families appear to be relegated to two areas: 1) Hotlines for reporting of radicalized children, 2) Resources for receiving support for how to counter radicalization. Current approaches therefore focus primarily on equipping parents with counter-narrative tools and strategies (El-Amraoui & Ducol, 2019;Ranstorp, Gustafsson, Hyllengren and Ahlin, 2016). But rather than being a mere vehicle for identification, reporting or delivering of counter-narratives, European interventions may benefit from integrating methods to strengthen parental bonds (Ranstorp et al., 2016). This could potentially be done through encouraging parental involvement in education activities (Speckhard, Shajkovci and Ahmed, 2018). In this regard, counter-radicalization may benefit from adopting the types of evidence-based parent training strategies that have been shown to successfully improve parental bonds and reduce the likelihood of a range of criminal outcomes .

Self-esteem and radical intentions
There are two opposing theories when it comes to the relationship between self-esteem and criminal outcomes. One theory posits that individuals with low self-esteem may be drawn to crime because they believe that it will enhance their self-esteem, in part by attracting acceptance and respect of those around them (Slaby & Guerra, 1988). The other theory posits that high self-esteem can spill over into narcissism, and that threats to self-esteem may be responded to more defensively, and aggressively, by those high in self-esteem (Baumeister, Smart and Boden, 1996;Bushman & Baumeister, 1998).
Similar opposing views have been put forth with respect to radicalization. One perspective holds that individuals with low self-esteem may seek out stronger connections to collectives who can provide a sense of identity, and thereby improve self-esteem (Kruglanski et al., 2018). Another view is that for those with a group-based identity and high-self esteem, frustration of goals or threats to the collective may be responded to defensively in order to maintain or affirm self-esteem (McGregor, Hayes and Prentice, 2015). Alternatively, insecure forms of high self-esteem, in which self-esteem is high explicitly by low implicitly, may seek certainty in radical belief systems and groups (Trip et al., 2019). In sum, the latter perspective views high-self esteem as a risk factor for radicalization when it is inflated, threatened, or unstable (Beelmann, 2020). Our results suggest that the overall nature of the relationship between self-esteem and radicalization is in line with this latter perspective.
The riskiness of self-esteem has already been noted in the context of counter-radicalization interventions, and in European contexts specifically. While there is some initial experimental evidence that shows that interventions aimed at improving self-esteem hold promise for reducing the likelihood of radicalization, these may only work when coupled with empathy training. These findings reflect the potential for improved self-esteem in and of itself to increase the likelihood of radicalization (Feddes, 2015;Feddes, Mann and Doosje, 2015).
It has previously been found that there is significant cross-regional heterogeneity in the relationship between self-esteem and criminal outcomes (Mier & Ladny, 2018). The results point to the possibility that these differences reflect the normative values placed on self-esteem in the different societies, or the degree to which individuals‹ identities are group-based (Turner, 1975). Similar to the relationship between self esteem and radicalization described above, research has found that the focus of western countries on promoting self-esteem as a normative value may inadvertently promoted narcissism; the type of inflated self-esteem that can increase the risk for criminal and radical outcomes. In segments of society in which relative standing is an important value, such as in Western Europe, this potential may be elevated. In order for self-esteem to be fostered in such a way so that it provides the psychological benefits that it promises whilst avoiding such negative consequences, parenting training may prove beneficial (Brummelman & Sedikides, 2020). Self-affirmation approaches may be especially useful as well since self-esteem, when threatened, can lead to radicalization as an expression of self-affirmation (Bélanger, 2021;Williams, 2020).

Employment and radical behaviors
The role of employment and poverty as a cause of terrorism is arguably one of the most extensively researched hypotheses in the literature. Perhaps more important than objective socio-economic related strain, unemployment can serve as a source of grievance and be a source of a lack of personal significance (e.g. Belanger et al., 2017Belanger et al., , 2019Krulanski et al., 2018). Additionally, unemployment can be accompanied by social strain, especially when the state of unemployment status is perceived to be the result of discrimination (Agnew, 2010(Agnew, , 2016. This may be especially important for minority groups, who figure prominently in the samples analyzed in this study. Does context matter? European-specific risk factors for radicalization One potential explanation for the differences between European and other democratic countries in the effects of unemployment may pertain to the relative degree of economic integration of minority groups (Meleagrou-Hitchens, Hughes, Clifford, 2019). In this regard, even though unemployment is generally higher in European countries than in other democratic countries such as the US, Canada, and Australia, unemployment gaps of specific groups, particularly Muslims and immigrants is greater in Europe (OECD, 2021).
This explanation is in line with evidence from macro-level economic models which have found that while there is little to no relationship between population-level employment and terrorism, there is a positive relationship with minority economic discrimination (Piazza, 2011) and relative deprivation (Varaine, 2020). In Europe in particular, there is a strong relationship with individual and country level unemployment and collective relative deprivation (Meuleman et al., 2020). Relative deprivation among the unemployed has been found to be associated with a range of factors pertaining to mental health and wellbeing (Walker & Mann, 1987). However, the strength of this effect is dependent on the prevalence of unemployment in society. So too, unemployment's impact on other risk factors, such as institutional trust, is greater in Western European countries than in the US, and this has been attributed in part to the differential rates of unemployment in these regions (Giustozzi & Gangl, 2020).
Another way that employment can effect radicalization is through its conditioning of social bonds and routine activities (LaFree & Miller, 2008;Windisch et al., 2016). Through its impact on routine activities, employment limits the amount of unstructured free time in which the individual can potentially engage with deviant elements or behaviors (Apel & Horney, 2017). As changes to routine activities also effect changes in socialization patterns, employment can foster changes in other attitudinal factors (Walters, 2016;Wolfe, McLean, & Pratt, 2017) such as legitimacy (Fagan & Tyler, 2005), and forms of perceived injustice and relative deprivation (Burchardt, 2005;Folger & Kass, 2000;Agnew, 2010Agnew, , 2016Rice, 2009).
In countries like the UK and Netherlands, providing at-risk groups with employment is part of primary and secondary level counter-radicalization. However, the rationale behind the inclusion of employment is that socio-economic factors, and a lack of social integration are the primary risk factors for radicalization (Van Dongen, 2016;b, 2012Thornton & Bouhana, 2017;HM Government, 2018). Some have previously criticized this view as undermining the potential benefits of employment by focusing on factors that are known to sometimes increase a sense of marginalization (Van Dongen, 2016).
In contrast to this perspective, countries such as Australia (Cherney, 2016) and the U. S. (National Academy of Science, 2017) treat employment more as a form of situational prevention. In line with the routine activities and social bonds perspectives, employment is more of a diversionary approach, and can serve as an off-ramp for promoting disengagement from radical trajectories (Simi et al., 2016). According to this perspective, while any effect on radical attitudes is desirable, this is not the purpose of employment, and nor is de-radicalization necessary for secondary and tertiary level interventions to succeed in promoting abstention or desistance from radical behaviours (Bartlett et al., 2010;Kruglanski et al., 2018). Given that the effects of unemployment are larger in Europe than in other democratic countries, European countries may benefit from adopting and integrating this view of employment into their counter-radicalization approaches. As Day and Kleinmann (2017) describe it: »We need to make employment a national security priority. We need to rethink »assimilation« to be less about mainstreaming cultural identity, and instead think about networks of would-be-radicals being absorbed and supported by healthcare, education, democratic representation, and respect for their beliefs. We need affective bonds that are stronger than theirs. Real CVE is about a stronger civic life, not merely »countering narratives.« (p. 21).

Limitations
While our study is the first to compare and contrast the magnitude of the effects for risk factors for radicalization, there are a number of limitations, both in terms of the data, the methodology, and interpretation. In terms of the data, as noted in the original review (Wolfowicz et al., 2021), there is little in terms of longitudinal study, meaning that the current estimates represent putative risk and protectives rather than risk factors. This means that we cannot make any claims about the causality of such factors as they pertain to radicalization. Methodologically, meta-analysis is limited to the availability of data. This means that we were limited by the number of studies available for each factor which prohibited looking at cross-national effects . This is important as there still exists a significant degree of heterogeneity between the different European societies (Kaasa, Vadi and Varblane, 2014). And when it comes to radicalization, and possible variation in the effects of risk factors, there is even evidence of within-region differences (Ellis et al., 2021), and more local-level differences (Miconi et al., 2020). As more research becomes available it may eventually be possible to carry cross-national analyses with a finer grain of analysis (Jugl et al., 2021).
With regards to interpretation, there is a potentially unlimited number of avenues for exploring societal and cultural differences that may explain the nature of the effects (Rutter, 1999). Such factors may include indices relating to societal individualism, power-dominance orientation, and masculinity. While the current study was focused on crossregional effects, these types of cross-cultural factors are deserving of specific attention in future research.

Conclusions
With the proliferation of counter-radicalization interventions in the west in recent years, there has been a debate as to the degree to which countries should target ›univer-sal‹ factors or contextually specific factors. Concurrently, there has been little evidence upon which identification of universal and context specific factors can be made. We found that for the overwhelming majority of factors for which evidence exists from different democratic regions, there is no significant inter-regional heterogeneity. That is, the current evidence points to these factors as having ›universal‹ relationships with radicalization in democratic countries. These include many factors already targeted by counter-radicalization interventions, such as: In-group identities, Experiences of discrimination, Perceived discrimination, and Anomic conditions (e.g. Social alienation). Additionally, for other individual background characteristics, including socio-demographic factors, personality related factors (e.g. Openness, Authoritarianism/Fundamentalism), current evidence suggests that they too have universal effects. As such, for interventions that target such factors, context may not matter. This means that it may also be realistic to draw lessons from other countries‹ approaches and experiences in the targeting of such factors. However, we also found evidence to support claims that certain factors may have context specific effects. In this regard we found a small number of risk factors whose relationship with radicalization was significantly larger in EU countries than in other democratic countries and vice versa. For these factors, context does appear to matter.
It is interesting to note that two of the factors that are commonly targeted by European interventions, namely integration and institutional trust, have relatively small (though significant) relationships with radicalization more generally in the Campbell review, but these relationships are even smaller in the European context. This raises concerns about the potential effectiveness of European inter-ventions that focus on such factors. Whilst these factors shouldn't necessarily be discounted, if anything it is non-European democracies that should consider focusing on integration in the context of counter-radicalization strategies. With regard to employment we found the employment was much more important in the European context, suggesting that employment-oriented interventions may be useful.
The results of this study indicate that while context rarely matters when it comes to risk factors for radicalization in democratic countries, sometimes it does. Unfortunately, counter-radicalization policies are often »shaped less by evidence-based research, and more so by local politics, history and culture« (Hardy, 2018:100). The results of this study indicate that European countries may benefit from dedicating greater resources to interventions and policies that target factors for which evidence exists that could identify them as context specific. These factors include both universal and specific risk factors in the European context.
Funding: Funding was received from EU grant no. 699824