Facing antisemitism in Europe: Individual and country-level predictors of Jews' victimization and fear across twelve countries

Rising antisemitism in the twenty-first century has alarmed Jewish communities and the general public, but antisemitic hate crime victimization remains understudied outside the US context. This study primarily relies on a comprehensive survey of 16,400 Jews across twelve European countries, supplemented with data from additional sources, to assess individual and country-level predictors of Jews' experiences and fears of antisemitic harassment and violence. Multilevel models indicate that young age, perceived discrimination, identity visibility, and identification with Israel are pronounced individual risk factors for victimization. On the country level, negative opinion of Israel and Muslim population share predict victimization, highlighting the role of a "new" or Israel-derived antisemitism in the twenty-first century. The factors most strongly associated with fear are young age, previous victimization, perceptions of an ambient antisemitic threat, and recent occurrence of fatal antisemitic violence. Overall, the findings underscore the importance of integrating general theory on hate crime and victimization with context-specific factors when seeking to understand the experiences of targeted groups.

Introduction

Hate crimes, broadly defined as prejudice-motivated acts of violence or hostility, are "message crimes" that not only inflict harm, stigma, and psychological stress on individual victims but also affect entire communities, instilling fear and threatening their sense of security and belonging (Barnes and Ephross 1994; Perry and Alvi 2012; Scheitle et al. 2022). Here, I explore a much discussed but understudied dimension of the hate crime problem, namely antisemitic harassment and violence as experienced and feared by Jews in Europe.

Anti-Jewish thought and practice are deeply rooted in two millennia of European history, with recurrent waves of hostility and persecution culminating in the Holocaust. After World War II, overt antisemitism became largely taboo in democratic societies, but this did not eradicate underlying prejudices. Despite a general decline in explicit antisemitic attitudes, verbal and physical attacks against Jews have increased in the twenty-first century (Enstad 2023). While a small minority, Jews today are often the victims of a disproportionate share of recorded hate crimes (Iganski and Sweiry 2016; Dodd 2023; FBI 2023). In Europe, this development has created a situation in which many Jews hide their identity in public (DPA 2017; JTA 2017) and even consider emigration because of not feeling safe in their country (Buck 2018).

Against this background, maintaining a hospitable environment for Jewish minorities has become a political concern of moral and historical significance and a matter of EU policy (European Commission 2021). Researchers have responded to this challenge by seeking to explain variation in antisemitic attitudes and recorded incidents, but much less attention has been devoted to the question of victimization and fear. Addressing this gap, I ask which factors are the most important in explaining Jews' experiences and fears of hate-motivated harassment and violence, using multilevel models to analyze data from a major 2018 survey of Jews across twelve EU countries (N = 16,395). In doing so, I connect the literatures on hate crime and antisemitism and derive a theoretical framework that sees victimization and fear as functions of both individual vulnerabilities and varying societal contexts.

Studying victimization is crucial as it directly assesses the impact of hate crime on the people targeted, potentially uncovering individual risk factors that are difficult to account for in incident-based research. Furthermore, incident data can be inconsistent due to variations in reporting and categorization practices across countries. Victimization surveys can offer a more reliable cross-national measure of hate crime, capturing the "dark figure" of experiences that were not reported to the police or other organizations and thus not recorded anywhere (Groves and Cork 2008).

Research on hate crime victimization, fear, and antisemitism

A key argument for the relevance of studying (and legally sanctioning) bias-motivated crime has been that such acts tend to harm victims more than non-bias crimes do, a proposition that is well-supported by evidence (Fetzer and Pezzella 2019; Díaz-Faes and Pereda 2022). Studies show that victims of hate violence are more likely than victims of other crimes to suffer symptoms of post-traumatic stress, including depression, anxiety, confidence loss, and fear (McDevitt et al. 2001; Iganski 2014). In the literature on hate crime victimization there are many studies investigating the impacts of hate crime, but, curiously, only a few that seek to identify determinants of victimization experience (for a review, see Díaz-Faes and Pereda (2022)). In a key study of hate crime victimization across fourteen European countries, Van Kesteren (2016) identified young age, migrant status, urban residence, and low income as the most important risk factors.

Extensive research on fear of crime has shown that fear is a social problem independently of crime experience and is influenced by various individual and contextual factors including age, gender, social disorder, and media consumption (Clemente and Kleiman 1977; Hale 1996; Henson and Reyns 2015). Fear of crime may exacerbate the impact of actual victimization, affecting individual quality of life as well as community trust (Jackson 2006). Fear of hate crime appears to affect people more severely than fear of non-bias-motivated crime (Díaz-Faes and Pereda 2022). In their study of fear of religious hate crime in the United States, Scheitle et al. (2022) found that previous victimization, indirect victimization (i.e., knowledge of other victims), and religious visibility were strong predictors of fear.

The topic of antisemitism has received little attention within the field of hate crime research. However, there is a growing multidisciplinary literature addressing the contemporary resurgence of antisemitism. This scholarship has predominantly focused on antisemitic attitudes (Bergmann 2008; Cohen 2018), incidents targeting Jews (Smith 2008; Feinberg 2020; Vergani et al. 2021), and antisemitic expressions in texts and cultural imagery (Becker and Bolton 2022). An important takeaway from this literature is that the frequency and severity of hate incidents can be high or increasing even in countries with low and declining levels of antisemitic attitudes, indicating that social desirability bias may mask higher levels of antisemitic sentiment in the population or that hate crime offenders are clustered in smaller ideological or religious subgroups (Jikeli 2017). Another consistent finding is that the Israel-Palestine conflict tends to trigger antisemitic incidents (LaFreniere Tamez et al. 2024). In general, these studies look at how non-Jews think about and act towards Jews; the victim perspective has received much less attention. A handful of studies do focus on Jews' perceptions and/or experiences of antisemitism, but these are either descriptive in nature or else do not aim to explain victimization as such, and are mostly confined to the US context (Dubow et al. 2000; Staetsky and Boyd 2014; Rebhun 2014; Kremelberg and Dashefsky 2016).

Hence, by investigating the determinants of antisemitic victimization and fear in Europe, the present study seeks to fill a research gap in hate crime scholarship as well as in the wider literature on contemporary antisemitism—a gap that is all the more significant given the current need to understand the challenge antisemitism poses to the safety of Jewish life in Europe.

Explaining Antisemitic Victimization and Fear: Individual Risk Factors and Societal Contexts

Antisemitism, commonly defined as a "persisting latent structure of hostile beliefs towards Jews as a collectivity" (Fein 1987), is both an instance of general outgroup hostility and a unique phenomenon with its own historical roots and contemporary manifestations. To account for this, I provide a theoretical framework that integrates general factors derived from the wider literature on hate crime and prejudice and context-specific elements derived from the literature on contemporary antisemitism. I first discuss explanations of victimization and fear based on the general hate crime literature, before moving on to antisemitism-specific factors.

Risk factors for hate crime victimization

The literature on hate crime victimization draws on multiple lines of criminological and sociological theorizing to account for individual variation in hate crime experience. Lifestyle-exposure/routine activities theory (LRAT) (Hindelang et al. 1978; Felson 1994; Cohen and Felson 2010), a framework originally developed to study crime victimization in general, is frequently applied in hate crime research. LRAT posits that for a crime to occur, there needs to be a motivated offender as well as a suitable target interacting in the absence of "capable guardians," i.e., people or mechanisms that could prevent the crime. It is predicted that young, urban, unmarried, and lower-income males are at higher risk of crime exposure because of how their routines, habits, and general way of life increase their proximity to potential criminal offenders. Several studies of hate crime victimization have found results that are consistent with LRAT (Wallengren and Mellgren 2015; Van Kesteren 2016; McNeeley and Overstreet 2018), and so we may hypothesize that:

H1: Being young, a big-city resident, unmarried, male, and having a lower income is linked to higher risk of antisemitic victimization.

Hate crime research has been paying increasing attention to intersectionality, seeking to understand how having multiple discriminated-against identities may affect victimization (Crenshaw 1989). Studies have shown that people belonging to more than one socially vulnerable group are more likely to experience hate crimes (Andersson et al. 2018; Macdonald et al. 2023). This may be because people who belong to multiple such groups are more visible targets for those who commit acts of prejudice and hate. Psychological mechanisms may also be involved, as increased subjective sensitivity, or "stigma consciousness" (Pinel 1999), can heighten one's awareness of potential discrimination and bias in interactions with others. This leads to the following hypothesis:

H2: Perceived discrimination on grounds such as gender, sexual orientation, or disability is linked to higher risk of antisemitic victimization.

The extent of ethnic minority group members' integration into majority society is expected to account for variation in hate crime experiences. While hate crime studies have linked immigrant status to a higher risk of victimization in European and US contexts (Van Kesteren 2016; McCann and Boateng 2022), the extent of ethnoreligious minority group members' integration into majority society likely also plays a role. Current theorizing suggests that integration may both reduce and increase exposure to prejudice and hate crimes, corresponding to what Rebhun (2014) calls the "descending" and "ascending" versions of the integration hypothesis. In the "descending" version, minority group members are expected to encounter less outgroup hostility from the majority as they become more integrated, aligning with the contact hypothesis (Allport 1954). According to the "ascending" version, the more highly integrated are more likely to face prejudice and stigmatization (Steinmann 2019). This "integration paradox" may stem from higher expectations for equal treatment among highly integrated minority group members or increased contact with members of the majority group, which leads to more frequent exposure to potential carriers of prejudice. From this, we can derive two hypotheses:

H3a: Integration is linked to lower risk of antisemitic victimization (the descending version).

H3b: Integration is linked to higher risk of antisemitic victimization (the ascending version).

As hate crimes are intended to denigrate outgroups, individual group identification has important implications for understanding victimization. Group identification is both a signal perpetrators rely on to identify targets and a factor influencing how minority group members perceive and interpret their experiences. In line with symbolic threat theory, when members of a minority group exhibit signs of strong group identification and cohesiveness, it can be perceived as a threat by the dominant group or other minorities, potentially leading to hostility or aggression (Stephan and Stephan 2000). Moreover, group identification can increase vulnerability by enhancing identity visibility through specific behaviors such as displaying group markers or attending religious or other communal events in public places (Feinberg 2020; Scheitle et al. 2022). A strong group identity may also increase psychological sensitivity to signals of prejudice in interactions with others (Operario and Fiske 2001). Studies indicate that people belonging to highly visible minority groups, such as those openly displaying religious symbols or openly identifying as part of the LGBTQ+ community, are at a greater risk of being subjected to hate crime victimization (Andersson et al. 2018). Accordingly, we can hypothesize that:

H4: The strength of Jews' identification with their group is linked to higher risk of antisemitic victimization.

Moving to country-level properties that may contribute to hate crime victimization, scholars has long argued that economic decline can drive hostile attitudes and behaviors toward out-groups (Hovland and Sears 1940). Arguments from relative deprivation, realistic group-conflict, and scapegoat theory suggest that competition over limited resources fuel intergroup conflict, including hate crimes, and especially so in times of economic decline when conditions of scarcity and unemployment exacerbate social tensions and prejudices (Bilewicz and Krzeminski 2010; Falk et al. 2011 Belgioioso et al. 2023).

H5: Country-level economic decline is linked to higher risk of antisemitic victimization.

Fear of hate crime

Fear of hate crime is closely related to, but distinct from, actual victimization. Fear encompasses the psychological and behavioral response to a perceived threat, which can be shaped by both direct victimization experience and signals about the prevalence and severity of threats in one's environment (Hart et al. 2022). Fear of crime in general has been studied extensively, but the literature on fear of hate crime victimization is thin and provides few strong theoretical expectations. For example, while fear of crime in general tends to increase with age and be stronger among women (Hale 1996), this may not be the case for fear of hate crime (Tiby 2001; Scheitle et al. 2022). Furthermore, while identity visibility can be expected to heighten fear, as being more visible means being a more identifiable target for potential hate crime perpetrators, at least one study has found that minority group members who openly displayed their identity were less afraid of being targeted, suggesting a resilience effect (Wallengren and Mellgren 2015).

The role of prior victimization may be more important. While the relationship between victimization and fear has been found to be weak when it comes to crime in general (Rader 2004), recent research suggests that this relationship may be stronger in the context of hate crimes (Scheitle et al. 2022). This could be due to the "message crime" aspect, which may heighten fear not only among those who have experienced such incidents directly, but also among those who have witnessed them or know someone who has been victimized.

H6: Victimization experience is linked to higher fear of victimization.

Beyond victimization experiences, perceptions of the broader social environment can also shape fear of victimization. This is captured in the concept of perceived ambient threat, which refers to the perceived risk of victimization based on cues in the environment, such as signals of increasing crime rates (Hale 1996). In the context of antisemitic hate crime, perceived ambient threat refers to Jews' perceptions of increasing antisemitic behaviors in their national environment, which could heighten fear independently of any victimization experience.

H7: Perceptions of ambient threat are linked to higher fear of victimization.

The context of antisemitism

Jews' experiences and fears of antisemitic victimization are likely determined not only by factors common to hate crime in general, but also by more context-specfic drivers. Most basically, we may expect that country-level antisemitic attitudes is a risk factor for victimization. A higher share of people with antisemitic attitudes in one's national environment increases the number of potential offenders. Moreover, in societies where hostile attitudes towards a minority group are more widespread, offenders may feel more morally justified in attacking this group due to a sense of approval from the social environment. While country-level antisemitic attitudes are therefore likely important, two caveats derived from the literature on contemporary antisemitism have implications for how to approach this. First, such attitudes come in different shapes, and second, they appear to be more salient in certain subgroups.

Antisemitic attitudes may be expressed as agreement with classical anti-Jewish stereotypes linked to ideas about nefarious conspiracies, hidden power, malignant intent, and so on—ideas that certainly continue to motivate antisemitic offenders. Yet not all antisemitism is expressed in this traditional way. Debates over a "new" antisemitism became salient following the surge in antisemitic incidents in the twenty-first century (Taguieff 2004; Wistrich 2010; Klug 2013). Proponents of newness emphasize the role played by the far left as well as Islamic extremism, and particularly their anti-Israel and antizionist sentiment and activism. Several studies have indicated a connection between anti-Israel and antisemitic attitudes (Kaplan and Small 2006; Cohen et al. 2009; Staetsky 2020), and so it can be expected that anti-Israel sentiment to some extent masks antisemitic sentiment and thus should be independently accounted for.

However, relying solely on mean country levels of antisemitic or anti-Israel attitudes could obscure significant variations within different subgroups. Antisemitic attitudes have repeatedly been found to be more widespread among certain subsets of European populations, in particular those with a far-right/nationalist political orientation (Mayer 2007; Staetsky 2020) and those with a Muslim religious orientation (Jikeli 2015; Kovács and Fischer 2021). Even in countries with low levels of overall antisemitic attitudes, there may be substantial ideological or religious subgroups that harbor such attitudes to a larger extent and thus display more active hostility towards Jews, and so the relative size of such subgroups may increase the risk of victimization.

These considerations lead to the following hypotheses:

H8: The level of classical antisemitic attitudes in one's country is linked to higher risk of antisemitic victimization.

H9: The level of anti-Israel attitudes in one's country is linked to higher risk of antisemitic victimization.

H10: The share of far-right voters in one's country is linked to higher risk of antisemitic victimization.

H11: The Muslim population share in one's country is linked to higher risk of antisemitic victimization.

There are also context-specific factors that may shape Jews' fear of antisemitic victimization. Here, I follow Scheitle et al. (2022) and propose that the strength of Jews' collective memory of trauma, varying across countries, plays a role in generating fear. Historical trauma can be passed down through generations and may contribute to a heightened sense of threat and fear among Jews today, even if they have not personally experienced antisemitic victimization. This is consistent with the literature on collective memory and intergenerational trauma, which suggests that historical events can have lasting effects on the perceptions and experiences of subsequent generations (Canetti et al. 2018). Finally, a more recent possible source of collective trauma is the occurrence of fatal antisemitic incidents. Only some European countries have witnessed murderous antisemitic violence in recent years. Such occurrences could serve as a signal or reminder of a serious threat and thus heighten levels of fear among Jews in those countries.

H12: Living in a country with stronger collective memories of trauma is linked to higher fear of antisemitic victimization.

Data

I draw on data from the EU Fundamental Rights Agency's FRA 2018 survey of Jews in twelve EU countries including Austria, Belgium, Denmark, France, Germany, Hungary, Italy, the Netherlands, Poland, Spain, Sweden, and the UK (N = 16,395). The survey focused on perceptions and experiences of antisemitism and was the second wave of an ongoing effort to document how antisemitism is affecting Jews across Europe (FRA 2019). The dataset is available for download through the GESIS data archive.¹

For small minority groups such as Jews in European countries, drawing representative samples is difficult and costly, especially when multiple nations are involved. For this reason, the FRA survey used an opt-in online approach with questionnaires publicized via Jewish communal organizations, Jewish media outlets, and social networks. The survey authors analyzed the resulting samples in light of extant benchmark data and estimates regarding the composition of the Jewish population. Most country samples were found to underrepresent the youngest age group and overrepresent older age groups. Moreover, most of the samples were found to underrepresent the communally unaffiliated segment of the Jewish population. Thus, the overall sample cannot be considered representative of the entire Jewish population in the given countries, but it does represent communally active Jews as well as those who may not be very involved in communal life but still remain close to it. To compensate for this, the survey authors computed weights based on census and Jewish community data about the composition of the Jewish community and population in each country (FRA 2018; Staetsky 2019).

To measure country-level contextual conditions, additional data sources were used. Data on antisemitic and anti-Israel attitudes were collected from the Anti-Defamation League's (ADL) global antisemitism surveys.² Data on far-right voting were taken from an unpublished dataset constructed by Anders R. Jupskås at the University of Oslo's Center for Research on Extremism (C-REX), and Muslim population estimates were gathered from a 2017 report by the Pew Research Center (Pew Research Center 2017). Data on economic conditions were retrieved from Our World in Data, drawing on Feenstra, Inklaar, and Timmer (2015) for GDP per capita figures and ILO estimates for unemployment rates.³ Figures for Jewish Holocaust losses are based on information from the Holocaust Encyclopedia (US Holocaust Memorial Museum 2018).

Measures

Victimization and fear of victimization

To quantify personal antisemitic victimization, I constructed two binary variables measuring whether respondents had experienced (1) harassment or (2) violence due to antisemitism in the past 5 years. Scoring 1 on either of these variables led to a score of 1 on the final antisemitic victimization variable, which was otherwise set to 0.

The harassment variable was constructed based on responses to two questionnaire items asked of respondents who had previously indicated being subjected to "offensive or threatening comments in person" at least once in the past 5 years for any reason. Those who responded "Yes, once" were then asked, "You said somebody has made offensive or threatening comments to you in person in the past five years. Did this happen BECAUSE you are Jewish?" Those who incicated having experienced more than one incident of such harassment received a similar follow-up question: "You said somebody has made offensive or threatening comments to you in person in the past five years. Did any of these incidents happen BECAUSE you are Jewish?" If respondents indicated at least one incident of antisemitic harassment through any of these items, they were coded as 1 in the harassment variable, otherwise they were coded as 0.⁴

The violence variable was similarly constructed, based on responses to two questions posed to respondents who had answered in the affirmative to the question, "In the past five years, has anybody ever physically attacked you?" Those who responded "Yes, once" were then asked, "And did this incident happen BECAUSE you are Jewish?" while those who indicated more than one experience of physical violence were asked, "And how many of these incidents in the past five years have happened BECAUSE you are Jewish?" If respondents indicated at least one incident of antisemitic physical attack through any of these items, they were coded as 1 in the violence variable, otherwise they were coded as 0.

This operationalization allows for a relatively precise measure of victimization compared to some previous studies that have employed more open-ended outcome measures (e.g., a single item asking whether one has "experienced antisemitism" in a given time period, which can mean many different things to different people) (Kremelberg and Dashefsky 2016).

I also measure family and indirect victimization. Family victimization is a binary variable created based on two items asking whether a family member or someone close had (1) "experienced verbal insults or harassment due to antisemitism" or (2) "experienced physical attacks due to antisemitism," both in the past 12 months, scored 1 for affirmative response to any of these and 0 otherwise. For the indirect victimization variable, respondents were scored 1 if they indicated having personally witnessed antisemitic incidents in which "other Jews [were] being verbally insulted or harassed" or "physically attacked" in the past 12 months, and 0 if not.

To measure fear of victimization, I created a scale based on the following items, which reflect both the affective and the behavioral dimension of fear:

1. 1). How worried are you about becoming a victim of antisemitic harassment in the next 12 months?

2. 2). How worried are you about becoming a victim of antisemitic violence in the next 12 months?

3. 3). How worried are you about a family member/someone close becoming a victim of antisemitic harassment in the next 12 months?

4. 4). How worried are you about a family member/someone close becoming a victim of antisemitic violence in the next 12 months?

5. 5). Do you ever avoid visiting Jewish events or sites because you do not feel safe as a Jew there or on the way there?

6. 6). Do you ever avoid certain places or locations in your local area or neighbourhood because you don't feel safe there as a Jew?

Response options were "not at all worried," "not very worried," "fairly worried," and "very worried" for items 1–4 and "never," "occasionally," "frequently," and "all the time" for items 5–6. A scale was computed by averaging the responses to the six items. This scale, an approximation of a continuous variable, measures the strength of respondents' fear of victimization. Thus, the scale should be unidimensional and internally consistent. Tests indicated high unidimensionality and internal consistency (α = .88; ω_t = .92).

Socio-demographic variables and discrimination experience

Socio-demographic variables included age, gender, residential location, marital status, and household financial status. Age is represented by a binary variable coded 1 for below 40 and 0 for above 40. Gender is a binary variable with male as the reference category, while financial status was measured with an item asking about the ability of one's household to make ends meet, rated on a six-point scale ranging from 1 ("with great difficulty") to 6 ("very easily").

Perceived discrimination was measured using a series of items asking respondents whether they had "felt discriminated against" in the past 12 months because of their age, sex, gender identity, sexual orientation, disability, or "another reason" (respondents were also asked whether they felt discriminated against on grounds of ethnic and religious identity, but these items were not considered here because of their proximity to the victimization outcome). A binary variable was computed, scored 1 for one or more kinds of perceived discrimination and 0 otherwise.

Integration and group identification

Individuals' degree of integration into majority society was measured using three variables: a nativity variable, coded 0 for foreign-born and 1 for native-born, and two items measuring individuals' sense of national and regional attachment. Respondents were asked to indicate to what extent they "feel attached" to their country or the region where they live, using a scale from 1 ("not at all attached") to 5 ("very strongly attached").

Following Rebhun (2014) and Scheitle et al. (2022), I employed several measures of both religious and non-religious Jewish identification. Two items measuring the strength of respondents' Jewish identity and religiosity were included, both rated on a 1–10 scale. I also included an identity visibility variable measuring whether one "never," "sometimes," or "always" publicly wears, carries, or displays visible signs of being a Jew.

In the context of exposure to antisemitism, the theoretically most important non-religious identification variable is attachment to Israel. I measure this by way of two composite indices. The first is an objective attachment to Israel index ranging from 0 to 4. Respondents were scored 1 point for having visited Israel, 1 point for having been born in Israel or lived there for more than one year, 1 point for having "some" or "many" relatives in Israel and 2 points for having "all or almost all" relatives there. The second index, subjective attachment to Israel, combines responses to two items: to what extent one feels attached to Israel (scale 1–5, from "not at all attached" to "very strongly attached") and the importance of supporting Israel to one's Jewish identity (scale 1–4, from "very unimportant" to "very important"). To adjust for the unequal response scales, these two items were normalized to a range of 0 to 1 and combined to create an index ranging from 0 to 2.

Perceived ambient threat

The perceived ambient threat variable in this study was constructed from seven different survey items, which measure respondents' perceptions of whether antisemitism has increased or decreased across various domains over the past five years. These domains include the internet/social media, vandalism, graffiti, cemetery desecration, politics, the media, and the street or public places. For each item, respondents indicated their perceptions on a five-point scale ranging from 1 ("Decreased a lot") to 5 ("Increased a lot"). An overall perceived ambient threat score was computed by averaging responses across all seven items, resulting in a scale from 1 to 5. This scale showed unidimensionality and good internal consistency (α = .87; ω_t a = .89).

To address concerns that the ambient threat variable might be measuring the same underlying construct as the fear of victimization outcome variable (i.e., a general fear of antisemitism), their dimensionality was examined using factor analysis. Results (see tables S1 and S2 in the supplementary materials) indicated that a two-factor structure fits the data better; the two variables appear to represent distinct constructs.

Country-level factors

Overall country levels of antisemitic attitudes were measured using the ADL antisemitism index. Index values were averaged across the 2014–2019 ADL Global 100 surveys, which asked respondents in nationally representative samples to respond "probably true" or "probably false" to eleven statements deemed to reflect antisemitic prejudice. Each country's index score represents the percentage of respondents indicating agreement with at least six statements. Anti-Israel attitudes were also measured using 2014–2019 ADL data. This measure is based on a basic favorability score, representing the percentage of respondents saying they have an "unfavorable opinion" of Israel.

The prevalence of far-right voting was measured by the average percentage of votes given to farright parties in national elections in the period 2012–2018, while the Muslim population variable represents the estimated percentage of Muslims as of 2016. Economic conditions were measured as the percentage change in GDP per capita (adjusted for purchasing power) from 2010 to 2018 and the absolute change in unemployment rates (International Labor Organization estimates) over the same period. Collective memory of trauma was measured using two variables: Holocaust mortality rate—the proportion of the pre-war Jewish population in each country that was killed during the Holocaust—and recent fatal antisemitic occurrence. The latter was a binary variable scored 1 for countries where such an attack had occurred in the 5 years preceding the survey, and 0 otherwise.

Analytic Approach

Because the FRA survey data consist of responses by individuals nested within countries, and given that significant between-country differences could be observed in the outcome variables, multilevel modeling appeared to be warranted (Gelman and Hill 2006). To test the appropriateness of a multilevel approach, I ran likelihood ratio tests to compare minimal baseline fixed-effects models and intercept-only mixed-effects models for both outcomes under study. In both cases, the mixed-effects models provided a better fit to the data (χ² = 205.9, df = 1, p < .001 for antisemitic victimization and χ² = 1868.9, df = 1, p < .001 for fear of victimization).

I employed logistic mixed-effects regression to analyze victimization and linear mixed-effects models for fear of victimization. Cluster-robust standard errors were incorporated to account for the potential correlation of residuals within countries. I report average marginal effects from the logistic models; i.e., the estimated change in the predicted probability of the outcome occurring for a one-unit increase in the predictor when other variables are held at their means or reference levels (Mood 2010).

An advantage of multilevel models is the opportunity to explore cross-level interactions. In addition to the main analyses, I conducted follow-up analyses to test whether particular country-level variables moderated the association between individual-level predictors and antisemitic victimization. These analyses included random slopes for the individual-level variables involved in the interaction, as recommended in the multilevel modeling literature (Heisig and Schaeffer 2019).

Below I report descriptive statistics in both unweighted and weighted forms. Model estimates reported here are unweighted, while weighted estimates are provided in the supplementary materials (tables S4–S6).

For data preparation and analysis, I used R (R Core Team 2023), supplemented with Stata's melogit and mixed commands for estimating models with cluster-robust standard errors. All code is provided as supplementary data.

Results

Table 1 presents a summary of all individual-level variables used in the analysis. An estimated 31 percent of Jews across these twelve countries reported experiencing antisemitic victimization in the preceding five years, with 4 percent of them subject to violent victimization. Fear of victimization, a variable ranging from 1 (low fear) to 4 (high fear), had a weighted average of 2.07. The influence of weighting was generally minimal, but it was more noticeable in the age variable due to the known underrepresentation of younger European Jews in the FRA samples. The slightly higher weighted means or percentages for the victimization variables likely reflect the younger age profile of hate crime victims.

Table 1.

Descriptive statistics for individual-level variables

Table 2 provides country-level weighted estimates of the 5-year prevalence of victimization and violent victimization, weighted mean scores on the fear-of-victimization scale, and scores on the country-contextual variables. 5-year victimization prevalence ranged from a low of 25 percent (UK and Italy) to a high of 48 percent (Belgium), while the estimates for violent victimization ranged from 1.3 percent (Spain) to 8.4 percent (Belgium).

Country-level bivariate correlations between the outcome and country-contextual variables are depicted in figure 1. While these correlations should be interpreted with caution due to the small number of countries, some tentative observations can be made. There appears to be a strong association between victimization prevalence and negative opinion of Israel. Fear of victimization was not linked to victimization prevalence at the country level but showed a strong correlation with the recent occurrence of fatal antisemitic attacks. In the multilevel models presented in the following sections, these and other country-level variables are further examined for their potential associations with individual-level outcomes.

Table 2.

Descriptive statistics for country-level variables

Figure 1.

Country-level bivariate correlations (p < .05).

Antisemitic Victimization

Models analyzing antisemitic victimization are summarized in table 3. The baseline model includes just the random intercept. Model 1 includes socio-demographic and integration variables, and Model 2 adds group identification factors. Finally, Model 3 incorporates a series of country-level variables. Figure 2 plots the coefficients from the full model.

The Intraclass Correlation Coefficient (ICC) for the baseline model shows that a small part (2 percent) of the overall variation in victimization can be attributed to differences between countries. As figure 3 shows, the data indicate that the odds of victimization were higher than average in Belgium, the Netherlands, and Germany, and below average in Hungary, France, the UK, and Italy. In the final model, the ICC is reduced to zero, indicating that all significant between-country variability is accounted for when including the full set of covariates.

Examining the socio-demographic predictors, young age stands out: those under 40 were substantially more likely to have experienced antisemitic violence or harassment, with a 12-percentage-point higher probability of victimization. Being female reduced the probability of victimization somewhat, as did living in a suburban area compared to a big city, being in a marriage or partnership, and having a better household financial situation. Moreover, perceived discrimination is positively and quite strongly associated with victimization: feeling discriminated against on grounds unrelated to ethnic/religious identity in the past year was linked to a 11-percentage-point increase in the probability of victimization.

Turning to the integration variables, no association with immigrant status was found for any outcome, but national and regional attachment were linked to lower victimization probabilities

Table 3.

Multilevel logistic regression models predicting Antisemitic victimization

and lower fear. This offers some support for the "descending" version of the integration hypothesis (integration reduces vulnerability to prejudice and discrimination).

Model 2 adds a series of group identification variables. Jewish identity strength and religiosity are both linked to higher probabilities of victimization. Moreover, identity visibility has a strong effect. "Sometimes" or "always" (vs. "never") wearing Jewish identity symbols is linked to a 6- and 13-percentage-point increase in the probability of victimization, respectively (this association held when removing respondents identifying as Haredi, i.e., strictly orthodox and highly distinguishable Jews). Model 2 further introduces two variables measuring the Israel dimension of Jewish group identification. Interestingly, while objective attachment to Israel (visits, residence,

Figure 2.

Predictors of victimization.

Figure 3.

Random effects of country on victimization (baseline model). Odds ratios with 95% CI.

relatives) did not affect the outcome, a one-standard-deviation increase in the subjective attachment index was linked to a 3.6-percentage-point higher probability of victimization.

Model 3 proceeds to add country-level variables. No association is found with levels of antisemitic attitudes or change in economic conditions, while far-right voting shows a slightly significant and minimal correlation, with a one-percent increase in such voting linked to a 0.1-percentage-point higher victimization probability.

Negative opinion on Israel and the share of the Muslim population emerge as stronger and highly significant correlates. Controlling for a range of covariates, a one-percentage-point increase in the share of a country's population with an unfavorable opinion of Israel is associated with a 0.8-percentage-point higher probability of antisemitic victimization. Similarly, a one-percentage-point increase in the Muslim population share is linked to a 1.2-percentage-point higher probability of victimization. Figure 4 displays these associations as predicted probability plots. While it should be kept in mind that the results cannot readily be generalized beyond the twelve countries included in the present sample, the findings nonetheless lend support to

Figure 4.

Predicted probabilities of victimization given varying levels of unfavorable opinion of Israel and Muslim population share (95% CI).

arguments emphasizing the importance of a "new" or Israel-derived antisemitism in twenty-first century Europe.

The victimization measure used here represents having experienced either antisemitic harassment or violence. Experiences of violence are much less common than harassment and may have different explanations. To investigate this, a separate model was estimated with violent victimization as the outcome. Results were similar overall (see Supplementary figure 1 and table S3). Far-right voting lost the small correlation that was present for the general victimization outcome, while Muslim population share was more strongly linked to violent victimization.

In follow-up analyses, I included a cross-level interaction term in Model 3 to examine whether the link between subjective attachment to Israel and victimization was influenced by the prevalence of negative opinion of Israel, based on the expectation that in contexts where negative views of Israel are more widespread, people identifying strongly with Israel might be more targeted. The result confirmed this expectation, as the association between identification with Israel and victimization depended entirely on the level of anti-Israel sentiment (for the interaction term, OR = 1.01, SE = .003, p = .024; see also Supplementary figure 2). Including this interaction rendered the subjective attachment variable non-significant, while the country-level unfavorability coefficient retained its strength. This consistent main effect indicates a powerful role of societal attitudes towards Israel in influencing antisemitic victimization risks at the individual level, independent of personal attachments and feelings towards Israel.

Fear of victimization

Three models analyzing Jews' fear of antisemitic victimization are presented in table 4, with coefficients from the full model (Model 3) plotted in figure 5. The baseline model without predictors shows that country differences in fear are more pronounced than those observed for the victimization outcome (ICC = 0.09). Figure 6 shows how fear varies between countries in the baseline model. Note that the difference between the most fearful country (France) and the least fearful (Hungary) approaches a full standard deviation.

Model 1 includes the same set of individual-level predictors as the full victimization model. Most of the variables that were associated with victimization are linked to fear in the same way. Living in a marriage or with a registered partner, however, was negatively associated with victimization but is positively associated with fear. This could be explained by a "parenting effect," whereby those in a relationship might express more fear due to concerns for the safety of their partner or children (Rader 2017).

Table 4.

Multilevel linear regression models predicting fear of victimization

Interestingly, the role of identity visibility is different in the case of fear. While identity visibility strongly predicted victimization, this factor is associated with lower levels of fear. In the full model, coefficients for "sometimes" and "always" wearing visible Jewish symbols are negative but non-significant. In the weighted model, however, these coefficients are significant at the 95 percent level (see table S6 in the supplementary materials). This might indicate a resilience effect, where those who openly display their Jewish identity are, or become, less fearful despite having a higher risk of victimization.

Model 2 adds victimization, family victimization, indirect victimization, and perceived ambient threat, which all correlate more strongly with fear than any of the other variables. Model 3

Figure 5.

Predictors of fear of victimization.

Figure 6.

Random effects of country on fear of victimization (baseline model). Estimates with 95% CI.

additionally incorporates two country-level variables. While Holocaust mortality rate has no effect, the recent occurrence of a fatal antisemitic attack is linked to higher fear, amounting to about 30 percent of a standard deviation. Though this estimate is less precise, it indicates that the impact on fear of living in a country with recent fatal antisemitic violence is comparable to that of having personally experienced antisemitic harassment or violence.

In a follow-up analysis, a cross-level interaction term was included in the full model to test whether the association between individual threat perceptions and fear was moderated by the recent occurrence of fatal antisemitic attacks in one's national environment. This interaction was significant and in the expected direction (b = .05, SE = .02, p = .013; see Supplementary figure 3). Moreover, it did not substantially alter the main effects of the two variables involved, indicating that objective events and subjective perceptions influence fear of hate crime both independently of each other and interactively.

When accounting for the full set of covariates in Model 3, the ICC is reduced to 0.05, which means that 5 percent of the variability in fear is attributable to country differences. Even so, meaningful country differences in fear do remain (see Supplementary figure 4). With a range of individual and country-level explanatory factors accounted for, the difference between the least fearful country (UK) and the most fearful ones (France, Germany, Sweden) remains substantial. It is likely that combinations of unmeasured context-specific conditions are at work here; exploring this further may require a case-based and qualitative approach.

Discussion

The aim of this study was to explain variation in antisemitic hate crime victimization and fear among Jews in Europe, using data from a 2018 survey of Jews across twelve countries supplemented with data on country-level conditions. The findings reveal a multifaceted picture. While socio-demographic factors, discrimination experience, and group identification accounted for substantial variation in victimization, country-level variables were also important. Most notably, unfavorable opinion of Israel and Muslim population share strongly predicted victimization. In contrast, "usual suspects" such as classical antisemitic attitudes, far-right voting, and economic decline were weakly if at all correlated with antisemitic hate crime experience. These results provide empirical support for the contention that so-called "new" antisemitism, in which hostility towards Israel blends with or masks antisemitism in far-left and Islamic milieus, is something to be taken seriously. The findings add depth to research from the US context that finds a consistent link between Israeli military conflicts and antisemitic hate crimes (Feinberg 2020; LaFreniere Tamez et al. 2024).

Young age, being male, urban, unmarried, and financially less well-off were all linked to higher risk of antisemitic victimization, supporting H1 and echoing prior research (Rebhun 2014). A more unexpected finding was that young age also predicted higher fear of victimization. This runs counter to the literature on fear of crime, which generally finds that fear increases with age (Hale 1996), and contradicts US studies finding that older Jews perceive antisemitism as more severe (Rebhun 2014). The result also differs from recent findings on fear of religious hate crime in the US context, in which no association with age was found (Scheitle et al. 2022), suggesting that the situation for young Jews in Europe might be uniquely vulnerable.

Perceived discrimination was linked to victimization and fear about as strongly as young age, supporting H2. Net of other factors, those who felt discriminated against on grounds unrelated to ethnicity or religion (such as age, gender, and sexual orientation) indicated antisemitic victimization experience (as well as fear of victimization) at higher rates than others. This result, which aligns with prior work on discrimination (Scheitle et al. Platt 2021), is open to different interpretations. On the one hand, the finding is consistent with an intersectionality framework, which posits that people who belong to multiple discriminated-against groups may face more frequent or intense identity-based victimization compared to those who belong to only one such group. On the other hand, the finding might also reflect variation in people's general sensitivity to signals of discrimination and prejudice.

While immigrant status did not affect victimization nor fear, those who felt more attached to their country and region were somewhat less likely to experience victimization and slightly less fearful, indicating some support for the proposition (H3a) that integration reduces vulnerability. Group identification showed more explanatory power. In support of H4, strength of Jewish identity, religiosity, and subjective attachment to Israel were all strongly linked to higher victimization, which confirms prior research (Rebhun 2014; Scheitle et al. 2022). This is likely a result of synagogue attendance and other kinds of publicly visible ritual observance, which increases distinguishability in the eyes of hate crime offenders. While identity visibility strongly predicted higher victimization, it negatively predicted fear. This correlation was small, but similar findings from other contexts suggest that it may reflect a resilience effect (Wallengren and Mellgren 2015), in which a stronger sense of group cohesion and identity confidence among those who openly display their identity can attenuate the sense of fear and worry.

Moving to country-level explanations, no association was found between economic decline and victimization, which was contrary to expectations (H5). Moreover, neither classical antisemitic attitudes nor far-right voting accounted for meaningful variation in victimization probabilities, which was contrary to H8 and H10.

By contrast, negative opinion of Israel and Muslim population share were both associated with higher risk of antisemitic victimization, supporting H9 and H11. This highlights the importance of the "Israel factor," an often hotly contested aspect of contemporary antisemitism. Notably, the association found between individual-level subjective attachment to Israel and victimization was entirely dependent on the country-level unfavorability towards Israel. Country-level opinion on Israel retained a strong association with victimization even when controlling for its cross-level interaction with subjective attachment to Israel. This suggests that societal attitudes towards Israel significantly influence antisemitic victimization, regardless of Jews' individual identification.

The strongest individual-level predictors of fear were prior experience of antisemitic victimization and perceptions of ambient threat (the belief that antisemitic behaviors had increased in recent years), which supports H6 and H7. On the country level, there was mixed support for H12, as collective memory of trauma predicted fear when measured as recent fatal antisemitic attack occurrence, but not when measured using Holocaust mortality rates. Moreover, the interaction between ambient threat perceptions and fatal attack occurrence contributed to increased fear, with the direct effects of those variables remaining significant. These findings constitute evidence of the in terrorem effect of hate crimes, instilling fear among individuals belonging to the targeted community (Perry and Alvi 2012).

One limitation of this study is the non-probability and self-recruited nature of the survey sample. This may introduce a degree of self-selection bias, as individuals who chose to participate in the survey may not be representative of the broader Jewish population. It is possible that those who have experienced antisemitic victimization were more likely to respond to the survey, potentially overestimating the prevalence of such experiences in the general Jewish population. Estimates of relationships between variables, however, should not be too strongly affected.

A second limitation is in the cross-sectional and correlational design. The underlying data come from a 2018 survey that represents a "snapshot" of Jews' perceptions and experiences at that time and in the twelve countries covered. Findings reported here do not necessarily hold for other national contexts. Whether they hold across time is an empirical question, and one which future research should attend to. By leveraging the two other waves of the FRA survey on Jews' perceptions and experiences of antisemitism, from 2012 and from 2023 (to date, these datasets have not been published), future research can exploit the advantages of a longitudinal design to examine causal relationships and better understand the dynamics between changing contextual conditions, hate crime victimization, and fear. Further advancing our knowledge of this problem can help inform strategies to maintain hospitable conditions for Jews and other targeted minorities.

Supplementary Material

Click here for supplementary material.