This is an open access article distributed under the terms of the Creative Commons
Attribution License (Bias-motivated intergroup violence has a long history (Bureau of Justice Assistance, 1997; Perry, 2003; Petrosino, 1999), yet hate crime (or bias-motivated crime1) constitutes a relatively new legal term. Hate crime legislation in the United States dates back to the 1980s (Gerstenfeld, 2017) but in several states and numerous countries around the world there is still no legal framework for addressing it (Bleich, 2011; eMORE, 2017). The topic gained interest in the social sciences following a wave of racist violence in Northern Europe in the 1990s (for more details see: Green et al., 2001) and remains relevant today, as we witness growing numbers of hate-motivated violence (Gerstenfeld, 2017). This interest is strengthened by the fact that hate crime leads to more severe physical (Levin & McDevitt, 1993) and psychological consequences for the victims, when compared to the effects of similar crimes committed without bias-motivation (Herek et al., 1997; Iganski & Lagou, 2015; McDevitt et al., 2001). The negative consequences of hate crime impact not only the immediate victim(s), but threaten their whole social group and may lead to indirect victimization (i.e., negative emotional and behavioral consequences for non-victimized members of a group that stem from the knowledge that ingroup members had been targeted by hate crime; Craig, 2002; Walters et al., 2020).
To better understand the causes and consequences of hate crime, most research to date has focused on three areas of inquiry (Craig, 2002; Green et al., 2001). The first grapples with the legal understanding of the term, the analysis of which is beyond the scope of this work. The second area focuses on the characteristics of hate crime perpetrators and factors that contribute to the occurrence of hate crime. Perpetrators have been categorized into four broad groups: thrill seekers, defenders of “territory”, retaliators, and those who perceive it as a “mission” to rid their community of unwanted people (Levin & McDevitt, 1993; McDevitt et al., 2002). It has also been established that members of hate groups, while generally more likely to engage in bias-motivated crime, are not responsible for all hate crime (Dunbar et al., 2005; Levin & McDevitt, 1993). The third area analyzes the consequences of hate crime for the victims. Extensive research confirms a broad range of negative consequences both for the targeted individual (Herek et al., 1999; Iganski, 2001; Iganski & Lagou, 2015) and their social groups (Bell & Perry, 2015; Craig, 2002; McDevitt et al., 2002). However, despite evidence showing that aggressive behavior in humans takes on different forms (Anderson & Huesmann, 2007; Card & Little, 2006; Craig, 2002) and that different groups of victims tend to become targets of different kinds of hate crimes (Martin, 1996; Messner et al., 2004), surprisingly little attention has been given to differentiating types of hate crime and factors that contribute to their relative prevalence (Dancygier & Green, 2010).
The current paper aims to fill this significant knowledge gap. Specifically, drawing from social psychological literature on individual and collective aggression (Anderson & Huesmann, 2007; Glick, 2002; Staub, 1989), intergroup threat (Stephan & Stephan, 2000; Stephan et al., 2009), and perceptions of outgroups (Cuddy et al., 2007; Fiske et al., 2002) we posit that at least two broad types of hate crimes, that we refer to as direct and indirect hate crime, need to be differentiated. By analyzing the largest hate crime database in Poland, we demonstrate that these two types of hate crime are perpetrated by different kinds of people, target different kinds of victims, and are differentially related to contextual variables (i.e., economic indicators and political climate). The practical utility of these findings lies in their potential to better inform hate crime prevention efforts and contribute to more tailored assistance to victims.
Hate Crime Research
Hate crimes are crimes in which the victim has been selected based on their (perceived) membership in a social group rather than due to their individual characteristics (Bleich, 2011; Green et al., 2001). They can be defined in a narrow or a broad way. Narrowly defined, hate crimes are only those acts that exhaust the legal definition of a crime in a given jurisdiction (i.e., they are a punishable offence in of themselves, e.g., battery) and are also characterized by the perpetrator’s bigoted motivation (FBI, 2018). The broader understanding does not assume that a hate crime must necessarily fulfill the definition of a crime, but understands hate crime as any discriminatory, hostile treatment of member(s) of a social group motivated by bias (Dancygier & Green, 2010; Paterson et al., 2018). As there is no specific hate crime legislation in Poland,2 for the purpose of this work we adopt the broader definition of hate crime.
Lack of legislation is just one of the limitations that burden investigations of hate crime. Depending on the country of interest, the definition of hate crime and the number and types of social categories (e.g., race, gender, sexual orientation) explicitly protected by the law differ, which makes cross-country comparisons difficult (Dancygier & Green, 2010). Even if hate crime legislation exists, people who are targeted are often unwilling to come forward and report it. They may believe that the police cannot or would not help them or fear reprisal from the offender(s) (Sandholtz, Langton, & Planty, 2013). They may fear secondary victimization or, in the case of lesbian, gay, bisexual, and transgender (LGBT) victims, unwillingly revealing their sexual orientation (Górska et al., 2016; Herek & Berrill, 1990). Victims targeted because of their immigrant status may be afraid of the language barrier or negative consequences of revealing their legal status (Pawlęga & Godzisz, 2016). The police may not qualify certain incidents as hate crime, as they might underappreciate the role of bias or even themselves harbor prejudice against the victims (Nolan & Akiyama, 1999). All this leads to vast disproportions between the official and unofficial hate crime statistics (e.g., in Poland: Pawlęga & Godzisz, 2016; in the UK: Paterson et al., 2018; in the U.S.: Sandholtz et al., 2013). In terms of methodological difficulties, hate crime is a topic that does not allow direct experimental testing, necessitating a focus on archival studies and analyses of existing data, with all their shortcomings (though see Heng et al., 2018 for a discussion of using archival data). Because of these reasons, our analyses of hate crime rely on a hate crime database - the “Brown Book” - compiled by a non-governmental organization in Poland (Iganski, 2011; Kornak, 2009, 2011; Kornak & Tatar, 2013).
Hate Crime as Intergroup Aggression
Aggression is defined as behavior that has the intention of harming another person, who is motivated to avoid the harm (Anderson & Bushman, 2002, p. 28; Allen & Anderson, 2017). It can range from mild (e.g., name calling) to severe (e.g., stabbing someone with a knife). On the severe end of this spectrum aggression is referred to as violence, i.e., aggressive behavior that aims to inflict severe physical harm (J. J. Allen & Anderson, 2017). Hate crimes constitute a specific type of intergroup aggression (Dunbar & Blanco, 2014) because the individual(s) (or groups) who perpetrate them act as representatives of their social categories and select their victims based on (perceived) group membership (Bleich, 2011; Green et al., 2001). In line with social identity theory (Tajfel & Turner, 1979), any interaction, even between just two people, is intergroup in nature if social rather than individual identities are activated. This makes hate crimes, even those committed by a single perpetrator against a single victim, inherently intergroup. Notably, intergroup contexts heighten group identification, making people more likely to comply with group norms (Reicher et al., 2008) but also more willing to aggress on behalf of their ingroups (Littman & Paluck, 2015).
Most psychological research on aggression focuses on its interpersonal aspects. This research has established that aggression is situation-specific, functional, and that it is not a unitary phenomenon (i.e., that different types of aggression exist; Anderson et al., 2003; Dodge & Coie, 1987). A crucial distinction differentiates direct and indirect aggression (Archer & Coyne, 2005; Björkqvist, 1994; Lagerspetz et al., 1988). Direct aggression pertains to acts of physical, verbal or non-verbal aggression, as well as actively destroying another person’s property (Björkqvist et al., 1994; Card & Little, 2006). It often entails a face-to-face interaction between the aggressor and the victim. Its primary purpose is to harm the victim and gain social status (Archer & Coyne, 2005). Indirect aggression aims to cause emotional and/or reputational harm, and its primary aim is the social exclusion of the target (Archer & Coyne, 2005; Björkqvist, 1994; Björkqvist et al., 1994).
Aggression fulfils two main functions: hostile and instrumental (Card & Little, 2006; Fite et al., 2010). Hostile (or reactive) aggression is impulsive and motivated by anger. It is a response to a perceived provocation and its main goal is to harm the victim (Wilkowski & Robinson, 2010). Instrumental (or proactive) aggression results from a cold calculation and entails the existence of additional goals (beyond harming another person) as well as a degree of premeditation. Proactive and reactive aggression differ in their predictors, correlates, and behavioral outcomes (Card & Little, 2006; Hubbard et al., 2010; James et al., 2020).
Similarly to the study of hostile and instrumental aggression, researchers of intimate partner violence (e.g., Johnson, 1995, 2006; Johnson et al., 2014; Stith et al., 2000) have argued that differentiating types of violence allows us to better understand the relative prevalence of male- vs. female-perpetrated violence in intimate relationships (Allen et al., 2009; Dobash et al., 1992; Frieze, 2005). In line with this reasoning, Johnson (1995, 2006) shows that the type of violence that men and women resort to differs. Men are more likely to use Intimate Terrorism, a type of violence focused on taking control over the partner, while both men and women are equally likely to resort to Situational Couple Violence, which does not entail coercive control, but frequently occurs in the context of a situational conflict between partners that escalates towards violence (see Straus & Gozjolko, 2014).
Unlike studies on individual-level aggression and intimate partner violence, research on intergroup aggression typically treats it as a unitary phenomenon (i.e., it examines predictors of its occurrence, without analyzing its forms or components). Most of the extant research on intergroup aggression focuses on the influence of group identification (Littman & Paluck, 2015; Reicher et al., 2008) and/or the role of the situation (e.g., Milgram, 1965) in shaping the occurrence of aggression or people’s willingness to engage in it. For instance, studies of genocide and ethnic cleansing typically explore their political and social predictors (Harff, 2003; Kopstein & Wittenberg, 2018; Reicher et al., 2008; Staub, 1989, 2003) rather than the different forms that such intergroup aggression might take. Given that the differentiation of forms of aggression and its underlying motivations contribute to better understandings of this phenomenon at the individual level, it seems inadequate that research on intergroup aggression lacks a similar distinction (Littman & Paluck, 2015).
Addressing this shortcoming, we introduce a more differentiated approach to hate crime. Inspired by research on individual-level aggression (Archer & Coyne, 2005; Björkqvist, 1994; Card & Little, 2006), we distinguish between direct and indirect types of hate crime. Regardless of their commonalities (i.e., the biased motivation and the fact that hate crime affects both the person/persons who become its targets as well as their whole social group), the two types of hate crime differ significantly in terms of their goals and form. The primary goal of direct hate crime is to hurt, often physically, members of an outgroup. Direct hate crime is most often a result of a perceived provocation, has a tangible target, and resembles reactive aggression at the individual level (see Dunbar, 2003). Examples of direct hate crime include physical violence (e.g., battery, murder) and destroying the outgroup’s property (for instance destroying a car that belongs to a minority member).
In contrast, indirect hate crime does not inflict physical harm on the target and its target is less concrete (e.g., the outgroup as a whole or its parts). Indirect hate crime is very frequently verbal and aims to communicate that the targeted outgroup (or its members) is unwelcome in a given place and/or not accepted in a given community; it may also communicate threats to the outgroup. Overall, indirect hate crime assumes a more instrumental character (see Dunbar, 2003 for a similar discussion at the individual level). Examples of indirect hate crime include hostile verbal expressions, hateful graffiti, and vandalization of the outgroup’s religious/sacred objects. Even though religious property is still property and as such its vandalization could be construed as a direct hate crime, we consider it an indirect hate crime because of its significant symbolic meaning. Vandalization of religious objects and places of worship (e.g., defacing Jewish tombstones, throwing pigs’ blood or heads into mosques; see: Allen, 2017) may not bring any significant physical damage to them, but it constitutes an attack on what the outgroup finds most sacred and may lead to severe negative psychological consequences (e.g., Vincent, 2009).
Below we propose and test a set of predictors of direct and indirect hate crime. The idea that differences in types of hate crime and their unique predictors should be investigated for a better understanding of this phenomenon has been articulated in earlier work (Dunbar, 2003; Green et al., 2001; Messner et al., 2004) but to our knowledge this is the first study to empirically test this idea.
Predictors of Direct and Indirect Hate Crime
Given the novelty of our approach, our main goals were to provide evidence that the two hypothesized types of hate crime can be differentiated and that they are differentially predicted by a set of variables established in the literature as predictors of hate crime and intergroup hostility. To this effect we decided to focus on three sets of predictors: perceptions of threat from the outgroup, perpetrators’ associations with hate groups, and the political and economic conditions of the society.
Threat
An extensive body of research demonstrates that perceptions of outgroups shape intergroup emotions and behavior (Cottrell & Neuberg, 2005; Cuddy et al., 2007; Fiske et al., 2002). In the context of intergroup aggression, a crucial variable shaping intergroup behavior is the perception of threat from outgroups (Littman & Paluck, 2015; Reicher et al., 2008). According to the intergroup threat theory (Stephan & Stephan, 2000; Stephan et al., 2009) outgroups may be seen as posing two distinct kinds of threat: realistic or symbolic. Investigation of the former has roots in the realistic group conflict tradition (LeVine & Campbell, 1972; Sherif et al., 1961) which sees competition over limited resources as the main source of intergroup hostility. Subsequent research emphasized perceptions of competition rather than actual competition as the key predictor (Esses et al., 2005). Consequently, realistic threat entails seeing outgroups as threatening the economic, political, or even physical well-being of the ingroup (Stephan et al., 2009). In contrast, symbolic threat pertains to perceptions of outgroups as different in terms of values, beliefs, or worldview and, for this reason, dangerous to the ingroup’s morals and beliefs (Stephan & Renfro, 2002).
Difficult life conditions, e.g., political or economic turmoil, have been shown to increase competition over scarce resources (i.e., increase perceptions of realistic threat) and relate to increased intergroup violence (Hovland & Sears, 1940; Staub, 1989, 2003). Additionally, perceived threats to the ingroup’s physical and economic well-being are seen as more severe than threats to group coordination and values (Cottrell & Neuberg, 2005). In line with these findings, we hypothesize (H1) that outgroups perceived as posing more of a realistic threat elicit more aggressive reactions and thus are primarily targeted by direct types of hate crime. Consequently, outgroups seen as posing more of a symbolic threat meet with “milder”, more indirect forms of hate crime.
Membership in Hate Groups
While considerable diversity exists among hate groups, all are characterized by bigoted ideology and some form of organization. Most (though not all) also espouse white supremacist views, right-wing ideology, and affiliate with a religion (Gerstenfeld, 2017, pp. 141-147). A plethora of research has focused on the role of hate groups (e.g., neo-Nazis, skinheads) as perpetrators of hate crime (Craig, 2002; Dunbar, 2003; Dunbar et al., 2005; Gerstenfeld, 2017; Green et al., 2001; Watts, 2001) which is intuitive as these groups openly declare their bigoted beliefs and willingness to aggress against outgroups that they deem undesirable. Acceptance of and engagement in aggression and violence, typical of hate groups, normalizes it in those groups (Littman & Paluck, 2015). This normalization, coupled with the dehumanization of outgroups (Haslam & Loughnan, 2014) and gradual desensitization to violence (Carnagey et al., 2007), creates a vicious circle and makes further engagement in violence among hate group members more likely. And when they do commit hate crimes, these tend to be more severe than crime perpetrated by non-members (Dunbar et al., 2005). While the presence of hate groups makes the occurrence of hate crime more likely (Adamczyk et al., 2014; Mulholland, 2013), it is important to note that despite hate group members’ and sympathizers’ propensities for hate crime, research shows that, among all hate crime offenders, members of hate groups typically constitute a minority (Craig, 2002; Levin & McDevitt, 1993). This means that, notwithstanding bias-motivation of all hate crime offenders, the majority of them cannot be categorized as members of hate groups.
In light of the above, it seems important to differentiate hate crime perpetrators who belong to or sympathize with hate groups and those who do not. Since hate groups tend to engage in more severe forms of bias-motivated crime (Dunbar et al., 2005), we hypothesize (H2) that perpetrators who belong to hate groups would be more likely to commit direct hate crimes, while other offenders (i.e., those not belonging to hate groups) should be more likely to commit indirect hate crimes.
Social and Political Context
The investigation of the effects of economic and political context on the dynamics of hate-motivated intergroup aggression has a tradition much older than the definition of hate crime. In their seminal study, inspired by the frustration-aggression hypothesis (Dollard et al., 1939), Hovland and Sears (1940) established that the deterioration of economic conditions was associated with increased numbers of the lynching of Black Americans in the South of the U.S. While the original results have been challenged (Green et al., 1998; Mintz, 1946) and subsequent studies of the link between economic conditions and intergroup violence gave mixed results (Green & Rich, 1998; Krueger & Pischke, 1997; Piazza, 2017), the role of difficult life conditions is nonetheless considered a risk factor for intergroup violence (Staub, 1989, 2003).
Besides indicators of economic prosperity, like unemployment or the GDP, socially shared attitudes towards members of various minority groups (e.g., ethnic, national, religious) might also constitute an important hate crime predictor. Prejudice is, after all, what puts “bias” in bias-motivated crime. Already in the 1950s, Allport (1954) argued that intergroup violence is simply an extreme form of prejudice. Contemporary accounts also see prejudice as a prerequisite for hate crime (Franklin, 2000; Green et al., 1999), but investigations of the relation between societal-level prejudice and hate crime remain sparse (Green et al., 1998; Sitzer & Heitmeyer, 2008).
There is some evidence that intergroup attitudes shared by the general public relate to bias-motivated crime. In a study of right-wing violence in Germany, Ohlemacher (1994) demonstrated that public attitudes towards foreigners were significantly correlated with anti-foreigner hate crime. This is in line with a long tradition of research on normative social influence which demonstrates people’s general propensity to act in ways that conform to what they believe is normative (Asch, 1955; Crandall et al., 2002; Deutsch & Gerard, 1955). Given that it is mostly right-wing ideology that fuels bias-motivated crime (Gerstenfeld, 2017), it is to be expected that the more a society as a whole moves towards right-wing political preferences, the more hostile it should become to various outgroups. One indicator of such hostility may be greater prevalence of hate crime.
Although the relations between societal-level economic indicators and attitudes, and bias-motivated violence are not always straightforward, for exploratory purposes we included three indicators that may be potentially important to differentiating direct and indirect hate crime. In line with the realistic group conflict theory, we expected (H3) that negative economic conditions (operationalized as decreasing GDP and growing unemployment) predict greater prevalence of direct hate crime. We also expected that higher levels of negative intergroup attitudes (H4) and greater prevalence of right-wing political preferences in a society (H5) are positive predictors of direct forms of hate crime as both of these indicators are associated with hate crime (e.g., Chakraborti, 2015; Müller & Schwarz, 2021).
Method
Research Context
The present study was conducted in Poland, a country characterized by relatively low levels of ethnic and religious diversity (Gudaszewski, 2015) and high levels of intergroup prejudice, when compared to other European states (Zick et al., 2011). In recent years Poland has also taken a sharp anti-democratic turn (Csaky, 2021), however, the data analyzed herein was collected before these political changes occurred. Given Poland’s limited anti-hate crime legislation (see Footnote 2) and the fact that hate crime is often underreported, even in jurisdictions with more adequate legislation (Górska et al., 2016; Herek et al., 1999), official hate crime statistics constitute an inadequate data source. In order to circumvent this limitation, our investigation utilized the largest (unofficial) hate crime database in Poland, the “Brown Book” compiled by the Never Again Foundation (Iganski, 2011; Kornak, 2009, 2011; Kornak & Tatar, 2013; https://www.nigdywiecej.org/en/). The “Brown Book” catalogues instances of hate crime as defined by the broad definition of the term cited above. The catalogue of hate crimes was created on the basis of local and national media reports that are monitored by employees of the Foundation and a network of trained volunteers. Each entry is based on a media report and/or eyewitness testimonies and cross checked with other sources (e.g., other media sources, other witnesses, relevant organizations representing minority groups). Unconfirmed events are not included in the database and misreported cases are corrected/removed as necessary. The reliability of data collected by the Never Again Foundation is widely recognized. For instance, it is used as a source of data on hate crime in Poland by the Office for Democratic Institutions and Human Rights (https://hatecrime.osce.org/), a monitoring body within the Organization for Security and Cooperation in Europe.
Our analyses were conducted on the database of incidents that occurred between 1991 and 20133. In total 3,153 events were recorded (see Figure 1). Besides a written description, characteristics of each event and the involved parties are coded by the “Brown Book” authors using a system of categories. Specifically, hate crime events are categorized into 16 unique types (e.g., “Use of physical violence against members of an outgroup, battery”, see Table 1 for the full list). Based on their social and/or political affiliations, perpetrators are categorized into eight types (e.g., “Members of particular hate groups/organizations”) and victim(s) are categorized into nine types (e.g., “National or ethnic minorities” [Jews, Roma, Germans, Ukrainians, etc.]; see Table 2 for details). Importantly, hate crimes are complex events and their descriptions often contain more than one type of violent activity described above. In order to preserve this complexity, when more than one category of perpetrator or victim were involved or the event consisted of different actions, all of these elements were represented in the coding system. For example when a Jewish person was verbally assaulted and beaten up by a skinhead, such an event was coded as containing both verbal and physical aggression, as well as one category of victim (national minority) and one category of perpetrator (member of a hate group).
Figure 1
Frequency of Hate Crime Events Recorded in the Dataset Per Year
Table 1
Types of Hate Crime Divided Into Direct and Indirect Hate Crime Categories
| Character of the event
|
|
|---|---|
| Indirect hate crime | Direct hate crime |
|
|
Table 2
Categorization of Perpetrators and Victims of Hate Crime
| Categories of perpetrators | |
|---|---|
| Hate group members | Others |
|
|
| Categories of victims | |
| Perceived as symbolically threatening | Perceived as realistically threatening |
|
|
aMost of the people in this category were racist soccer hooligans attacking or shouting racist slurs at people of color (including players).
Measures
Event-Level Independent Variables
Characteristics of the perpetrators and victims4 of a given hate crime were used as predictors of the type (direct versus indirect) of hate crime.
Characteristics of the Victims
The “Brown Book” distinguishes between nine types of hate crime victims (e.g., ethnic and national minorities [such as Jewish people, Roma people, Ukrainians, Belarussians]; political adversaries). Based on the integrated threat theory (Stephan & Renfro, 2002) we split these distinct groups into two broad categories that represent perceived threat that is predominantly symbolic versus realistic. The first category (symbolically threatening groups) consisted of all political and religious groups within Polish society that are defined primarily by their distinct values and worldviews and the latter category (realistically threatening groups) consisted of all of the ethnic and national minorities who are often seen as competing for scarce resources with the ethnic majority members (i.e., ethnic Poles). In the following analyses, the proportion of victims belonging to groups perceived as realistically threatening to all victims in a given crime was used as an indicator of the type of victim. This proportion had a range from 1 (if all victims of a given crime belonged to groups perceived as realistically threatening) to 0 (if all victims of a given crime belonged to groups perceived as symbolically threatening). Overall 59.6% of all hate crime instances were directed at groups seen as realistically threatening (see Table 2).
Characteristics of the Perpetrators
The “Brown Book” distinguishes between eight types of perpetrators. In order to simplify our analyses and to verify H2, we categorized them as either members of hate groups (e.g., skinheads) or those who did not belong to such groups. For a more conservative test of our hypothesis, perpetrators categorized as “other” and “unknown” in the original source were included in the non-hate group category. As noted above, some events were perpetrated by both hate group members and non-members. The former were involved in 68.4% of the catalogued events (see Table 2) and they were sole perpetrators of 57.8% incidents. In the analyses presented below a proportion of hate group members/sympathizers among perpetrators to all perpetrators of a given hate crime was used as an indicator of the type of perpetrator. This proportion had a range of values from 1 if only hate group members/sympathizers were involved in a given crime to 0 when only non-hate group members perpetrated a given hate crime.
Contextual-Level Independent Variables
In order to account for the role of the economic, political, and intergroup context at the country level in shaping the nature of hate crimes, a number of country-level variables were included in the analyses.
Economic Situation
The economic situation in the country was approximated using the unemployment rate (Statistics Poland, 2019) and the annual GDP growth data as reported by the World Bank (2018).
General Political Climate
The political climate was operationalized as the average declared political preference (left- versus right-wing). The data came from the Centre for Public Opinion Research (CBOS), a Polish polling agency which has conducted large, annual, representative sample surveys of political attitudes among Poles since 1989 (Kazanecki, 2015).
Intergroup Attitudes
We also created an index of negative intergroup attitudes. The index was based on CBOS’ annual representative sample survey of attitudes towards other nations and towards ethnic and national minorities in Poland (Omyla-Rudzka, 2019). The index was created as the average percentage of respondents who declared negative attitudes toward two most salient and strongly disliked minorities in Poland: the Roma and Jewish people. It is noteworthy that attitudes towards these two groups (unlike most others who were sometimes added and sometimes omitted from the surveys) have been measured in all CBOS surveys, making them reliable indicators of the changes in intergroup attitudes over time.
Dependent Variable
The main dependent variable was an index of relative prevalence of indirect vs. direct type of hate crime elements. As indicated above, each event was categorized as exemplifying at least one of 16 hate crime types. In order to create the index, four competent judges (scholars specializing in the study of intergroup relations with at least a master’s degree in social science) received definitions of direct and indirect hate crime and were asked to categorize the 16 types of events differentiated in the “Brown Book” as representing either one or the other. The inter-rater reliability was high: Krippendorff's Alpha for nominal variables calculated using bootstrap with 1000 resamples was .87 with 95% CI [.77, .96] (Hayes & Krippendorff, 2007). Any inconsistencies were resolved using majority categorization (i.e., we categorized a given hate crime as direct or indirect if at least 50% + 1 judges suggested to do so). The final categorization of events as direct or indirect hate crime is presented in Table 1.
Because the numbers of recorded hate crime incidents differed vastly between different years (see Figure 1) and our interest was in the type rather than number of attacks, we did not focus on the raw numbers of hate crimes. The final index was calculated as a proportion of indirect hate crime elements to all hate crime elements in each of the described hate crime events. For instance, if a given event had three elements (battery, verbal abuse, and hateful graffiti) the value of our indicator for this event would be 0.66 (i.e., 2/3 elements constituted indirect violence). The average value of this indicator was M = 0.55 (SD = 0.48), which shows that overall indirect hate crime was slightly more prevalent than direct hate crime.
Analytic Strategy
In order to analyze the relation between the independent (event-level and contextual-level) variables and the relative prevalence of indirect (vs. direct) type of hate crimes we tested several multilevel models, using robust maximum likelihood estimators. We used the stepwise strategy recommended by Hox et al. (2017) and assessed differences between models using the log-likelihood test (Satorra & Bentler, 2010).
Results
In the first step we computed a null model (see “Null model” in Table 3) to examine the proportion of variance in the dependent variable that can be attributed to second-level (contextual) variables. This simple two-level model consisted only of the clustering variable (year) and the dependent variable (type of hate crime) and allowed us to assess the amount of variance in the dependent variable that can be attributed to either of the two levels. The intra-class correlation was ICC = 0.12, indicating that 12% of the variance in the dependent variable can be attributed to clustering at Level 2.
Table 3
Multilevel Regression Models for Indirect Violence Nested in Years of Observation
| Variable / Model | Null Model | Fixed Level 1 predictors |
Fixed Level 2 predictors |
Random model: Victims |
Random model: Perpetrators |
Random model with interactions |
|---|---|---|---|---|---|---|
| Intercept | 0.53 (0.03)*** | 0.62 (0.02)*** | 0.62 (0.01)*** | 0.63 (0.01) | 0.62 (0.01)*** | 0.64 (0.01)*** |
| Level 1 predictors (event-related) | ||||||
| Victims (0 = only symbolic threat; 1 = only realistic threat) | 0.43 (0.04)*** | 0.43 (0.04)*** | 0.43 (0.04)*** | 0.42 (0.04)*** | 0.42 (0.03)*** | |
| Perpetrators (0 = other; 1 = hate groups) | -0.22 (0.03)*** | -0.21 (0.03)*** | -0.20 (0.02)*** | -0.26 (0.03)*** | -0.23 (0.02)*** | |
| Level 2 predictors (contextual variables) | ||||||
| Neg. int. attitudes | -0.37 (0.09)*** | -0.49 (0.18)** | -0.29 (0.09)** | -0.39 (0.12)*** | ||
| Political attitudes | -0.10 (0.04)* | -0.12 (0.05)* | -0.04 (0.07) | -0.09 (0.05)† | ||
| Unemployment | 0.01 (0.002)*** | 0.01 (0.003)*** | 0.01 (0.002)*** | 0.01 (0.002)*** | ||
| GDP growth | 0.003 (0.005) | 0.002 (0.006) | 0.003 (0.006) | 0.004 (0.006) | ||
| Cross level interactions | ||||||
| Victims x Neg. int. attitudes | 0.93 (0.20)*** | |||||
| Perpetrators x Political attitudes | -0.29 (0.07)*** | |||||
| Random part | ||||||
| Within year variability | 0.20 (0.005) | 0.138 (0.006) | 0.14 (0.006) | 0.13 (0.006) | 0.13 (0.006) | 0.13 (0.006) |
| Between year variability | 0.03 (0.009) | 0.006 (0.003) | > 0.001 (0.001) | 0.001 (0.001) | 0.002 (0.002) | 0.001 (0.001) |
| Variance of slopes Victims | 0.02 (0.005) | 0.003 (0.003) | ||||
| Variance of slopes Perpetrators | 0.01 (0.006) | 0.005 (0.003) | ||||
| Deviance | 3817.64 | 2018.01 | 1991.73 | 1943.86 | 1974.37 | 1903.45 |
Note. Neg. int. attitudes = negative intergroup attitudes. SE of estimate is presented in parentheses.
†p < .08. *p < .05. **p < .01. ***p < .001.
In the second model (“Fixed Level 1 predictors” in Table 3) we added event-level predictors: the type of victim group (a proportion with the value of 1 if all victims belonged to groups perceived as realistically threatening and 0 if all victims belonged to groups perceived as symbolically threatening) and the type of perpetrator (a proportion with the value of 1 if only hate group members were involved and 0 if only non-hate group members were involved). The results show that realistically threatening groups were more often targets of indirect hate crime whereas perpetrators belonging to or sympathizing with hate groups were more often involved in direct hate crime. A substantially lower deviance for this model confirmed a better data fit when compared to the null model. Additionally, the amount of variance explained due to clustering remained significant.
In the third model (“Fixed Level 2 predictors” in Table 3) we included all Level-2 predictors (i.e., the contextual-level variables). We found a significant positive effect of unemployment and significant negative effects of negative intergroup attitudes and right-wing political preferences. Specifically, in years with higher unemployment, there was a higher prevalence of indirect hate crimes, whereas years with higher levels of societal negative intergroup attitudes and more right-wing political preferences were characterized by a greater prevalence of direct hate crimes. Adding Level-2 predictors decreased the variability between years to a level of non-significance. This suggests that a proportion of between-year differences in type of hate crimes can be explained by contextual variables. We also detected a substantial decrease in deviance, which suggests a good fit of the model to the data.
In the next steps we added a random component for the type of victims and perpetrators (in two independent models; see “Random model: Victims” and “Random model: Perpetrators” in Table 3) using a variable-by-variable strategy in order to avoid overparameterization (Hox et al., 2017). Results show that for both the type of victim and the type of perpetrator there is significant variability of the slope between the years. We tested cross-level interactions between explanatory contextual-level variables and event-level variables and found two significant interactions: between negative intergroup attitudes and type of victims, and between right-wing political preferences and type of perpetrator. We included these interaction terms in the final model (“Random model with interactions”, in Table 3). This model, inclusive of the random part and cross-level interactions, turned out to have the best fit to our data. Most of the effects are similar to those described in previous models. However, the effect for contextual-level political preferences was no longer significant (even though a statistical trend towards significance was still noticeable).
In order to interpret the interaction effects two separate plots were created. Results showed that when right-wing political preferences were stronger in society, hate group perpetrators were more involved in direct hate crimes. No effect was observed when right-wing political preferences were relatively less prevalent in society (see Figure 2).
Figure 2
Cross Level Interaction Between Right-Wing Political Preferences (RW att) and the Type of Perpetrators (Hate Group Members vs. All Perpetrators); Dependent Variable: Type of Hate Crime (Indirect vs. All Hate Crimes)
Note. Dotted lines indicate 95% confidence intervals (UL = upper limit; LL = lower limit).
An analysis of the second interaction term showed that when negative intergroup attitudes were more widespread in society, outgroups perceived as more realistically threatening were targeted with hate crime of a more indirect type, however, when negative intergroup attitudes were relatively low the effect was opposite and those same outgroups were more likely to be victims of more direct types of hate crime (see Figure 3).
Figure 3
Cross Level Interaction Between the Index of Negative Intergroup Attitudes (Xeno) and the Type of Victim Group (Realistically Threatening vs. All Victims); Dependent Variable: Type of Hate Crime (Indirect vs. All Hate Crimes)
Note. Dotted lines indicate 95% confidence intervals (UL = upper limit; LL = lower limit).
Discussion
The main purpose of the current study was to introduce a more fine-grained approach to analyzing and understanding hate crime. Drawing insights from the psychology of violence (Anderson et al., 2003; Archer & Coyne, 2005) and the psychology of intergroup relations (Littman & Paluck, 2015; Reicher et al., 2008; Stephan et al., 2009) we argued that at least two distinct types of hate crime, direct and indirect, may be differentiated and that the relative prevalence of either type could be predicted by characteristics of perpetrators and victims of a given hate crime as well as by country-level contextual variables reflecting the state of the economy, political climate, and the level of negative intergroup attitudes in society. We conceptualized direct hate crimes primarily as acts of physical aggression (battery, murder, sexual violence) as well as destruction of the victim’s property. Our understanding of indirect hate crime sees it as various forms of verbal abuse, the vandalization of religious objects, and the propagation of bigoted ideology. Following categorization by four competent judges, we divided all instances of hate crime in a large Polish hate crime database into direct and indirect categories and used a multi-level modelling approach to verify a set of predictions regarding their relations with event-related and contextual variables.
The results show that indirect hate crime was slightly more prevalent than direct hate crime (50.4% vs. 40.7%). Nearly 70% of recorded hate crime perpetrators were members or associates of hate crime groups and about 60% of hate crime victims belonged to groups perceived as realistically threatening the ethnic majority. Moreover, groups seen as posing more of a realistic threat (Stephan et al., 2009) were more likely to become targets of indirect hate crime. Conversely, groups that differ from the majority population primarily in terms of values and worldviews (i.e., those who are perceived as symbolically threatening) were more likely to be targeted by direct hate crime. This result was inconsistent with our expectation (as well as previous research, e.g., Cottrell & Neuberg, 2005) that groups perceived as symbolically threatening would be targeted with more indirect hate crime while those seen as realistically threatening would be targeted with more direct hate crime (H1). In line with H2, indirect hate crime was more prevalent among perpetrators who did not belong to hate groups whereas perpetrators belonging to hate groups were more likely to engage in direct hate crime. All of these effects were relatively strong and remained significant in the final model, indicating that they operated independently of the contextual variables included in the analyses.
Changes in the GDP did not relate to the type of hate crime but greater unemployment in a given year was related to greater prevalence of indirect hate crime, which contradicted our predictions (H3). In line with our hypothesis (H4) greater prevalence of negative intergroup attitudes in society was associated with more direct forms of hate crime. Societal-level political preferences were only marginally related to a greater prevalence of direct hate crime in the final model, providing somewhat weak support for H5.
Results of our analyses support the utility of differentiating direct and indirect hate crime and contribute to the literature in several ways. To our best knowledge, the present research is the first to analyze different types of hate crime, relations among characteristics of the victims, the perpetrators, and the social context, and the characteristics of hate crime events. We did so by using a series of multilevel models and provide evidence that the two hypothesized types of hate crime may be reliably differentiated and that they are predicted by a set of explanatory variables.
We also demonstrated that hate group members were more likely to engage in direct hate crime (when compared to general population perpetrators who are more likely to commit indirect hate crimes). This is in line with research showing that hate groups openly proclaim violence against groups they deem undesirable (Carnagey et al., 2007), subscribe to ideologies that dehumanize potential targets (Haslam & Loughnan, 2014), and commit more violent hate crimes than people not affiliated with hate groups (Dunbar et al., 2005). Interestingly, nearly 70% of all hate crimes reported by the “Brown Book” were perpetrated by members of hate groups. This contrasts with the extant literature that claims that only a minority of bias-motivated crimes are perpetrated by members of hate groups (Levin & McDevitt, 1993). A possible explanation for this outcome lies in the breadth of the definition of hate groups used in our work. Due to their violent tendencies and ties to hate groups (Garland & Rowe, 2001; van Limbergen et al., 1989), we classified soccer hooligans as members of hate groups. However, such categorization is not typical in hate crime research. This is because hooligans’ hateful rhetoric is seen as a byproduct of violent soccer culture rather than a goal in and of itself (van Limbergen et al., 1989), which is characteristic of the more established hate groups. We claim that in contemporary Poland (Pankowski & Kornak, 2013) and other parts of Europe, right-wing ideology has become a crucial part of hooligans’ identities (Garland & Treadwell, 2010; Nielsen, 2013), thus justifying their inclusion in the hate group category.
The results of our analyses also showed that groups perceived as realistically threatening were more likely to be targeted with indirect types of hate crime. While this was not in line with our predictions, a possible explanation for this effect is that Poland is a very ethnically homogeneous country–over 96% of the population is ethnically Polish (Gudaszewski, 2015). Very low levels of diversity mean that those who intend to harm realistically threatening minority groups may not be able to physically attack members of these groups (because they are simply difficult to come across) and thus resort to more indirect forms of hate crime, e.g., hateful graffiti. This contention necessitates further research, possibly using interviews with hate crime perpetrators.
At the societal level, negative intergroup attitudes and right-wing political preferences were linked to greater prevalence of direct hate crimes. Additionally, a cross-level interaction showed that when right-wing political preferences were widespread, hate group members were more likely to be involved in direct hate crimes. We believe that this could be interpreted as an example of societal level radicalization. Specifically, when social norms become more hostile (i.e., when the societal-level intergroup attitudes become more negative and right-wing political preferences more prevalent), the extreme part of the political scene may radicalize even further. Psychologically this mechanism might be understood as an attempt to protect optimal distinctiveness: when the mainstream society moves to the right, hate groups (that are predominantly right-wing) go even further with their actions in order to remain distinct (Brewer, 1993). The finding that intergroup attitudes and political orientation impact the form of hate crime is also in line with social psychological literature on the relation between social norms and violence which shows that both pro-bullying and anti-bullying classroom norms impact bullying roles taken up by students (Salmivalli & Voeten, 2004). In total, around 10% of variance in bullying behavior is explained by the classroom context (Salmivalli, 2010). Similarly, neighborhood street norms predict violence beyond individual attitudes and the relation between individual attitudes and violence is stronger in neighborhoods with a greater prevalence of violent social norms (Stewart & Simons, 2010).
A significant cross-level interaction showed that more negative intergroup attitudes in society relate to groups perceived as realistically threatening becoming likely targets of indirect hate crime. At the same time, when the level of negative intergroup attitudes decreases, realistically threatening groups become targets of more direct hate crime. We suggest that these results might be related to shifting social norms and to the limited “availability” of groups seen as realistically threatening in Polish society. Overall, indirect hate crime is more likely to be perpetrated by people who do not belong to hate groups. When social norms shift towards greater intergroup hostility (stronger negative intergroup attitudes) it seems to result in an overall increase in hate crime. Specifically, when controlling for direct hate crime perpetrated by hate groups members, it can be seen that indirect violence toward groups seen as realistically threatening perpetrated by people who do not belong to hate groups is also on the rise. This finding is in line with results showing a positive relation between negative intergroup attitudes (islamophobia) and presence of hate speech (Winiewski et al., 2017).
Contrary to our predictions, a less prosperous economy, approximated by higher rates of unemployment (but not GDP which was not a significant predictor) was related to a greater prevalence of indirect hate crime. This effect could be explained by the relative deprivation theory (Pettigrew et al., 2008). Specifically, we believe that the increase in individual frustration (due to high unemployment) in connection with hostile intergroup attitudes and conservative, right-wing political ideology may relate to an increase in the perpetration of hate crime by ordinary, non-ideologically engaged citizens who typically resort to more indirect forms of hate crime.
The presented study is not without limitations. First, the “Brown Book” is not an exhaustive catalogue of all hate crime events. It was created by a non-governmental organization whose monitoring procedures rely on a network of voluntary informants. This means that we cannot fully rule out that there exists a bias in terms of some crimes being more likely to be recorded than others. Having said that, the system of data collection that the “Brown Book” relies on is standardized (i.e., volunteers systematically monitor local and national press and cross-check reports) and provides a more reliable picture than official hate crime statistics. Second, the numbers of crimes recorded in the early 1990s (i.e., at the beginning of the cataloguing effort) were lower than those recorded in the late 1990s and in the 2000s, making analyses using raw numbers challenging. In order to circumvent the problem of unequal frequencies we focused on the relative prevalence of different types of hate crime rather than absolute numbers. With this approach, only one basic assumption needs to be met: the procedure for collecting and processing data needs to be constant. We have no reason to doubt that this assumption was met, given the descriptions of the data collection provided in the “Brown Book” itself (Kornak, 2009, 2011; Kornak & Tatar, 2013). Additionally, in our analyses we focused on the relations between relatively objective characteristics of the discussed events and contextual variables that were collected independently of the "Brown Book" (e.g., GDP, political attitudes). Lastly, the presented analyses focused on only one cultural context and although they provide strong evidence for the value of differentiating direct and indirect hate crime, this evidence should be treated as an initial validation of the proposed theoretical framework. Future research is needed to demonstrate that direct and indirect hate crime provide a useful framework to be applied in other cultural contexts as well.
In conclusion, research presented in this article constitutes the first analysis of direct and indirect hate crime. We provide evidence that the relative prevalence of the two types of hate crime is significantly shaped by the type of target group, type of perpetrator (hate group member versus not), and a host of societal-level variables. We hope that following this approach will contribute to a better understanding of hate crime as a social phenomenon, as well as to hate crime prevention and more adequate victim support.