A configurational perspective of hate crime policies

Abstract

A configurational perspective is adopted to analyse the contributory factors of hate crimes in the USA. Through a fuzzy-set qualitative comparative analysis, multiple nonlinear combinations of the policies leading to high or low hate crimes are revealed. The results provide a complementary and substitutive multiple-policy strategy to policymakers in navigating a complex, unstable environment. Specifically, the study reveals different pathways to achieving the same outcome, and certain casual relationships are found to be asymmetric and inconsistent across outcomes. Several important findings are noted from our analyses. The magnitude of hate crimes in a state is found to be monotonically related to the percentage of immigrants in the population and to labor market freedom. However, such monotonic results do not necessarily extend to other variables. For example, expenditure on education and median household incomes are associated with both high and low hate crimes.

INTRODUCTION

In this paper, the contributory factors of hate crimes in the USA are examined. The Uniform Crime Reporting (UCR) program of the Federal Bureau of Investigation (FBI) defines hate crime as a committed criminal offense that is motivated, in whole or in part, by the offender’s bias(es) against factors such as race, ethnicity, religion, sexual orientation, and disability.¹ Hate crimes can take many forms, such as vandalism, assault, harassment, or even murder. The key element that distinguishes a hate crime from a regular crime is the motivation behind it, which is based on the offender’s hatred or prejudice towards the victim’s identity or membership in a particular group. Hate crimes not only harm individual victims but also create a climate of fear and intimidation within the targeted communities. They are considered serious offenses and may be subject to enhanced penalties under the law.

In the literature, several factors have been hypothesized as having a causal relationship with hate crime. One factor that has been studied in literature is the role of education. This hypothesis originates from the work of Becker (1968). The opportunity cost of criminal activity is the wages one can earn legally. Educated persons are likely to have a higher wage, and therefore have a lower incentive to commit crime. Krueger and Pischke (1997) studied the causes of ethnic violence in Germany in the 1990s and found no connection between the average education level in a county and the magnitude of violence against foreigners. Jensen et al. (2020) considered a sample of 689 violent crime offenders and 277 non-violent offenders who were all accused of hate crime between 1990 and 2018 in the USA and found that 76 per cent of the total number of offenders in the sample had low levels of education. Further, 58 per cent of the violent offenders had low levels of education. Dancygier (2023) examined the profile of supporters of hate crimes in Germany and found that people with low levels of education are more likely to support or justify hate crimes. Overall, the literature does not find any evidence of a negative relationship between education levels and hate crimes. The total expenditure on education and average math scores are controlled for in this study. Test scores control for the quality of education. Therefore, total education expenditure and average math score together control for the quantity and quality of education.

A second possible factor that affects hate crime is the condition of the labor market. In the literature, this aspect has mostly been captured by the unemployment rate. Krueger and Pischke (1997) and Green et al. (1998) concluded that unemployment has no relationship with hate crimes. In contrast, Gale et al. (2002) and Ryan and Leeson (2011) concluded that unemployment increases hate crime. Overall, there is mixed evidence about the relationship between hate crime and unemployment. Unemployment captures the probability of finding a job. While this is important, there are also other aspects of the health of the labor market such as the ease of hiring or firing workers, and the rigidity of hours. These additional factors also affect the well-being of workers. Therefore, instead of the unemployment rate, the health of the labor market is measured using Labor Market Freedom developed by the Heritage Foundation.

Another factor that can affect hate crime is income. Dancygier (2023) found widespread support for hate crimes across all income levels. However, overall prior research has not found any strong link between average income and hate crime in a geographical area. In this study, the median household income in a state was considered. It captures the quality of life in a state. In addition, it allows us to find out if fluctuations in income have any relationship with hate crimes.

Prior research has also considered the proportion of immigrants in the population as a possible factor in hate crimes. Espiritu (2004) found that a higher proportion of immigrants results in a higher incidence of hate crimes. A report prepared by the Department of Justice in 2014 found an increase in hate crimes against Hispanics in the mid-2000s, but a decrease thereafter.² In recent times, particularly after the coronavirus disease 2019 pandemic, there has been an increase in hate crimes against Asians. As for the reasons, one is that “….perpetrators resent the growing economic power of a particular racial or ethnic group and engage in scapegoating; others react to a perceived threat to the safety and property value of their neighbourhood” (Bureau of Justice Assistance 1997). This is also consistent with the racial threat hypothesis (Blalock 1967) which suggests that as the population of a minority group increases, the majority group perceives this growth as a threat. This perceived threat can be economic, political, or criminal, leading the majority group to implement stricter social controls to maintain their dominance. This theory has been used to explain various social phenomena, including higher rates of incarceration and stricter law enforcement in areas with larger minority populations (Wang and Todak 2024). In recent work, Cikara et al. (2022) found that members of the largest minority group in an area are more likely to be victims of hate crimes. The reason is that members of the majority community feel most threatened by members of the largest minority group.³

The role of addictive substances is also considered. Harlow (2005) examines the National Crime Victimization Survey between July 2000 and December 2003 and finds that 30.6 per cent of the hate crime victims perceived the perpetrators to have taken either drugs or alcohol. Jensen et al. (2020) found that 28 per cent of the total number of offenders in their sample had a record of substance abuse. Hamad (2019: 8) reviewed the literature on hate crime and described the profile of a typical perpetrator. One of the characteristics of a typical perpetrator is substance abuse. Overall, there is evidence that substance abuse is a possible factor in hate crimes. Therefore, the role of one type of substance abuse, that is, alcoholism is examined. Specifically, alcoholism serves as a measure for the proportion of the population using addictive substances during the commission of a hate crime A measure called “Binge Drinking Score” developed by the Centers for Disease Control and Prevention (CDC) assigns a score to each state depending on how well they are doing in the context of binge drinking, with a higher value indicating a higher health status/quality of life in the state with regard to binge drinking is used.

Another stream of research and factor to consider is whether terrorism and hate crime are closely related. Enders and Sandler (2002) define terrorism as “the premeditated or threatened use of extra-normal violence or force to obtain a political, religious, or ideological objective through the intimidation of a large audience.” Hate crimes and terrorism are violent crimes that are motivated by prejudice or hatred and therefore have some degree of commonality. However, there is a difference of opinion regarding their closeness, with one group of researchers perceiving them as “close cousins” and the other group as “distant cousins.” Krueger and Maleckova (2002: 120) considered terrorism and hate crime to be “close cousins.” Mills et al. (2017) consider hate crimes by non-extremists, hate crimes by far-rightists, and terrorism, and found that an increase in any type of such crime causes an increase in each of the others in the same area. They conclude that these three are “close cousins.”

However, there is an opposing viewpoint also. Hate crimes are typically motivated by a desire to harm or intimidate a specific group of people, while terrorism is motivated by a desire to achieve a political or ideological goal. Hate crimes typically target individuals or small groups, while terrorism typically targets larger populations. Deloughery et al. (2012) argue that if terrorism and hate crime are indeed closely related, they should have similar triggers. However, they find that while terrorism causes hate crime, the converse is not true. This is because terrorism is often perpetrated by a member of a weak or marginalized group, while hate crimes are typically perpetrated by a member of a strong group. Therefore, they conclude that terrorism and hate crime are not close cousins. In conclusion, while hate crimes and terrorism are undoubtedly cousins, it is not certain that they are close cousins. Therefore, several factors such as lack of political rights that affect terrorism but have not been considered in the literature on hate crimes are excluded.

In an assessment of the use of complexity theory in public policy, Cairney (2012) made two important observations. First, public policymaking analysis does not simply constitute individual elements. They exhibit interdependence and interact with one another to produce the observed outcomes. Second, policy implementation environments are not uniform, stable, or permanent. Therefore, the effectiveness of a policy may be contextual. This calls for a multiple-policy strategy to allow pivoting. Further, the complexity of cause–effect relations in public policymaking cannot be adequately conceptualized, modeled, or analysed (even with a sufficiently large sample size) with a traditional statistical framework, such as regression, which measures the net effects of independent variables. While the introduction of interaction and moderation variables can accommodate certain conditions, it also rapidly increases the number of estimated parameters, especially in high-order interactions. This renders a meaningful interpretation almost impossible. Qualitative Comparative Analysis (QCA) can meet these challenges due to its natural ability to conceptualize research models from a configurational (combination of conditions that lead to an outcome) perspective. QCA studies the interconnected dynamics of a complex system, in which the impact of one condition on the outcome of interest is also dependent on other conditions. A change in one condition can trigger changes in other conditions, eventually changing the entire solution structure.

QCA provides the lens through which the complex causality in hate crime policymaking where numerous policies interplay simultaneously in causing the presence (or absence) of a particular outcome can be deciphered. This causal–effect relation has three attributes: conjunctural, equifinal, and asymmetric (Ragin 1987). Conjunctural causation, the first attribute, reveals if and how conditions interdependently lead to an outcome through a configurational perspective. This contrasts with multivariate statistical techniques where interaction terms must be introduced, leading to an increasingly challenging interpretation once the order of interactions exceeds two. Equifinality, the second attribute, allows alternative conditions or a combination of conditions leading to the same outcome. This naturally provides multiple sufficient solutions. On the contrary, multivariate statistical techniques produce a single causal condition which is insufficient for complex problems thus requiring a multiple-policy approach. Asymmetry, the third and final attribute, recognizes that the conditions causing the failure of a policy are not necessarily the inverse of the conditions causing the success of the same policy. This means the former must be assessed separately in analysing policy failure.

With world problems becoming increasingly complex, there has been exponential growth in QCA-based studies since 2007, spreading to a wide variety of disciplines. Out of the 469 articles surveyed by Roig-Tierno et al. (2017), politics, business and economics, and sociology (also the top three) account for 54 per cent of the total. This growing popularity manifests QCA’s effectiveness in determining the contexts in which a cause or a combination of causes influences a particular outcome for a wide variety of complex political, social, and business issues. As with any technique, there are certain assumptions and limitations that need to be acknowledged for fuzzy-set qualitative comparative analysis (fsQCA) (see Wagemann and Schneider 2007).

RESEARCH DESIGN, METHODOLOGY, AND ANALYSIS

Background

Charles Ragin, an American sociologist, developed QCA in the 1980s to address the inadequacy of traditional multivariate statistical techniques in assessing cause–effect relationships for problems of complex patterns but of a small sample size. QCA integrates the best features of the case-oriented approach with the best features of the variable-oriented approach (Ragin 1987). The former which generally focuses on a small number of cases produces in-depth accounts of cases but then these findings are not generalizable and therefore lack a systematic process for cross-case comparison. In contrast, the latter typically analyses large samples but lacks detailed information about individual cases. QCA remediates this split because it facilitates within-case complexity assessment and systematic cross-case comparison. Ragin’s initial intent was to analyse a small sample of macro-level cases but then QCA was extended to a large sample of survey data involving organizations and individuals (e.g. Greckhamer et al. 2013; Emmenegger et al. 2014). The major phases of our QCA analysis are detailed in the sections below.

Cases, outcomes, and conditions

A QCA study has three fundamental elements—cases, outcomes, and conditions. Different from quantitative research, cases in the QCA study must be selected purposefully based on the study focus. Our key outcome variables of interest are education measured by total expenditure in education and average math score, immigration based measured by % Foreign-Born population, health status measured by binge drinking score, quality of life measured by median household income, and labor market conditions based measured by Labor Market Freedom.

The research model is depicted in Fig. 1. The aim is to understand the nonlinear relationships and interdependencies of causal factors/policy instruments impacting hate crime through the three tenets of QCA—conjunctural causation; equifinal causation; and asymmetric causation, identifying the necessary and sufficient conditions that lead to the outcomes.

In a QCA study, cases must be comparable but also exhibit sufficient variety with regard to the outcome and the conditions. Following this principle, the cases using information from a variety of sources such as the Bureau of Labor Statistics, Centers for Disease Control and Prevention, Census Bureau, and the Department of Justice were built. A total of 1,100 cases were collected, spanning 2000–2021, and covered 50 states in the USA. A detailed discussion of the variables and related sources can be found in Table 1.

Some variables such as total expenditure on education, average math score, median household information, % Foreign-Born and total population are based on objective facts and are less susceptible to systematic measurement errors. Some other variables such as labor market freedom and binge drinking score are based on subjective judgments. They are however obtained from reputable agencies, and we are not aware of any major deficiencies of their data. The number of hate crime incidents has elements of subjectivity because it depends on the judgment of law enforcement agencies about the motive of crime. Other researchers (such as Deloughery et al. 2012) have also expressed concerns about the quality of this variable. Nevertheless, our source is the best available on hate crimes in the USA.

The summary statistics of our outcome and causal variables are reported in Table 2.⁴ Notice that total expenditure on education is a highly dispersed variable because the standard deviation is larger than the mean. In our methodology, group membership depends on the percentile ranking of an observation and the results are therefore unaffected by the degree of dispersion.

Calibration

Analysis of the casual complexity can be conducted with a set-theoretic approach using either crisp-set (csQCA) or fuzzy-set (fsQCA), two commonly applied forms in the literature. In csQCA, membership scores—“1” for full membership or “0” for full non-membership—are assigned to each case, indicating whether they are a member of the sets. The values of conditions are also dichotomous with “0” indicating fully absent and “1” fully present. Many applications like ours, however, have conditions or outcomes with a spectrum of values and this type of “fuzzy” data calls for the use of fsQCA where calibration is required.

Calibration is arguably the most important step in the fsQCA analysis. The process assigns set-membership scores to indicate the degree to which cases are members of sets or the degree a condition is present or absent. Membership scores must be assigned to each case for both the conditions and outcome. As pointed out by Ragin (2008), a fuzzy membership score represents a truth value, not a probability, of a condition or outcome. Arellano et al. (2021; Appendix C.5) outline three approaches to calibration: (1) manual attribution using “substantive and theoretical knowledge”; (2) direct approach using three anchor points; and (3) indirect approach using regression techniques. These three approaches apply different rules to translate data into a set membership score. While approach (1) is preferred because it is based on theoretical reasons, approach (2) is widely adopted in the absence of theoretical knowledge, relying on the distributional information of the data. Approach (3) is not commonly applied in the literature.

The direct method through within-sample distribution is used to establish the anchor points for full membership, the crossover point, and full non-membership scores (Ragin 2008; Fiss 2011; Rubinson et al. 2019) for the quantitative variables that form the focus of our research endeavor. Following Pappas et al. (2017) and Pappas and Woodside (2021), the 80th, 50th, and 20th percentiles are used as the thresholds for full membership, crossover point membership, and full non-membership, respectively, to account for skewness. Since our data does not follow a normal distribution and is skewed, these thresholds are appropriate for our study.

The variables were calibrated using R (Thiem 2018) and Table 3 summarizes the calibration results. As indicated in Table 3, one condition required reverse coding to facilitate result interpretations.

Truth tables

After calibrating the variables, the truth table is generated. As the fsQCA program drop cases with a 0.5 membership score (because they mean neither in nor out of the set), we follow the suggestion of Fiss (2011) and add a constant of 0.001 to the membership scores of the causal conditions that are below 1. Using the fsQCA software (Ragin and Sean 2016), separate analyses are conducted to identify the causal recipes leading to low hate crime and high hate crime. The two truth tables (Tables 4 and 5) follow next.

The truth table computes all possible combinations of conditions that may occur. There are six conditions under study, which leads to (2⁶), that is, 64 rows of possible configurations. The fsQCA program assigns each case to a truth table row based on its membership score and the number of cases for each row is also presented in the truth table. Some rows might have a zero count, indicating none of the cases are explained by these configurations. The minimum acceptable solution number is set to three; configurations with a lower count are excluded from further analysis. The outcome value is determined by the raw consistency of each truth table row. While 0.75 is the minimum acceptable threshold (Ragin 2008), a more stringent threshold of 0.8 is set (Fiss 2011). The truth table rows with at least 0.8 raw consistency are assigned an outcome value of 1 and a value of 0 otherwise. This leads to 55 rows and a total of 1,146 cases in all of the two truth tables.

RESULTS AND DISCUSSION

The Standard Analysis in the fsQCA program is employed to minimize the final truth table, allowing all conditions present or absent to contribute to the outcome. The program employs the Quine–McCluskey algorithm to reduce the truth table rows to more simplified combinations and produce three solutions: complex solution, parsimonious solution, and intermediate solution. Fiss (2011) recommends integrating the parsimonious and intermediate solutions for a better interpretation of the findings. The results are presented in the Ragin and Fiss (2008) format (see Tables 7 and 8), showing the intermediate solutions with parsimonious solutions marked as core conditions. The Overall Solution Coverage ranges from 0.299 to 0.5425 and the Overall Solution Consistency from 0.7538 to 0.7787. They are in line with fsQCA studies on ethical issues. Table 6 summarizes these two metrics on four Journal of Business Ethics studies published in 2019 and 2021. Overall, the results indicate that each outcome can be achieved via multiple pathways, but each path exhibits its unique characteristics. In the next two subsections, the configurations of the causal conditions that are sufficient for the two outcomes shown in Tables 7 and 8 are discussed.

Configurations for low hate crime

Table 7 shows three sufficient configurations that produce low hate crime.

Both the presence and absence of a Binge Drinking Score were found to lead to Low-Hate Crime, depending on how it interacts with other conditions. The presence of Labor Market Freedom and absence of Average Math Scores and per cent of Foreign Born are noted to be necessary conditions for all three solutions leading to low hate crime. In the introduction, the possible impact of Labor Market Freedom on hate crime was discussed. A high degree of freedom in the labor market means a high level of competition in the labor market. In such markets, it is easy to hire and fire workers. It is obvious that ease of hiring means that workers can switch jobs easily. Also, ease of firing means that firms will be able to maintain profitability, which helps workers in the long run. The presence of Labor Market Freedom leads to overall prosperity and reduces the incentive to commit hate crimes. It was also mentioned in the introduction that immigration can lead to an increase in the frequency of hate crimes simply because there are more targets. Therefore, a low proportion of Foreign-Born residents leads to low frequency of hate crimes.

Solution # 2 was found to have the largest raw coverage, indicating that it is the most impactful solution leading to low hate crime. In addition to the three necessary conditions mentioned above, the presence of Median Household Income and the absence of alcoholism contributes to the low hate crime rate in this solution. The former means that relatively richer states have low hate crimes. This is primarily because rich states have more resources, and consequently better law enforcement. This results in lower violent crime in rich states.⁵ Total Education Expenditure is a do not care condition in this configuration.

Overall, the three solutions show that there are multiple pathways that result in low hate crime.

Configurations for high hate crime

Table 8 shows ten sufficient configurations that produce high hate crimes.

Asymmetric causality is evidenced by the asymmetrical solution structures for low hate crime and high hate crime. It is interesting to note that there are more ways to fail than to succeed. Only three pathways lead to low hate crime but ten to high hate crime. High hate crime is found to be associated with the presence of Median Household Income (except in solutions 3 and 8, both of which have low raw coverage). This is likely because of low-quality data from poor states. Indeed, a problem of hate crime data is the possibility of underreporting (Deloughery et al. 2012; Pezzella et al. 2019), and it is reasonable to conjecture that this problem is exacerbated in poor states. Therefore, rich states may appear to have more hate crimes simply because they are more efficient in documenting these crimes. The following table (Table 9) shows the annual average incidents of hate crimes per capita between 2000 and 2021 in different regions of the USA:

According to the above table, the number of hate crimes per capita is lowest in the South, and highest in the Northeast. The Northeast has high average incomes, while the South has low average incomes. Not surprisingly it is found that high hate crime is associated with high median household income. There are strong reasons to believe that hate crimes were significantly underreported in the South compared to other parts of the country. For example, according to the FBI’s own report,⁶ 38 police departments in large cities (population over 100,000) of the South either did not file any report on hate crimes or reported zero hate crimes in 2020. In comparison, there were only 21 such police departments in the rest of the country. Therefore, more research is required in the future to determine the quality of hate crime data from different regions in the country. We therefore suggest that one should be cautious in accepting this result, and that is a limitation of the study.

Solution # 6 has the largest raw coverage, indicating that it is the most impactful solution leading to the high hate crime. In this solution, the factors are the presence of total education expenditure, Median Household Income, % Foreign-Born, and absence of Average Math Score. Labor Market Freedom and Binge Drinking Score do not play a role in this configuration. The complexity of policy analysis, motivating the use of fsQCA is evident in the ten very different pathways to high hate crime.

Policy implications

Our findings broadly support the following:

Education policies have a complicated relationship with hate crimes. Interestingly, the presence of Total Expenditure on Education is associated with low hate crime and high hate crime, that is, states that spend extensively on education could be anywhere on the spectrum. A similar ambiguity can be noted in regard to the Average Math Score. The absence of an Average Math Score is associated with low hate crime as well as high hate crime. Thus while the presence of education expenditure puts states anywhere on the spectrum of hate crime, the absence of policies focused on quality achieves the same outcomes. The impact of education policy on hate crimes does not change monotonically and is truly asymmetric.

The percentage of immigrants has a major impact on hate crimes. While the absence of % Foreign-Born is associated with low hate crime, its presence is associated with high hate crime. This implies that hate crimes are relatively frequent when there is a significant proportion of immigrants. Therefore, the percentage of immigrants has a monotonic impact on hate crimes. This is because immigrants are often targets of hate crimes. Rich states are associated with both high hate crime and low hate crime. Rich states can spend more on policing which explains why they are associated with low hate crime. The association of rich states with high hate crime is more difficult to explain. Our conjecture is that this is an artifact of low-quality data from poor states. The presence of Labor Market Freedom is associated with low hate crime. However, its absence is associated with high hate crime. Thus, improved labor market conditions reduce hate crime. This implies that by developing better labor market regulations, governments not only promote businesses but also help in reducing hate crimes.

Finally, the impact of substance abuse via alcoholism is considered. There is no clear guidance that can be deciphered from a policy standpoint. The efforts by a state to curb alcoholism (or substance abuse generally speaking) are not necessarily associated with high hate crime. Also, its absence is not necessarily associated with low hate crime.

CONCLUSIONS AND FUTURE RESEARCH

In this study, fsQCA is used to identify multiple, nonlinear combinations of the policies leading to the failure and success of community conditions measured by hate crime. The results show that fsQCA with its configurational perspective is an excellent policy evaluation tool; an outcome can be achieved through different pathways, for example, three configurations lead to low hate crime, while ten configurations lead to high hate crime. There is significant variation in the combinations of policies that result in the same outcome. For example, one pathway (solution 2) to low hate crime requires the presence of median household income and the absence of total education expenditure, while the other two pathways require exactly the opposite.

It is found that certain causal relationships can be asymmetric and inconsistent across outcomes. For example, the presence of median household income is associated with both low hate crimes and high hate crimes. The results provide a complementary and substitutive multiple-policy strategy to policymakers in navigating a complex, unstable environment.

Understanding the causes of hate crimes is crucial for law enforcement agencies for several reasons:

• a. Improved Identification and Reporting: Knowledge of the underlying causes helps officers recognize hate crimes more accurately. This leads to better reporting and data collection, which is essential for understanding the scope and nature of these crimes (US Department of Justice 2022). We have pointed out earlier that there are concerns in the literature about the quality of data on hate crimes.

• b. Effective Prevention Strategies: By understanding the motivations behind hate crimes, law enforcement can develop targeted prevention strategies. This might include community outreach programs, education campaigns, and partnerships with local organizations to address the root causes of hate (US Department of Justice 2024).

• c. Policy Development: Insights into the causes of hate crimes can guide policymakers in creating laws and regulations that address these issues more effectively. This can include hate crime legislation, funding for support services, and initiatives to promote social cohesion (National Institute of Justice 2023).

Good community conditions embodied by reduced hate crimes are a complex and challenging decision that requires careful consideration. The study provides a blueprint for policymakers.

Acknowledgement

We have cited all the research articles and other relevant sources that our work builds upon.

Conflict of interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Funding

We do not have any funding to disclose.

Ethics approval and consent to participate

No human subjects or animals were involved in this research.

Consent for publication

Both authors have given their consent to publish this manuscript.

Data availability

All data and materials will be made available upon request.

REFERENCES

Footnotes