The Difference “Hate” Makes in Clearing Crime An Event History Analysis of Incident Factors

Journal of Contemporary Criminal Justice2014, Vol. 30(3) 268 –289

© 2014 SAGE PublicationsReprints and permissions:

sagepub.com/journalsPermissions.nav DOI: 10.1177/1043986214536663

ccj.sagepub.com

Article

The Difference “Hate” Makes in Clearing Crime: An Event History Analysis of Incident Factors

Christopher J. Lyons1 and Aki Roberts2

AbstractStudying hate crime clearance rates provides an opportunity to uncover the factors that influence police effectiveness for a relatively new legal category—one that was designed ostensibly to protect minorities, and that may pose unique challenges for police reporting, defining, and investigation. Using multiple years (2005-2010) of data from the National Incident-Based Reporting System (NIBRS), we estimate event history models to compare the incident-level predictors and relative probability of arrest for hate and nonbias crimes. As an aggregate category, we find hate crimes are less likely to clear than nonbias crimes. However, the most prototypical hate crimes—White-on-non-White incidents motivated by racial and ethnic bias—are as likely to clear as the most successfully cleared nonbias crimes. Our results suggest that only hate crimes that fit popular constructions of “normal victims and offenders” receive investigative outcomes comparable with otherwise similar nonbias offenses.

Keywordshate crime, clearance, law enforcement

Since the term hate crime first appeared in the legal arena in the late 1970s and early 1980s (Grattet, Jenness, & Curry, 1998), hate or bias crimes1 have come to refer to criminal offenses motivated at least partially by offender prejudice against a victim’s putative group membership. As a legal concept, hate crime distinguishes between

1University of New Mexico, Albuquerque, USA2University of Wisconsin-Milwaukee, USA

Corresponding Author:Christopher J. Lyons, Department of Sociology, University of New Mexico, MSC05 3080, 1 Albuquerque, NM 87131-0001, USA. Email: [email protected]

536663 CCJXXX10.1177/1043986214536663Journal of Contemporary Criminal JusticeLyons and Robertsresearch-article2014

at UNIV OF NEW MEXICO on August 27, 2015ccj.sagepub.comDownloaded from

mailto:[email protected]

http://ccj.sagepub.com/

Lyons and Roberts 269

crime motivated by bias and otherwise similar crime. The majority of U.S. states have enacted some version of hate crime legislation, including laws that explicitly penalize bias-motivated crimes, laws that increase civil liability for hate crime offenders, and laws that require officials to collect hate crime statistics (Grattet et al., 1998; Soule & Earl, 2001). In addition to state law, the U.S. Congress passed the federal Hate Crime Statistics Act in 1990, mandating the Attorney General to collect data on crimes moti-vated by animus (King, 2009; Nolan, Akiyama, & Berhanu, 2002).

At one level, the rapid diffusion of hate crime legislation in the last three decades indicates widespread success of the antiviolence movement. Via lobbying and political consciousness-raising efforts and sponsoring hate crime legislation, groups such as the Anti-Defamation League of B’nai Birth, the Southern Poverty Law Center, and the National Gay and Lesbian Task Force “collectively redefined age-old conduct and constructed new portraits of victimization” (Jenness & Broad, 1997, p. 31). However, official recognition from the state in the form of social policy is not the only goal of social movements. The antiviolence movement also seeks the proper mobilization and enforcement of the law—including the allocation of police effort to investigate and clear reported hate crime incidents by arrest. Modern hate crime policy represents an attempt to engage the legal system in combating criminal acts motivated by prejudice (Grattet et al., 1998; Levin, 1999), yet we are only beginning to explore the effective-ness of police responses to reported hate crimes.

Clearance refers to the investigation of an incident resulting in the arrest of a sus-pect, and as such, is often used as a measure of law enforcement effectiveness (Paré, Felson, & Ouimet, 2007; Riedel, 2008).2 Social scientists recognize that low or declin-ing clearance rates, especially for serious crimes, may have deleterious consequences for law and society more broadly. Poor clearance rates may reduce the deterrent effect of the criminal justice system, public trust in the police, and morale among police officers, and may also traumatize victim’s families or increase fears of victimization (Riedel & Jarvis, 1998). Studying hate crime clearance rates provides an opportunity to uncover the factors that influence police effectiveness in responding to a relatively new legal category—one designed ostensibly to protect minorities, and that, as many have noted, may pose unique challenges for police reporting, defining, and investiga-tion (Berk, Boyd, & Hamner, 1992; Cronin, McDevitt, Farrell, & Nolan, 2007; Jacobs & Potter, 1998; Martin, 1995, 1996). Given the conceptual ambiguities and controver-sies facing this relatively new legal category of crime, clearance rates offer some indi-cation of the institutionalization and “settling” (Phillips & Grattet, 2000) of the concept in law enforcement.

Some work has begun to examine law enforcement responses to hate crimes (e.g., Cronin et al., 2007; Garofalo & Martin, 1993; Grattet & Jenness, 2008; King, 2007, 2008; Martin, 1995, 1996; Wilson & Ruback, 2003), yet researchers have rarely stud-ied the factors leading to hate crime clearance by arrest (see Stacey, 2008, for an exception). Furthermore, researchers have yet to explore the relative likelihood of clearance for both bias and nonbias crimes across multiple jurisdictions and states. Most research on conventional crime conceptualizes clearance as a function of (a) the quality or quantity of evidence produced by a case above and beyond police control



270 Journal of Contemporary Criminal Justice 30(3)

and/or (b) the decision to allocate resources by law enforcement (Litwin, 2004; Roberts & Lyons, 2009). The relative impact of evidentiary versus discretionary factors on clearance rates may differ for hate and non-hate crimes. Specifically, hate crimes may generate different quality or quantity of evidence for police investigation, or may be subject to distinct pressures or motivations for police responses. On one hand, hate crimes may be disadvantaged relative to nonbias crimes due either to additional hur-dles in investigating hate crime incidents or to the devaluation of prototypically low status hate crime victims. On the other hand, police may treat bias crime with more urgency than otherwise similar nonbias offenses, given claims as to the increased seri-ousness of bias crimes to both individual victims and the larger community (Gerstenfeld, 2004; Lyons, 2008; Wisconsin v. Mitchell, 1993).

Despite their centrality for understanding the difference that “hate” makes for law enforcement efficacy, these questions remain largely unexplored. As an initial foray into these questions, we examine the relative clearance processes for bias-motivated crimes and nonbias conventional crimes using data from the 2005 to 2010 National Incident-Based Reporting System (NIBRS), which provides detailed incident-level characteristics of hate and non-hate crimes. We also take fuller advantage of NIBRS data by appropriately modeling time to clearance via event history analysis (Roberts & Lyons, 2009).

Evidence, Discretion, and Crime Clearance

Two perspectives offer contrasting though not necessarily mutually exclusive explana-tions for the processes that lead to successful clearance. The first prioritizes nondiscre-tionary, evidence-based features of incidents over extralegal factors. This view contends that elements outside police control, such as characteristics that influence the availability and quality of investigative evidence leading to an arrest, primarily deter-mine the likelihood of clearance. The relative success of the police in clearing bias versus nonbias crime may depend on the degree to which the characteristics and evi-dence-generating aspects of hate crime events differ from conventional crimes (see Messner, McHugh, & Felson, 2004).

Legal factors thought to determine the nature of evidence for investigation include weapon type, the relationship between victim and offender, and concomitant offenses. Research consistently shows that the use of a firearm versus a contact weapon decreases the odds of clearance because of the lack of physical contact between victim and offender and the diminished potential for DNA or other evidence, whereas increased familiarity between offender and victim improves odds of clearance (Addington, 2006; Alderden & Lavery, 2007; Lee, 2005; Litwin & Xu, 2007; Puckett & Lundman, 2003; Regoeczi, Kennedy, & Silverman, 2000; Riedel & Rinehart, 1996; Roberts, 2007). Indeed, the decline in homicide clearance rates in the United States during the postwar period coincided with increases in the proportions of gun and stranger homicides (Riedel & Jarvis, 1998). Furthermore, lower homicide clearance rates compared with other industrialized countries such as Japan may be due to the greater prevalence of guns in American society (Roberts, 2008). Whereas some




research suggests that homicides committed concurrently with another felony can also lead to lower clearance rates, presumably because such incidents tend to have a ran-dom “hit and run” nature that impedes investigation (Litwin, 2004), more recent research finds that concomitant offenses involving non-lethal incidents may actually lead to more information for police investigation (Roberts & Lyons, 2009). In addition to these characteristics, researchers argue that offense seriousness primarily drives law enforcement responses (Gottfredson & Hindelang, 1979), with more serious offenses—in terms of underlying criminal offense, multiple victims and offenders, and victim injury—receiving more witnesses, better cooperation with officials, and the most effi-cient police responses. More serious offenses may also afford less opportunity for discretion based on extralegal factors (Kalven & Zeisel, 1966).

To the degree that we can adjust for relevant case characteristics associated with the quantity and quality of evidence available to police, the evidentiary perspective implies we should observe few if any differences in the probability of clearance for bias and nonbias crimes. However, insofar as the early stages of investigation differ for bias and nonbias crimes, later clearance probabilities may diverge even after accounting for case characteristics. Police may face unique challenges in identifying and accurately classifying hate crimes compared with non-hate crimes (Nolan et al., 2002), due to the “ambiguity of applying legal definitions to cases, uncertainty regarding bias motiva-tion, and infrequency of reported events to law enforcement” (Cronin et al., 2007, p. 213). Thus, the process by which agencies identify hate crime incidents often requires the collection of additional evidence. In an informative account of reporting practices across eight police agencies, Cronin et al. (2007) note that many agencies conduct follow-up investigation of potential bias crimes—including interviewing victims, wit-nesses, and suspects—to ensure that agencies correctly classify an incident as bias motivated. This additional investigative information required at early stages for clas-sification may later contribute to the efficacy of clearance by arrest. Different report-ing infrastructure and practices may therefore “advantage” incidents classified as hate crimes for successful clearance.

Whereas evidence-based perspectives emphasize the quality and quantity of infor-mation available to police, a second perspective on crime clearance focuses more on police discretion in the allocation of investigative resources. As LaFree (1989) notes, “evidence [leading to an arrest] does not create itself. Its generation requires human labor and interpretation” (p. 105). According to the discretionary view, clearance is partly a product of extralegal factors, such as the gender, age, and race or ethnicity of the victim and offender, that influence the seriousness with which officers, agencies, and witnesses view particular incidents. As hate crimes often involve minority vic-tims, police discretion in allocating effort to clear crimes may operate either to disad-vantage or benefit hate crime relative to conventional crime clearance.

Disadvantaging Hate Crimes

On one hand, victim-devaluation perspectives, derived in part from Black’s (1976) theory of law and social reaction theorists (e.g., Swigert & Farrell, 1977), suggest that




the “quantity of law,” one aspect of which is the allocation of effort and resources to clear a case, varies with the status of parties. According to these perspectives, incidents involving lower status victims will receive less law. Here, perceptions of seriousness on the part of the police and the general public play an important role, either because of overt victim devaluation and diminished investigative effort or because third parties are less likely to come forward to provide quality evidence for incidents involving lower status victims (Cooney, 1994). Research on crime clearance has focused mostly on victim status while neglecting the potential joint effects of victim and offender characteristics. However, as research on other criminal justice outcomes suggests (e.g., Jacobs, Qian, Carmichael, & Kent, 2007), law enforcement’s perceptions of seri-ousness may also be a function of the combination of victim and offender status, including racial and ethnic identity (Black, 1976).3 Numerous scholars assert that crime incidents with White victims and non-White offenders, often referred to as “upward deviance,” attract the most police attention and effort (Jacobs et al., 2007). It follows that these crimes should be the most likely cleared (Roberts & Lyons, 2009). In contrast, incidents between non-White victims and White offenders, termed “down-ward deviance,” are likely viewed as the least serious (Jacobs et al., 2007), and should therefore be the least likely cleared. Following these assertions, racial and ethnic hate crimes that often involve downward deviance (Whites targeting racial or ethnic minor-ities) may be particularly disadvantaged relative to nonbias incidents and other hate crimes.

Advantaging Hate Crimes

Social movement discourse on the relative seriousness of hate crime incidents pro-vides an alternative to devaluation perspectives. Proponents of hate crime legislation argue that hate crimes should be distinguished from otherwise similar nonbias crimes primarily because they pose more of a threat to society (Grattet et al., 1998; Jenness & Broad, 1997). First, hate crimes may cause greater physical or psychological harm to individual victims (Herek, Cogan, & Gillis, 2002). Second, hate crimes may not just target the individual victim but, symbolically, the victim’s entire social group as well, thus singling out whole communities for exclusionary violence with the potential to exacerbate long-standing intergroup tensions and provoke retaliatory unrest among communities (Craig, 1999; Wisconsin v. Mitchell, 1993). Thus, even relatively minor hate crime incidents may disproportionately affect communities by raising levels of fear and hostility (Martin, 1995). Third, the historical legacy of victimization of tradi-tionally marginalized groups may compound the harm of hate crime. This legacy includes not only extralegal discrimination by members of the general public but also victimization by mainstream institutions, including law enforcement (Boeckmann & Turpin-Petrosino, 2002; Comstock, 1991).

Paralleling social movement discourse, research indicates that survey respondents perceive hate crime incidents as more serious and deserving of special attention than otherwise similar nonbias crimes (Lyons, 2006, 2008; Steen & Cohen, 2004). These perceptions may translate into increased sensitivity to hate crimes among witnesses




and law enforcement personnel. Insofar as police perceive hate crimes as more serious or deserving of special attention, they may allocate extra resources to investigate hate crime incidents. This may be true especially for incidents that mirror perceptions of prototypical hate crimes. Research indicates that event prototypes involving “normal” victims and offenders can shape law enforcement responses (Sudnow, 1965). Expectations regarding prototypical victims and offenders also influence perceptions of prejudice and discrimination (Inman & Baron, 1996). Offenses matching typical event schema for hate crimes might be seen as more serious by law enforcement, whereas incidents that do not match event prototypes for hate crime might be pursued with less vigor. For example, White-on-non-White incidents most directly trace popu-lar discursive constructions of racially and ethnically motivated hate crime—that is, the victimization of the weak (non-Whites) by the powerful (Whites; Lyons, 2008). Social-psychological research indicates that asymmetrical discriminatory incidents in which the offender holds status advantage over the victim are viewed as more preju-diced, negative, and serious than the same behaviors when performed by minorities against majorities (Lyons, 2006; Rodin, Price, Bryson, & Sanchez, 1990). In line with popular framing of hate crime legislation, incidents in which the offender holds status advantage over the victim are also more likely to be identified as hate crimes (Lyons, 2008). Thus, in contrast to the general predictions of a devaluation perspective, down-ward deviance for hate crimes—especially for racial and ethnic hate crimes—may in fact receive more attention from law enforcement and clear at greater rates than non-bias crimes.

Data and Method

Using NIBRS data, the current analysis examines differences in clearance by arrest between bias and nonbias violent incidents. We estimate multivariate survival models that control for key incident characteristics that may account for differences in the raw clearance probabilities of bias versus nonbias crimes. In subsequent survival models, we disaggregate hate crimes based on the underlying bias motivation (racial, ethnic, religious, sexual orientation, and disability), and further explore how victim and offender racial combinations in hate crime incidents influence clearance odds relative to nonbias crimes.

We focus on law enforcement efficacy after an incident is reported to or by an agency. Thus, we largely bracket important concerns about the processes related to reporting crime, hate crime or otherwise. However, we caution that an unknown quan-tity of crimes either never comes to the attention of the police, or is not reported by police to the FBI. This may be especially likely for hate crimes, as hate crimes argu-ably face numerous additional impediments to reporting at the individual and agency levels (Cronin et al., 2007). The official hate crime data we use may or may not repre-sent the larger universe of all committed or reported hate crimes. Furthermore, we acknowledge a potential source of bias in our results if incident factors related to either hate crime or conventional crime clearance also affect the likelihood of reporting, defining, or classifying incidents.




Data

The current analysis uses data from NIBRS. NIBRS is the largest scale incident-level data set that includes a variety of offense types collected by the FBI. For each crime incident in participating police agencies, NIBRS provides crucial information for the current analysis: bias motivation based on race, ethnicity and national origin, religion, sexual orientation, or disability, and whether an arrest was made within the arrest-reporting period (ending December 31 of the following year), and, if so, the number of days between incident and arrest. NIBRS also reports on situational characteristics such as weapon type and victim–offender relationship, and victim and offender demo-graphics. Because NIBRS records information on all offenses, victims, and offenders associated with the incident, we can determine the presence or absence of concomitant crimes and code for multiple victims and offenders.

Due to the relative infrequency of bias crimes, we pool 2005 to 2010 NIBRS data to examine bias and nonbias violent crime incidents that were cleared or not cleared by arrest. We first select incidents involving homicide, kidnapping, forcible sexual offense, aggravated assault, simple assault, or intimidation. For multiple offense incidents, we use the most serious offense associated with the incident when select-ing violent incidents. We then classify incidents according to whether at least one of the offenses associated with the incident was coded by the agency as motivated by bias. This results in 6,871 incidents classified as bias motivated. Because of the very large number of nonbias violent incidents (5,505,562), we randomly select a smaller sample of nonbias incidents equal to 5 times the number of bias incidents (34,355) via PROC SURVEYSELECT in SAS. The final data therefore include 41,226 vio-lent incidents (6,871 bias and 34,355 nonbias incidents) across 36 states and the District of Columbia.4

Hate crime incidents vary by type of motivating bias, including race, national ori-gin, religion, ethnicity, sexual orientation, and disability. Wilson and Ruback (2003) found that type of bias crime determined police responses to hate crime incidents across Pennsylvania counties. For instance, anti-White incidents were more likely to be cleared than anti-Black incidents, suggesting that police may invest more effort in hate crimes targeting majority members (Wilson & Ruback, 2003). We thus use explore how type of bias and the race and ethnicity of victims and offenders influence the likelihood of clearance relative to nonbias crimes.

Event History Analysis

We use event history rather than logistic models for both statistical and substantive reasons. Statistically, event history models provide more accurate estimates than logis-tic regression approaches by incorporating variation in time to clearance and utilizing more of the potential information in the data (Regoeczi, Jarvis, & Riedel, 2008). Our data indicate substantial variation in time to clearance: Seventy percent of cleared inci-dents cleared within the same day of the incident, but more than half of the remaining cleared incidents took a week or more. Event history analyses explicitly incorporate




“censoring” of uncleared incidents, recognizing the possibility that a crime incident could still clear after the end of the arrest-reporting period.5 By considering both time to clearance (for cleared incidents) and censoring (for uncleared incidents), the event history approach improves the accuracy of the estimates of independent variables’ effects on clearance (Allison, 1995). Substantively, longer clearance times have poten-tially negative consequences, such as greater opportunity for an at-large offender to commit more crimes and increased fear of crime and negative perceptions of police performance. For these reasons, researchers increasingly use information on time to clearance when investigating the likelihood of arrest (Addington, 2007; Lee, 2005; Regoeczi et al., 2008; Roberts, 2007; Roberts & Lyons, 2009, 2011).

Our relatively long study window allows us to view the days to clearance or censor-ing as a continuous measure. We thus use a continuous event history approach and analyze the data via the Cox model, which combines the proportional hazards model with partial likelihood estimation methods (Allison, 1995). Previous clearance studies have used the Cox model (e.g., Regoeczi et al., 2008; Roberts, 2007; Roberts & Lyons, 2009, 2011). Unlike other event history methods that assume a particular probability distribution for event times, the Cox approach is very flexible and does not specify precisely how the hazard rate—in effect, the instantaneous probability of an event occurring—depends on how much time has passed since the incident (Allison, 1995). Formally, the Cox model describes the relationship between the covariates and the hazard rate.6

Parameter estimates from SAS PROC PHREG use the information in the observed cleared incidents and time to censoring for uncleared incidents. Note that in interpret-ing these estimates, the hazard rate is inversely related to time to clearance: High hazard rates imply a relatively short expected time to clearance, whereas low hazard rates indicate a relatively longer expected time to clearance. The hazard ratio exp(βi) reports the estimated difference in the hazard resulting from a difference of 1 unit in X (or comparing a variable category to the reference), and is therefore similar to the odds ratio in logistic regression. Tied event times occur when two or more incidents clear in the same number of days. The observed number of days to clearance is a grouping of continuous, but unknown, times to clearance, so if the full precision were observed, there would be a precise ordering of these seemingly tied event times. As the ordering of event times is used to obtain parameter estimates, some procedure for addressing the tied events times is needed. Allison (1995) discusses various approaches for this; here, we use the EXACT method in PROC PHREG.

Dependent Variable

Time to clearance or censoring represents our dependent variable. We measure time to clearance as the number of days between incident and arrest; for uncleared (censored) incidents, time to censoring is the number of days from the incident to the end of the arrest-reporting period. As mentioned earlier, NIBRS reports arrest information during the NIBRS year and the following year, so the maximum value for the dependent vari-able is 730 days.




Independent Variables

The main independent variable for the current study indicates whether the incident is motivated by bias. Depending on the models, we use a broad categorization (bias vs. nonbias), more detailed categorizations (racial, ethnic or national origin, religious, sexual orientation, or disability vs. nonbias), or bias motivation combined with vic-tim’s race or with victim–offender racial combinations. We control for situational fac-tors of crime incidents and demographic characteristics of victims and offenders that may influence the likelihood of clearance by arrest by affecting either the quality of evidence available to police or the quantity of law allocated to particular cases. Situational characteristics include weapon type, victim–offender relationship, con-comitant offense, and seriousness of the incident. We categorize type of weapon into firearm, non-firearm weapon, and no weapon. When both firearm and non-firearm weapons are present for a single incident, we prioritize the presence of a contact weapon given the potential for additional evidence. We classify the victim–offender relationship as family, friend/acquaintance, stranger, and unknown.7 For cases with multiple victims and/or offenders, we use the closest victim–offender relationship. Concomitant offense is a dichotomous indicator of the presence or absence of a con-comitant offense. Seriousness of an incident is measured by its legal offense category (homicide, kidnapping, forcible sexual offense, robbery, aggravated assault, simple assault, and intimidation), presence or absence of multiple victims and/or offenders, and victim injury (categorized as no injury, one or more minor injuries, and one or more major injuries/death).

Victim demographic characteristics include gender, age, and race. For gender, we use three categories due to the possibility of multiple victims: female, which may include multiple females; male, which may include multiple males; and multiple gen-ders, indicating at least one female and male. Due to the increased attention that young victims may receive, we compare child victims less than 13 years old with non-child victims. When multiple victims have different ages, we code the incident as involving a child victim if at least one victim is a child. Victim’s race is dichotomous: White or non-White. In the case of multiple victims with different races, we categorize the inci-dent as involving a non-White victim if at least one victim is non-White. Analyses also include offender’s gender and race using the same categorization as for victims’ demo-graphic characteristics. As we pooled multiple years of NIBRS, we also include dummy variables for years.

Multiple Imputation

A total of 18.4% of the incidents we examine have missing information on at least one of the independent variables. In NIBRS, missing data may reflect information that is either unknown to police or information that was known to investigators, but not recorded officially. As with much official crime data, missing information is almost surely not missing completely at random. In our data, hate crimes contain more miss-ing offender characteristics than nonbias incidents as they more often involve




offenders unknown to the victims. Furthermore, we find more missing data among uncleared than among cleared incidents, especially for offender characteristics. Due to this kind of systematic patterning, dropping incidents with missing data may distort analytic results (Riedel & Regoeczi, 2004). We therefore use multiple imputation tech-niques to address missing values for independent variables. Multiple imputation uses the observed data to impute missing values, creating multiple imputed data sets so that variability in the imputed values across these data sets can indicate uncertainty in the imputation process (Allison, 2001). With 10 such data sets created in IVEWARE soft-ware (Raghunathan, Lepkowski, van Hoewyk, & Solenberger, 2001), PROC PHREG estimates for each data set are combined in PROC MIANALYZE. The final parameter estimates are averages across the imputed data sets, and estimated standard errors include variability across the imputed sets as well as the average of the standard errors across the data sets.

Results

Table 1 presents summary statistics for all bias incidents recorded in NIBRS from 2005 to 2010, along with a random sample of nonbias incidents during the same time period (including imputed data). We begin by comparing the raw clearance rate for hate crimes represented as one aggregate category with the clearance rate for the ran-dom sample of nonbias crimes. Table 1 reveals that as an aggregate category, bias crimes are less likely than nonbias crimes to result in arrest. Whereas 41.4% of non-bias crimes lead to an arrest, only 31.6% of hate crimes are cleared—a 9.8% difference in the raw probability of clearance. For cleared incidents, average time to clearance was slightly longer for bias incidents (10.37 days) than for nonbias incidents (9.53 days).

The descriptive statistics in Table 1, however, also reveal that bias crimes and con-ventional crimes differ on a number of incident-level characteristics related to the odds of clearance. Compared with nonbias crimes, the underlying criminal offense in bias crimes is much more likely to involve intimidation and less likely to involve simple assault. Bias crime incidents are also less likely to implicate the use of a weapon, espe-cially a firearm, and more likely to include male and non-White victims, result in no injury to the victim, and to be perpetrated by offenders who are White and unknown to the victim. Previous research suggests these factors contribute to clearance likeli-hoods, raising the possibility implied by evidentiary perspectives that the raw differ-ences in clearance between bias crimes and conventional crimes may be due to underlying differences in the incidents themselves.

Table 2 shows the multivariate results investigating the difference in clearance haz-ard rates between bias and nonbias incidents. Model 1 in Table 2 examines the differ-ence between bias and nonbias incidents while controlling only for year. Consistent with the descriptive results, the estimated parameter for bias is statistically significant and negative in Model 1. The clearance hazard rate for bias incidents is 26.6% less than that for nonbias crimes incidents. In addition to year, Model 2 of Table 2 controls for other incident-level characteristics that, under the evidentiary and discretionary




Table 1. Descriptive Statistics, NIBRS 2005-2010.

Bias incidents (n = 6,871) Nonbias incidents (n = 34,355)

Frequency % Frequency %

Cleared 2,170 31.6 14,231 41.4Bias motivation Racial 4,065 59.2 Ethnic 938 13.7 Religious 512 7.5 Sexual orientation 1,245 18.1 Disability 111 1.6 Weapon type Firearm 251 3.7 2,197 6.4 Other weapon 3,494 50.9 23,801 69.3 No weapon 3,126 45.5 8,357 24.3Victim–offender relationship Family 316 4.6 8,467 24.6 Friend/acquaintance 2,908 42.3 17,265 50.3 Stranger 1,944 28.3 3,709 10.8 Unknown 1,703 24.8 4,914 14.3Concomitant offense 735 10.7 2,335 9.7Legal offense category Homicide 14 0.2 106 0.3 Kidnapping 22 0.3 411 1.2 Forcible sexual offenses 61 0.9 1,785 5.2 Robbery 268 3.9 2,440 7.1 Aggravated assault 1,196 17.4 4,864 14.2 Simple assault 2,513 36.6 18,842 54.8 Intimidation 2,797 40.7 5,907 17.2Multiple victims 1,475 21.5 5,475 15.9Multiple offenders 1,562 22.7 4,899 14.3Victim injury No injury 4,493 65.4 19,132 55.7 Minor injury 1,929 28.1 13,259 38.6 Major injury/death 449 6.5 1,964 5.7Victim gender Male 4,218 61.4 13,025 37.9 Female 2,005 29.2 18,410 53.6 Female and male

victims648 9.4 2,920 8.5

Victim age Child 422 6.1 2,374 6.9 Non-child 6,449 93.9 31,981 93.1

(continued)




Bias incidents (n = 6,871) Nonbias incidents (n = 34,355)

Frequency % Frequency %

Victim race White 3,776 55 22,407 65.2 Non-White 3,095 45 11,948 34.8Offender gender Male 5,534 80.5 25,620 74.6 Female 1,025 14.9 7,120 20.7 Female and male

victims312 4.5 1,615 4.7

Offender race White 4,679 68.1 19,602 57.1 Non-White 2,192 31.9 14,753 42.9

Note. NIBRS = National Incident-Based Reporting System.

Table 1. (continued)

Table 2. Event History Models of Bias and Nonbias Crime Clearance, NIBRS 2005-2010.

Model 1 Model 2 Model 3

B (SE) Hazard ratio B (SE) Hazard ratio B (SE) Hazard ratio

Bias incident1 −0.309 (.023)** 0.734 −0.100 (.026)** 0.905 Bias motivation2

Racial −0.005 (.031) 0.995 Ethnic −0.010 (.057) 0.990 Religious −0.294 (.093)* 0.745 Sexual orientation −0.356 (.056)** 0.700 Disability −0.407 (.177)* 0.666Weapon type3

Firearm −0.166 (.055)** 0.847 −0.164 (.055)* 0.849 Other weapon 0.184 (.034)** 1.202 0.184 (.034)** 1.202Victim–offender relationship4

Family 0.320 (.046)** 2.077 0.327 (.046)** 2.088 Friend/acquaintance 0.075 (.032)* 1.627 0.082 (.031)* 1.634 Unknown −0.411 (.102)* 0.663 −0.409 (.101)* 0.664Concomitant offense 0.339 (.030)** 1.404 0.340 (.030)** 1.405Legal offense category5

Homicide 1.107 (.133)** 3.025 1.113 (.133)** 3.043 Kidnapping 0.505 (.080)** 1.657 0.508 (.080)** 1.662 Forcible sexual

offense0.022 (.062) 1.022 0.031 (.062) 1.031

Robbery 0.143 (.067)* 1.154 0.144 (.067)* 1.155 Aggravated assault 0.727 (.050)** 2.069 0.729 (.050)** 2.073 Simple assault 0.611 (.045)** 1.842 0.616 (.045)** 1.852Multiple victims 0.221 (.039)** 1.247 0.222 (.039)** 1.249Multiple offenders −0.007 (.112) 0.993 −0.007 (.112) 0.993

(continued)




perspectives, may influence clearance. Although the coefficient for bias remains sta-tistically significant and negative in Model 2, adjusting for key incident factors sub-stantially reduces the estimated difference in the clearance hazard rates of bias and nonbias crimes. In Model 2, we estimate hazard rates to be 9.5% less for bias than for nonbias incidents. Comparing Models 1 and 2, incident characteristics appear to account for about 64% (1 − [9.5/26.6]) of the bias/nonbias difference in hazard rates estimated in Model 1. Figure 1 illustrates the separate clearance functions for hate and nonbiased crimes, holding all other covariates at modal values. As implied by the lower clearance hazard rate for hate crimes, the survival (non-clearance) curve for hate crimes is above that for nonbias crimes. The difference between the curves at a par-ticular time t indicates the difference in the estimated probability of remaining uncleared at t for hate crimes relative to nonbias incidents.

As “bias” represents a broad aggregate category, Models 1 and 2 mask potentially important heterogeneity in types of bias crime. In Model 3, we disaggregate bias crimes into five general categories based on the classification of offender bias. Model 3 suggests that not all bias crimes are treated equally in relation to nonbias crimes. Specifically, only incidents classified in terms of religious, sexual orientation, and dis-ability bias are less likely to clear than nonbias crimes. In contrast, ethnically moti-vated incidents and the largest category of reported bias crimes—racially motivated incidents—are, on average, as likely to clear as nonbias crimes.

We next explore how victim and offender race or ethnicity influences the clearance hazard rates for bias crimes. In Table 3, we focus on racially and ethnically motivated crimes for which the racial and ethnic status of victims and offenders is most perti-nent.8 In Model 1, we find that racial and ethnic bias crimes targeting non-White

Model 1 Model 2 Model 3

B (SE) Hazard ratio B (SE) Hazard ratio B (SE) Hazard ratio

Victim injury6

Minor 0.132 (.018)** 1.141 0.134 (.018)** 1.143 Major/death 0.071 (.042)† 1.074 0.074 (.042)† 1.077Victim gender7

Female 0.034 (.021) 1.035 0.031 (.021) 1.031 Female and male 0.184 (.116) 1.202 0.178 (.116) 1.195Child victim −0.228 (.035)** 0.796 −0.235 (.035)** 0.791White victim 0.051 (.023)* 1.052 0.070 (.023)* 1.073Offender gender8

Female −0.087 (.047)† 0.917 −0.089 (.047)† 0.915 Female and male −0.429 (.513) 0.651 −0.429 (.512) 0.651White offender 0.225 (.022)** 1.252 0.214 (.022)** 1.239

Note. Models include year dummies; n = 41,226; missing values imputed; standard errors are given in parenthesis. NIBRS = National Incident-Based Reporting System. Referents: 1 = nonbias crimes; 2 = nonbias crimes; 3 = no weapon; 4 = stranger; 5 = intimidation; 6 = no injury; 7 = male victim(s); 8 = male offender(s).†p < .1. *p < .05. **p < .001.

Table 2. (continued)




victims are as likely to clear as the most successfully cleared nonbias incidents—that is, nonbias crimes against White victims. In contrast, racial and ethnic bias crimes involving less prototypical White victims have about a 10% lower clearance hazard rate than nonbias crimes against Whites. Whereas bias crimes targeting non-Whites are more likely to clear than bias crimes involving White victims, Model 1 indicates the opposite pattern for nonbias crimes. For nonbias crimes, incidents involving non-White victims have about an 11% lower clearance hazard rate relative to incidents involving White victims.

We take the analysis a step further in Model 2 by considering victim–offender race or ethnic combinations. White-on-non-White incidents reflect the most common images of racial and ethnic bias crimes yet are also as instances of “downward” devi-ance that may receive less law. In contrast to the general predictions of a devaluation perspective, we find that White-on-non-White bias incidents are among the most likely victim–offender combinations to clear. The hazard clearance rate for White-on-non-White racial or ethnic bias crimes is indistinguishable from that of White-on-White nonbias crimes. However, all other (less prototypical) victim–offender combinations for racial and ethnic bias crime are among the least likely incidents to clear, with hazard clearance rates ranging from about 21% to 12% lower than the White-on-White nonbias referent.

In additional models not shown, we find that the racial and ethnic status of victims and offenders is less consequential for non-racial or ethnic hate crimes motivated by religious, sexual orientation, or disability bias. These incidents are less likely to clear than nonbias crimes when involving either White or non-White victims and none of the four victim–offender race or ethnic combinations result in similar clearance odds for these incidents relative to nonbias crimes.

Figure 1. Survival functions for bias and nonbias crimes, adjusting for incident factors.




In terms of control variables, the results reported in Models 2 and 3 generally are consistent with expectations and prior research. Briefly, homicides and aggravated

Table 3. Event History Models of Racial/Ethnic Bias by Victim and Offender Race, NIBRS 2005-2010.

Model 1 Model 2

B (SE) Hazard ratio B (SE) Hazard ratio

Victim race by racial/ethnic bias1

Non-White victim/bias −0.009 (.037) 0.992 White victim/bias −0.107 (.040)* 0.899 Non-White victim/nonbias −0.120 (.026)** 0.887 Racial dyads by bias2

White-on-non-White/bias 0.021 (.043) 1.021 Non-White-on-non-White/bias −0.242 (.091)* 0.785 Non-White-on-White/bias −0.214 (.069)* 0.808 White-on-White/bias −0.137 (.058)* 0.872 White-on-non-White/nonbias 0.018 (.048) 1.018 Non-White-on-non-White/nonbias −0.310 (.020)** 0.733 Non-White-on-White/nonbias −0.128 (.031)** 0.880Weapon type3

Firearm −0.154 (.056)* 0.857 −0.152 (.056)* 0.859 Other weapon 0.185 (.034)** 1.203 0.184 (.034)** 1.202Victim–offender relationship4

Family 0.323 (.046)** 1.381 0.338 (.045)** 1.403 Friend/acquaintance 0.077 (.036)* 1.080 0.086 (.032)* 1.090 Unknown 0.408 (.098)* 1.504 −0.402 (.096)* 0.669Concomitant offense 0.338 (.030)** 1.403 0.337 (.030)** 1.401Legal offense category5

Homicide 1.084 (.138)** 2.956 1.086 (.139)** 2.963 Kidnapping 0.499 (.081)** 1.648 0.498 (.081)** 1.645 Forcible sexual offense 0.039 (.062) 1.040 0.038 (.062) 1.039 Robbery 0.135 (.067)* 1.145 0.128 (.067)† 1.137 Aggravated assault 0.730 (.049)** 2.075 0.729 (.049)** 2.073 Simple assault 0.622 (.044)** 1.863 0.622 (.044)** 1.863Multiple victims 0.224 (.041)** 1.251 0.226 (.041)** 1.254Multiple offenders −0.004 (.114) 0.996 −0.004 (.115) 0.996Victim injury6

Minor 0.138 (.018)** 1.148 0.139 (.018)** 1.149 Major/death 0.073 (.043)† 1.075 0.073 (.043)† 1.076Victim gender7

Female 0.034 (.021) 1.035 0.036 (.021)† 1.037 Female and male 0.179 (.116) 1.196 0.176 (.117) 1.193Child victim −0.237 (.035)** 0.789 −0.240 (.035)** 0.786Offender gender8

Female −0.084 (.045)† 0.919 0.084 (.045)† 1.088 Female and male −0.420 (.500) 0.657 −0.333 (.462) 0.717White offender 0.181 (.024)** 1.198

Note. Models include year dummies; n = 41,226; missing values imputed; standard errors are given in parenthesis. NIBRS = National Incident-Based Reporting System. Referents: 1 = White victim/nonbias; 2 = White-on-White/nonbias; 3 = no weapon; 4 = stranger; 5 = intimidation; 6 = no injury; 7 = male victim(s); 8 = male offender(s).†p < .1. *p < .05. **p < .001.




assaults have higher clearance hazards rates than simple assaults, the presence of a contact weapon and injury improves the clearance hazard rates as does knowing the offender, and incidents involving concomitant offenses or multiple victims are more likely to clear than single-offense or single victim incidents. We note that level of injury correlates strongly with type of crime in our models. However, the results for Tables 2 and 3 are quite similar if we exclude either injury level or crime type; we leave both in our models given the differences between bias and nonbias crimes noted in Table 1.

Conclusion

Using multiple years of data from the NIBRS, we set out to investigate the difference “hate” makes when it comes to law enforcement efficacy in clearing incidents by arrest. Specifically, we compared the relative probability of arrest for hate crime and nonbias crimes. Bias crime legislation distinguishes criminal incidents motivated by offender bias from otherwise similar crimes, yet we know little about the relative effi-cacy of police investigation for these crimes versus conventional, nonbias crimes. Given the well-noted conceptual ambiguities associated with the recent legal con-struct, police may face additional hurdles in investigating hate crime incidents. Furthermore, the prototypical hate crimes involve lower status victims who may be devalued by police and the general public. Victim devaluation and investigational challenges may produce lower clearance rates for bias crimes relative to nonbias crimes. Alternatively, insofar as perceptions of hate crimes mirror social movement discourse as to the increased seriousness of bias crimes to both individual victims and the larger community (Gerstenfeld, 2004; Lyons, 2008; Wisconsin v. Mitchell, 1993), the police may treat bias crime with more urgency.

Event history analyses reveal that hate crimes, as an aggregate category, are less likely to clear than nonbias crimes. Although underlying differences in case character-istics associated with the quality and quantity of evidence account for much of the difference in hazard clearance rates, even net of incident controls, bias crimes are about 9.5% less likely to clear. These general patterns indicate that hate crimes are somehow disadvantaged in crime clearance. Yet, additional analyses reveal that clear-ance hazard rates differ by hate crime type. In particular, only non-race and non-ethnic hate crimes are less likely to clear. We find additional nuances when examining the race or ethnicity of victims and offenders. Although we uncover no conditions under which hate crimes are more likely to clear than the most successfully cleared nonbias crimes, we find no differences in the clearance hazard rates between these nonbias crimes and the most prototypical hate crimes. Specifically, only White-on-non-White incidents motivated by racial and ethnic bias are equally likely to clear as White-on-White nonbias crimes.

These later results suggest that police and witnesses may perceive some hate crimes as more serious and worthy of investigative effort than others. For hate crimes that fit popu-lar constructions for victims and offenders—in particular, racially or ethnically motivated incidents of downward deviance where Whites target non-Whites—perceptions are most




likely to align with social movement discourse as to the distinctive seriousness of crimes motivated by bias. Less prototypical hate crime incidents, however, may receive less attention, either because police and witnesses devalue these incidents or because these incidents generate less quality evidence. Given the extra hurdles involved in hate crime investigation (Cronin et al., 2007), police may be motivated to focus on “normal” inci-dents that they or the larger community deem most worthy of investigational effort. The role of witness and perhaps even victim cooperation is also key to the generation of evi-dence (Cooney, 1994). Victims and witnesses may be more likely to cooperate with the police or hold authorities accountable for investigation when communities see incidents as serious and worthy of attention.

We find that the combination of victim and offender race or ethnic status matters only for racially and ethnically motivated hate crimes. Given the history of racial con-flict and violence in the United States, these crimes are central foci of the antiviolence movement (Levin, 1999). However, we caution that NIBRS data include no informa-tion on relevant statuses for religious, sexual orientation, or disability hate crimes. We therefore are unable to assess whether event prototypes also influence clearance out-comes for these other types of hate crime. More detailed information on the religious identities of victims and offenders, for example, would allow researchers to consider how various victim–offender combinations influence clearance rates for religiously motivated hate crimes.

Our study represents a preliminary foray into hate crime clearance research, yet much more work is needed to more completely understand the efficacy of law enforce-ment responses to this new category of crime. Perhaps, most importantly, research should attempt to clarify the mechanisms leading to clearance processes for hate crimes and conventional crimes. We assume that perceptions of seriousness on the part of law enforcement officials and witnesses and additional investigative challenges rep-resent key processes distinguishing clearance outcomes for hate crimes and conven-tional crimes, yet NIBRS data do not permit us to measure these mechanisms directly. Although we are able to control for an array of important case characteristics, we acknowledge that other unobserved variables may help explain the net relationships we report here. Future research should investigate mechanisms further, using data that provide more detailed information. For example, data from police reports or inter-views may allow researchers to include the level of witness cooperation in their analy-ses, although it is difficult to collect such information for as many jurisdictions as are present in NIBRS data. We also call for additional detailed qualitative research along the lines of Cronin et al. (2007) to elucidate the processes and unique hurdles involved in investigating hate crimes.

NIBRS represents one of the few data sources that allow researchers to examine hate crime clearance across multiple agencies; nonetheless, NIBRS coverage is not nationally representative. Only a few more than half of the states currently participate in NIBRS, and not all agencies in participating states report data. In general, large police departments are less likely to adopt NIBRS than smaller agencies (Addington, 2006; Maxfield, 1999). Although NIBRS participation is increasing, we cannot be certain that the patterns we report generalize to all, and especially larger, agencies.




Finally, as a first step our research focuses only on incident-level determinants of clearance by arrest. Future research should use a multilevel framework to explore the role of agency context as well. Grattet and Jenness (2008) note that the relatively decen-tralized system of American policing “grants the agency significant freedom to develop and implement (or not) its own approach to crime control based on its assessment of the nature of its specific community problems and its organizational commitments to addressing such problems” (p. 506). As the growing research on conventional crime clearance demonstrates, both incident and contextual factors determine the probability of arrest (Eitle, Stolzenberg, & D’Alessio, 2005; Litwin, 2004; Roberts, 2007). Similarly, existing research on police responses to hate crime indicates important variation “in how policing of hate crime unfolds across types of police personnel, police units, jurisdic-tions, and polities” (Jenness, 2007, p. 154). Future research could explore, for example, how agency-level policies and characteristics, including the presence of a dedicated hate crimes unit and community policing initiatives, as well as the social, political, and demo-graphic features of cities, influence hate crime clearance. Although researchers confront nontrivial methodological challenges in combining NIBRS data with administrative data on police agency characteristics due to restricted sampling (Addington, 2006), this line of research would contribute to our understanding of the complex processes leading to the production of law enforcement efficacy for crimes of bias.

Acknowledgment

We thank John Roberts, the editors, and anonymous reviewers for helpful comments.

Declaration of Conflicting Interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) received no financial support for the research, authorship, and/or publication of this article.

Notes

1. We use the terms “hate” and “bias” interchangeably throughout.2. Although relatively rare, crime incidents may also clear by “exceptional means” when

arrest of a suspect is not possible due to reasons beyond police control, such as death of the offender, prosecution declined, extradition denied, or victim’s refusal to cooperate with investigation.

3. We do not claim to test fully Black’s (1976) assertions about social status and the quan-tity of law. Black refers to status as a multidimensional construct, and although he views racial identity (as well as gender and age) as a rough proxy for socio-economic status, data simply cannot appreciate the complexity of status along numerous social dimensions. Nonetheless, we follow previous research (e.g., Jacobs, Qian, Carmichael, & Kent, 2007; Roberts & Lyons, 2009) that examines the consequences of victim and offender racial combinations for criminal justice outcomes.




4. The 36 states include Alabama, Arizona, Arkansas, Colorado, Connecticut, Delaware, Georgia, Idaho, Illinois, Iowa, Kansas, Kentucky, Louisiana, Maine, Massachusetts, Michigan, Mississippi, Missouri, Montana, Nebraska, New Hampshire, North Dakota, Ohio, Oklahoma, Oregon, Rhode Island, South Carolina, South Dakota, Tennessee, Texas, Utah, Vermont, Virginia, Washington, West Virginia, and Wisconsin.

5. The arrest-reporting period for National Incident-Based Reporting System (NIBRS) data covers the NIBRS year and the following year. Thus, for example, in 2005 NIBRS data the end of the arrest-reporting period was December 31, 2006. Two years was the maximum time between incident date and possible clearance date.

6. Writing the clearance hazard rate for incident i at time t as h (t), with covariates Xik and an

unspecified time dependence indicated by α(t), the model is log ( ) ( ) .h t t Xi ik kk

K

= +=∑α β1

7. We adopted the NIBRS categorization of victim–offender relationship. “Family” includes the victim as child, in-law, stepparent, stepchild, stepsibling, and other family member of the offender. “Friend and acquaintance” includes offender as friend, acquaintance, neigh-bor, babysitter, boyfriend, girlfriend, child of boyfriend/girlfriend, ex-spouse, employee, employer, and otherwise known to the victim.

8. For the analysis examining the effect of bias combined with victim race on clearance, we necessarily omit the separate victim race variable. Similarly, in the analysis with bias motivation combined with victim–offender racial combination, we exclude the victim and offender race variables.

References

Addington, L. A. (2006). Using National Incident-Based Reporting System murder data to eval-uate clearance predictors. Homicide Studies, 10, 140-152.

Addington, L. A. (2007). Hot vs. cold cases: Examining time to clearance for homicide using NIBRS data. Justice Research and Policy, 9, 87-112.

Alderden, M. A., & Lavery, T. (2007). Predicting homicide clearances in Chicago: Investigating disparities in predictors across different types of homicide. Homicide Studies, 11, 115-132.

Allison, P. D. (1995). Survival analysis using the SAS system: A practical guide. Cary, NC: SAS Institute Inc.

Allison, P. D. (2001). Missing Data. Thousand Oaks, CA: Sage Publications.Berk, R., Boyd, E., & Hamner, K. (1992). Thinking more clearly about hate-motivated crimes.

In G. Herek & K. Berill (Eds.), Hate crimes: Confronting violence against lesbians and gay men (pp. 123-143). Newbury Park, CA: Sage.

Black, D. J. (1976). The behavior of law. New York, NY: Academic Press.Boeckmann, R., & Turpin-Petrosino, C. (2002). Understanding the harm of hate crime. Journal

of Social Issues, 58, 207-225.Comstock, G. D. (1991). Violence against lesbians and gay men. New York, NY: Columbia

University Press.Cooney, M. (1994). Evidence as partisanship. Law & Society Review, 28, 833-858.Craig, K. M. (1999). Retaliation, fear, or rage: An investigation of African American and White

reactions to racist hate crimes. Journal of Interpersonal Violence, 14, 138-151.Cronin, S. W., McDevitt, J., Farrell, A., & Nolan, J. (2007). Bias-crime reporting: Organizational

responses to ambiguity, uncertainty, and infrequency in eight police departments. The American Behavioral Scientist, 51, 213-231.




Eitle, D., Stolzenberg, L., & D’Alessio, S. (2005). Police organizational factors, the racial com-position of the police, and the probability of arrest. Justice Quarterly, 22, 30-57.

Garofalo, J., & Martin, S. (1993). Bias motivated crimes: Their characteristics and the law enforcement response. Carbondale: Center for the Study of Crime, Delinquency and Correction, Southern Illinois University.

Gerstenfeld, P. B. (2004). Hate crimes: Causes, controls, and controversies. Thousand Oaks, CA: Sage.

Gottfredson, M., & Hindelang, M. (1979). A study of behavior of law. American Sociological Review, 44, 3-18.

Grattet, R., & Jenness, V. (2008). Transforming symbolic law into organizational action: Hate crime policy and law enforcement practice. Social Forces, 87, 1-28.

Grattet, R., Jenness, V., & Curry, T. (1998). The homogenization and differentiation of hate crime law in the United States, 1978 to 1995: Innovation and diffusion in the criminaliza-tion of bigotry. American Sociological Review, 63, 286-307.

Herek, G., Cogan, J., & Gillis, R. (2002). Victim experiences in hate crime based on sexual orientation. Journal of Social Issues, 58, 319-339.

Inman, M., & Baron, R. (1996). Influence of prototypes on perceptions of prejudice. Journal of Personality and Social Psychology, 70, 727-739.

Jacobs, D., Qian, Z., Carmichael, J., & Kent, S. (2007). Who survives on death row? An indi-vidual and contextual analysis. American Sociological Review, 72, 610-632.

Jacobs, J., & Potter, K. (1998). Hate crimes: Criminal law and identity politics. New York, NY: Cambridge University Press.

Jenness, V. (2007). The emergence, content, and institutionalization of hate crime law: How a diverse policy community produced a modern legal fact. Annual Review of Law and Social Science, 3, 141-160.

Jenness, V., & Broad, K. (1997). Hate crimes: New social movements and the politics of vio-lence. New York, NY: Gruyter.

Kalven, H., Jr., & Zeisel, H. (1966). The American jury. Boston, MA: Little Brown.King, R. (2007). The context of minority group threat: Race, institutions, and complying with

hate crime law. Law & Society Review, 41, 189-224.King, R. (2008). Conservatism, institutionalism, and the social control of intergroup conflict.

American Journal of Sociology, 113, 1351-1393.King, R. (2009). Hate crimes: Perspectives on offending and the law. In M. Krohn, A. Lizotte,

& G. Penly Hall (Eds.), Handbook on crime and deviance (pp. 525-549). New York, NY: Springer.

LaFree, G. (1989). Rape and criminal justice: The social construction of sexual assault. Belmont, CA: Wadsworth.

Lee, C. (2005). The value of life in death: Multiple regression and event history analyses of homicide clearance in Los Angeles county. Journal of Criminal Justice, 33, 527-534.

Levin, B. (1999). Hate crimes: Worse by definition. Journal of Contemporary Criminal Justice, 15, 1-21.

Litwin, K. (2004). A multilevel multivariate analysis of factors affecting homicide clearance. Journal of Research in Crime & Delinquency, 41, 327-351.

Litwin, K., & Xu, Y. (2007). The dynamic nature of homicide clearances: A multilevel model comparison of three time periods. Homicide Studies, 11, 94-114.

Lyons, C. (2006). Stigma or sympathy? Attributions of fault to hate crime victims and offend-ers. Social Psychology Quarterly, 69, 39-59.




Lyons, C. (2008). Individual perceptions and the social construction of hate crime. The Social Science Journal, 45, 107-131.

Martin, S. (1995). A cross-burning is not just an arson: Police social construction of hate crimes in Baltimore County. Criminology, 33, 303-326.

Martin, S. (1996). Investigating hate crimes: Case characteristics and law enforcement responses. Justice Quarterly, 13, 455-480.

Maxfield, M. (1999). The National Incident-Based Reporting System: Research and policy applications. Journal of Quantitative Criminology, 15, 119-149.

Messner, S., McHugh, S., & Felson, R. (2004). Distinctive characteristics of assaults motivated by bias. Criminology, 42, 582-618.

Nolan, J., Akiyama, Y., & Berhanu, S. (2002). The Hate Crime Statistics Act of 1990. American Behavioral Scientist, 46, 136-153.

Paré, P., Felson, R., & Ouimet, M. (2007). Community variation in crime clearance: A mul-tilevel analysis with comments on assessing police performance. Journal of Quantitative Criminology, 23, 243-258.

Phillips, S., & Grattet, R. (2000). Judicial Rhetoric, meaning making, and the institutionaliza-tion of hate crime. Law & Society Review, 34, 567-606.

Puckett, J., & Lundman, R. (2003). Factors affecting homicide clearances: Multivariate analysis of a more complete conceptual framework. Journal of Research in Crime & Delinquency, 40, 171-193.

Raghunathan, T. E., Lepkowski, J. M., van Hoewyk, J., & Solenberger, P. (2001). A Multivariate Technique for Multiply Impuring Missing Values Using a Series of Regression Models. Survey Methodology, 27, 85-96.

Regoeczi, W., Jarvis, J., & Riedel, M. (2008). Clearing murders: Is it about time? Journal of Research in Crime & Delinquency, 45, 142-162.

Regoeczi, W., Kennedy, L., & Silverman, R. (2000). Uncleared homicides: A Canada/United States comparison. Homicide Studies, 4, 135-161.

Riedel, M. (2008). Homicide arrest clearances: A review of the literature. Sociology Compass, 2, 1145-1164.

Riedel, M., & Regoeczi, W. (2004). Missing Data in Homicide Research. Homicide Studies, 8, 3, 163-192.

Riedel, M., & Jarvis, J. (1998). The decline of arrest clearances for criminal homicide: Causes, correlates, and third parties. Criminal Justice Policy Review, 9, 279-306.

Riedel, M., & Rinehart, T. (1996). Murder clearances and missing data. Journal of Criminal Justice, 19, 83-102.

Roberts, A. (2007). Predictors of homicide clearance by arrest: An event history analysis of NIBRS incidents. Homicide Studies, 11, 82-93.

Roberts, A. (2008). Explaining differences in homicide clearance rates between Japan and the United States. Homicide Studies, 12, 136-145.

Roberts, A., & Lyons, C. (2009). Victim-offender racial dyads and clearance of lethal and non-lethal assault. Journal of Research in Crime & Delinquency, 46, 301-326.

Roberts, A., & Lyons, C. (2011). Hispanic victims and homicide clearance by arrest. Homicide Studies, 15, 48-73.

Rodin, M., Price, J., Bryson, J., & Sanchez, F. (1990). Asymmetry in prejudice attribution. Journal of Experimental Social Psychology, 26, 481-504.

Soule, S., & Earl, J. (2001). The enactment of state level hate crime laws in the United States: Intrastate and interstate factors. Sociological Perspectives, 44, 281-305.




Stacey, M. (2008). Likelihood of arrest in bias crime incidents: The effects of bias units. Paper presented at the annual meeting of the Academy of Criminal Justice Sciences, Cincinnati, OH.

Steen, S., & Cohen, M. (2004). Assessing the public’s demand for hate crime penalties. Justice Quarterly, 21, 91-124.

Sudnow, D. (1965). Normal crimes: Sociological features of the penal code in a public defender office. Social Problems, 12, 255-276.

Swigert, V., & Farrell, R. (1977). Normal homicides and the law. American Sociological Review, 42, 16-32.

Wilson, M., & Ruback, R. (2003). Hate crime in Pennsylvania, 1984-99: Case characteristics and police responses. Justice Quarterly, 20, 373-396.

Wisconsin v. Mitchell, 508 U.S. 476. (1993).

Author Biographies

Christopher J. Lyons is an Associate Professor of Sociology at the University of New Mexico. His research explores the structural foundations of community and neighborhood violence, often from a multilevel framework. Recent articles have examined the political context of the immigration and crime relationship and neighborhood investment patterns and violence.

Aki Roberts in an Assistant Professor in the Department of Sociolgy at the University of Wisconsin-Milwaukee. Her interests include quantitative methods, crime clearance, National Incident-Based Reporting System (NIBRS) data, Japanese crime, policing, police networks, and motor vehicle theft.



The Difference “Hate” Makes in Clearing Crime An Event History Analysis of Incident Factors

Documents