Racial Bias in Case Processing: Does Victim Race Affect Police Clearance of Violent Crime Incidents?

Abstract

Prior studies have illustrated racial differences in perceptions of police legitimacy. African‐Americans’ views, however, appear to be complex, shaped by perceptions of over‐enforcement of crimes committed by African‐American offenders coupled with under‐enforcement of crimes involving African‐American victims. Using data from the 2002 National Incident‐Based Reporting System, we examine whether victim race (alone, and in combination with offender race) affects police case clearance of four types of violent criminal incidents (homicide, aggravated assault, rape, and robbery) as a potential explanation of African‐Americans’ reduced levels of support for the police. Results suggest that the race of the victim, particularly in combination with the race of the offender, is related to police clearance of violent criminal incidents, but that this relationship is not as strong as those between agency, offense type, and situational characteristics of the incident. Implications for research and policy on police—community relations are discussed.

Keywords:

• African‐Americans
• police
• legitimacy
• NIBRS

Acknowledgments

This research was supported by funding from the American Statistical Association/Bureau of Justice Statistics Small Grant Program and Faculty Mentoring Program, Georgia State University. The authors would like to thank Lynn Addington and John Jarvis for their assistance in clarifying issues associated with the NIBRS data. The authors would also like to thank Finn‐Aage Esbensen, Chris Melde, Brad Brick, the anonymous reviewers, and the JQ Editorial staff for providing suggestions on a previous version of this manuscript.

Notes

1. Portions of this study were presented at the 2007 Annual Meeting of the Academy of Criminal Justice Sciences, Seattle, Washington. Points of view in this document are exclusively those of the authors and do not necessarily represent the official position of the U.S. Department of Justice or the American Statistical Association.

2. For excellent reviews of the findings and limitations related to police clearance of homicide cases, see Wellford & Cronin (1999) and the May, 2007, issue of Homicide Studies.

3. According to the FBI, crimes may be “cleared” through an arrest or through “exceptional means.” Specifically, “[a]n offense is cleared by arrest, or solved for crime reporting purposes, when at least one person is (1) arrested, (2) charged with the commission of the offense, and (3) turned over to the court for prosecution (whether following arrest, court summons, or police notice)” (U.S. Department of Justice, 2004b, p. 79). Additionally, “[g]enerally, an offense can be exceptionally cleared when it falls into one of the following categories … (1) Suicide of the offender …, (2) Double murder …, (3) Deathbed confession …, (4) Offender is killed by police or citizen …, (5) Confession by an offender who is already in law enforcement custody or serving a sentence …, (6) Offender is prosecuted by state or local authorities in another city for a different offense or is prosecuted in another city or state by the federal government for an offense which may be the same …, (7) Extradition denied, (8) Victim refuses to cooperate in the prosecution …, (9) Warrant is outstanding for felon but before being arrested the offender dies …, (10) The handling of a juvenile offender either orally or by written notice to parents in instances involving minor offenses such as petty larceny; no referral is made to juvenile court as a matter of publicly accepted law enforcement policy …” (U.S. Department of Justice, 2004b, p. 80–81). Since crimes cleared through “exceptional means” rarely result from law enforcement intervention, NIBRS researchers (e.g., Litwin, 2003) have typically focused on crimes cleared by arrest. We follow these prior works by focusing on crimes cleared through arrest, but use the terms “crimes cleared through arrest” and “clearance” interchangeably, while acknowledging the technical differences in the methods of clearance.

4. We acknowledge, however, that problems associated with the quality of data collected by law enforcement agencies remains an important concern (O’Brien, 1985) remain. For example, Davies (2007) presents evidence that the FBI‐collected homicide data continue to be shaped by the local political environment (e.g., media attention, mayoral and city council priorities, police resources and practices, and prosecutorial practices and climate). The effects of these factors were muted, however, in Davies’ quantitative analyses. Additionally, Neapolitan’s (2005) examination of the quality of race‐specific data collected by the FBI in terms of reporting cards presented by local agencies to the FBI and information contained in the Supplementary Homicide Reports (SHR) remain inconsistent across agencies. While NIBRS was intended to increase data quality and consistency, Addington’s (2004b, 2006) description of NIBRS data collection procedures and comparisons with the SHR illustrate that despite the increased rigor associated with NIBRS participation, problems associated with missing data (particularly those associated with circumstances and victim–offender relationships) remain. We encourage future researchers to continue exploring different sources of data with different sets of limitations, while also examining data quality issues associated with each source.

5. This was carried out in a series of steps. First, we linked the three batch headers using the ORI (i.e., reporting agency) variable. Next, we added the administrative segment using the ORI and Incident Number. Third, we added the offense segment using the ORI, Incident Number, and UCR Offense Code. Fourth, we added the property segment using the ORI, Incident Number, Type of Property Loss, and Property Description. Fifth, we added the victim segment using the ORI, Incident Number, and Victim Sequence Number. Sixth, we added the offender segment using the ORI, Incident Number, and Offender Sequence Number. Seventh, we added the arrestee segment using the ORI, Incident Number, and Arrestee Sequence Number. We excluded the three “windows” segments‐incident, property, and arrestee‐as information included in these segments generally contain “updates” to the files originally submitted to the FBI (U.S. Department of Justice, 2004a, p. 11). This resulted in a merged data file with a total of 6,436,004 incidents.

6. As police practices affecting case clearance may be different in urban and rural areas (Weisheit, Falcone, & Wells, 1999; Weisheit, Wells, & Falcone, 1994) and have been found to vary across regions (Maguire, 2002; Maguire, Kuhns, Uchida, & Cox, 1997; Maguire & Mastrofski, 2000), we restricted the analyses to cities larger than 100,000 resulting in 1,448,269 observations. The data were also restricted to incidents involving one or more of the following offenses: murder/non‐negligent manslaughter, negligent manslaughter, forcible rape, robbery, and aggravated assault. Each of these offenses is a serious violent crime typically involving face‐to‐face interaction between offenders and victims. As they are often classified as “high profile” (Corsianos, 1993; Puckett & Lundman, 2003), they will typically receive more attention from police (Corsianos, 1993) potentially resulting in a higher likelihood that police will clear the case. These offenses are, however, still heterogeneous enough to provide important variation in seriousness according to both public perception (Wolfgang, Figlio, Tracy, & Singer, 1985) and criminal statutes, allowing us to examine the possibility that factors affecting police case clearance will be conditioned by the offense itself. The offense‐type restriction resulted in a file with 95,046 observations. A criminal incident may have a single offender and a single victim but there also is the possibility of multiple offenders and/or victims. Given that prior research has illustrated the unique nature of crimes involving multiple offenders and multiple victims (Short & Strodtbeck, 1965), we restricted the analytical file to include only those cases involving a single offender and a single victim. A total of 49,639 observations met these restrictions.

7. The restriction to white and African American individuals was due to representation of each race category in the data file. We recognize this is a significant limitation in the current study. We also agree with Martinez (2007) that Latinos’ experiences are notably sparse in the literature and that future studies should be more inclusive in their examinations. We note, however, prior works illustrating the unique historical experiences between African Americans and the police (Bass, 2001a, 2001b; Johnson, 2003; Websdale, 2001) may produce different findings for different racial/ethnic groups within the United States.

8. As noted by one anonymous reviewer, each of these four offenses is a serious felony for which clearance is in agencies’ “best interests.” We agree that clearing serious felonies is, indeed, in the best interest of police departments. We suggest that these four offenses do, however, present differential degrees of “seriousness,” thus presenting leeway in the way offenses are handled. The quandary, however, is that police records of less serious offenses have been found to have questionable reliability and validity (O’Brien, 1985). We would expect that less serious offenses would present more divergence in clearance than the four offenses examined herein. After all, the history of law suggests that many of these least serious offenses (such as vagrancy and loitering) were developed to exert social control over African Americans presence in public spaces (Bass, 2001a, 2001b). Additionally, these offenses provide officers/agencies the greatest discretion in how they are addressed. We restricted our analyses to serious felonies (ranging in degree of seriousness) based on three assumptions: (1) the most serious offenses are the ones with which citizens are generally most concerned; (2) the quality of police data is better for serious offenses as opposed to less serious offenses; and (3) if differences were detected in these most serious offenses (where officers/agencies have the least discretion), the strength of the under‐enforcement and/or over‐enforcement perspectives would have particularly important implications. We acknowledge that our choice of offenses presents a conservative estimate of differences in under‐ or over‐enforcement and encourage future research to examine differences in clearances of less serious incidents, such as simple assault and enforcement of city ordinances.

9. Stranger and unknown relationship types are designated as such in the NIBRS codebook. Cases involving acquaintances in this study include the relationship types in the NIBRS codebook grouped as “Within the family” and “Outside the family but known to victim.”

10. No injury and minor injury are coded as such from the NIBRS codebook. Major injuries in this study consist of apparent broken bones, other major injuries, possible internal injury, loss of teeth, severe laceration, and unconsciousness.

11. Personal weapons, according to the NIBRS codebook, consist of “hands, feet, teeth, etc.” Other weapons include “other weapons” and “blunt objects.” Gun consists of “firearm,” “handgun,” “rifle,” “shotgun”, and “other firearm.”

12. Cramér’s V is a symmetric measure of nominal by nominal association. In 2 × 2 tables, −1 ≤ V ≤ 1 where V is obtained by: (n ₁₁ n ₂₂ – n ₁₂ n ₂₁)/(rm ₁ rm ₂ cm ₁ cm ₂)^½ where “rm” and “cm” denote the row marginal and column marginal, respectively. When the dimensionality exceeds 2 × 2, 0 ≤ V ≤ 1 where V is obtained by: {(χ²/n)/min(I −1, J − 1)}^½ where “I” and “J” denote the number of rows and columns, respectively (see StataCorp 2005, p. 392).

13. The probability value associated with each obtained Pearson Chi‐Square Test of Independence statistic is not reported in this study as the data used in this study were selected in a nonrandom manner. The use of the probability value associated with the Pearson Chi‐Square Test of Independence statistic assumes random sampling (see e.g., Mattson, 1984, pp. 169–170).

14. The offense of murder was excluded on the basis of insufficient data representation and variability. We refer readers to Roberts (2007) for a specific examination of the effects of victim race on homicide case clearance.

15. The estimation sample was constructed on the basis of sufficient data representation by reporting agency (i.e., ORI). Some agencies lacked adequate variability and/or representation. The following agencies, grouped by crime type, comprise the aggregated estimation sample: Aggravated Assault (CO0010100;CO0210100;CO0210100;CT0009300;IA0570100;IA0770300;KS0460500;KS0460600;MA0071800;MA0146000;MI2539800;MI4143600;MI5076500;MI5080600;OH0570200;OH0770100;OHCIP0000;SC0400100;TN0190100;TN0330100;TN0470100;TN0630100;TNMPD0000;TX1880100;TX2270100; UT0180300;VA0990000;VA1030000;VA1110000;VA1160000;VA1170000;VA1200000;VA122000 VA1280000). Rape (CO0010100;CO0210100;CO0300400;CT0009300;IA0570100;MA0071800;MI2539800;MI5080600; OH0570200;OH0770100;OHCIP0000;SC0400100;TN0190100;TN0330100;TN0470100;TN0630100;TN MPD0000;TX1880100;TX2270100;UT0180300;VA1030000;VA1110000;VA1160000;VA1170000;VA1200000;VA1220000;VA1280000). Robbery (CO0010100;CO0210100;CT0009300;IA0570100;IA0770300; KS0460600;MA0071800;MA0146000;MI2539800;MI4143600;MI5080600;OH0570200;OH0770100;OHCIP0000;SC0400100;TN0190100;TN0330100;TN0470100;TN0630100;TNMPD0000;TX1880100;TX2270100;UT0180300;VA0990000;VA1030000;VA1110000;VA1160000;VA1170000;VA1200000;VA1220000;VA1280000).

16. In a departure from previous NIBRS studies, the standard error and the p‐value associated with each parameter estimate are not reported. The omission of the standard errors and p‐values is motivated by the fact that the selection of the cases in the NIBRS sample is nonrandom. That is, it is first not possible to know the probability of NIBRS reporting by the agency. Second, even if the probability of reporting at the agency level were known, this study employs a restricted analytical sample to accommodate exploration of the research question regarding victim race and clearance. Standard errors and p‐values of regression estimates play no role in a nonrandom sample. The emphasis here is the magnitude of the relationship between each regressor and police case clearance. The magnitude of the relationship is determined via the odds ratio and the percentage difference in the predicted probability.

17. The offense of murder was again excluded on the basis of insufficient data representation and variability.

18. Cases where the victim–offender relationship was unknown were approximately 3% less likely to be cleared than cases involving strangers.

19. We do not report victim/offender race interactions in these analyses due to the low number of interracial incidents when disaggregated by offense type.

20. As raised by an anonymous reviewer, the differences in clearance across offense types and jurisdictions may be due to differences in police expertise and/or resources. We agree that this is a plausible explanation worthy of additional study. Since the purpose of the current study is primarily descriptive in its examination of the effect of victim race on police clearance as a potential explanation of African Americans’ relatively unfavorable view of the police, we do not specifically focus on reasons for regional variations or some of the other control variables. We hope that future research tackles this and other related topics.