Racial Context and Crime Reporting: A Test of Black’s Stratification Hypothesis

Abstract

Contextual factors that contribute to race differences in reporting crime to the police are an important element in Donald Black’s theory of the behavior of the law, yet few studies have investigated whether these differences vary depending on social context. The present study investigates whether the relationships between victim and offender race and the reporting of crime are moderated by the level of racial stratification in a given place as Black’s stratification hypothesis would predict. Using victim survey data from 40 metropolitan areas, as well as data from other sources, we find results that are consistent with Black’s stratification hypothesis, namely, that victim and offender race are more strongly associated with the reporting of crime in those metropolitan areas where the gap in economic status between blacks and whites is larger and the groups are more residentially segregated. The theory, however, is unable to account for the high rates of reporting of black-on-black assaults found across the 40 metropolitan areas. The question of how the needs of black victims may outweigh their reluctance to call the police is an important issue for future research.

Methodological Appendix

Study Period (1996–2004)

Data were restricted to this period so that we can use the data to determine how an incident became known to the police (whether the police were called by the victim or a member of the victim’s household, or whether the police discovered the incident in other ways). To do so, the MSA data needed to be linked to the national NCVS files to obtain variables not available in the MSA file. This linking is not permitted for data prior to 1996 because of changes in the Census Bureau’s scramble routines that identify specific incidents.

MSAs in the NCVS MSA Sample

The names of the MSAs are: Anaheim-Santa Ana, CA; Atlanta, GA; Baltimore, MD; Boston, MA-NH; Charlotte-Gastonia-Rock Hill, NC-SC; Chicago, IL; Cincinnati, OH-KY-IN; Cleveland-Lorain-Elyria, OH; Columbus, OH; Dallas, TX; Denver, CO; Detroit, MI; Fort Lauderdale, FL; Fort Worth-Arlington, TX; Houston, TX; Kansas City, MO-KS; Los Angeles-Long Beach, CA; Miami, FL; Minneapolis-St. Paul, MN-WI; Nassau-Suffolk, NY; New York, NY; Newark, NJ; Norfolk-Virginia Beach-Newport News, VA-NC; Oakland, CA; Orlando, FL; Philadelphia, PA-NJ; Phoenix-Mesa, AZ; Pittsburgh, PA; Portland-Vancouver, OR-WA; Riverside-San Bernardino, CA; Sacramento, CA; St. Louis, MO-IL; San Antonio, TX; San Diego, CA; San Francisco, CA; San Jose, CA; Seattle-Bellevue-Everett, WA; Tampa-St. Petersburg-Clearwater, FL; Washington, DC-MD-VA-WV; and West Palm Beach-Boca Raton, FL.

Sample of Assault Victimization (N = 5,191)

The analysis excluded the following incidents:

• (1) Non-reportable incidents. We excluded incidents that were non-reportable (16% of incidents) because the police were at the scene, the offender was a police officer, the reporter was a third party (not part of the victim’s household), or the police learned of the incidents “some other way” (not by victim reporting).

• (2) Incidents with victims who were neither black nor white. Blacks and whites were the victims of the majority of cases and there were insufficient numbers of incidents (4% of incidents) involving victims of other races to support separate analysis.

• (3) Incidents with the race of the offenders unknown (6% of incidents). Comparing known and unknown offender race incidents showed no statistically significant difference in the extent of weapon presence and physical injury. The demographic and socioeconomic characteristics of the victims were also similar. Because the probability of missing offender race is not associated with the probability of calling the police (we found no statistically significant difference in the reporting rates), deleting the incidents with unknown offender race does not bias the regression estimates (Little 1992).

• (4) Incidents occurring in a city or county different than the residence of victims. We imposed this data exclusion rule so that the MSAs in which the respondents lived were the relevant context for the reporting of crime. About 85% of assaults occurred in the same city or county as the residence of victims. In unreported analysis, we also included incidents that did not occur in the same city or county, but occurred within 50 miles from the residence of victims to allow for the possibility that the incidents may still have occurred within the MSA boundary. This analysis covered 94% of assaults and the findings were essentially the same.

Measures of Variables

Victim Race :

1 = black victim; 0 = white victim

Offender Race :

1 = black offender, and 0 = white (nonblack) offender

Approximately 19% of assaults involved multiple offenders, and 85% of multiple-offender incidents were committed by offenders of the same race. Offender race was coded “1” if any of the offenders were reported to be black. In preliminary analysis, we considered an alternative coding for black offender that included multiple-offender incidents in which most of the offenders were black. The results were the same.

About 15% of incidents involved offenders of a race reported to be other than black or white. We found there to be no statistically significant difference in the likelihood of reporting between incidents involving white offenders and offenders of some “other” race, but significant differences between black offenders and offenders of some “other” race. Thus, the two categories of “white” and “other” race were combined and contrasted to incidents involving “black” offenders. Excluding offenders of some “other” race did not change the results.

Percent Black :

Percent population that is non-Hispanic black

Black-to-White Economic Power :

A composite measure of blacks’ economic power relative to that of whites (α = .89)

The estimates for 1995–1999 were derived by linear interpolation using the 1990 and 2000 data, and the 2000 value was used to proxy values in 2001–2003.

Black Absolute Economic Power :

A composite measure of blacks’ absolute economic power (α = .91)

The estimates for 1995–1999 were derived by linear interpolation using the 1990 and 2000 data, and the 2000 value was used to proxy values in 2001–2003.

Black-to-White Voter Turnout :

Ratio of black-to-white voter turnout rate

Using data from the CPS from 1994 to 2002 (US Census Bureau, various years), this variable was calculated as the percentage of voting-age blacks who voted in the previous November election, divided by the percentage of voting-age whites who voted. Because the CPS voting data are only available in even-numbered years, we lagged this variable so that the most recent voting data were used to explain the reporting of crime (e.g., the 2002 voting ratio was used to explain reporting in 2003 and 2004).

Black Elected Officials :

Ratio of % black elected officials to % black voting-age population

This variable was calculated using data from multiple sources. First, the annual number of black elected officials was obtained from the 1995–2002 roster data from the Joint Center for Political and Economic Studies (Joint Center, various years). Second, the number of elected officials in each MSA was estimated using data from the 1992 Census of Governments (US Census Bureau 1995; for similar estimation method, see Stults and Baumer 2007). The 1992 Census is the last Census to include data on popularly elected officials. Although the number of elected officials may have changed in later years, the changes were expected to be small (e.g., from 1992 to 2007, the number of county and city governments had remained relatively constant and increased by less than 1%; in 1992, there was only a 3% increase in the number of city and county elected officials from that reported in 1987). Third, estimates of the voting-age populations were derived from the Census Bureau (US Census Bureau 2008). Because black elected official data are not available after 2002, the information in 2002 was used to assess the reporting of crime in 2003 and 2004.

Residential Racial Integration :

A composite measure of black-white residential integration (α = .98)

This variable was coded by reversing the direction of black separateness scores. Black separateness scores are the sum of standard scores on five segregation indices: the index of dissimilarity (D), the isolation index ( _b P* _b ), the correlation ratio (V), the absolute clustering index (ACL), and the spatial proximity index (SP). Because segregation indices are not available on an annual basis, we constructed the separateness scores for 1995–1999 by using linear interpolation of the data from the 1990 and 2000 data. For incidents that occurred in 2001 and onwards, the 2000 black separateness score was used.

Weapon Presence :

1 = offender had a weapon; 0 = no

Physical Injury :

1 = victim was physically injured; 0 = no

Central City :

1 = victim lived in a central city area of the MSA; 0 = no

Near Home :

1 = incident occurred within a mile from the victim’s home; 0 = no

Public Place :

1 = incident occurred in a public place; 0 = no

Northeast :

1 = incident occurred in the Northeast; 0 = no

Midwest :

1 = incident occurred in the Midwest; 0 = no

South :

1 = incident occurred in the South; 0 = no

West :

1 = incident occurred in the West; 0 = no

Incident Year :

Year of incident (0 = 1996; 1 = 1997; and so on)

Victim Age :

Age of the victim (in years)

Female Victim :

1 = yes; 0 = no

Victim Married :

1 = yes; 0 = no

Victim Income :

Level of victim household income (1 to 14)

Victim Education :

Level of victim education (0 to 18)

Victim Employed :

1 = victim had a job at the time of the incident; 0 = no

Multiple Offenders :

1 = crime was committed by more than one offender; 0 = no

Underage Offender :

1 = any of the offenders were under age 18; 0 = no

Female Offender :

1 = any of the offenders were female; 0 = no

Offender Relative/Acquaintance :

1 = any of the offenders were a relative or acquaintance of the victim; 0 = no

Urbanism :

Sum of standard scores for the natural log of the population size and the natural log of the population density (α = .74)

Socioeconomic Disadvantage :

Index of MSA socioeconomic disadvantage (α = .87)

This variable was calculated by summing the standardized values on median family income (with sign reversed), the poverty rate, the percentage of children under 18 years of age living with a single parent, and the unemployment rate. Because socioeconomic data are not available annually, the index scores for 1995–1999 were estimated with linear interpolation using the 1990 and 2000 census data; the 2000 scores were used to proxy the scores in 2001–2003.

Percent Hispanic :

Percent population that is Hispanic

Racial Diversity :

The multi-race entropy score of the MSA

The entropy score (E) was calculated using the Census Bureau’s annual population estimates (1995–2003) for six racial/ethnic groups (non-Hispanic whites, non-Hispanic blacks, non-Hispanic Asians and Pacific Islanders, non-Hispanic American Indians and Alaska Natives, non-Hispanics of other races, and Hispanics). The calculation formula is $E = \sum\limits_{i = 1}^{6} {(\prod_{i} } )\ln [1/\prod_{i} ]$, where $\prod_{i}$ indicates the proportion of group i population to MSA population (Massey and Denton 1988). Conceptually, diversity and segregation are two related but different concepts. A place, for example, can have a high diversity score if all racial groups are present, but also be very segregated if all groups live in their own neighborhoods and do not interact with one another.

Index Crime Rates :

UCR Part I offenses (less arson) per 100,000 population

To reduce year-to-year fluctuations, we pooled three-years data to calculate this variable (e.g., the average crime rate for 1994, 1995, and 1996 was used for analyses of the reporting of incidents in 1996; the average crime rate for 1995, 1996, and 1997 was used for reporting in 1997, and so on).

Black-to-White Arrest Ratio :

Black arrest rate divided by white arrest rate

Like index crime rates, this variable was calculated by pooling three-years data. The black arrest rate was estimated by total number of UCR-reported arrests for violent crimes for blacks (murders, rapes, robberies, and aggravated assaults) divided by the MSA black population; the same procedure was used for the white arrest rate. Because of missing data, it was necessary to impute the arrest rates in year t from 1996 to 2003 in Washington, DC and the MSAs in Florida by using the corresponding arrest rates in 1995 multiplied by the index crime rate in year t divided by the index crime rate in 1995. We assume that as crime rates change, so do the arrest rates.

Black Representation in Police :

Ratio of % black sworn officers to % black population

Data on police officers were obtained from the 1997, 2000, and 2003 LEMAS (US Department of Justice 2006). The LEMAS is a national survey of police organizations conducted every 3–4 years by the Census Bureau for the Bureau of Justice Statistics. It collects data from all police departments that employ 100 or more sworn officers, plus a nationally representative sample of smaller agencies. To obtain the measure of percent black officers, we aggregated data from agencies within the MSA core counties (for similar use of the LEMAS data, see Eitle et al. 2005; Schnebly 2008). The 1997 LEMAS data were linked to the 1996–2000 NCVS incidents, while data from the 2000 LEMAS were linked to the 2001–2003 incidents, and data from the 2003 LEMAS were linked to the 2004 incidents. Because change in percent black officers from 1997 to 2003 was small for all MSAs in our sample (the average yearly change was about 3%), our findings were not affected by the linking procedure (e.g., the results were the same when the 2000 LEMAS data, not those in 1997, were linked to the 2000 incidents).