Saturday, June 27, 2015

The National Crime Victimization Survey Validates the Uniform Crime Report's Data on Black Crime

In this post I am going to show that both the National Crime Victimization Survey and the Uniform Crime Report report highly similar violent crime rates among blacks. This suggests that the findings of the Uniform crime report on racial crime disparities in general are valid and not explainable by police bias or by reporting bias among police stations.

Two Measures of Crime

The National Crime Victimization Survey (NCVS) is a survey carried out yearly by the Department of Justice in which a random sample of approximately 90,000 households and 160,000 individuals are asked about their experience with crime over the last 6 months. The response rate for this survey is typically 80% or higher (1). Participants are asked if they have been the victim of a violent crime in the last 6 months. If they have then they are asked to answer various questions about the crime and the perpetrator of said crime. These bi-yearly interviews are combined on a yearly basis. The results are then weighted to eliminate bias in the sample based on demographic variables like sex and age and then used to estimate national crime rates. 

The Uniform Crime Report (UCR) is a aggregation of data sent to the FBI every year by police stations all around the country (2). Not all police stations send in this data, but the UCR  manages to get information for police stations which have jurisdiction over 277 million Americans ( aprox 94% of the total population). The data the FBI compiles includes information on the demographics of who is arrested every year. 

Until 2008, both the UCR and the NCVS contained information on the race of criminals in America. (However, both surveys failed to separate Hispanics and Whites). However, when Eric Holder took over the Justice Department he had the NCVS stop releasing this data to the public (3). The UCR still releases data on incarceration rates separated by race and for decades it has been showing that Blacks commit more violent crime than Whites. Some have argued that this data actually reflects bias in who the police arrest, or bias in which police stations report to the FBI, rather than actual differences in criminality between the races. In this post I am going to provide evidence that this is false by showing that the NCVS reports the same racial disparity in crime than the UCR does. I will do this by comparing the results of the NCVS and the UCR on rates at which blacks and whites commited rape, assault, and robbery, the three largest categories of violent crime in both surveys, (there are no victims of homicide to interview) for the years 2000-2008 (4-21). 

Racial Difference in Crime Rates

The first step in doing this is the calculate the number of rapes, assaults, and robberies, committed by blacks and by whites for each year. In the NCVS we can see that tables 40 and 46 give you is the total number of single offender and multiple offender crimes committed each year and the proportion of those crimes that were committed by blacks and by whites. To find the total number of each criminal act committed by each race you must multiply the total number of single offender crimes committed by the proportion that were committed by the race in question and then add that to the total number of multiple offender instances of the same crime that were committed multiplied by the proportion of said acts that were committed by the race in question. The UCR just gives you number of crimes committed by each race in table 43. However, we must make sure and add together "aggravated assault" and "other assault" in order to compare our numbers to the NCVS's assault categories which includes all (non sexual) forms of assault. 

Once we have the number of rapes, assaults, and robberies, committed by each race we can determine how frequently each crime occurred among each race. We do this by diving the total population size of each race during each year, taken from the census (22), by the number of crimes they committed. For instance, in 2008 there were 247221954 White people in America and white people committed 2209699 assaults. This means that there was one assault committed for ever 112 white people. It should be noted that this isn't the same thing as saying that 1 in 112 white people committed an assault because a single white person could have committed multiple assaults and therefore accounted for the 1 assault per 112 white people for several hundred people. (Note: difference in total number of crimes recorded by each survey reflect the fact that the UCR doesn't cover the whole country.)

NCVS

UCR


To figure our the racial disparity between these rates we divide the white rate by the black rate. For instance, the NCVS shows that the White robbery rate of 2008, 1 per 1811 people, divided by the Black rate of one per 167 people is 11. This means that, per capitia, Black people committed 11 times as many assaults as White people in 2008.
NCVS






UCR


We then can measure how different the racial disparities reported by the NCVS and the UCR are by subtracting the NCVS disparity from the UCR disparity. A positive difference will indicate that the UCR over-estimates black crime relative to the NCVS. As can be seen below, most of the differences are actually negative suggesting that the UCR underestimates black crime relative to the NCVS. In general, the two surveys match up very closely. The average differences are -0.47 for rape, .58 for assault, and -2.29 for robbery.




Racial Differences in Proportions

Another way of comparing the two surveys is to see if the NCVS and the UCR both report that blacks commit roughly the same proportion of each crime. In the case of the NCVS we find the proportion of a crime that blacks have committed by diving the total number of crimes committed in a given year by the total number of crimes committed by blacks which, as explained above, we get by combining proportions of single offender and multiple offender crimes on tables 40 and 46. Once again, the UCR just gives us the proportions on table 43. 


Such an analysis shows that the UCR tends to report that blacks make up a somewhat higher proportion of violent criminals than the NCVS. 



However, a closer look at the NCVS numbers reveals that often times the race of the offender is written down as "mixed" or "unknown". I think that many of these mixed and unknown offenders are black and that, as a result, the NCVS under-estimates the proportion of violent crime committed by blacks. We can get around this (and test this hypothesis) by simply subtracting all the crimes committed by people who are neither white nor black from both the UCR and the NCVS and then seeing if blacks make up a similar proportion of the remaining criminals in each survey. As can be seen below, they do. 



This remains true if we aggregate the crime data for 2000-2008 and produce smaller charts that make the degree to which these surveys agree more obvious: 



Conclusion

In conclusion, both the NCVS and the UCR report very similar racial differences in violent crime. Because of this, it is highly unlikely that the UCR's numbers can be explained by police bias. Instead, the most likely explanation for the UCR numbers is that blacks really do commit far more crime than whites. Since police are demonstrably not bias when arresting people for most violent crime it is reasonable to infer, until evidence to the contrary is presented, that they are also not bias in their arrests for crimes like homicide and drug use. 

Sources:

1. Data Collection: National Crime Victimization Survey (NCVS)
2. Appendix IV – The Nation's Two Crime Measures
3. Obama’s Justice Dept. Limits Publication of Interracial Crime Stats
4. 2008 Crime in the United States table 43
5. 2007 Crime in the United States table 43
6. 2006 Crime in the United States table 43
7. 2005 Crime in the United States table 43
8. 2004 Crime in the United States table 43a
9. 2003 Crime in the United States table 43
10. 2002 Crime in the United States table 43
11. 2001 Crime in the United States table 43
12. 2000 Crime in the United States table 43
13. Criminal Victimization in the United States, 2008 Statistical Tables
14. Criminal Victimization in the United States, 2007 Statistical Tables
15. Criminal Victimization in the United States, 2006 Statistical Tables
16. Criminal Victimization in the United States, 2005 Statistical Tables
17. Criminal Victimization in the United States, 2004 Statistical Tables
18. Criminal Victimization in the United States, 2003 Statistical Tables
19. Criminal Victimization in the United States, 2002 Statistical Tables
20. Criminal Victimization in the United States, 2001 Statistical Tables
21. Criminal Victimization in the United States, 2000 Statistical Tables
22. Population Estimates Vintage 2008: National Tables

2 comments: