Institutional racism in arrest rates

In the United States, the relationship between race and crime has been a topic of public controversy and scholarly debate for more than a century. The crime rate varies between racial groups. Most homicides in the United States are non interracial, meaning the perpetrator and victim are of the same race.

Race and crime in the United States- from Wikipedia

Crime data sources
The Uniform Crime Reports represent the primary source of data used in the calculation of official statistics regarding serious crimes such as murder and homicide, which is supplemented by the information provided through the NCVS and self-report studies, the latter being the best indicator of actual crime rates for minor offenses such as illegal substance abuse and petty theft. These crime data collection programs provide most of the statistical information utilized by criminologists and sociologists in their analysis of crime and the extent of its relationship to race.

These programs are the best data we have on actual crime in the Unites States.

Uniform Crime Reports (UCR)
the Uniform Crime Reports (UCR) program is a summary-based reporting system that collects data on crime reported to local and state law enforcement agencies across the US. This report includes: murder and non-negligent homicide; non-lethal violent crimes comprising robbery, forcible rape and aggravated assault. 94% of the total population is under jurisdiction. The data the FBI compiles includes information on the demographics of who is arrested every year.

Limitations of the UCR include


 * The UCR only relies upon local law enforcement agency crime reports, the UCR program can only measure crime known to police and cannot provide an accurate representation of actual crime rates.
 * The UCR tracks crime for the racial category of "White" to include both Hispanic and non-Hispanic ethnicities.

National Crime Victimization Survey (NCVS)
The National Crime Victimization Survey (NCVS), is a national, representative, bi-annual survey done by the Department of Justice. It covers the frequency of crime victimization and the characteristics and consequences of victimization. The primary purpose behind the NCVS program is to gather information on crimes that were not reported to police, though information is also collected on reported crimes. The survey collects data on rape, assault, robbery among other things. 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 race of the offender of said crime. Interracial crime reporting discontinued during the Obama administration after 2008. These bi-yearly interviews are combined on a yearly basis, weighted to eliminate bias in the sample based on demographic variables like sex and age and then used to estimate national crime rates. There is no apparent incentive to lie on these surveys.

Limitations to the NCVS program include


 * The NCVS relies upon the recollection of the individuals surveyed, the NCVS cannot distinguish between true and fabricated claims of victimization, nor can it verify the truth of the severity of the reported incidents. (for example, if the victim is too traumatized to report)

National Incident-Based Reporting System (NIBRS)
The National Incident-Based Reporting System (NIBRS) is an incident-based reporting system used by law enforcement agencies in the United States for collecting and reporting data on crimes. Local, state and federal agencies generate NIBRS data from their records management systems. Among the many data tables relating to crime, it including the following tables: “Offenders, Race by Offense Category, 2018” and “Arrestees, Race by Arrest Offense Type, 2018”. From these two tables, data on police reports of crimes and arrests of suspects segregated by race and crime category can be found.

Limitations to the NIBRS Program may include


 * The program makes no distinction of Hispanics and whites
 * The program covers twenty percent of the United States

In order to see if blacks are over or underrepresented in arrest rates given where they should be, we can analyze the disparity between offender and arrest rates.

The US population in 2018 US Census Bureau[1][a]. The table’s title is “Detailed Race” and the TableID is C02003. This data does not differentiate between Hispanic whites and non-Hispanic whites (or Hispanics of any other race). Normally this is a problem, but FBI UCR data also does not distinguish the two groups, so it ends up being beneficial for facilitating the analysis.

Every year, the FBI publishes the Uniform Crime Reporting (UCR) program data[2][a], and according to their website, the data comes “from more than 18,000 city, university and college, county, state, tribal, and federal law enforcement agencies voluntarily participating in the program.” This data covers every police agency in the US. Within the UCR is the National Incident-Based Reporting System (NIBRS), which “captures details on each single crime incident—as well as on separate offenses within the same incident—including information on victims, known offenders, relationships between victims and offenders, arrestees, and property involved in the crimes.” The NIBRS contains many data tables relating to crime[3][a], including the following tables: “Offenders, Race by Offense Category, 2018” and “Arrestees, Race by Arrest Offense Type, 2018”. From these two tables, data on police reports of crimes and arrests of suspects segregated by race (with no distinction of Hispanic status) and crime category can be found.

I used R 3.6.1 with the help of the tidyverse package to produce the following graphs.

Results – All Crime
The simplest analysis to run first is to aggregate all crime and just look at offenses and arrests by race. The next two graphs take the racial proportions of all crime and divide them by the racial proportions of the US population for offenses and arrests respectively. A ratio above 1 (the red line) means that racial group is over represented among crime offenses/arrests given their population size.

Looking at the three biggest racial groups (whites, blacks, asians), the results are not surprising. Blacks are massively over represented among crime offenses and arrests, whites are slightly under represented, and asians are extremely under represented.

If there is racial bias against a certain group in police arrests, then one would expect the ratio of arrests to offenses to be above 1, meaning that they are over represented among those arrested given their reported offense rate.

The opposite of what is expected is found. Blacks are less likely to be arrested than expected given their rate of crime offenses reported. On the contrary, whites and asians are slightly over represented, although it is probably not a significant difference from a ratio of 1. American Indians/Alaskan Natives are definitely over represented.

By aggregating all crime, it does not seem like blacks are more likely to be arrested for crime given that they have been reported of it by the police.

Results – Desegregating by Crime
More insight can be made by splitting the data by the type of crime. Given that the majority of racial tensions right now are between blacks and whites and that they constitute the two biggest racial groups perpetrating crime, the following section only focuses on those two races.

The following two graphs show the same data as the first two graphs in this article but separated by category of crime.

Perhaps an easier way to display this information is just simply as a ratio between the black and white rate of crime per category.

Some of these numbers are certainly shocking. Blacks are 11 times more likely than whites to be reported for a robbery (given their US population). Similarly, blacks are 7 times more likely than whites to be arrested for homicide and robbery.

Similarly to the graph #3, taking a ratio of arrests to offenses would give an indication of racial bias in the arresting process of law enforcement. For entertainment purposes, the full name of the following graph would be: bar graph of the ratio between the black-white ratio of population-adjusted crime arrests and the black-white ratio of population-adjusted crime offenses separated by category of crime and color-filled by the total number of arrests of blacks for each category of crime. Whew.

The red band in the middle represents an arbitrary 0.95-1.05 range of insignificance (I don’t currently know how to calculate error bars for odds ratios). Given this, blacks experience significant racial discrimination in arrest rates for Drug/Narcotic Offenses, Embezzlement, Extorion/Blackmail, Gambling Offenses, and Stolen Property Offenses. On the contrary, whites experience significant racial discrimination in arrest rates for Animal Cruelty, Arson, Assault Offenses, Burglary/Breaking & Entering, Destruction/Damage/Vandalism, Fraud Offenses, Human Trafficking Offenses, Larceny/Theft Offenses, Motor Vehicle Theft, Pornography/Obscene Material, Prostitution Offenses, Robbery, and Sex Offenses.

The color of each bar represents the total amount of arrests of blacks for each crime category. This shows that the three biggest causes of arrests for blacks are Assault Offenses, Drug/Narcotic Offenses, and Larceny/Theft Offenses. Only drug offenses seem to be racially biased against blacks, while assault and theft seem favorable for blacks.

On the other side of the spectrum, Gambling Offenses and Extortion/Blackmail are the 5th and 3rd least common causes of arrests for blacks respectively (out of 23 crime categories).

Conclusion
Given this data, there is no evidence of anti-black bias in arrest rates given police report rates. If anything, there is a pro-black bias. This conclusion holds with the majority of crime categories.

However, there are many obvious limitations with this analysis. The most obvious is that if racial bias in policing exists, then the police would also over report crime offenses for blacks given their actual crime rate. This is a fair critique, although it is unusual that there would be a noticeable pro-black bias when considering arrest rates given the prior offense report rates. Arguing for the existence of anti-black bias in policing should include an explanation for this discrepancy, since the logical continuation of that argument would probably predict that police would either be anti-black or neutral in arresting.

In the future, data from the National Crime Victimization Survey (NCVS)[4][a] could be used to provide a different perspective. The BJS reports data “from a nationally representative sample of about 240,000 interviews on criminal victimization, involving 160,000 unique persons in about 95,000 households. Persons are interviewed on the frequency, characteristics, and consequences of criminal victimization in the United States.” The NCVS provides extrapolations of crime in the US from a sample of victims rather than law enforcement. This could eliminate any potential racial bias in police reporting of crime.

However, certain problems with the data make it hard to combine NCVS data with the FBI data. For one, the data may be unreliable, although the direction is uncertain. Second, the data does segregate Hispanics from the other races, which actually complicates things since the FBI data doesn’t. To work with the two sources of data, a lot of assumptions about demographics and crime rate would have to be made to split the Hispanic data between the other races. I did some off-the-cuff data transformation and found that a lot of Hispanics were not accounted for between whites, blacks, and asians, leading to weird results. I might tackle this again in the future but I’m not too confident. Third, the data does not separate by category of crime and only includes violent incidents, which restricts the depth of analysis that can be done. Additionally, definitions of violent incidents with the FBI’s definitions would have to be matched.

In the future, I might incorporate the racial demographics of the prison population separated by crime to compare incarceration rates to offense and arrest rates. However, the data might run into the same problems as the NCVS data as listed above.

I previously wrote an article exploring racial bias in police brutality from a global context. You can find that article here. Comment any suggestions or critiques of my articles.

Analysis of actual violent crime rates compared to arrest rates
If the National Crime Victimization Survey (NCVS) is the best data we have on actual crime and the Uniform Crime Reform (UCR) is the best data we have on reported crime, we can compare the racial breakdown of offenders to the racial breakdown of the prosecuted regarding rape, robbery, and assault to see what the real crime rate of these racial groups are. Additionally, we can see if the police arrest racial groups at disproportionate rates. The page, "Racial disparities in crime" calculates the racial breakdown of rapes, assaults, and robberies, committed by blacks and whites in 2008 using “NCVS 2008”.

Analysis of arrest rates
D’alessio and Stolzenberg 2003reviews 335,619 incidents of forcible rape, robbery, and assault across 17 states during 1999 found that the odds of arrest for white offenders is approximately 22% higher for robbery, 13% higher for aggravated assault, and 9% higher for simple assault than they are for black offenders. An offenders race plays no noteworthy role in the likelihood of arrest for the crime of forcible rape. These findings suggest that the disproportionately high arrest rate for black citizens is most likely attributable to differential involvement in reported crime rather than to racially biased law enforcement practices (Race and the Probability of Arrest, Florida International University, 2003)

Beaver et al 2013 is an analysis that reveals that African American males are significantly more likely to be arrested and incarcerated when compared to White males. This racial disparity, however, was completely accounted for after including co-variates for self-reported lifetime violence and proxies for courtroom behavior. Implications of this study are discussed and avenues for future research are offered.(National Longitudinal Study of Adolescent Health)