I want to offer up some straightforward data on how Americans experience interactions with law enforcement and the degree to which that experience varies by race. My goal here was to try to put my hands on the best data I could find, and to look at those data while trying to impart as little spin as possible. You’ll find a list of the sources I used at the end of this post.
Key Measure for Understanding the Role of Race
The main question of interest is whether outcomes of interactions with law enforcement officials differs by race. We might, for example, want to understand whether African Americans are more likely than white Americans to be arrested. The most simplistic approach would be to compare the number of African Americans who are arrested to the number of white Americans who are arrested.
The problem with this measure is that there are more white Americans than African Americans, so we should expect there to be more arrests of white Americans than African Americans, and that might remain the case even with a bias against African Americans.
To overcome this problem, we can compare the fraction of all of the arrests that are of a specific race to the fraction of the general population that are that race. This ratio – which we will call the Race Representation Ratio – will be 1.0 if the fraction of arrests for that race is the same as the fraction of the total population for that race.
For example, in 2016 the FBI Uniform Crime Report documents a total of 8,421,481 arrests. Of these, 5,858,330 were white and 2,263,112 were African American. That means that about 27% of all arrests were of black people. At that time period, the black population represented about 12.3% of the US population, the black RRR for arrests is 2.2 (= 27% divided by 12.3%).
Using the Race Representation Ratio helps us to understand the degree to which race is playing a role in various encounters and outcomes associated with law enforcement. Specifically:
- If the RRR is less than 1.0 for a race, that race is underrepresented in the outcome being analyzed. In this case, that race is protective.
- If the RRR is equal to 1.0 for a race, that race is perfectly represented in the outcome being analyzed. In this case, that race is irrelevant.
- if the RRR is greater than 1.0 for a race, that race is overrepresnted in the outcome being analyzed. In this case, that race is endangering.
One nifty feature about the RRR is that you can use it to convert the overall chance of an event for the entire population to the chance of that event for a race. If, for example, the overall chance of being arrested in a year is 3%, then the overall chance of a black person being arrested in a year would be approximately 6% (= 3% x 2 RRR).
An important technical note: as we will see, much of the data are from the 2015-2016 time period, and nearly all of it involves residents 16 years or older. Accordingly, I used population figures from that time period. Specifically,
- Population 16 years and older: 253,587,400
- Fraction of that population identified as African American: 12.25%
- Fraction of that population identified as white: 64.99%
Let’s now shift gears and look at the black and white RRRs for various interactions with the police.
The Role of Race in Various Interactions by Police
I’ve calculated the RRR for both African Americans and whites for a spectrum of interactions with police. At one end of this spectrum are interactions that are relatively innocuous (e.g., police officer offers assistance to someone with car trouble); at the other end are killings of civilians by police.
- Resident-initiated contacts (e.g., when civilian reports a possible crime and an officer responds)
- Police-initiated contacts, not including traffic stops, street stops (which include interactions with people in public or a parked car), or arrests — these include officers stopping to assist a civilian with car trouble, serving a summons, etc.
- Traffic stops focused on the driver
- Traffic stops focused on the passenger
- Street stops
- Arrests
- Killing of civilians by police
(NOTE: We will also take a look at the number and fraction of police that die in the line of duty each year.)
The following chart shows the RRRs for both blacks and whites along this spectrum of encounters. (Remember that when the RRR > 1, the race is over-represented in the encounter, i.e., relative to the fraction of the overall population that race represents.)
The single most important takeaway from this chart is that as the severity of the interaction increases (from civilian initiated encounters to being killed by a police officer while the suspect is fleeing), the RRRs for African Americans increase:
- African Americans are under-represented in civilian-initiated encounters and low-intensity police-initiated encounters (e.g., assisting with car trouble)
- African Americans are roughly proportionally represented in traffic stops
- African Americans are over-represented for street stops, arrests, and being killed by police
For example, among the 2,000 people who have been shot and killed by the police while fleeing since 2015, 28% were black and 41% were white. Although more white people are killed fleeing than black people are, African Americans are far over-represented (RRR = 2.3) and whites are far under-represented (RRR = 0.6).
Police Get Killed Too
According to the FBI’s Uniform Crime Report, about 100 police officers die on the job each year. Roughly half (average of 51 from 2015 to 2019) of these deaths are “felonious” (i.e., police officers were killed by suspects); the other half (average of 47 from 2015 to 2019) were accidental deaths (e.g., being struck by a vehicle during a traffic stop).
The number of officers killed “feloniously” in the line of duty should be understood in light of the overall number of officers. According to the National Law Enforcement Memorial Fund, there are “800,000 sworn law enforcement officers now serving in the United States.” Other estimates put the number closer to 650,000 full-time law enforcement officers. If we use the lower estimate of the total number of officers (650,000), the chance that an officer will be feloniously killed in the line of duty is approximately 1 in 13,000 per year.
We can do a similar calculation for African Americans overall. Roughly 240 African Americans are killed by police each year. The number of black Americans 16 years and older is approximately 31 million (according to the Bureau of Justice Statistics). This means that the chance that an African American will be killed by a police officer is approximately 1 in 130,000… about ten times less than the risk born by police officers. (Note that the chance of dying at the hands of a police officer is not evenly spread among all African Americans; we could expect that number to be higher for urban black males. Nonetheless, a similar argument could be made about police — their deaths are not evenly spread among all officers.)
A Deeper Look at Arrests
Not all arrests are the same. The FBI’s Uniform Crime Report provides more detailed information about various types of arrests. At the highest level, arrests are placed into one of three categories:
- Violent crime (murder, non-negligent manslaughter, rape, robbery)
- Part I property crime (burglary, larceny-theft, motor vehicle theft, arson)
- All other crimes
The following chart shows the RRRs for black and white Americans for each of these categories of arrest.
Note that a similar pattern appears within arrest categories as appeared across the larger spectrum of encounters with police officers: the more severe the encounter, the more over-represented are African Americans (RRR > 1) and the more under-represented are white Americans (RRR < 1).
The following chart shows black and white RRRs in detail for various types of arrests. Note that other than arrests for DUI, liquor laws and drunkenness, African Americans are over-represented (RRR > 1) and white Americans are either properly represented (RRR = 1) or under-represented (RRR < 1).
Summary
Here are the main takeaway points:
- In general, the more serious the interaction, the more likely African Americans are to be over-represented. This is true for police killings, arrests and other encounters.
- The general trend toward greater over-representation by African Americans as the encounters become more serious may mirror a greater tendency for situations to escalate when police interact with African Americans than when they interact with white Americans. Specifically, if interactions between police and white Americans are more likely to de-escalate than they are between police and African Americans, a similar pattern would result.
- I haven’t looked deeply at issues of conviction, sentencing, imprisonment and parole by race. This is probably worthy of further analysis.
As the data on which I relied are solely descriptive, it is not possible to assign causality. That said, the over-representation of African Americans across the spectrum of serious encounters with police officers reveals a significant, persistent effect of race on these interactions, one which is clearly unfavorable to African Americans.
Sources
Encounters with Police
I used data from the Bureau of Justice Statistics for the analysis of police encounters and race. Specifically, encounter data came from the following tables: cpp15t01.csv, cpp15t03.csv and cpp15t04.csv.
Arrests
I relied on the FBI’s Uniform Crime Report for data on arrests by type and race (Tables 21A, 21B and 21C).
Civilian Killings by Police
An investigation by the Washington Post in the fall of 2014 (in the wake of Michael Brown’s death) revealed that the FBI was significantly under-counting killings by police. That’s because there is no centralized database of such deaths; the FBI relies on the thousands of law enforcement agencies to voluntarily report officer-involved shootings that lead to the death of civilians.
Starting in 2015, the Washington Post began to assemble its own database of shooting deaths by police officers. The database is quite detailed and includes the race of those killed by police, whether the person killed was fleeing, and other information. Unfortunately, the database only tracks people shot and killed by police; deaths due to other causes (as in George Floyd’s murder) are not included.
Fortunately, another group (Mapping Police Violence) includes deaths of civilians at the hands of police regardless of the method by which the civilian died. Because the overwhelming number of civilian deaths at the hands of police are by gunshot, the results of the analysis using either of these databases are nearly identical. For example, the Black RRR for overall killings by police using the WP database is 1.96; using the MPV database it’s 2.07.
The RRRs for police killings of civilians presented in the analysis above rely on the WP database.