This issue is a flashpoint today in America, and these shootings are frequently reported with benchmarks demonstrating disparity. But which benchmarks are the most valid? And are they the best -- or only -- way to characterize the problem? A new study takes a different approach.
Stop and frisk, was a controversial police program in New York where “suspicious” individuals were stopped without any other probable cause and questioned by police. Very heated arguments centered around what was “suspicious” and the role bias played in that judgment. But how do you calculate racial disparity in these situations? A new study looks at this question in an evenly more emotionally charged concern, fatal officer-involved shootings. (FOIS).
In calculating a disparity, you first establish a benchmark and then make a comparison. The reference most easily at hand are populations aggregates, comparing the percentage of civilians of a particular race killed to that races overall contribution to the population. In the researcher’s introduction, they point to Baltimore, with a Black community of 12% and FOIS of 26%; there is a disparity. But is that a valid benchmark? Not all citizens have equivalent exposure to police, especially in situations involving violence and firearms. Using benchmarks based on racial “contributions” to violent crime, the calculated disparity may disappear or reverse. It becomes a problem of the most pertinent denominator. There are lots of benchmarks to choose from, and their advocates can be passionate.
“What factors predict the race of a person fatally shot by police?”
To avoid the controversy inherent in the choice of benchmarks and gain new insight, the researchers utilized regressions, trying to identify what factors were most predictive of the race of civilians killed in police confrontations. They began by constructing their own database based on FBI Uniform Crime Reports, augmented with local reporting of events and Freedom of Information Act requests to law-enforcement agencies. Their study reported on the analysis of 917 instances of FOIS in 2015 . Here are the basics.
- 56% of cases involved a single officer, 39% involved 2 to 4 officers.
- Officers were overwhelmingly male at 96%
- Officers averaged ten years of experience on the job.
- Compared to the racial percentages in law enforcement officers, more Caucasian officers, fewer African-American officers, and equivalent Hispanic officers were involved in FOIS.
Using regressions of the victim’s race against all the other officer and civilian predictors they found
- “…in the typical shooting, we did not find evidence of anti-Black or anti-Hispanic disparity.” Using averages calculated across the entire data set the regression showed that Black civilians were 6.67 times less likely to be shot, Hispanic civilians 3.33 times less likely to be killed than Caucasians civilians.
- “…the race of the person fatally shot closely followed race-specific homicide rates.” Roughly 44% of the variance in the race of the civilians mirrored the percentage of county-level , race-specific violent crime. 43% of the variance could be attributed to the racial composition in the county-level population.
- The number of officers involved, their race, gender, and experience did not seem to be a factor.
- Some characteristics made civilians more likely to be killed. For Black civilians younger age, being armed, mental health issues, suicidal intent or attacking the officers were “risk factors;” for Hispanic civilians, only mental health seemed to be a factor.
Where does that leave us?
This is a “wicked” problem. I wrote about this study because it goes against a conventional media narrative, and frankly, I was interested in how many outlets will pick up and report their findings. FOIS are a problem readily distorted by a media heuristic of “if it bleeds, it leads.” But the belief that FOIS clearly reflect disparities of police behavior does not seem supported by this study.
Bias can be either a noun or a verb. Disparity speaks to bias as a noun and is a necessary but insufficient measure of bias understood as a verb. There are certainly examples of specific officers and groups of officers that demonstrate bias’s verb form, for example, the latest revelations regarding some police groups within Facebook. This study did not address intent; it is unlikely that any study can. As with everything that we humans touch, it is more complex and nuanced.
I cannot imagine being in this situation, so I find it difficult to put myself into the shoes of the police or the civilians. As a physician who has dealt with life and death and seen an innocent moment spiral away into an unmitigated disaster, I recognize the complexity. FOIS leave no one undamaged, and we need to do all we can to reduce the occurrence. Simplistic slogans may heighten awareness, but solutions based on simplistic assumptions will result in little change or unintended consequences – I guess that is why it is a “wicked” problem.
 42 cases were removed that did not involve shootings or had incomplete data.
 Knowing the location of the FOIS, enabled researchers to use racial demographics at the county level, a finer measure than state or national demographics.
Source: Office characteristics and racial disparities in fatal officer-involved shootings PNAS DOI:10.1073/pnas.1903856116