Police Killings, Gender and Race, Part 2

by Mike Kimel

Police Killings, Gender and Race, Part 2

A few weeks ago, I had a post looking at police killings of civilians. Using statistics on police shootings, homicides, and demographics, I reached the controversial conclusion that:

contrary to popular perception, the data seems to show that Black people are less likely to be met with lethal force than non-Black people would be for behavior that genuinely constitutes a threat toward police personnel. At the same time, police are, on average, harsher toward Black people than toward non-Black people because the police are more likely expect a threat from Black people than from non-Black people. (Again, due to the high murder rate in general, and high rate of killing police.) Put plainly: Black people are subject to more low-level police attention (i.e., traffic stops, searches), but less high-level police attention (i.e., actions that reduce violent crime).

When the numbers point one toward a conclusion that neither fits the popular perception nor the popular mood, it is natural to take another look and see if one can find more information that either disproves one’s conclusions or supports them. In that spirit, I stumbled on this paper by Min-Seok Pang and Paul A. Pavlou.

Their paper states:

To theorize the role of technology use, we draw upon signal detection theory (Green and Swets 1988, Wicknes 2002, MacMillan 2002, Correll et al. 2002, 2014) to propose a simple, stylized model for a police officer’s decision to pull the trigger. We model that when deploying deadly force, the officer takes two factors into consideration – (i) a risk that a suspect poses an imminent, life-threatening danger to bystanders and/or the officer herself and (ii) a perceived risk that she would be held accountable for the death. Based on this model, we derive how technology use influences fatal shootings by police officers. First, technology use for intelligence analyses and access help reduce the ambiguity in the degree of violence of the suspect perceived by an officer. Second, the use of evidence gathering technologies, such as wearable body cameras, is likely to help the officer justify her shooting, making her less reluctant to deploy fatal force.

Their assumption (i) reads a lot like this from my earlier post:


As noted above, we expect that if police are rational, they will react to perceived threats upon their own lives. If one group contains individuals who are more of a threat to police (and other civilians) than other groups, then that group will be met with lethal force by police more frequently. If a group is less of a threat to police (and other civilians), police will be less likely to use lethal force on that group. In the end, use of lethal force by the police against any given group will approximate the threat posed by that group.

However, I did not explicitly state Pang & Pavlou’s assumption (ii) in my post, which weakened my conclusion.

Here’s the key point from Pang & Pavlou:

Our empirical analysis produced several interesting findings. First, we found that in police departments that conduct statistical analyses of digitized crime data, there are 2.15% fewer fatal shootings, substantiating our theoretical prediction that criminal intelligence can prevent police officers from using lethal force. Similarly, the use of smartphones by officers for intelligence access is related to 2.72% fewer deadly shootings. We obtained similar results from the alternative data from killedbypolice.net and the FBI. Surprisingly, we found that the use of wearable video cameras is associated with a 3.64% increase in shooting-deaths of civilians by the police. We explain that video recordings collected during a violent encounter with a civilian can be used in favor of a police officer as evidence that justifies the shooting. Aware of this evidence, the officer may become less reluctant to engage in the use of deadly force. We conducted more in-depth analyses with incident circumstances (e.g. whether a subject was armed) and demographics of victims (e.g. race, age), and we obtained more intriguing findings. Notably, the above-mentioned effect of technology use on fatal shootings is more pronounced for (a) African American or Hispanic victims than Whites or Asians and (b) for armed suspects than unarmed civilians.

They authors tiptoe around their own findings here and at a couple other points in the paper. I can understand that – I spent a heck of a lot of time word-smithing when I wrote my earlier post as well. However, there is one point (deep in the weeds in the middle of the paper – far from the abstract and the conclusion) where their findings are plainly stated:

The coefficients of statistical analyses, smartphone, and video cameras are more negative and significant when shooting victims are African Americans or Hispanics (Column 5), 6 male (Column 6), and younger than 31 years old (Column 8). 7 Surprisingly, the impact of all three technologies is found to be insignificant for White and Asian deaths (Column 4), even though these two groups constitute 51.4% of the shooting victims by the police in 2015.

So, in short,
a. Use of statistics and smartphones leads to police killing fewer Black and Hispanic males, particularly those under 31 years of age.
b. Use of body cameras lead to police killing more Black and Hispanic males, particularly those under 31 years of age.

The mechanism for a. is that “police officers are able to obtain realtime intelligence on potential suspects (e.g. how violent they would be or what kind of weapons they would use) and take appropriate measures to subdue them without lethal force.” I think that is partly true or at least ambiguous. More precisely, realtime intelligence makes it easier to determine whether a suspect is likely to be a threat at all. In other words, it makes it increases the likelihood that a police officer will find out, before it is too late, whether someone dressed like a gangbanger is a poser or a danger.

Issue b. is a lot more interesting and has more important ramifications. The authors find that given “an expectation that video recordings would be used as substantiating evidence, police officers become less hesitant to deploy lethal force when they perceive the presence of deadly threats from suspects, leading to more correct hits.”

In other words, when a suspect constitutes a genuine threat to the police or other civilians, a cop with a body camera is more likely to react with lethal force. This is because when a threat does exist, the body camera is likely to provide evidence of its existence after the fact.

“Surprisingly, the impact of all three technologies is found to be insignificant for White and Asian deaths (Column 4), even though these two groups constitute 51.4% of the shooting victims by the police in 2015.”

These effects do not require, or even allow for the preface “[s]urprisingly” if one does the math in my earlier post. Better information in the hands of the police would, in general, improve the application of deadly force by police. Fewer false negatives and false positives would mean that, as per the data I provided, police would be less likely to shoot at individuals who are less of a threat, and more likely to shoot at individuals who are more of a threat. Having body cameras simply reduces the threat to a police officer’s career to reacting appropriately to the threat level presented. In fact, having body cameras increases the threat to a police officer’s career to reacting inappropriately. This is true whether the officer shoots a non-threatening civilian or whether the officer negligently allows a threatening individual to follow through on those threats if the opportunity was there to prevent them.

But Pang & Pavlou also miss a key point when they state this:

This study offers crucial implications for policymakers and practitioners in law enforcement. In response to nationwide attention on the police use of lethal force, a number of police departments are considering increased use of wearable body cameras, hoping that this approach will ultimately reduce deaths of civilians by the police (Ariel et al. 2015). We provide empirical evidence that demonstrates otherwise; the use of body cameras by officers is associated with more deaths of civilians.

The bolding is mine – and it is incorrect, or at least incomplete. The use of body cameras by officers is, indeed, associated with more deaths of civilians at the hands of police. But the effect on civilians as a whole depends on whether the killings by police are of a threatening person or of a non-threatening person. As I noted in my post, the current status quo means that

…the most violent members of the Black community are, on average, more likely to be allowed to go about their business unimpeded than the most violent members of the White, Asian, and Native American communities.

These violent individuals are the very people who are most responsible for the terrible death toll in the Black community. Getting this right is, quite literally, a life or death issue for a very large number of people. Getting it wrong (in however well-meaning a way) will only make things worse.