On Triggernometry show Roland G. Fryer Jr. makes the claim

I thought I was going to be able to show that the police were biased very easily, and then we gathered literally millions and millions of data points on lethal use of force. what we found was Zero racial differences and that is the part that made people really really upset

He seems to refer to paper 'An Empirical Analysis of Racial Differences in Police Use of Force'. In which he writes:

VI. Interpretation
A number of stylized facts emerge from the analysis of the preceding sections. On non-lethal uses of force, there are racial differences – sometimes quite large – in police use of force, even after controlling for a large set of controls designed to account for important contextual and behavioral factors at the time of the police-civilian interaction. As the intensity of use of force increases from putting hands on a civilian to striking them with a baton, the overall probability of such an incident occurring decreases but the racial difference remains roughly constant. On the most extreme uses of force, however – ocer-involved shootings with a Taser or lethal weapon – there are no racial differences in either the raw data or when accounting for controls.

On Harvard blog Justin M. Feldman makes the opposite claim:

There is racial bias in shootings by police

Is there evidence for racial bias in police shootings in America when accounting for controls?

  • 1
    "when accounting for controls" is a big methodological choice. If (plausible but I have no proof) African-American are more at risk of being controled and if (plausible but I have no proof) the risk of being shot given that one is controled is equal across racial groups, then the answer to your question would be "no" but the risk of being shot would be higher for A-A than for the general population.
    – Evargalo
    Commented Jun 17 at 7:53
  • @Evargalo That is part of the claim so I had to include it.
    – pinegulf
    Commented Jun 17 at 7:56
  • @Evargalo You comment gives the impression that you don't understand what the word "controls" means in the context of statistics. Commented Jun 27 at 13:00
  • @Acccumulation : then my comment gives a false impression of this statistician who deals with controls all the time.
    – Evargalo
    Commented Jun 27 at 13:33
  • @Evargalo In the context of statistics, the entire dataset is controlled. Individual observations aren't controlled, and certainly people are not controlled. Commented Jun 29 at 3:57

2 Answers 2


Is there racial bias in police shootings in America?

This is a loaded question (pun intended) because different people will interpret the question differently. At what levels of policing does racial bias occur? And at which of those levels (if any) is racial bias a problem? 538 actually wrote an article about this very issue.

Fryer's finding

On the most extreme uses of force, ...there are no racial differences in either the raw data or when accounting for controls.

This finding could be rephrased as "after the police has identified a person as a suspect, the police use extreme force at an equal rate between black and white suspects." This suggests that "suspects" do not experience racial discrimination. Based on the data recorded, this is true.

Systemic Racial Bias

But that is not the only way to define racial discrimination. The other factor is the rate that suspects are correctly identified for each racial group. Or to put it another way, how accurately do the police label people from different racial groups as "suspects"? If there is bias among the police force in treating black people as "suspects" at disproportionally high rate, then you'd expect more black people to experience extreme force as a proportion of the total population (not just the "suspect" population).

This infographic from the 538 article explains this very well:

538 infographic about police bias


Basically, racial bias is complicated, and the data that is collected is not at a high enough level to say definitively that there is no racial bias. Police shootings get a lot of attention, but by then multiple police interactions have already happened, any number of which could be biased. It can be simultaneously true that officers treat "suspects" the same regardless of race while also treating "people" differently when it comes to classifying them as suspects. My intuition is that the police likely identifies people as suspects incorrectly, and that disparity leads the black Americans to experience police shootings as a disproportional (biased) rate as compared to white Americans, but we would need better data on the more mundane police interactions to know for sure.

  • I'm unfamiliar with 538. It seems to be part of ABC news, but beyond that it's question mark.
    – pinegulf
    Commented Jun 20 at 15:06
  • 3
    They started with statistics-based political forecasting, but expanded briefly to other analytics stuff, including sports forecasting. Then ABC acquired them.
    – ryanyuyu
    Commented Jun 20 at 16:30
  • 1
    Personal issues (i.e. allegations of sexual misconduct) aside, Fryer Jr. is a fairly respected scholar whose research is probably correct insofar as it goes, but I do have to wonder about his motives considering that he is going on provocateur shows like "Triggernometry" to focus on that part of it while ignoring the other factor that you mention, the likelihood of racial bias in whom police choose to interact with in the first place. While I hesitate to accuse him of dishonest argumentation, a lot of those who use his research certainly are.
    – Obie 2.0
    Commented Jun 22 at 0:11
  • 3
    After all, for non-police murders, no one is trying to figure out whether the murderers exhibited bias starting from the point that they chose a target. No one says "Sure, Dylan Roof attacked a majority-Black church, but once he was there, he would have killed anyone he saw, so how biased was he, really?" Only police get this benefit of the doubt, but that kind of begs the question, since it assumes their pre-homicide targeting to be unbiased and legitimate by default.
    – Obie 2.0
    Commented Jun 22 at 0:17

Accounting for contextual variables recorded by police officers who may have an incentive to distort the truth is problematic. Yet, whether or not we include controls does not alter the basic qualitative conclusions. And, to the extent that there are racial differences in underreporting of non-lethal use of force (and police are more likely to not report force used on blacks), our estimates may be a lower bound. Not reporting officer-involved shootings seems unlikely.
— 'An Empirical Analysis of Racial Differences in Police Use of Force', page 5, Roland G. Fryer, Jr

This is about controlling any bias that might affect the reporting of data. Any such bias would reduce the values in the final result, not exaggerate them.

  • For non-lethal incidents, the actual differences in racial bias will be greater than what the data suggests (and those differences really are significant).

  • For lethal incidents, the data and the observed differences will be much more reliable (and those differences really are negligible).

I.e. according to this analysis, the result of race having no effect on fatal incidents highly corresponds to reality.

Justin M. Feldman's report though doesn't even contain the word "fatal". Instead it uses "shot", a term that will include many non-fatal incidents, which the Fryer report agrees are racially biased.

Feldman seems to be using a straw-man argument against the truth.
Even the title of his report, "Roland Fryer is wrong: There is racial bias in shootings by police", itself contains the fallacy, incorrectly implying that Fryer claimed there wasn't bias in shootings.

  • 1
    "straw-man argument against the truth." A link to source of the truth would be useful.
    – pinegulf
    Commented Jun 18 at 6:04
  • 1
    If Feldman is talking about shootings, while Fryer is talking about deads, they can be both right at the same time. That's not the same thing as a strawman argument.
    – Rekesoft
    Commented Jun 18 at 10:25
  • 1
    @Rekesoft. If they were two independent reports, I'd agree. But in this case Feldman's report is a specific disagreement targeted at Fryer's. See my updated last paragraph. Commented Jun 18 at 12:37
  • The quote from Fryer says "Not reporting officer-involved shootings seems unlikely," but you summarize that there is likely underreporting of non-lethal but not lethal use of force. Maybe that's from a different part of the paper, but since the answer is saying Feldman isn't comparing apples to apples it might be good to clarify this. After all, shootings aren't always fatal and fatalities aren't always caused by shootings. Commented Jun 26 at 19:13

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .