Photo by Markus Spiske on Unsplash

Police Shooting Research and the Conditional Probability Mistake

Photo by Markus Spiske on Unsplash

Police Shooting Research and the Conditional Probability Mistake

Yesterday, Andy Wheeler posted a summary of the problems with recent studies about officer-involved shootings, including one my colleagues and I published in February 2017. As usual, Andy’s criticisms were thoughtful and spot on. And I hope I can take him up on that conference beer soon, even though he’s at a new job.

That said, I do want to push back just a little about the motivation for our paper. Brad and I had the idea for it because journalists and activists were getting the denominator wrong when they tried to interpret the (at the time) new WAPO & Guardian data.

I remember daily conversations with Brad and Geoff about what we could and could not learn from these new crowdsourced data in the absence of a meaningful benchmark. Ultimately, we felt it was informative to determine whether - strictly in this universe we had data on (i.e., fatal shootings) - race was correlated with having been unarmed while controlling for threat. At the very least, we thought it was a more reasonable way of analyzing the data than how it was being used at the time.

We really tried to be transparent about the limitations of the data:


Of course, with the benefit of hindsight, we’d be even more careful.

I agree wholeheartedly with Andy that the question of whether “the police” are racially biased is a facile question. But I also recognize it’s a question that journalists, activists, and others routinely try to answer simply by looking at raw numbers reported by WAPO or Mapping Police Violence (or at most, benchmarking them against the general population). Consistent with the first retort Andy said he often hears, I feel it is our responsibility as social scientists to learn what we can from the data we have, and push for better data so that we can continue to learn more.


Since then, I’ve written extensively about the importance of using better benchmarks to understand disparities in policing. See for example, here, here, and here. It’s complex, to say the least.

And now, thanks to Knox, Lowe, and Mummolo’s fascinating forthcoming paper, I realize it’s even more complex than a lot of us thought.

Lastly, Andy hit the nail on the head here:


I got some pretty nasty emails after that CPP paper, and occasionally still have my motives questioned by folks on both sides. It’s unfortunate, because I think that sort of thing scares a lot of really sharp people away from doing this research.

Justin Nix
Associate Professor of Criminology and Criminal Justice

My research interests include police legitimacy, procedural justice, and officer-involved shootings.