Thread by @PeterMoskos

I'm trying to understand this story, and I'm having a tough time. Presumable somebody uses this does, but I've never used the National Vital Statistics System for police info. I actually didn't even know the name.
nytimes.com/2021/09/30/us/…

It's certainly not an official count of police related deaths in my world. So I'm reading this going, researchers discover a bunch of police-related data is bad? Good (to discover it's bad). Make it better. But I wouldn't have thought it was a count, much less an undercount.

It was bigger news when I learned how much the UCR was undercounting deaths at the hands of police, because I was using UCR and though I knew the data wasn't good, I didn't want to believe how bad it was. It was the best we had at the time.

But I still question the new findings: "[over 39 years] nearly 31,000 Americans were killed by the police, with more than 17,000." Now way. That's still way off. It's way too low.

We know that about 1,000 people /yr are shot and killed by police. Probably about another 300 are killed or die. But here data gets messy. People do die in police custody. They can overdose or have a heart attack. There's usually multiple factors involved in such deaths.

Also, the # of people killed by police has decreased since 1980, based on every city that keeps data. So I don't know how the article says, "The annual number of deaths in police custody has generally gone upward since 1980." I don't buy it. Says who? This bad data? Maybe ODs.

17,000 (old figure) is so absurdly low any researcher with a basic knowledge of the topic should not touch that data set. But even 31,000 (new figure) is way too low for 39 years. I mean, it's got to be at least or around 1,300 a year. Because that's what he have now.

Wild but somewhat conservative complete guess? I'd say something close to 60,000 have died in police custody between 1980 and 2018. With probably more than 40,000 shot and killed. But we don't know.