Under the Death in Custody Reporting Act, states who receive certain federal funds are required to report any death of someone who is detained, under arrest, in custody or incarcerated to the attorney general. Summary death counts have been made available to the public, but new reporting from The Marshall Project shows that the underlying data is plagued with inaccuracies. Anna Flagg is a data reporter and helped cover this story. She joins us to share more on what is wrong with the data and some of the inaccuracies the publication found.
Note: The following transcript was transcribed digitally and validated for accuracy, readability and formatting by an OPB volunteer.
Dave Miller: This is Think Out Loud on OPB. I’m Dave Miller. Under the 25-year-old Death in Custody Reporting Act, states are required to report any time someone dies in custody and report that to the U.S. Department of Justice. The data is available to the public. But a new article from The Marshall Project found that the data is incomplete and plagued with inaccuracies. Anna Flagg is a data reporter who worked on this story, and she joins us now. It’s great to have you on Think Out Loud.
Anna Flagg: Thank you for having me.
Miller: You started with a pretty shocking fact that in this federal database, George Floyd’s murder was not categorized as having been perpetrated by a police officer. What does it say?
Flagg: Yeah, that’s right. So under the Death in Custody Reporting Act, the manner of death is labeled in one of a few categories and one of those is death attributed to use of force by law enforcement or corrections officer – which is what happened, in reality, to George Floyd. But in the data, his death is mislabeled as a homicide, which, in their parlance, is an incident between two or more incarcerated individuals resulting in a death or a killing at the hands of someone other than a police officer. That’s kind of according to their definition for this data. So, for one of the most famous cases of police use of force in history, the government data does not call it a police use of force.
Miller: How is this one point, among tens of thousands of entries in this database, illustrative of a larger picture?
Flagg: Well, I think this is a death that led to international protests and a murder charge, a murder conviction for the police officer. And even in this case, the data is not accurate in the government records. So imagine how many other errors there are in less well-known cases that we don’t know about, right? So, we took a look at the raw data here and compared it to other data sets that are compiled by research organizations and academics. We found just a ton of gaps, inaccuracies, outdated information, blanks, underreporting. We found a lot of problems with the data.
Miller: And I should say that in custody deaths, it includes people who die in county jails and county lockups, in state prisons, and also, as in the case of George Floyd, during arrests. How is the reporting supposed to work? I mean, who is supposed to record these in the first place and then who is supposed to aggregate it or get it to the feds?
Flagg: So the way the program is implemented right now, it’s the job of each state to collect deaths data from all the local law enforcement agencies in their state. They collect that and then they submit it to the Department of Justice, the agency called the Bureau of Justice Assistance or the BJA. And that’s the way it’s implemented right now. It’s been implemented in different ways in the past and it’s been more successful at different times in history, but that’s how it’s supposed to work right now.
And I think one major issue with that system, as it is at this moment, is that it requires every state to sort of come up with its own system for collecting the data from the agencies and submitting it. And that could be anything from setting up like a web portal, or it could be collecting an Excel spreadsheet or kind of distributing an Excel template, or it could be a phone call, someone calls the state administering agency and tells them about a death. It’s just very inconsistent, very different across different states and that’s led to even more problems with the data.
Miller: You found at least 681 deaths that were simply missing from the federal count nationwide. But 450 of those, like two-thirds or so, were from one state, Louisiana, alone. Why is that?
Flagg: Yeah, so I don’t think that’s necessarily an indication of a higher number of deaths that are missing in Louisiana, so much as the fact that there is a comparison data set available in Louisiana. So Andrea Armstrong and other researchers from Loyola University maintain their own database of deaths in custody that is completely independent from the DCRA federal government data collection. And so because of that we were able to compare the data collected by the federal government to the data collected by these researchers and find that there were these hundreds of death records that were absent. But I think if there were more resources in other states similar to that one, then we would find many, many more missing deaths.
Miller: Where do you think the reporting system is most likely to break down?
Flagg: As you mentioned before, there’s three large buckets of deaths. Those are state prison deaths, deaths that occur in local jails, and then deaths that occur in arrest-related interactions with law enforcement. And I think that that last category, the third one, is kind of the most difficult category to collect data from because there’s like 18,000 different law enforcement agencies in the country.
So the task of every state coordinating with all the agencies in their state and kind of gaining compliance from them – not just compliance but like awareness of the reporting requirement, and then compliance to actually report that data and submitting that successfully to the Department of Justice – I think that’s what has proved most challenging so far.
Miller: I mentioned that one number that you were able to find of clear examples when deaths were missing from the federal account, a number that as you noted, would be much, much higher if every state were like Louisiana and had researchers who gave really reliable data that you could then cross check.
But you also reported that even for deaths that were in the federal system, the majority had important information that was missing. And I want to dig into some of those unfilled categories so we can understand what’s at stake.
So for about one in six entries, the manner of death was marked as “unavailable” or “investigation pending.” Is that the case even for deaths that go back to 2019 or 2020?
Flagg: The data set that we were looking at was specifically from 2019 through 2023, so that kind of covers all of that. And yeah, I think that’s pretty much a correct result for that entire time span. One out of every six does not have a manner of death specified. And that was a really striking finding to me just because, of course, the reason to collect deaths in custody data is to try to reduce deaths in custody in future and try to prevent similar deaths from happening again, try to find patterns for the root causes. But how can you possibly expect to figure out how to reduce deaths in custody if you don’t even know how people are dying?
Miller: You also note, to that end, that one in 10 entries specified no location of death, so it’d be hard to find geographic patterns; even more than that, didn’t specify the race of the decedents, so it’d be hard to see demographic patterns. Eight hundred entries didn’t include the law enforcement agency responsible at the time of death, so it would be hard to say this sheriff’s office has an unusually high number, so the feds couldn’t look into that. What do the numbers look like in Oregon, in particular?
Flagg: Looking a little bit at the Oregon information … And again, this is from an incomplete data set that was just briefly kind of leaked online and made available, so certainly, part of the problem here is that we don’t feel like we have a complete data set. These results are just kind of what we can see in that data set, that we know is flawed. But just kind of assuming all that, it looks like there were 292 deaths in the four-year time period that we’re looking at there, between fiscal year 2020 and 2023.
And most of those are in state prisons – about 65% of them – 17% were in jails, and 12% were in interactions with the state or local law enforcement. And in Oregon state data in particular, it looked like 20% of cases had “unavailable” or kind of “investigation pending” filled in as a manner of death. So for Oregon, in at least one of five cases, you don’t have that manner of death information to work with.
Miller: But as you said, and the caveat is really important, short of going through every single media report or calling up every single law enforcement agency in the state, we cannot know how accurate or how complete the Oregon data in that federal database is.
I want to turn to the federal piece of this though, more squarely. What kinds of penalties are states supposed to face if they don’t accurately and adequately report this data?
Flagg: When DCRA was first passed in 2000, it had a compliance requirement that was tied to truth in sentencing grants and that was supposed to be kind of an incentive for people to report, to do this data reporting. But over the years, it appeared that the reporting rates were not high at all, in particular for police departments. So again, the state prison deaths and the local jail deaths were getting reported at much higher rates than the deaths that occurred in arrest situations.
So in 2014, when the law was reauthorized, they added in a penalty where a certain amount of your, it’s called Justice Assistance Grant or JAG funding, federal funding, can be withheld from a certain state if the agencies in that state do not fully report their deaths in custody data. So that’s kind of the penalty aspect that legislators added to DCRA in order to try to give the Department of Justice a way to enforce this law. That was back in 2014.
The problem is though, the Department of Justice has never used this penalty against a single state, despite all of these long-standing problems with the data that have been documented. The department just hasn’t chosen to ever exercise this penalty to enforce the law.
Miller: Before we say goodbye, I do want to turn to one seeming bright spot. You found that Washington state is actually doing something well. What did they pass there in 2021?
Flagg: Yeah, Washington state passed their own state-level deaths in custody reporting law. And they’ve put together an entire kind of working group with a pretty deep-detailed process for collecting death in custody data. They collect information well beyond what’s required by the federal DCRA law. And because of that, they’re able to bring together people from the Department of Corrections, medical people and people from the governor’s office to analyze the root causes that are leading to deaths in prisons and identify patterns and make recommendations to try to prevent avoidable deaths in future.
Miller: Has that happened? Has the data been used in a way that seems similar to what the federal government hoped with or at least the people who passed that law hoped would happen at the federal level?
Flagg: Yeah, I think they’ve used the data to find situations that were leading to higher levels of death that could be prevented. I think there was one in particular that they talked to us about, where there was a certain area and a facility where a lot of suicides were happening. And they put a barrier there and then they set up some mental health support and some hotlines. I mean, a lot to be done in situations like that, but they were kind of taking action based on patterns identified in the data, in order to reduce deaths going forward. And that’s really the purpose of collecting this data.
Miller: Anna, thanks very much.
Flagg: Thank you for having me.
Miller: Anna Flagg is a data reporter for The Marshall Project. She worked on a story that came out recently about incomplete and inaccurate information in the federal database about in custody deaths from all around the country.
“Think Out Loud®” broadcasts live at noon every day and rebroadcasts at 8 p.m.
If you’d like to comment on any of the topics in this show or suggest a topic of your own, please get in touch with us on Facebook, send an email to thinkoutloud@opb.org, or you can leave a voicemail for us at 503-293-1983.
