Evaluating Crime Reporting in Washington, DC

Please just read the article.

My recent piece on DC got a lot of attention for some reason and I assume it’s because people enjoy reading mundane assessments of crime in random US cities. I’m really excited to put up huge numbers on a future assessment of Tulsa’s crime trends! Joking aside, one of the issues raised by President Trump and many others in the comments centers around the accusations in this piece from the local NBC affiliate published last month. From the piece:

The Metropolitan Police Department confirmed Michael Pulliam was placed on paid administrative leave in mid-May. That happened just a week after Pulliam filed an equal employment opportunity complaint against an assistant chief and the police union accused the department of deliberately falsifying crime data, according to three law enforcement sources familiar with the complaint.

The union claims police supervisors in the department manipulate crime data to make it appear violent crime has fallen considerably compared to last year.

My general thoughts on the matter are that extraordinary claims require extraordinary proof and there’s not much publicly available proof just yet. I’ll note that DC reported a 7 percent increase in aggravated assaults to the FBI last year, so if numbers are being intentionally manipulated to show a drop in assaults reported to the FBI then they did a really poor job. In other words, I have much more confidence in the FBI figures than the public website figures (I’ll get more into that in a bit).

I’ll also note that I’m firmly in the camp that the drop in violent crime on the public website is likely overstated. As I said last week:

Measuring DC’s overall violent crime trend is a bit more complicated. Violent crime in DC is down a huge amount from where it was in the early 1990s and down a good bit from where it was a decade ago. Violent crime in DC fell in 2024 after increasing in 2023, that much is clear from DC’s reporting to the FBI. The exact picture for 2025 is a bit murkier though violent crime is likely falling this year. DC’s open data is likely overstating the degree of the city’s decline in violent crime right now though.

I’m just a keyboard jockey sitting in New Orleans, so I’m completely unable to evaluate the assertion that crime data was intentionally manipulated. What I can do is evaluate the different data reporting streams to decide what appears accurate and what prompts some skepticism.

It’s worth noting that I talk about unintentional crime data reporting issues ALL THE TIME in this newsletter. Moreover, FBI reporting almost always differs from the preliminary figures put out by agencies. This is sometimes because an agency’s initial figures will get revised before they’re final. Other times it’s because an agency’s raw data may count incidents while the violent crime figures reported to the FBI count victims (so a triple homicide will show up as one homicide incident). Or — as is the case in DC — the agency may not report public figures exactly in line with what they’ll eventually report to the FBI.

With all of that in mind, here’s what I think regarding the validity of the data:

First, DC’s murder and carjacking data appears to be very solidly reported and there’s no reason to suspect the large declines being reported in both categories are anything but large declines. DC’s murder data as reported to the FBI is nearly identical to what’s reported on the public website.

There are some small discrepancies between what the FBI reports and what DC reports on its public website, but that’s to be expected. (Note that DC didn’t report data to the FBI for 8 months in 2021 due to the NIBRS transition, so those are going to be missing from any graph having pre-2022 data).

DC’s law enforcement open data is also matched by other sources, such as the Gun Violence Archive and the DC Health Department’s Firearm Injury Dashboard which shows a 25 percent drop in firearm injury visits from DC residents through May 2025 compared to the same five months in 2024.

So DC’s murder trend is pretty well known and we can feel confident in the assertion that gun violence and murder are falling fast after having surged between 2020 and 2023. Murder has fallen a lot in DC since peaking in 2023. It is down a ton from where it was in the 1990s. Murder still remains too high — much higher than it was at the low point in 2012 — and is a problem worth continued attention and solutions. All of these things are true.

I also trust the carjacking data. While there appears to be a pretty clear discrepancy in how overall robberies (more on that in a second) are being reported, the carjacking figures reported to the FBI via NIBRS match up very strongly to the counts published on the city’s dashboard through 2024.

If the carjacking figures — which make up a reasonably small portion of all robberies — were suspect then I would expect to see the same discrepancy with them that we see with the rest of robberies. We don’t, we actually see slightly more carjackings reported publicly than to the FBI, but the difference is small and evens out over time.

This is what we expect to see when comparing FBI and agency data and suggests that the data being submitted to the FBI and the data on the dashboard are more or less the same. Note that I only went back to 2022 because DC made the switch to NIBRS in 2021 (the FBI’s old system couldn’t capture carjackings).

You can do the same exercise with the motor vehicle theft trends as well to see exactly how they should be matching up. The result is more or less identical trends:

These trends stand in pretty sharp contrast to the overall robbery statistics which show a clear disagreement between the data being submitted to the FBI recently and the data shown on the publicly available website:

Robberies reported on the public website tracked robberies reported to the FBI very closely until late 2023. Robberies have fallen a ton in DC since 2023 regardless of which metric you use, but the decline on the public website was about twice as large as the drop reported to the FBI¹.

What about assaults? Those are harder to tease out because DC is reporting assaults with a dangerous weapon and not aggravated assaults. There are typically 2 to 3 times more of the former than the latter in a given year and they don’t always move in lockstep. Comparing the two directly doesn’t really paint an obvious picture of, well, anything.

You can also compare the two assault types by indexing to each’s January 2022 total and you do see broad agreement through 2022 and most of 2023 before they split…

But if you zoom out to go back to 2017 then you seen that the 2022 to 2023 period is the outlier and there’s nothing really unusual about what’s being reported right now relative to what gets reported to the FBI. Certainly nothing that screams deliberate underreporting of assaults.

DC’s open data allows us to evaluate crime trends by district, so if a single or group of district commanders were artificially lowering crime counts then it might show up here.

But nothing really stands out as all districts are more or less seeing consistent drops of late citywide. The nearly uniform drop across districts suggests a single commander is not manipulating statistics, but (again) I’m unable to draw any firm conclusions from a keyboard in New Orleans.

Where does all of this leave us?

I mentioned in my earlier piece that I do not find the publicly reported assault figures to be particularly accurate. I do not, however, think Washington DC is underreporting data to the FBI, otherwise the robbery/assault reported to the FBI would match the public data’s drop and underreporting would be much harder to tease out. You don’t publish a 27 percent drop in assaults on your website but report a 7 percent increase to the FBI if you’re deliberately underreporting them both publicly and the FBI.

While there’s no way to prove or disprove this from where I sit, there’s a clear divergence between what is being reported in the public data specifically as it relates to robberies and assaults and what is being reported to the FBI. A healthy skepticism of the overall public violent crime data remains warranted with more trust in the murder/carjacking and property crime data being reported.

Data reporting issues happen all the time with crime data, so perhaps I’m less phased by this likely occurring in DC than others are. It’s fully plausible that this is a data reporting issue with a perfectly reasonable and benign explanation. It’s also possible that the claims of manipulation have some merit, time will tell on that front.

Crime data is imperfect and imprecise, especially when it’s preliminary like this. My advice is to think of trends more broadly to account for the possibility of errors and revisions. In the case of DC, my advice would be to ignore the violent crime bottom line in the public data — relying instead on the FBI data through 2024 — while concentrating on the more trustworthy murder/gun violence, carjacking, and property crime trends from the public data to discuss 2025’s crime trends.

Ultimately, my bottom line is the same one I wrote about in my initial piece: “Even if we can’t say for certain from the publicly available data how far (violent crime’s) falling, there’s no reason to suspect the overall trend being reported isn’t correct.”

1. I casually mentioned robberies as a measure that was relatively trustworthy in my initial piece before doing a deep dive. Obviously that’s not the case upon much closer examination.