Exploring The Difference Between FBI Murders and CDC Homicides

A crime data mystery!

One of the questions I’ve been thinking about without much luck in recent years has been why the gap between the FBI’s murder estimate and the CDC’s homicide count don't match up as well as they used to. Last year, for example, the FBI reported a 14.9 percent drop in murder while the CDC reported an 11.7 percent drop. The same issue occurred again in 2023 when the CDC reported an 8 percent drop in homicides compared to the FBI reporting a nearly 12 percent drop in murder. In 2019, by comparison, the FBI reported a 0.3 percent drop in murder while the CDC reported a much closer 1.6 percent drop.

To dig into this problem, I grabbed murder data from the FBI’s Crime Data Explorer and homicide data from CDC’s WONDER. Every murder is a homicide but not every homicide is a murder, so the number of homicides reported by the CDC should be a bit higher than the number of murders initially reported by the FBI in a given year. As you can see in this graph, the share of murders initially estimated by the FBI per homicide reported by the CDC has fallen from steadily right around or just below 0.9 for most of the 2010s to 0.84 in 2023 and 2024.

But the FBI revises previous year numbers in the years after publishing them, and using the revised murder estimates published in the FBI’s 2024 report shows that the growing gap is largely being fixed in later FBI revisions to their initial estimates.

Every year from 2010 to 2020 fits around that 0.9 murders per homicide, but it falls a bit in 2022, even more in 2023, and a good bit in 2024. This seems to reflect a growing issue with the FBI’s initial estimates which appears to be getting corrected as time goes on and revised estimates are issued.

To be clear, the FBI is still accurately portraying the nation’s murder trends each year, but its initial estimates have been less accurate than later revisions in recent years using the CDC homicide counts as a benchmark.

And much of the cause for this problem centers around one state.

To see this, I compared the number of homicides for each state from 2022 to 2024 as published by the CDC with the number of murders reported to the FBI in each state over that span. Can you see which state stands out?

Or looking at it another way, here’s the difference between the CDC’s homicide rate reported in each state in 2024 versus that state’s estimated murder rate according to the FBI that year. In Texas, for example, the CDC’s homicide rate was 0.77 homicides per 100,000 higher than the FBI’s estimated murder rate.

One state stands out again, can you spot it?

I’ve written before about how the FBI’s estimative process doesn’t always do particularly well when a lot of data is missing and that seems to explain a decent chunk of the problem here.

It’s not just Mississippi though.

I’ve also written about the problems Illinois has had with reporting accurate data to the FBI, a problem that really started nearly 40 years ago. In Louisiana, neither New Orleans nor Baton Rouge — the state’s two largest cities which also produce the most murders — reported data to the FBI after the state did not allow non-NIBRS agencies to report (Baton Rouge transitioned in mid-2024). The estimates for Tennessee, Georgia, and Alabama were also decently far below what the CDC homicide data suggests they should have been.

But Mississippi really stands out.

A large part of the problem is that murder in Jackson has absolutely exploded post-pandemic. The Jackson Police Department has not reported data to the FBI since 2019, but combining police department data with an excellent dashboard kept by local news station WLBT shows just how awful the city’s trend has gotten albeit with substantial improvement over the last few years (and homicide in Jackson is down 45 percent this year).

Note that in this graph that the 2020 to 2024 counts are for all homicides rather than just murders because there’s no publicly available record of Jackson murders in those years that I could find, so it’s probably slightly larger than murder rate alone.

Because Jackson hasn’t been sending data to the FBI for the last five reports, this enormous surge in murders has not been formally counted in the US crime estimates. Instead, the FBI has estimated Jackson’s count each year like it does for every city that doesn’t report in a given year. Here’s how the FBI’s estimation procedure is supposed to work:

If an agency has supplied less than 3 months of data, the FBI computes estimates by using the known crime figures of similar areas within a state and assigning the same proportion of crime volumes to nonreporting agencies. The estimation process considers the following: population size covered by the agency; type of jurisdiction, e.g., police department versus sheriff’s office; and geographic location.

So if an agency is missing completely then they use the crime rate of other similarly sized agencies in that state to estimate how many crimes that agency had. There’s a problem with Mississippi though. Jackson has right around 150,000 people in it. The state’s next biggest city is Gulfport (population of around 70,000), but Gulfport hasn’t reported data to the FBI since 2021.

Southaven is the only other city in Mississippi with more than 50,000 people, but Southaven has just barely more than 50,000 people and reported 1 murder in 2024. If you used Southaven as the stand-in city, and I don’t know for sure that that’s what the FBI is doing in this case, then you’d apply a murder rate of 1.7 per 100k (Southaven’s rate) to Jackson’s population of 141,000 and get 2.5 murders for Jackson in 2024 (WLBT suggests a slightly larger 100ish murders in 2024 for comparison).

It’s not exactly clear what methodology the FBI uses in a situation like this, but only a handful of cities larger than 50,000 had murder rates even half as high as Jackson’s last year. There’s not really any city in the US that was all that close to Jackson’s murder rate in 2024, so not having Jackson’s data makes it much harder to match the CDC’s homicide counts.

Louisiana faced a similar issue in 2024 without New Orleans and Baton Rouge’s reporting, but Louisiana has two cities (Lafayette and Shreveport) bigger than 100,000 with somewhat relatable (albeit smaller) murder rates which should make accurately estimating the state’s murder rate much easier. Plus nearly 80 percent of Louisiana’s population reported data to the FBI in 2024 compared to less than half of Mississippi’s.

The big question, at least in my opinion, isn’t why there is a growing gap between the FBI and CDC murder/homicide estimates, but rather why are the FBI initial estimates becoming predictably lower than their more final tallies?

The FBI is sort of fixing these problems with later revisions that increase the number of murders estimated in previous years, but the process is opaque and it’s not clear exactly what is being revised. Yes, the number of murders in 2023 was revised up in 2024 to more accurately reflect what the CDC data suggests occurred. But those revisions weren’t a result of Mississippi’s number getting corrected, instead it was mostly larger states having their murder estimates revised upwards.

So why is the FBI’s murder estimate consistently showing larger declines than the CDC’s homicide count? A good bit of it is explained by Mississippi underreporting murders much more than other states. But Mississippi doesn’t account for all or even a majority of the FBI’s initial undercount.

Ultimately, a lack of further insight into how the initial and revised estimates for Mississippi and other states missing large numbers of murders are being made makes firm answers to this question difficult to come by.

The FBI’s estimates are directionally and proportionally correct — murder really did fall a lot in 2024. But digging deeper into this issue really drives home just how inexact our estimates are and how much uncertainty is needed when discussing what is happening nationally.