The 10 Do's and 500 Don'ts of the FBI's New Monthly Crime Data

Crime data reporting is changing, here's how to take advantage.

The FBI has started releasing monthly crime data on the Crime Data Explorer which absolutely could b an invaluable resource for crime data needs. It’s also data that data is preliminary, subject to change, unaudited, and full of potential pitfalls.

With this data release, my mind was drawn to a favorite Simpsons episode where Bart joins Springfield’s equivalent to the Boy Scouts in order to get a pocket knife. Before getting said knife, however, he has to read a book which advises him “Don't do what Donny Don't does."

The FBI’s monthly data can be confusing if you’re not careful, so here are some things to do and not do along the way. Remember: definitely don’t do what Donny Don’t Does with FBI monthly data.

Don’t think the data is final.

Monthly crime data has the potential to be wildly misleading if users are not careful. Not every agency reports complete data every month, there are frequent data reporting errors that will eventually get caught and ironed out, and the most recent month of data will usually be an undercount.

Sometimes those undercounts will be more severe than others and it’s not always easy to tell what is a big vs a small undercount. This is especially true when looking at national or statewide data though many individual agencies will also have fairly large recent undercounts. Here, for example, are aggravated assaults per month in the CDE right now from 2023 through July 2025.

The June data is a pretty clear undercount and the July data is an enormous undercount. That’s to be expected! It will get fixed as time goes on, but you have to be able to separate the good from the bad with this data. Some agencies are reporting full July data, some are reporting only through June, and some aren’t reporting at all in 2025.

Open data from New Orleans provides a good example of why this problem exists. You can take the share of vehicle burglary Calls for Service with a Report To Follow disposition — meaning someone called 911, an officer showed up and usually agreed that a crime occurred — and figure out the share of those incidents with a written report in the city’s Records Management System (RMS).

Nearly all (97 percent) vehicle burglaries in the last two weeks of July with an RTF disposition have a report in the RMS compared to just 60 percent of the incidents in the last two weeks. This issue will resolve naturally, it just takes a few weeks before reporting becomes stabilized.

The Real-Time Crime Index accounts for this problem to develop a national trend by lagging by 45 to 50 days and auditing out agencies that are clearly underreporting. That’s reasonably easy to do when you’ve got a sample of 400-500 agencies, much harder to do when your sample measures is 15,000 agencies.

Don’t think the data is always correct.

Historical reporting for some agencies isn’t always going to be correct, that’s just a fact of the US crime system. Take auto thefts in Los Angeles:

LAPD is reporting current data to the FBI (good), but the agency also really struggled with the NIBRS transition. There is no 2021 data because of NIBRS, and some crime types (namely theft and motor vehicle theft) weren’t fully recorded when the NIBRS switch was turned on in May 2024.

We account for this in the RTCI by auditing agencies and not including obvious underreporting wherever we can identify it. Fortunately, with Los Angeles and other California agencies we can get full monthly reporting through the end of 2024 through the California Department of Justice.

Do look at thefts as a good way to evaluate an agency’s completeness.

I like to look at thefts because they’re the most common type of crime. If thefts are being obviously underreported than other stuff probably is as well. Here’s Louisville thefts through June (they haven’t reported July data). That huge drop off in the most recent month is probably a reporting artifact.

And here’s Bellevue, Washington showing even more obvious underreporting so far in 2025.

The RTCI gets Louisville data from the agency’s open data feed and the CDE may not be a good source for more real-time reporting. We try to be conservative in terms of which agencies get included in the RTCI, so Bellevue isn’t included in our sample because of concerns about underreporting.

Don’t expect every agency to be there.

Louisiana did not report data for non-NIBRS agencies to the FBI in 2024 so New Orleans — my hometown — doesn’t have any data past 2023. Non-NIBRS agencies as a whole likely won’t be submitting monthly data to the FBI, but nearly 90 percent of the country is covered by a NIBRS agency now so non-participating agencies are outliers at this point.

Don’t forget more crimes may be reported because more agencies are reporting.

We should expect the number of crimes reported via a NIBRS-based system to go up as more agencies switch to NIBRS. This shouldn’t be as much of a problem going forward since non-NIBRS agencies are the exception rather than the rule, but that wasn’t the case just a few years ago.

Do be wary of agencies reporting quarterly, semiannual and annually.

NIBRS doesn’t suffer from this problem, but some agencies haven’t always report monthly data to the FBI. This is especially problematic with historical data, like in this graph of homicides in New York City from 2002 to 2013:

No, murder in New York City didn’t speak every third month, but NYPD reported three months of data all at once without differentiating to the FBI how many offenses were in each month.

And then there’s Florida which used to be a crazy mishmash of quarterly, semiannual and annual reporting and is now…less crazy but still kind of nuts.

Don’t use homicide and murder interchangeably.

All murders are homicides but not all homicides are murders. The FBI’s main menu gives you homicide per month counts but you can scroll down to offense types and grab murders per month if you like.

Do find novel ways to use the data.

The monthly data includes the master files for the current YTD which includes things like the Supplementary Homicide Report file. That means that we can more or less make a rolling graph like this tracking the share of homicides committed via firearm from the 1980s through just a few months ago.

Don’t use monthly data to draw broad conclusions about nationwide crime.

The CDE’s monthly data is great, but the quirks of how much is getting reported monthly and how much of that is underreported screams caution. This is especially true when you’re using it to evaluate national trends because the CDE’s monthly data will likely always show a large drop.

This, for example, is not the nation’s monthly murder trend by month:

The RTCI accounts for this by creating an apples-to-apples comparison that accounts for potential underreporting by auditing and weeding out agencies that are clearly incomplete. Creating a relatively consistent sample from month to month also enables a reliable understanding of the national trend that should mimic the FBI’s full year report that will be released in a year or so. The CDE’s monthly data can’t do that and people will struggle to use it for such purposes.

Don’t be surprised if the recent data being sent to the FBI doesn’t match what’s published by an agency.

Agencies don’t always report complete data to the FBI every month though the problem should be corrected when it’s ultimately due. Take thefts in Escondido, California. Comparing what the agency has published each month through June 2025 via a regional crime reporting system called ARJIS to the FBI data shows this point.

Both ARJIS and the FBI are in agreement through 2020 when NIBRS was likely implemented. There are small differences between 2021 and 2024 with a large discrepancy in December 2021 that I’m guessing was used to catch up. Overall, there aren’t really any problems until 2025 when the ARJIS figures continue on their merry way while the FBI counts fall off a cliff. I would fully expect this problem to get sorted out, but it highlights a downside risk to relying just on CDE data.

Do still be wary of using 2021 data.

The transition to NIBRS was a nightmare for crime data and a lot of agencies still haven’t reported their 2021 data to the FBI. Some agencies that didn’t report in 2021 have while others only partially reported. Ignore 2021 crime data when it’s not available, it’s crime data’s gas leak year.

Don’t get confused by the FBI’s national percent change calculations.

The CDE’s front page has a percent change graphic that seems straightforward but really isn’t.

First off, this is not a YTD measurement. The FBI describes it as: “The data used to produce these preliminary trends are from all law enforcement agencies that submitted at least 6 common months of complete offense reports within the two consecutive 12-month windows.”

Confused?

It is lagged 3 months, which is smart, but even then it’s comparing May 2024 through April 2025 with May 2023 through April 2024. The sample of agencies includes all agencies that had at least 6 months in common for each reporting period. So not every agency has reported 12 months of data over that stretch.

For example, the calculation for Anchorage, Alaska — which has only reported 3 months of data in 2025 so far — is May 2024 through March 2025 vs May 2023 vs March 2024. An agency that hasn’t reported any 2025 data will still show up based on May to December 2024 vs 2023.

And the monthly data is not audited. LAPD is reporting a 52 percent drop in property crime to the FBI due to NIBRS transition issues — shown above — which accounts for nearly a full percentage point of the national drop in property crime as calculated by the FBI.

There’s also Oakland, which erroneously reported an enormous surge in aggravated assaults in 2023. They haven’t reported any 2025 data so Oakland’s 46 percent drop in violent crime between May 2024 and April 2025 is only comparing May to December 2024 with May to December 2023.

The direction of crime shown in the FBI’s calculations is all correct, but the collection manner does not allow for an accurate understanding of the nation’s crime trends specifically this year. The FBI is not saying that violent crime is down 6.6 percent this year compared to last, and I’m definitely not looking forward to having to explain this fact over and over again when people ask.

Do appreciate how far crime data has come since the NIBRS transition.

Finally, monthly data from the FBI has the potential to be great if users are careful. We’re using this data with the RTCI to reach our goal of evaluating national crime trends from a sample of 500+ agencies nationwide. It's a far cry from 2021 when only 65 percent of the country reported even annual data.

None of this would have happened without the NIBRS transition. NIBRS forced agencies to report data in ways that allow it to be evaluated for completion and submitted to the FBI with lightning speed (for crime data). The transition wasn’t exactly smooth, but US crime data is in a much better place in 2025 compared to where it was just four years ago.