For this conversation, we talk about why murder has fallen so dramatically over the last few years, how you build a sustainable organization that spans the political spectrum, and how CCJ brings together experts on such a wide range of topics covering everything from crime trends to the future of AI in criminal justice.

Adam Gelb has been working for a more just and effective criminal justice system throughout a varied 40-year career. Before founding the Council on Criminal Justice in 2019, Gelb was an award-winning crime reporter at The Atlanta Journal-Constitution, staff to the U.S. Senate Judiciary Committee, crime policy director for the lieutenant governor of Maryland, director of the Georgia Sentencing Commission, and led public safety initiatives at the Pew Charitable Trusts. Gelb speaks frequently with the media about national trends and state innovations and advises policy makers on strategies that are grounded in facts, evidence, and fundamental principles of justice.

Please give the episode a listen and rate it if you're so inclined (5 stars and up please!).

Apple

Spotify

Amazon

And while you’re here, be sure to check out these other recent great episodes:

Baltimore Mayor Brandon Scott

Filmmakers Ferne Pearlstein and Bob Edwards

Professor and researcher Ian Adams

Politics podcaster Galen Druke

Leave a comment

Last year’s trends are firmly established and not really worth revisiting — at least in any way that’s interesting to me the analyst and you the reader. The FBI hasn’t come out with any formal 2025 reports yet and it’s still slightly too soon to be seriously reading the tea leaves (though that didn’t stop me from trying last week).

Still, this is the 200th edition of this newsletter, so I wanted to do something a little different. I remembered that I made a quiz on this newsletter when it was still a baby and figured that might be a fun way to pass the time and see how closely you all are paying attention! 10 questions plus a bonus.

The answers are at the bottom, so let me know how you do in the comments! Anything less than 10 of 10 is unacceptable.

Question 1:

Question 2:

Question 3:

Question 4:

Question 5:

Question 6:

Question 7:

Question 8:

Question 9:

Question 10:

Bonus Question!

Answers:

1 - 10% - The FBI reported a 9.2 percent motor vehicle theft clearance rate in 2024 which is not great! The national motor vehicle theft clearance rate has typically hovered around 15 percent but fell all the way to 8 percent in 2023. It probably cleared 10 percent in 2025 based on the FBI’s monthly data, but not by much. Still not great!

2 - 1989 - The National Incident-Based Reporting System (NIBRS) first started in 1989 though only around 1/3 of the nation’s law enforcement agencies were using it as of 2015. That figure was closer to 90 percent in 2024 (87.2 percent to be exact) with both NYPD and LAPD becoming compliant in recent years.

3 - 1930 - The first version of the Uniform Crime report as we know it today was published in 1930. The first version of the UCR published by the FBI didn’t come out until August 1930. Prior to that, the UCR was the brainchild and responsibility of the International Association of Chiefs of Police (IACP).

4 - NCVS - The Bureau of Justice Statistics runs the massive nationwide survey called the National Crime Victimization Survey or NCVS. They have a handy dashboard that makes it super easy to break down their survey data.

5 - 2014 - This was sort of a trick question because I asked about the number of murders, not the rate. There were right around 14,000 murders in the US in 2025 (my best guess!) and, while we can’t say when the murder rate per 100k was lower, that’s roughly in line with the number of murders in 2014 (14,177 per the FBI).

6 - 30% - According to the National Crime Victimization Survey, just 30.5 percent of property crimes in 2024 and 29.9 percent in 2023 were reported by victims to police. The share varies by type of offense, with only about a quarter of thefts but 80 percent of auto thefts being reported each year. Have to have that insurance payout on stolen cars.

7 - 7 million - The FBI reported 6.7 million fewer property crimes in 2024 than they reported in 1990 (5,986,400 versus 12,655,486) and that gap likely exceeded 7 million in 2025. Burglary has fallen by 75 percent in the last 25 years while theft and motor vehicle theft have both fallen by around 50 percent.

8 - Found not guilty at trial - A case can be considered cleared if an agency has enough evidence to make an arrest but the offender dies, the victim refuses to cooperate, or prosecution is declined for reasons other than lack of probable cause. What happens in a trial is immaterial to a case’s clearance status as far as the FBI is concerned.

9 - Clarksburg, West Virginia - CJIS is housed in West Virginia, which you would know if you’d listened to my podcast episode with FBI Assistant Director Timothy Ferguson.

10 - Monthly - The IACP started UCR with monthly reporting with the first issue covering 400 cities with a population of around 20 million. The FBI took over in September 1930 and within a few years they started quarterly reporting, then semiannual and finally annual reports. Quarterly reports were begun again by the FBI in 2020 and they are now reporting monthly data again, a change that started in 2025.

Bonus - Vince Young was the GOAT.

How did you do?

Leave a comment

To better understand this amazing turnaround, I talked to Baltimore Mayor Brandon Scott. Mayor Scott shares his journey to becoming mayor, the city's innovative strategies, and his insights on reducing gun violence and transforming public safety in Baltimore. This is an episode about how data-driven approaches, community engagement, and leadership are shaping one city’s future.

Brandon M. Scott is the 52nd Mayor of Baltimore, and the youngest person to hold the position in more than 100 years. Mayor Scott is committed to ending gun violence, restoring the public’s trust in government and creating a brighter, better and more equitable Baltimore for all.

Please give the episode a listen and rate it if you're so inclined (5 stars and up please!).

Apple

Spotify

Amazon

And while you’re here, be sure to check out these other recent great episodes:

Filmmakers Ferne Pearlstein and Bob Edwards

Professor and researcher Ian Adams

Politics podcaster Galen Druke

Arnold Ventures Executive Vice President Jennifer Doleac

Leave a comment

It’s so early that the Mets aren’t even disappointing their fans yet, so take it with a grain of salt. But. The evidence so far is that the trends of 2025 have largely continued into the first few months of 2026.

There are four different data sources I want to rely on to preliminarily assess murder and gun violence trends in the US.

First, let’s start with the Real-Time Crime Index. The RTCI sample for January 2026 was released last week showing murder down 30 percent in a sample of nearly 500 agencies covering around 100 million people nationwide. Longtime readers may recognize that the sample is considerably smaller than in recent months (last month’s sample had around 100 more agencies covering 15 million more people).

I’ll be honest, putting together a sample for January data is really difficult. It’s early in the year so YTD is pretty meaningless1. It’s also cold so crime counts are usually their lowest or second lowest (after February) of the year in most cities. It’s also a new year so many agencies haven’t restarted their reporting practices yet — at least one state is sending us their data this week so we will update the data with a somewhat larger sample.

The result is a smaller, less reliable sample than we’ll get in a few months. The RTCI is a data point but not a particularly reliable one at the moment.

The trend is a little clearer in the below table of 30 police departments with the most murders in 2025 that also had data through February 2026. Those agencies were down close to 30 percent in January (again, not predictive) and are down more than 20 percent so far in 2026.

It’s too early to say what this will mean for the national change in 2026, but it is suggestive that murder was down a good bit in the first two months of 2026. These cities will be predictive of the national change later in the year, so keeping tabs on them is a good way of evaluating the trend ones the leaves become more tea-like later in the year.

The Gun Violence Archive is another source pointing to the continued drop in gun violence in 2026. Shooting victims were down 12 percent nationally through February in the GVA and the data as of mid March suggests the drop is becoming less steep.

Finally, I built a shooting tracker that provides updated data on shooting victims and incidents from 25 agencies (GIVE’s 28 agencies counted as one). The tracker scrapes data daily from each of these sites and updates as best it can. This isn’t perfect, but it’s a good way of evaluating changing gun violence trends in an automated and hopefully easy to use fashion.

Anyhow, unsurprisingly, the shooting tracker finds shootings down nearly 20 percent in the places that have easily accessible (and scrapable) shooting data.

So, there we have it. We can start evaluating national crime trends in a more predictive fashion once May/June get here. In the meantime, we can say from the very early, very preliminary data that murder and gun violence are still down at the start of 2026 though there is some evidence that the decline is starting to slow. Stronger conclusions will have to wait a few months.

New on the Jeff-alytics Podcast

Ferne Pearlstein and Bob Edwards made a documentary more than a decade ago called “The Last Laugh” which talked about how comedy can tackle a tragedy as immense as the Holocaust and what lessons that might suggest for effectively communicating about crime. This is a bit of a leap for me creatively — as they might say in the biz — but it’s a conversation that I really wanted to have and one that I really enjoyed.

Apple

Spotify

Amazon

And while you’re here, be sure to check out these other recent great episodes:

Professor and researcher Ian Adams

Politics podcaster Galen Druke

Arnold Ventures Executive Vice President Jennifer Doleac

FBI Assistant Director Timothy Ferguson

Leave a comment

Talking to them and trying to tease out what lessons it might have for understanding and communicating about crime was a conversation that I really wanted to have and one that I really enjoyed. I think you will too.

Ferne Pearlstein is an award-winning producer, director, cinematographer, and editor, based in New York. She is a winner of the Sundance Cinematography Prize, and a member of the Documentary Branch of the Academy of Motion Pictures Arts & Sciences. In addition to her features SUMO EAST AND WEST and THE LAST LAUGH, which she produced, directed, shot, and edited, and both of which premiered at Tribeca Film Festival and aired nationally on PBS”s Independent Lens series, she is currently producing JACK WHITTEN: A COSMIC SOUL, directed by Yoruba Richen.

Robert Edwards is writer, filmmaker, and the author of Resisting the Right: How to Survive the Gathering Storm, published by OR Books, about how to resist right wing authoritarianism in the US. He wrote and directed the feature films LAND OF THE BLIND, starring Ralph Fiennes and Donald Sutherland, and WHEN I LIVE MY LIFE OVER AGAIN, (aka ONE MORE TIME) starring Christopher Walken and Amber Heard. Most recently he co-produced and co-directed the feature documentary DEATH & TAXES with Justin Schein, and writes The King’s Necktie, a weekly blog on politics.

ASK E. JEAN, director Ivy Meeropol’s feature documentary about E. Jean Carroll, which premiered at the 2025 Telluride Film Festival, and on which Ferne a writer, editor, and supervising producer, will be in theaters this spring.

Catch it here, below, or wherever you get your podcasts!

Apple

Spotify

Amazon

And while you’re here, be sure to check out these other recent great episodes:

Professor and researcher Ian Adams

Politics podcaster Galen Druke

Arnold Ventures Executive Vice President Jennifer Doleac

FBI Assistant Director Timothy Ferguson

Data from the FBI shows good news though — these types of incidents appear to be falling. The big story here isn’t just that these types of incidents are falling, but that crime data has advanced to be able to measure flash mob shopliftings (among scores of other offense types) as trends develop.

The first source that tipped me off to the trend was a report that the FBI published last year titled “Reported Flash Mob Shoplifting Incidents: 2020‒2024”.

These “flash mob” shoplifting incidents have become more common in recent years, and they receive disproportionate coverage in the media when they do occur. They are exceedingly rare though, with these types of incidents making up around 0.1% of all shoplifting incidents reported to the FBI between 2020 and 2024.

The number of flash mob shoplifting incidents reported to the FBI rose in 2024, but that’s largely a factor of improved reporting as agencies like NYPD and LAPD became NIBRS compliant. Among agencies that had consistent data from 2020 to 2024, however, flash mob shopliftings fell a bit in 2024.

The FBI report points to one factor that is notnot surprising: offending in these incidents skews younger. A plurality of offenders in these incidents are under 20 and around 65 percent of people arrested between 2020 and 2024 in a flash mob shoplifting were under 30 compared to around 36 percent of people arrested in all other shopliftings.

There are a few aspects of this report that are especially noteworthy. First, I’m not entirely sure why, but I chuckle at the idea of traveling back in time and telling J. Edgar Hoover that his FBI will one day be writing public reports about flash mob shopliftings.

Second, it’s noteworthy that the FBI is writing these reports for public consumption in the first place. This is one of more than a dozen special reports the FBI has released since the start of 2025 covering a large variety of topics and points to the Bureau’s efforts to better communicate crime data and trends.

Finally, and most importantly, everything in the report can be replicated by the public with a little gumption and know-how. Indeed, with the FBI reporting monthly data we can extend the findings of this report out to the end of 2025 and say with confidence that the downward trend is continuing. The ability to do this is pretty revolutionary in how we interact with crime data.

To do that, I grabbed the FBI’s NIBRS master files and, using their definition of flash mob shoplifting, took every incident between 2022 and 2025 of a flash mob shoplifting. I removed agencies (like LAPD) that weren’t reporting to NIBRS in 2022 to get a consistent reporting population. Doing this shows the number of these incidents peaked in 2023/204 and came down in 2025. Even if we assume that December’s figures are underreported, the trend of a peak in 2023/2024 and a steady decline holds.

I stopped the analysis in December 2025 because of fears of underreporting — the FBI has data available through February 2026. Shopliftings are almost certainly badly underreported to the police, but I’m guessing that more high profile shopliftings like these get reported with much more frequency.

So, good news that these types of offenses are falling, and even better news that we are able to measure and report out complex offense types like this in something approximating real time. The NIBRS transition was an enormous struggle and created confusion, misinformation, and an asterisk year. Now that we are 5 years removed from that transition, however, there is enormous improvement and the benefits of switching to NIBRS are plainly obvious.

New on the Jeff-alytics Podcast

Have you heard about the police department in Utah where report drafting AI interpreted footage from an officer’s body camera of The Princess and the Frog playing in the background of an incident to mean the officer had morphed into a frog? AI has come a long way in the last few years but it still isn’t perfect. Within AI is the potential for revolutionary disruption of traditional processes, but there is also the danger of relying too heavily on a tool that is only right most of the time for efforts that require perfection or near perfection.

For this conversation, I turned to Ian Adams. Ian is an assistant professor of criminology and criminal justice at the University of South Carolina. He has written extensively about AI and gives a terrific perspective on the role of AI in policing. Check it out below or wherever you get your podcasts!

Apple

Spotify

Amazon

And while you’re here, be sure to check out these other recent great episodes:

Politics podcaster Galen Druke

Arnold Ventures Executive Vice President Jennifer Doleac

FBI Assistant Director Timothy Ferguson

Leave a comment

AI has come a long way in the last few years but it still isn’t perfect. Within AI is the potential for revolutionary disruption of traditional processes, but there is also the danger of relying too heavily on a tool that is only right most of the time for efforts that require perfection or near perfection.

For this conversation, I turned to Ian Adams. Ian is an assistant professor of criminology and criminal justice at the University of South Carolina. Before taking his PhD in political science at the University of Utah, he was a police officer and police labor executive. His research is focused on policing, broadly construed, with a focus on behavior and technology.

Ian has also researched and written extensively about AI, and today’s conversation is all about the uses of AI in policing, the potential/actual pitfalls, and where this technology might be heading in the world of criminal justice.

If you’re interested in some extra credit work, two papers related to this topic you should check out are:

1. Adams, I. T., Barter, M., McLean, K., Boehme, H. M., & Geary, I. A. (2024). No man’s hand: Artificial intelligence does not improve police report writing speed. Journal of Experimental Criminology. https://doi.org/10.1007/s11292-024-09644-7

2. Adams, I. T., McLean, K., & Alpert, G. P. (2026). Improving police behavior through artificial intelligence: Pre-registered experimental results in two large US agencies. Criminology, 0(0), 1–15. https://doi.org/10.1111/1745-9125.70028

To get in touch or peruse peruse different papers/projects/dashboards, Ian’s website is ianadamsresearch.com.

Catch this great episode here, below, or wherever you get your podcasts!

Apple

Spotify

Amazon

And while you’re here, be sure to check out these other recent great episodes:

Politics podcaster Galen Druke

Arnold Ventures Executive Vice President Jennifer Doleac

FBI Assistant Director Timothy Ferguson

Orleans Parish District Attorney Jason Williams

Leave a comment

So, why do I think there were 14,000 murders in the United States last year? Well, first off, I’m not nearly as confident in the precise figure of 14,000 as I am in that general ballpark. I need a few hundred murders on either side of that figure as my margin of error because my guess is, in fact, a guess.

Three organizations - my organization’s Real-Time Crime Index, the Council on Criminal Justice and Major Cities Chiefs Association have released data from variously sized samples of mostly cities. Sampling cities is a tried and true method for understanding our trends, but there is some uncertainty that comes along with it. The RTCI has murder down 18.5 percent in 2025, CCJ has it at 21 percent, and MCCA has it in the middle at 19.3 percent.

One way to guesstimate last year’s murder total is to take the 2024 figure combine it with the expected percent change, and boom. The FBI reported 16,935 murders in 2024, and if we lower 2025’s total by 18.5, 19.3 or 21 percent then we get a range of 13,379 to 13,802 murders in 2025.

But, remember, the FBI will revise up 2024’s total and the decline (however large it is) will be relative to the revised total, not the current total. So, let’s assume that the FBI’s revised figure goes to a 17,575 which would be an increase of 640 — the 2023 total was revised upwards by 655 in 2024 and the 2022 total was revised upwards by 620 in 2023. A decline in the range of 18.5 to 21 percent puts the range from that revised total at 13,884 to 14,323.

The RTCI gives another possible way to address this. Going back to 2018 (and excluding 2021 due to the NIBRS transition), the current RTCI sample has represented 51.5 percent of murders estimated by the FBI to have occurred nationally. If you divide the 7,205 murders in the current sample in 2025 by 51.5 percent then you get…just about 13,978 murders in the US last year.

Now, this calculation is based on the non-revised 2024 murder figure and it also assumes that the current sample isn’t underreported (it probably is slightly), but the basic calculation once again puts the US murder count in 2025 at right around 14,000.

Finally, the FBI’s Crime Data Explorer has a solution as well!

There are 12,702 agencies covering nearly 81 percent of the US population that has reported 12 complete months of data to the FBI right now. We can assume that it’s something of an undercount, but it shouldn’t be a huge undercount given that final year data is due to the FBI in less than a month.

Those agencies have reported 11,824 murders so far in 2025 and they reported 14,031 through December 2024 — 82.9 percent of the FBI’s currently estimated 16,935 murders for that year. Now, we don’t know exactly how the other agencies will shake out, so there’s just as much uncertainty with this methodology as with the first two I used. But if we divide 11,824 by .829 we get…14,271 murders in the US in 2025.

None of these methodologies are particularly sophisticated and there should be a good margin of error attached to either side of my guess of 14,000 murders. If the FBI reports 13,800 or 14,350 murders in 2025 then I’m going to call this guessing a roaring success!

More important to me, however, is that this exercise shows just how far things have come as far as understanding our crime trends as they develop.

Gone are the days of creating a Google sheet to track murder in 25 cities that publish data and making a wild guess. Will this prediction come true? We’ll find out when the FBI publishes their 2025 estimates in a few months, but we’ll really find out in summer 2027 when the 2025 figures are likely to be revised upwards.

Note: I rewrote the top paragraph of this piece a bit because it didn’t properly capture the seriousness of the subject matter.

New on the Jeff-alytics Podcast

I talked with Galen Druke, founder of GDPOLITICS.com, about the potential role that the nation’s crime trends might play in the upcoming midterm elections. This episode’s mission was simple: do the most terrifying thing imaginable for your intrepid podcast host and try to talk about the role of politics on how we talk about and perceive crime.

Check it out below or wherever you get your podcasts!

Apple

Spotify

Amazon

And while you’re here, be sure to check out these other recent great episodes:

Arnold Ventures Executive Vice President Jennifer Doleac

FBI Assistant Director Timothy Ferguson

Orleans Parish District Attorney Jason Williams

Share

There’s a national election coming up in November and we discuss whether the nation’s crime trends will play a role or if the fact that crime is dropping will make it a political non-issue for the midterms. The discussion also touches on the potential for bipartisan collaboration in criminal justice reform and the evolving landscape of political communication. I don’t delve into politics often so it was good to have a knowledgeable guide like Galen for this terrific conversation.

Galen Druke is the host of the GD POLITICS podcast, which aims to explain politics and the world with curiosity, rigor and a sense of humor. He previously covered elections for 10 years at FiveThirtyEight, where he hosted their popular politics podcast.

Catch this great episode here, below, or wherever you get your podcasts!

Apple

Spotify

Amazon

And while you’re here, be sure to check out these other recent great episodes:

Arnold Ventures Executive Vice President Jennifer Doleac

FBI Assistant Director Timothy Ferguson

Orleans Parish District Attorney Jason Williams

Leave a comment

The clearance rate data for 2025 is still very preliminary, but the monthly data is a pretty good predictor of where things will turn out in the formal Crime in the US release. Right now, nearly 36 percent of robberies that have been reported to the FBI for 2025 have been cleared which, if it holds up, would be the highest national clearance rate for robberies since 1965.

For context, my Saints went 6-11 in 2025 which works out to a nearly identical 35.3 percent winning percentage. So, robberies may have been cleared at the highest rate in 60 years last year, but it still wasn’t great!

Robberies do, however, stand out among other crimes with respect to their clearance rates relative to pre-COVID levels. Clearance rates for most crime types fell enormously in 2020 and 2021 but have returned to their historical norms. Robberies, by contrast, are far exceeding previous levels. Indeed, the robbery clearance rate in 2025 was likely the highest since the 1960s when clearance rate stats were notoriously unreliable.

There are 768 agencies that fit the Real-Time Crime Index inclusion criteria (cities of 50,000 or more, counties of 100,000 or more) that have complete data for both 2023 and 2025 in the Crime Data Explorer. The data for 2025 is still preliminary, so this might be an overstatement, but 619 (80 percent) of those 768 agencies had fewer robberies in 2025 than 2023 while 70 percent of them had a higher robbery clearance rate in 2025 than in 2023.

These figures are subject to change, but only 3 of the 95 agencies that reported 250 or more robberies in 2023 reported more robberies in 2025.

I thought of and largely ruled out a few possible explanations for this change in robbery clearance rates.

A Change in Reporting.

It’s possible that some definitional or reporting system change is causing this increase. The rape clearance rate fell from around 50 percent in 1990 and 2000 to around 30 percent in 2020 and 2025, but that was almost certainly due to a definitional change in how rapes were reported starting in 2013/2014.

Robbery has definitionally changed from a violent crime under the Summary Reporting System to a property crime under the National Incident-Based Reporting System (my understanding of the reason for this change is that classifying robberies as a property crime enables police departments to capture data on what property was stolen). This does mean that the number of robbery incidents rather than the number of victims — like in other violent crimes — gets counted by police departments.

This shouldn’t have any impact on how clearances get counted, but it did come to mind as maybe a possible explanation. It would be easier to buy as an explanation if the robbery clearance rate had started rising when the NIBRS switch occurred, but it fell through 2022 and then started rising.

A Quirk of Monthly Reporting

The FBI’s switch to monthly reporting is terrific, but it comes with the quirk that some agencies don’t report all their data on a monthly cadence. In recent years (since the NIBRS switch) that has generally meant that December clearance rates are lower than other months, but that isn’t the case in the 2025 data. Still, this is the first time we have had preliminary monthly data and it's possible that the grain of salt that we're currently taking needs to be larger, only time will tell.

Agencies still have time to report data, but the effect on the overall annual cc, clearance rate will be minimal even if the same pattern holds and December falls below the rest of the year’s rate. If you add 2,500 offenses to December’s robbery total and don’t add a single clearance then the national robbery rate falls from 35.8 percent to 35.1 percent, still extremely high by historical standards.

A Weird Agency

There’s no Coffee County in the 2025 robbery data as far as I can tell. Every month from nearly every agency is higher than it has been.

A New York/Chicago Thing

New York City and Chicago can do screwy things to national clearance rates. They have thousands of robberies (and other crimes), but don’t always report clearances. Chicago reported crime data but not clearances every year from 1995 to 2020. New York missed 2003 to 2012.

The good news is that both agencies are reporting via NIBRS now so we’ve got their data now.

The other good news is that they sort of cancel each other out in terms of the impact of non-participation on the national clearance rate. New York City reported a 59.9 percent robbery clearance rate in 2025, a very large and impressive improvement from a 46 percent robbery clearance rate in 2024.

Chicago also improved! To the highest reported rate in decades! But it was only from 5 percent in 2024 to 8 percent in 2025 and, as the chart above shows, Chicago is missing many years of clearance rate data.

Both cities have been included in the national clearance rate count for several years now and that certainly doesn’t explain the massive jump pretty much everywhere, so the answer to this question is probably not just a Chicago/New York thing.

Fewer Crimes = Higher Rate

This one likely has some truth to it. In general, less crime does typically lead to higher clearance rates, and 2025 likely had the lowest robbery rate ever recorded in the US (data available back to 1960).

But the connection isn’t ironclad and it isn’t like fewer crime inherently means a higher clearance rate in an agency. When we compare how robbery clearance rates have changed compared to the actual number of robberies in the 95 agencies that had 250+ robberies in 2023, we see maybe some relationship but not an incredibly strong one at the agency level.

And on a national level, decreasing robberies has never been strongly connected with increasing robbery clearances. The last time the US robbery rate fell by a similar amount (from 95.4 robberies per 100k in 2017 to 84.9 per 100k in 2018), the national robbery clearance rate reported by the FBI only rose from 29.7 percent to 30.4 percent.

Police Are Solving More Robberies

As Sherlock Holmes said, "When you have eliminated the impossible, whatever remains, however improbable, must be the truth.”

Nearly 40 percent (38 of the 92) that had fewer robberies in 2023 than 2025 actually had more robberies reported as cleared.

Clearances can be made by exception, but that is rarely the case in robberies. Take Arizona, for example, which provides an easy portal for breaking down clearances by type. The share of clearances by arrest jumped up from 2023 to 2025 while the other type of exceptional clearances barely budged.

Taking another look at that graph of robbery clearance rates over time but focusing just on the last 25 years shows a steady increase since 2000 that was interrupted by the pandemic.

Perhaps whatever was driving that increase prior to 2020 — be it technology like DNA, cameras, new investigative techniques, more resources, or all of the above — is returning to the pre-existing trend. Perhaps surging robbery clearance rates speaks to larger issues of community trust in law enforcement being lost in 2020/2021 before being regained in the last few years.

Whatever the cause, it’s a good thing to see. But, much like my Saints going from 5 to 6 wins, improving robbery clearance rates from 30 percent to 35 percent highlights just how much work remains to be done on this issue as a nation.

New On The Jeff-alytics Podcast

This week’s I talked with Jennifer Doleac, Executive Vice President of Criminal Justice at Arnold Ventures, about how data and research can dramatically improve outcomes in the criminal justice system. Jen has a new book out called The Science of Second Chances and we talk through what the book tells us about how policymakers (and everyone really) can build a better approach to reducing. Give it a listen below or wherever you get your podcasts!

Apple

Spotify

Amazon

And, while you’re here, be sure to check out these other great recent episodes of the pod!

FBI Assistant Director Timothy Ferguson.

Orleans Parish District Attorney Jason Williams.

Jens Ludwig from the Chicago Crime Lab.

Leave a comment

Jen is one of the most thoughtful scholars on how we can solve criminal justice problems with research-backed policies, and this is a conversation I’ve been looking forward to having ever since launching the podcast.

We talk about her transition from academia to Arnold Ventures and the importance of evidence-based policy in criminal justice reform. We also discuss the need for leniency in handling first-time offenders, the significance of swift and certain consequences in deterring crime, and the role of DNA technology in solving crimes. It’s a great conversation about following research to improve policy decisions.

Jennifer Doleac is an economist and the Executive Vice President of Criminal Justice at Arnold Ventures, a philanthropy focused on evidence-based policy in the US. She is the author of THE SCIENCE OF SECOND CHANCES, and the host of Probable Causation, a podcast about research related to crime and criminal justice policy. She is a leading expert on the economics of crime, and a vocal proponent of using rigorous research to inform policy.

Give it a listen here on Substack, below, or wherever you get your podcasts!

Apple

Spotify

Amazon

And, while you’re here, be sure to check out these other great recent episodes of the pod!

FBI Assistant Director Timothy Ferguson

Orleans Parish District Attorney Jason Williams

Jens Ludwig from the Chicago Crime Lab

Leave a comment

The issue became apparent when looking at burglary offenses and clearances in the Crime Data Explorer. You can see the monthly counts of each in the below graphs (note the different scales along the Y-axes).

Can you spot the problem? It pops out a little bit more when comparing the clearance rate (clearances divided by offenses) of burglaries in the US per month over time.

There are big jumps in reporting in December each year from 2017 through 2020 which is a feature of many agencies (especially Florida) reporting all their data all at once. It only marginally impacts the rate though because they’re reporting both offeneses and clearances all at once. This issue gets ironed out as NIBRS came online in 2021 and agencies started reporting more consistently.

The real issue is in August 2019. There were 11,801 burglaries cleared out of 90,333 in July 2018 good for a clearance rate of 13 percent. In August, however, there were 19,349 clearances reported — 7,500 more than in July — compared to 90,019 offenses, a nearly 22 percent clearance rate).

The cause of this anomoly isn’t immediately obvious.

There was no change in reporting in places like New York City or Los Angeles or Chicago that can occassionally impact national numbers. There was not some enormous busted burglary ring or other cause leading to a one-time surge in police effectiveness in solving burglaries.

It took some digging, but eventually I figured out the cause: Coffee.

Actually, it was Coffee County, Georgia — population ~25,000. The Coffee County Sheriff’s Office reported around 3 burglaries cleared per month against nearly 17 burglaries per month between January and July 2019 — good for a 16 percent clearance rate which was above the national clearance rate of 14 percent that year.

In August, however, Coffee County reported 7,261 burglaries cleared versus 17 offenses, a 42,711 percent clearance rate which was well above average that year. That one county in Georgia was responsible for around 38 percent of all burglary clearances reported to the FBI in August 2019.

The problem wasn’t just with burglaries. Coffee County reported 5,000 motor vehicle theft clearances in September 2019 compared to just 1 such offenses which works out to a workmanlike 500,000 percent motor vehicle theft clearance rate that month. The below graph of burglary and motor vehicle theft clearances per month from 2017 to 2024 in Coffee County shows the absurdity of those two datapoints.

Removing Coffee County’s data from the national totals would lower the national burglary clearance rate in 2019 from 14.3 percent to 13.6 percent. The national motor vehicle theft clearance rate would fall from 13.5 percent to 12.8 perent.

It’s impossible to say what caused these errors, but they aren’t the only clearance errors in the data. In June 2015, the town of Blairsville (that’s in Georgia, population around 600) reported 2 thefts and 3,400 theft clearances — a 170,000 percent theft clearance rate.

Calhoun County (also in Georgia) reported 1,002 aggravated assaults cleared in December 1990 compared to 0 actual offenses. West Bloomfield Township, Michigan reported 1 murder and 19 murders cleared in March 1990.

The most egregious error in historical clearance rate data comes from the Oconee County Sheriff’s Office (this will shock you, but also in Georgia) which in July 2013 reported 3 motor vehicle theft offenses and 10,000 cleared — a 333,333 percent clearance rate. Then they reported 1 motor vehicle theft offnese in October 2013 versus 2,000 cleared. Oconee’s error was nearly entirely responsible for the national motor vehicle theft clearance rate rising from 11.9 percent in 2012 to 14.2 percent in 2013 — it fell back to 12.8 percent in 2014.

There aren’t any incredibly obvious errors in the most recent 2024 data though Rochester Hills, Michigan (city of about 75,000) is interesting. The local police reported a big surge in burglaries in 2023 with 108 reported between September and December compared to just 26 over the first 8 months of the year. Most of these burglaries occurred at rental storage facilities per the state’s crime dashboard, and many were then solved in 2024 with 58 cleared in April and another 20 in September. The result is that Rochester Hills reported a 48 percent burlgary clearance rate in 2022, it fell to 9.7 percent in 2023, but then erupted to 233 percent in 2024.

Some editions of this newsletter are dedicated to intense analyses of complex datasets which hopefully move the ball forward on understanding national crime and policing trends. Other pieces are quirky looks at the always-present flaws in our crime data. This one is much more of the latter than the former.

Crime data should be always taken with a grain of salt, and hopefully this piece has driven home just how much salt to take when digesting your clearance rate data.

New On The Podcast

This week’s I talked with FBI Assistant Director Tim Ferguson about how the FBI produces the crime data that makes this newsletter possible. It’s a rare, inside discussion about the people behind the Uniform Crime Report. Give it a listen below or wherever you get your podcasts!

Spotify

Apple

Amazon

Also check out these recent great guests!

Jason Williams on balanacing fairness and justice in New Orleans.

Jens Ludwig on reducing gun violence in America.

Jerry Ratcliffe on bringing science to policing.

Leave a comment

There have been massive advances in how the FBI collects and publishes crime data over the last few years and today I am talking with the man in charge of that operation: FBI Assistant Director Tim Ferguson. Tim runs the FBI’s Criminal Justice Information Services Division which, among other things, runs the FBI’s Uniform Crime Report. CJIS has made dramatic leaps in producing fast, accurate crime data that makes it possible to better understand our crime trends in near real-time.

Tim shares his unexpected journey from a small farming town in Illinois to a leadership role within the FBI, detailing his diverse experiences in law enforcement, including violent crime and narcotics investigations. He emphasizes the importance of stepping outside one’s comfort zone and embracing new challenges, which ultimately led him to his current position at CJIS, where he oversees a vast array of services and technologies that support law enforcement across the nation.

This is a special, inside look at how the crime data sausage gets made.

Check it out here, below, or wherever you get your podcasts!

Spotify

Apple

Amazon

And be sure to check out these other recent great episodes of the Jeff-alytics podcast!

Jason Williams

Jens Ludwig

Jerry Ratcliffe

I’ve written before about how 2025’s murder rate might be the lowest ever recorded, but CCJ’s report drives home how vague the term “ever recorded” can be. There’s a good bit of appropriate analytic uncertainty in the CCJ report that got lost on the other side of the report.

The White House picked up on the CCJ report (sans the estimative caveats), saying “A new report shows the murder rate across the nation’s biggest cities plummeted last year to its lowest level since at least 1900, marking the biggest one-year drop in recorded history.”

They weren't alone in removing uncertainty from the report. For example, CNN also stripped out the caveat and nuance in this chyron.

Murder did really fall a ton last year and it fell at the fastest rate ever recorded to likely the lowest murder rate ever recorded. But we don’t know what the murder rate in 2025 was, and that raises an important question: Can we say when the last time that murder was this low in the United States?

The first step towards answering that question is to guesstimate what the nation’s murder rate was in 2025 as a starting point for comparing other years. An exact murder rate is unknowable at this point, and it’s hard to say from current data (hence the caveats in my and CCJ’s analyses).

We do know that the lowest murder rate ever recorded since the FBI started using a consistent methodology is 4.44 per 100k in 2014. Anything below that would be the lowest ever recorded.

I would ballpark the 2025 rate at somewhere around 4.2 (+/- 0.2) per 100k and feel reasonably confident that the FBI’s 2025 crime report will put it around there when released later this year. Where did I come up with that range?

Well, the FBI estimated a murder rate of around 5 per 100k in 2024, but the FBI’s 2024 estimate is almost certainly lowballing the murder rate. I have talked about the FBI’s revision process a lot, but the bottom line is that 2024’s murder estimate is probably too low by around 700 or so murders.

This will get updated when the 2025 estimates are released sometime this Fall, and that’s important becuase when we talk about a 20 percent drop in murder in 2025 we are referencing a 20 percent decline relative to the FBI’s revised murder rate, not the current one that they released last year.

My guess is that the FBI will revise up 2024’s murder rate to around 5.2 per 100,000 based on their recent revision pattern of increasing the previous year’s count up by 700 or so. The FBI initially reported 19,252 murders in 2023 when they initially reported it but increased that count to 19,902 in 2023 when 2024 figures were released in 2025.

So let’s assume as similar revision for 2024 and count down from there as our starting point .

From there, let’s say murder fell 18 percent in 2025. That’s conservative but not overly so. The Real-Time Crime Index has murder down 19.8 percent through November but we have to account for underreporting potentially making the decline slightly smaller as time goes on and agencies complete their reporting. Additionally, the FBI will revise up their initial 2025 estimate — released August to October 2026 — when they release 2026 data in 2027.

All of which goes to say that I expect the eventually be slightly smaller than the enormous one we have right now though it should stay historically large.

Doing all the math with an 18 percent drop in 2025 to a slightly revised up 2024 rate, therefore, gives us a murder rate of 4.2 per 100,000 last year. I won’t be shocked if it’s a few points on either side of that though, but it’s a good strating point.

With last year’s murder rate guesstimated we can start to try and answer the question: when was murder this low?

The FBI’s national estimates have been largely standardized since around 1960 though the quality of some of that data wasn’t super great in the 60s. Still, I’ve already shown that a murder rate of 4.2 per 100,000 would be the lowest ever formally recorded by the FBI going back to 1960.

The modernity of the FBI’s national estimates mostly dates back to their 1958 report which streamlined the offenses being reported and created true national estimates for the first time. We can guesstimate murder rates before that, but I don’t have much confidence in them and neither should you.

In 1954, the FBI reported 6,850 murders and a murder rate of 4.2 per 100,000 based on data from around 75 percent of the population including nearly 90 percent of cities.

Murder was roughly even in 1955 per the FBI, rose around 1.8 percent in 1956, and fell very slightly in 1957.

If you use the FBI’s estimates from 1957, before they changed up their estimative procedure in 1958, there were 6,920 murders in the US. Doing back-of-the-napkin math on that figure would put the national murder rate that year at right around 4 per 100,000 based on reporting from 83 percent of the population and estimates from non-reporting arears.

When the FBI went back with a new methodology in 1958, however, they looked at data from 1957 again and estimated that the actual 1957 murder rate was at 4.7 per 100,000 with more than 8,000 murders now estimated for 1957.

So that 1954 estimate of 4.2 per 100,000 — which used the same methodology as the original 1957 estimate — is probably not super reliable.

There are a few other years where the national murder rate would at or be slightly lower than 2025’s 4.2 per 100,000, but the reliability of data for those years is extremely low.

Indeed, I think we should ignore all murder estimates prior to 1958. If we want to go by just murders from the FBI then we can go back to the early 1930s with estimates that are increasingly less reliable the farther we go back.

I think the safest thing that we can say is that the 2025 murder rate was likely the lowest ever recorded. Three years — 1954 to 1956 — stand out as possibly equal or lower, but the FBI’s methodology those years wasn’t nearly as reliable as it was a few years later.

But what about the data back to 1900?

The US did not formally start systemically measuring murder as a crime until 1930 and there aren’t reliable national estimates until 1960ish. It’s not correct to say anything about the US murder rate prior to 1900 — which isn’t what CCJ is doing.

The CDC has death rate data by cause back to 1900 which you can access (it’s a bit obnoxious) here for data on the national homicide rate that far back. The problem, of course, is that homicides and murders are not synonymous. Every murder is a homicide but not every homicide is a murder.

The FBI's murder rate and CDC's homicide rate almost always move in the same direction with the homicide rate usually a small bit higher than the murder rate. The difference has gotten slightly larger in recent years (and the 2021 FBI estimates are still very flawed), but the trends are identical.

The CDC has relatively complete homicide data back to the 1930s, but reporting was incomplete prior to that. Moreover, research has shown that it was incomplete with a bias towards low homicide states artificially lowering the reported homicide rate before 1930.

The US homicide rate in the early 1900s is reported at right around 1 per 100,000, but the actual rate was probably closer to 7 or 8 per 100,000 according to researchers who have estimated the early 20th century rate. That data was used in the CCJ report, and by other homicide scholars and is generally considered far more reliable than the official data.

It’s hard to say exactly where 2025’s homicide rate from the CDC will end up. Data through midyear points to a bit more than a 15 percent drop in the national homicide rate in 2025 which would come out to a rate of just around 5 per 100,000. That would be roughly tied with 2014 (4.98 per 100k) as the lowest national homicide rate since the 1950s when it reached 4.5 per 100k in 1955 and 1957.

So what can we say with confidence?

2025 featured the lowest murder rate ever recorded, but our records on murder only really start in 1960.

The drop in 2025 was the largest one year drop ever recorded and that’s true of both murder and homicide. The CDC’s homicide data back to 1900 puts the 2025 decline as the largest ever at 15 percent, supplanting a 14 percent drop in 2002 (because of 9/11) and 13ish percent drops in 1935 and 1945.

2025 was likely not the lowest homicide rate ever recorded but it was likely right around the lowest homicide rate in 70 or so years. The lowest homicide rate ever recorded was in 1903 (1.1 per 100,000) though that was derived from data from just 10 states (CT, IN, MA, ME, MI, NH, NJ, NY, RI & VT) plus DC and research shows it was a bad undercount.

In my opinion, we can’t really say when the US murder rate was this low prior to 1960. The FBI recorded a similarly low murder rate in 1954 but that was a flawed figure that is inconsistent with our current estimative methodology.

It’s plausible that we would have had a lower murder rate at some point in the 1950s with a consistent methodology extended to then. It’s also plausible that 2025 would have had the lowest murder rate of any year since before the turn of 20th Century with a consistent methdology.

As is frequently the case with flawed crime data, we just can’t say for sure.

New On The Jeff-alytics Podcast

This week’s I talked with Orleans Parish District Attorney Jason Williams about the challenges of balancing crime reduction with justice in a place that has all to often had neither. We also talk about how the innovative “N.O.D.I.C.E” program addresses the geographic footprint of crime in New Orleans. Jason emphasizes the importance of data-driven decision-making, community engagement, and the myriad of challenges stemming from non-unanimous juries in Louisiana. Give it a listen below or wherever you get your podcasts!

Spotify

Apple

Amazon

Also check out these recent great guests!

Jens Ludwig on reducing gun violence in America.

Jerry Ratcliffe on bringing science to policing.

Dara Lind on crime and immigration.

Leave a comment

Jason led the New Orleans City Council’s Criminal Justice Committee when I was a public safety consultant for the council, so it was a great joy to get to talk to him about his work in New Orleans as he approaches the end of his first term.

In this conversation, we discuss the challenges of balancing crime reduction with justice and how the innovative “N.O.D.I.C.E” program addresses the geographic footprint of crime in New Orleans. Jason emphasizes the importance of data-driven decision-making, community engagement, and the myriad of challenges stemming from non-unanimous juries in Louisiana.

The Honorable Jason Rogers Williams was sworn in as District Attorney of Orleans Parish on January 11, 2021. He is a proud Tulane University Law School Alumni. After graduating law school, he started his own law practice focused on criminal defense, Jason Rogers Williams and Associates. After winning a series of high-profile criminal cases, the Louisiana Supreme Court appointed Jason as a judge in New Orleans’ Criminal District Court. In 2014, he saw an opportunity to make government work for the people of New Orleans and ran for City Council. He was a council member for six years, four of which he served as Council President, before being elected District Attorney on a promise of sweeping justice reforms. Jason is a proud and devoted father to three children: 5-year-old Xavier, 17-year-old Graham, and 23-year-old Zoe. He lives in uptown New Orleans with his wife, Liz, founder of the Center of Resilience, his youngest son, and a very sweet but very bossy pit bull, Jean Louise.  

Spotify

Apple

Amazon

If you like the show, feel free to like, subscribe, and leave a glowing review. And be sure to give recent episodes a listen!

Jens Ludwig

Jerry Ratcliffe

Dara Lind

It’s a good group to present in front of, and, having done this presentation every year for most of the last decade, I’ve enjoyed reflecting on just how much the story being told in my presentation has changed over the last decade or so of presenting. From historic lows in murder in New Orleans in 2019 to skyrocketing crime and response times in 2022 and finally back to historic lows in murder and improved response times in 2025.

There is perhaps no crime type that better illustrates the ups and downs over the last decade than carjackings. They below graph of carjackings in New Orleans rolling over 12 months tells quite a tale. Carjackings in New Orleans surged in 2015/2016, came down through 2019, skyrocketed between 2020 and 2022, and have fallen like a rock ever since.

To put this into a bit more context, there were an 53 carjackings in January 2022 and just 4 in January 2024 and 2025 in New Orleans. There were only 57 carjackings in New Orleans in all of 2025, only 4 more than the January 2022 total.

Carjackings are a bit tricky to measure because formal measurement is a product of NIBRS. There aren’t a ton of agencies that were publishing carjacking data prior to the big surge in 2020, so finding monthly data on carjackings going back a decade or more can be challenging. But the trend is very similar pretty much everywhere where we have data. There’s Chicago:

Washington, DC (didn’t start measuring until 2018):

And New York City:

That carjackings fell a ton in 2025 is evident in data from 34 large agencies that publish aggregated data (note that I don’t quite have full 2025 data for a few agencies so they are compared versus the same months with complete data in 2024).

And the drop is even wilder when comparing 2025 carjacking totals in those cities with 2022:

All of which raises the very good question: why is this remarkable trend occurring?

My answer is that I do not know. Carjackings don’t have high clearance rates and, while there is evidence that carjacking clearance rates improved in 2025 in places with such data, this likely mostly reflects fewer incidents rather than a change in policing leading to dramatically more arrests.

One possible answer in my opinion is that society — especially younger people — are healing from the shock that was COVID. Carjacking offenders tend to be younger than offenders of other crimes which is clear from the below arrestee data from a few big cities that publish it in sufficient detail through their state UCR programs.

Data on arrestees in Houston and Philadelphia — which admittedly is just a two city sample — suggests a massive drop in offending among young people with no real change in offending for people 25 and up.

I’m sure that the actual explanation to this positive trend is more than just “the kids are alright”, but something influenced younger people to commit tons of carjackings a few years ago and is no longer as big an influence. A stronger explanation is beyond the scope of this newsletter, but regardless of the cause, it’s nice to be able to write about a positive trend every once in a while.

New On The Jeff-alytics Podcast

Check out my most recent podcast episode with the Chicago Crime Lab’s Jens Ludwig. This is a great conversation about what causes a surprisingly high share of gun violence and what concrete steps cities can do to reduce it.

Spotify

Apple

Amazon

Leave a comment

Today, Jens Ludwig discusses his extensive research on gun violence, particularly focusing on youth violence and the role of the University of Chicago Crime Lab. He emphasizes the importance of using data and behavioral science to understand and address the root causes of gun violence.

Jens also critiques media representations of gun violence and shares insights from his book 'Unforgiving Places,' which explores the complexities of gun violence in America. He offers practical policy recommendations for reducing gun violence, highlighting the need for community engagement and innovative intervention programs.

Jens Ludwig is the Edwin A. and Betty L. Bergman Distinguished Service Professor at the University of Chicago and Pritzker director of the University of Chicago Crime Lab. He is an elected member of the National Academy of Medicine, and author of “Unforgiving Places: The Unexpected Origins of American Gun Violence” (University of Chicago Press, 2025).

Spotify

Apple

Amazon

If you like the show, feel free to like, subscribe, and leave a glowing review. And be sure to give recent episodes a listen!

Jerry Ratcliffe

Dara Lind

William Reynolds

The main idea was that murder is not a great statistic because there is a lot of luck and randomness that determines the share of shooting victims who die in any given year in any given city. It’s far better, therefore, to track shootings in order to determine whether a change in murder is due to a change in luck or an actual change in gun violence.

It’s with that in mind that my interest gets piqued when I run across a tweet like the one below.

There are a few layers to evaluate here.

The one that most immediately comes to mind is the idea that the national murder decline today is driven by improvements in emergency trauma care rather than declines in shootings. That’s a relatively easily debunked idea though I see it pop up from time to time on social media. Here are shootings rolling over 12 months from the Gun Violence Archive:

Line went up, line has very clearly been going down for the last few years.

There is absolutely evidence that advances in medical science are improving the survivability of gun violence, but cities with shooting data don’t show any real change in the share of shooting victims who die over the last 15 or so years (GIVE description and data here).

New Orleans has improved a little bit, which you can absolutely argue is a testament to improvements in trauma care, but the actual driver of the city’s reduction in murder — from 266 in 2022 to 121 last year — has been a dramatic drop in shootings.

It is hard to say exactly why shooting survivability has not changed much over the last 15 years despite massive improvements in medical science. The most plausible answer in my opinion comes from a terrific piece by Justin George — then of the Baltimore Sun — that looked at how the deadlier firearms and ammunition make Baltimore particularly dangerous.

One can convincingly argue that the proliferation of deadlier firearms and ammunition has offset some of the gains of emergency trauma care this century. In 2014, for example, 14 percent of “crime guns”” traced by the ATF were associated with .22 caliber weapons and 19 percent were associated with larger, and presumably more deadly, 9mm ammunition — the two most common caliber types assoiated with firearm traces. By 2023 (the ATF didn’t release 2024 data), the share of .22 caliber had fallen to 7 percent while the share of .9mm had risen to 46 percent.

So, my theory is that medical science has improved this century but those improvements have likely been offset by changes in firearm availability.

But that’s not exactly what the tweet that inspired this piece is getting at.

The tweet in question is quoting a tweet that references a 2006 artile titled “Medical advances mask epidemic of violence by cutting murder rate.” That article is about a 2002 study titled “Murder and Medicine: The Lethality of Criminal Assault 1960-1999”.

The 2002 study contends that advances in medical techniques between 1960 and the end of 1990s dramatically reduced the lethality of aggravated assaults and lowered the US murder rate far below where it would be without those advances. The argument goes, therefore, that the United States reaching roughly the same murder rate in 2025 as it had in 1960 is not all that impressive because the country has much more survived violence.

I don’t totally buy it.

My questions about the linked study start early. From the study: “Starting from the basic view that homicides are aggravated assaults with the outcome of the victim’s death, we assembled evidence from national data sources to show that the principal explanation of the downward trend in lethality involves parallel developments in medical technology and related medical support services that havesuppressed the homicide rate compared to what it would be had such progress not been made.”

But this is not true. Homicides are not aggravated assaults with the outcome of the victim’s death. The FBI defines aggravated assault as “an unlawful attack by one person upon another for the purpose of inflicting severe or aggravated bodily injury. The UCR Program further specifies that this type of assault is usually accompanied by the use of a weapon or by other means likely to produce death or great bodily harm. Attempted aggravated assault that involves the display of—or threat to use—a gun, knife, or other weapon is included in this crime category because serious personal injury would likely result if the assault were completed.”

If you shoot me with a gun and I survive then it’s an aggravated assault. If you threaten to beat me with a baseball bat then it could be classified as an aggravated assault. Those aren’t the same thing though.

The paper uses homicides and aggravated assaults to create a “lethality rate”. This is the number of homicides divided by the number of homicides plus aggravated assaults in a given year. The implication is that falling lethality sine the 1960s is responsible for the murder rate not being substantially higher than it ended up being.

This paper was written in 2002 but we can pretty easily extend the “lethality” calculation out towards today. Doing so shows basically no change between the late 1980s and today in “lethality” with a slight rise this century.

The paper also breaks down “lethality” by weapon type to show a pretty clear across-the-board drop between the 1960s and 1999.

Here is the rub though: it’s impossible to say how much of the drop in “lethality” since the 1960s is real and how much is a construct of poor reporting way back when.

I wrote about this issue in 2024, but the 1960s and part of the 1970s were a Wild West of crime data reporting practices. As a DOJ OJP report from 1999 notes, only 10 states had state programs submitting data to the FBI as of 1970 and it wasn’t until the late 1970s that 40+ states were submitting data.

What this means in practice is that what got reported was what was decided by individual agencies with no training, standardization, or oversight from state agencies ensuring consistency between agencies. Aggravated assaults by firearm were not even tracked until 1964 and it wasn’t until the late 1970s that at least 80 percent of the country was covered by an agency reporting at least one firearm assault.

That aggravated assaults were likely underreported by police throughout the 1970s and 1980s (and, by logic, almost certainly the 1960s) is a central tenet of a 2015 paper from Janet Lauritsen. She argues that “Although long-term trends in robbery, burglary and motor vehicle theft in the UCR and NCVS are similar, this is not the case for rape, aggravated assault, or a summary measure of serious violence. NCVS trends in serious violence are more highly correlated with homicide data than are UCR trends suggesting that the NCVS is a more valid indicator of long-term trends in violence for crimes other than robbery.”

The below graph pretty clearly proves her point:

So, getting back to the point of this piece — are improvements in trauma care responsible for falling murder in America right now?

We can show pretty convincingly with shooting data that murder has fallen the last years because shootings have fallen dramatically — not major improvements in trauma care. We can also show that shootings have not become less lethal over the last 10 to 15 years suggesting that any gains from better trauma care — which have undoubtedly occurred — may have been offset by changes in firearm availability.

That is about as far as we can go with confidence though because shootings were not captured as a statistic until recently. New York City goes back to the early 1990s and that’s about as far back as any agency goes. Chicago goes back to 2010, Philadelphia goes back to 2015 — at least publicly.

In New York there isn’t any relationship between the “lethality rate” and the share of shootings that end in a fatality over the last two decades which is what the former is supposed to be measuring.

Aggravated assaults are simply too poorly reported historially to draw firm conclusions about the long term. I would guess that improvements in trauma care have likely helped make shootings more survivable today than they were 50 to 60 years ago. But a lack of standardization of reporting practices among law enforcement agencies in the 1960s and 1970s and little to no data on shootings until relatively recently lay doubt into whether more people actually survived shootings in the 1960s and 1970s than today.

My takeaway — as is frequently the case — is that people should avoid reading strong conclusions into crime data that can be explained by changes in reporting practices. Murder is falling now because fewer people are getting shot. Going farther back to evaluate the impact of improved trauama care relative to the 1960s is, in my opinion, not advised.

New In The Jeff-alytics Podcast

I talked with criminologist Jerry Ratcliffe about how to use science to improve policing. We talk about the history of using research in policing, why you should be very careful about Year-to-Date comparisons, and much more! Check it out below or wherever you get your podcasts! Also be sure to check out his podcast Reducing Crime!

Apple

Spotify

Amazon

Leave a comment

This is a fun conversation with one of the most engaging and thoughtful voices out there on how we can use research and science to improve policing outcomes. Jerry also hosts a terrific podcast of his own called Reducing Crime so be sure to check that out!

Jerry Ratcliffe is a former British police officer, college professor, and host of the Reducing Crime podcast. He works with police agencies around the world on crime reduction, leadership, and criminal intelligence strategy. ​After an ice-climbing accident ended a decade-long career with London’s Metropolitan Police, he earned a first class honors degree and a PhD from the University of Nottingham. He has published over 100 research articles and eleven books, including “Reducing Crime: A Companion for Police Leaders” and “Evidence-Based Policing: The Basics”. Ratcliffe has been a research adviser to the FBI and the Philadelphia Police Commissioner, an instructor for the ATF intelligence academy, and he is a scientific advisor to the International Association of Chiefs of Police. He is a Professor of Practice and the Faculty Director of the Master of Applied Criminology and Police Leadership in the Department of Criminology at the University of Pennsylvania.

Check out this episode here on Substack or wherever you get your podcasts!

Spotify

Apple

Amazon

And be sure to check out some recent great episodes!

Dara Lind

Alex Piquero

William Reynolds

I often find lists like this are overly complicated, prone towards conflating data reporting errors with “safety”, and fail to adequately contextualize the data. This list from Wallet Hub, however, is of course no exception.

To begin with, the list purports to “determine where Americans can feel most secure — in more than one sense — WalletHub compared more than 180 cities across 41 key indicators of safety.”

Their formula is easy to understand:

In order to determine the “safest” cities, the analysis takes 41 metrics across three broad categories: Home & Community Safety (17 metrics, 60 points), Natural Disaster Risk (6 metrics, 20 points), and Financial Safety (18 metrics, 20 points). Each of those 41 metrics are graded on a 100-point scale with 100 being the best. Also, weights are added to each metric for good measure.

The result is a ranking of 182 cities from “safest” to…not.

A general list of sources is provided, but — as is usual with lists like this — the sources just tell us a bunch of organization they got data from with no easy way to check the math. They mention the FBI, for example, but don’t say what year of crime data they’re using and how they handle missing or incomplete data.

For me, the biggest blinking red light that something doesn’t add up in WalletHub’s rankings is that Hurricane Storm-Surge Risk Index Score carries the same weighting as the Hail Risk Index Score. Hail storms can be bad! But they are not city-altering.

New Orleans is among the most at-risk places for natural disasters, a risk every New Orleanian acknowledges every June through November. We haven’t gotten a second dog (or a cat) because we don’t have the space in the minivan to evacuate with another animal — that’s a pretty good indicator about the fear/risk of natural disasters there.

I know it seems like I’m here to tell you that my hometown of New Orleans is not the most dangerous place in America, but I am extraordinarily wary of any list that ranks New Orleans 71st out of 182nd in terms of natural disaster risk. Any ranking that puts Shreveport as a riskier place than New Orleans for natural disasters is immediately suspect.

Going through the Home & Community Safety metrics raises some questions. The heaviest weighted metrics are terrorist attack over the last 10 years and murder rate. It’s unclear to me why terrorist attacks — events that can be poorly defined, extraordinarily rare, and unlikely to be repeated in the same place — are weighted so heavily as a determinate of community safety. Is a city that had a terrorist attack in 2016 really less safe now?

Murder rate at least makes sense as a metric for evaluating “safety”, but any reliance on FBI crime data is going to have issues. It’s already dated, so if you want to rank cities by most/least dangerous in 2025 or 2026 you inherently are using 2024 data at best.

New Orleans is particularly problematic in this situation because New Orleans did not report data to the FBI in 2024. If you decide to use 2023 data for New Orleans then you are using a substantially higher murder rate (53 in 2023 vs 33 in 2024 vs even lower in 2025) which ignores the massive drop in murder there over the last 3 years. New Orleans had the 3rd highest murder rate of the cities ranked using 2023’s rate, but the city had the 9th highest using 2024’s rate.

Jackson, MS probably had the nation’s highest murder rate for any city of over 100k in 2024, but Jackson has not reported data to the FBI since 2019 when the murder rate was considerably lower. There are 15 cities in the rankings that didn’t have 2024 data or don’t have a dedicated police department reporting data (Columbia, MD, for example, is reported as part of the Howard County Sheriff’s Office).

Next is the rate of assaults. This seems straightforward, but whether this means aggravated or all assaults (simple + aggravated) is not said. Reviewing the data strongly implies that they mean aggravated assaults, but the list of safest cities doesn’t exactly match the FBI’s rates.

Pembroke Pines had the second lowest aggravated assault rate in 2024 but comes in at 4th lowest on WalletHub’s list. Savannah, Georgia and Worcester, MA had the lowest rates but aren’t ranked at all by WalletHub.

The other problem with ranking cities like this is that what gets counted as an aggravated vs simple assault can vary wildly by city. Virginia Beach reported more than 28 simple assault for every aggravated assault in 2024 while Oakland reported a nearly 1 to 1 ratio.

Memphis did indeed have 52 times the rate of aggravated assaults as Virginia Beach did in 2024, but Memphis only had 3 times the simple assault rate. This doesn’t mean that either city is inherently reporting more correctly, but it does drive home how problematic some of these numbers can be. There is likely a difference in how these two cities classify offenses that probably does not reflect the level of “safety”.

This is exactly why the FBI is so insistent that people fight the urge to rank cities. From the FBI:

“Each year when Crime in the United States is published, many entities—news media, tourism agencies, and other groups with an interest in crime in our Nation—use reported figures to compile rankings of cities and counties. These rankings, however, are merely a quick choice made by the data user; they provide no insight into the many variables that mold the crime in a particular town, city, county, state, region, or other jurisdiction. Consequently, these rankings lead to simplistic and/or incomplete analyses that often create misleading perceptions adversely affecting cities and counties, along with their residents.”

The WalletHub ranking also uses rape and theft as crime categories but apparently does not use robberies, burglaries or motor vehicle thefts for some unknown reason.

Moving on, law enforcement employees — not officers — are included but it’s not clear whether that is a sign of safety or lack thereof. A higher rate of officers is generally correlated with a higher violent crime rate. Dinging Irvine, CA for having the third lowest officers per capita (cities 100k+ in 2024) doesn’t make a ton of sense.

Three other Home & Community Safety metrics stand out as particularly egregious inclusions. The first is including hate crimes. If you’re familiar with my analyses, you may recall that I have thoughts about hate crimes data (expressed previously here, here, and here.) Reporting can be incredibly spotty and varied between departments and years, and there is never a way to know the extent to which more reported hate crimes reflects better community reporting to police, better reporting by police to the FBI, or more actual crimes.

Another flawed metric that stands out is Perception of Safety. This is defined as being “based on perceptions of visitors of Numbeo website in the past 3 years. If the value is 0, it means it is perceived as very low, and if the value is 100, it means it is perceived as very high.”

New Orleans receives millions of visitors every year, but Numbeo’s perceptions of crime in the city is based on 147 contributors who agreed to fill out a survey on crime presumably over the last 3 years. And this metric is given equal weight in determining how safe a city is with assaults and thefts.

Then there’s the Share of Sheltered Homeless which is only given half weight. What is the degree of “safety” provided by a city with a small, mostly unsheltered homeless population versus a city with a larger, mostly sheltered population?

I have issues with other metrics, but those are the ones that stuck out the most. I could dig deeper into the Financial Safety category, but it is just window dressing as the Home & Community Safety largely determines a city’s overall “Safety” ranking, especially in the 100 or so cities with the most crime.

Ultimately what we are left with is an impossibly complex web of flawed individual metrics that say little about these cities other than that the cities reporting more crime are considered less safe than the cities with more crime.

Lists like this drive home why I prefer to compare a city’s trends against itself rather than other cities. Is New Orleans “safer” than Memphis or Oakland or St Louis? That’s an arbitrary and impossible to answer question. Is New Orleans or Memphis or Oakland or St Louis getting safer? That one we can answer pretty confidently.

Which city is the safest or least safe in America is simply an unanswerable question and lists like this one use complexity of adding more metrics to obfuscate this fact in my humble opinion.

New In The Jeff-alytics Podcast

I talked with immigration policy expert Dara Lind about the connection (or lack thereof) between immigration and crime. We chat about what people tend to get wrong and the criticality of historical context for understanding the present moment. Check it out below or wherever you get your podcasts!

Apple

Spotify

Amazon

And be sure to check out other recent episodes:

Atlantic County Prosecutor William Reynolds

Former Bureau of Justice Statistics Director Alex Piquero

Leave a comment