Crime At Midyear: Sizable Declines in Violent and Property Crime Across a Large Sample So Far in 2024
Estimating crime trends as we turn to the second half of 2024.
There’s a scene I like in Michael Bay’s Pearl Harbor1 where Dan Akroyd’s senior naval intelligence officer is explaining the challenge of using Japanese signals intercepts to assess what the Japanese fleet is likely to do. Akroyd’s character summarizes the problem in one of my favorite the only good scenes in the movie.
Akroyd: “The intercepts have missing words and garbled lines, so to explain the decrypts, we have to try to interpret what we think they're trying to do.”
Skeptical Admiral: “Interpret? You mean guess.”
Supportive Admiral: “They use their informed intuition, sir.”
Akroyd: “We guess.”
That’s basically the world of analyzing crime data. If we had complete data from all 18,000+ agencies nationwide then we wouldn’t have to guess at the nation’s crime trends. But in the absence of that data we are forced to guess as to the trends using the available imperfect data and our informed intuition about historical norms.
We can use a sample of cities with available data to overcome the lack of complete data. This sample will not be precise in terms of describing national trends but it should be accurate within a few percentage points, and the larger the sample then the more accurate the guess.
To guesstimate crime trends for the first half of 2024 I was able to gather data from more than 190 cities with available data through at least May. In that group of cities, murder is down around 19 percent, violent crime is down 7 percent, and property crime is down over 10 percent compared to 2023.
A few things to note about this methodology:
This is a large sample. I gathered data on YTD murder, violent, and property crimes (based on Uniform Crime Report Part I definitions) from cities with data at least through the end of May. The sample covers around 60 million people which is roughly the same size as the actual attendance at 18,000 Atlanta Falcons home games. There are 38 states represented in the sample.
It is reasonably late in the year so a sample of this size should be reasonably predictive of the year-end trend. If you had data from every city of 250,000 or more through June (which we don’t) then your average miss on the national year-end change in murder, violent, and property crime would be about +/- 2 percent. There’s a healthy margin of error on any assessment of a sample like this at this point in the year, but the change found lies well outside that margin of error.
The data is not perfect. There’s no data on violent and property crimes available for Los Angeles, some places may be slight undercounts due to underreporting the most recent full month to their State UCR program, and some places don’t publish their public data to FBI standards (a handful of cities only publishes felony assault or grand larceny, for instance). Moreover, the property crime decline may be overstated given previously gathered evidence that the decline in auto thefts is strongest in larger cities which are obviously overrepresented here. So let’s add a healthy margin of error on either side to the declines. Doing so accounts for our uncertainty but doesn’t change the expected trends.
This is just a guess. It’s a strongly informed guess but it’s a guess that could plausibly be wrong (either because the sample doesn’t match the nation’s wider trend or because something happens over the next six months). The data also is not exact. I used incident-level data for some agencies which won’t mimic reported offenses with 100 percent accuracy. In addition, some agencies only report NIBRS data so I had to translate it to UCR Part I based on the offense types which adds inexactitude to the totals.
You can see the full sample in this spreadsheet with links to source data as best I could (note that these are all unofficial numbers and a snapshot in time, so they may differ from what’s available now). Murder and violent crime are down or even relative to 2023 in around 70 percent of the cities while property crime is down or even in around 75 percent of the cities.
That murder is down more in this sample than in the bigger YTD sample makes sense considering we don’t have Los Angeles crime data beyond murder (so LAPD isn’t included in the above 190 city sample). Including just Los Angeles murders (+2 percent as of late June) would bring down the sample’s change to -18.3 percent which is pretty close to what our big YTD sample is at right now.
This sample does not mean that murder will fall 20 percent while violent crime will fall 7 percent and property crime will fall 11 percent nationally. It does mean that murder is down a lot — historically so — while violent crime and property crime appear to be falling at healthy clips so far this year.
All of this matches to various degrees what shows up in our YTD murder sample, the (admittedly flawed) FBI quarterly data, Major Cities Chiefs Association data, and Gun Violence Archive data.`
Speaking of the big YTD sample, murder is now down around 18 percent in nearly 270 cities with available data — including over 230 cities with data through at least the end of May. Of note, the sample’s decline has clearly settled down from where it was 3 months ago when it was around -20 percent. A 20 percent decline in murder nationally would be immensely large (more than twice as large as any other recorded one year decline), and I mostly expected some regression towards the historical norm.
You can see the change in the YTD murder sample at the end of each of the last four months:
March -20.4 percent.
April - 20 percent.
May -18.6 percent.
June -18 percent.
The change is slight but it’s definitely there. That said, murder is still down at a historically large pace and one of the major questions for the second half of 2024 is whether it can eclipse 2023’s possibly historically rate of decline.
There is still plenty of time in the year, however, for these trends to regress towards a much smaller decline. Plus the data collection process is flawed enough to add a few percentage points in either direction to any assessment of the current trends.
All of which brings me back to an earlier exchange in that one good Pearl Harbor scene.
Skeptical Admiral: That's not exactly hard evidence, Captain Thurman.
Dan Akroyd: Well, Admiral, if I had hard evidence we'd already be at war.
If we had complete data from 18,000 agencies through May or June then we wouldn’t need to guess as to the nation’s crime trends. The choices are either to wait until October 2025 when the FBI’s 2024 data is formally released or to use large samples of cities (along with other sources) to produce an informed guess of the trends that should mimic the national estimates pretty closely.
I would argue that there’s value in evaluating the likely scope and direction of our crime trends even if the assessment carries a fair amount of uncertainty. There’s still time left in the year for these trends to change, but violent crime and property crime are most likely falling in the United States as of midyear with a potentially historic decline in murder on the table for the second year in a row.
I like it solely as a former intel analyst. The moviegoer in me doesn’t like anything from that movie.
I think you underestimate the quality of your guesses, and possibly, overestimate the quality of the move.
Thanks, as always, for the updated data aggregation, Jeff.