How 'Human Error' Likely Increased the Reported US Violent Crime Rate in 2023

How one "anomaly" changed the trajectory of reported violent crime in California and nationally.

Did violent crime rise or fall in California in 2023?

The FBI estimates that there were 203,749 violent crimes in California in 2023, up from 196,252 in 2022. That’s 7,497 more reported violent crimes last year, good for a 3.8 percent increase meaning that as long as the next sentence doesn’t start with ‘but’ this will be my shortest newsletter yet.

But it’s not quite that simple for one main reason:

Oakland.

The Oakland Police Department reported 15,372 violent crimes in 2023 which is a measly 140 percent increase from the 6,395 violent crimes reported by Oakland in 2022 per the FBI. The 15,000 or so violent crimes would be more than 50 percent higher than Oakland’s previously reported most violent year (1992), so either Oakland had an insane surge in violent crime last year or we are looking at a data reporting error.

A deeper dive into the available data suggests OPD’s reporting to the FBI had issues in 2023. OPD’s year-end crime report shows violent crime rose in 2023, but the rise was a far more muted (though obviously still high) 21 percent. OPD’s publicly reported violent crime figures don’t always match up to what is reported by the FBI — which is very normal for a lot of cities — but the OPD reported values are usually pretty close to what the FBI reports.

Below I’ve charted violent crimes reported publicly by OPD on the Y—axis and what has been reported by the FBI on the X-axis for every year from 2010 to 2023 (excluding 2021 due to the NIBRS transition).

The 2023 count is an obvious and enormous outlier.

The FBI data allows us to break down offenses by month and offense type for Oakland going back decades which further highlights the 2023 outlier. Other types of reported violent crime rose in Oakland last year as would be expected during a sizable 20-something percent increase in violent crime citywide, but aggravated assault was up an astounding amount due to two outlier months.

Aggravated assault shot up incredibly in January and October 2023 but otherwise was reasonably normal. Zooming out even further shows just how ridiculous Oakland’s January and October aggravated assault totals are compared to every reported month since 1960. December 2003 had the previous high — when Oakland was clearly going through some reporting issues — but even that pales in comparison to the two 2023 outliers.

The Oakland Police Department acknowledged the error a few months ago after the San Francisco Chronicle reported on the issue. OPD released a statement saying “At this time, we believe the anomaly is due to human error” though the nature of this error is not fully clear beyond that statement.

Correcting Oakland’s error alone would change the FBI's estimated national violent crime decline in 2023 from -3.04 percent to -3.76 percent. It would also lower the FBI's estimated US violent crime rate by around 1 percent, from nearly 364 percent 100k to around 361 per 100k, the latter of which would be the lowest violent crime rate reported since 1970.

But it doesn’t quite work like that.

Mistakes happen every year with the FBI data in terms of both over and underreporting. My guess is that these mistakes have relatively smaller overall impacts from year to year which tend to even out over time. It’s hard to collect precise crime data from tens of thousands of agencies each year!

My assumption is that the errors tend to even out between over and underreporting over the long term, so there's no point in trying to “fix” the data ourselves. That said, a big US city reporting 9,000 more violent crimes to the FBI than were recorded by the agency is likely a rare occurrence. So while Oakland’s overcount for 2023 is not ideal, it also shouldn’t change our impressions of the national trends from year to year.

It's likely rare that a single human error commands such a large potential change in a given year, but it speaks to why we shouldn't see much difference in a rate of 361 per 100k compared to 364 per 100k. We have precise counts in sports, but we don't really in crime data. It's sufficient to say that reported violent crime fell in 2023 and was roughly equal to the 50 year low reached in 2014.

Which brings us back to the original question: did violent crime rise in California in 2023?

Had Oakland reported a 21 percent increase in violent crime — as suggested by the agency’s 2023 end of year report — rather than the 140 percent increase reported by the FBI then Oakland would have reported nearly 9,000 fewer violent crimes in 2023 than was reported. Doing the math, replacing Oakland’s reported increase with the expected one one would give California 133 fewer violent crimes in 2023 than 2022.

Of course, doing a deep dive into every single agency in California may have found other — undoubtedly smaller — errors elsewhere which could wipe out the ever-so-slight decline.

In my view, we certainly don’t really have evidence that violent crime fell in California in 2023, but a human data error is likely responsible for the rise. It’d probably be most accurate to say, therefore, that violent crime in California it was about even in 2023 rather than up or down.

Underreporting and overreporting of crime to the FBI is not unique to Oakland in 2023. I wrote about how Greenwood, South Carolina reported way more murders than likely occurred in 2022. Interestingly, the FBI reduced Greenwood’s 2022 murder tally from 58 as reported last year to 29 when updating this year’s report (which is still almost certainly an over-count!). And my hometown of New Orleans ended up underreporting rape in 2022 due to a miscommunication with the state UCR program.

These are just a few examples of both over and underreporting crime which speak to why I try to discuss crime data through a lens of accuracy rather than precision. Hopefully this piece provides some evidence for why the FBI's annual estimates are, in fact, called estimates.

They may be well-informed and they may be an accurate reflection of reported crime in a given year, but they are not precise. Remembering that is critical to effectively using crime data.