The Curious Case of Greenwood, South Carolina
Did a town of 20,000 really have the nation's highest murder rate in 2022?
There were 7,220 police departments representing populations of 1,000 or more in 2022. Cities with traditionally high murder rates like Birmingham, New Orleans, and St Louis reported murders rates of 70+ per 100,000. Many other cities of below 10,000 people reported a handful of murders leading to outsized murder rates of 100 or more.
But no city in America with more than 1,000 people reported a higher murder rate in 2022 than Greenwood, South Carolina (population of roughly 22,000).
Greenwood reported (the operative word in this sentence) an astounding 58 murders in 2022 which meant an even more astounding murder rate of around 262 per 100k. That total came on top of 20 murders in 2021 giving the city 78 murders reported in 2021 and 2022 which is more murders than the city reported between 1993 and 2020 combined.
Did Greenwood, South Carolina really have such an astoundingly high murder rate in 2022?
A deeper dive into the data suggests probably not.
The first clue is that the South Carolina state UCR program reports 29 homicides in Greenwood in 2022. That’s exactly half of 58, so Greenwood somehow reached 29 murders to report (still astoundingly high!) and then that figure got doubled for some unknown reason.
South Carolina’s state UCR program has a handy report that lets us dig deeper into Greenwood’s 2022 murders. Greenwood reported 14 of the 29 murders in July 2022, so I dug into that month to highlight what I suspect is the problem.
Below is what you get when you select Greenwood PD and July 2022. If you count up the unique ages in each incident then you get to 14 murder victims, all via firearm.
There probably weren't 14 murders in July 2022 though. The mid-July shooting death of a 29 year old mother shows up in numemrous news articles as well as the reported data, but a review of Greenwood news articles didn’t show any other July 2022 murders. I would’ve expected nonstop news stories had a baby, an 8 year old, and six teenagers been shot and killed in a town of 20,000 over the course of a month.
Another clue that this is a reporting error is the fact that the one murder that is verifiable through news articles is the only single victim incident that month. All of the other reported murders had multiple victims including a double murder, a triple murder, and two murders with four victims each.
Finally, the Gun Violence Archive shows one shooting victim in July 2022 in Greenwood, a fatal shooting of a 29 year old woman.
Taken together, the evidence is pretty strong that there was one murder victim in Greenwood in July 2022 rather than 14. There are a handful of other homicides reported by Greenwood in 2022 which are verifiable in GVA data.
There’s also a pregnant woman who survived a shooting but lost her baby which matches to one of the murders of a child under 7 days old counted by Greenwood. But the FBI’s user manual is clear that should not be counted as a murder.
Of the eight fatal shootings listed in GVA data for Greenwood in 2022, six appear to match the FBI’s definition of a murder. That there were no non-gun murders in the city's reported data implies that the city had around six murders rather than 29 or 58 in 2022.
How did Greenwood get from 6 to 29 to 58? I can only guess, but I’d guess that data was inputted incorrectly in 2021 and 2022 and nobody caught the mistake in either year.
Rather than wildly speculate on what caused the mistakes, I think it’s worthwhile to consider what these mistakes mean for crime data consumers.
First, it shows that even published crime data on the FBI’s website can be flawed. If you see a city with an implausible increase or decrease from one year to the next then usually it’s due to incomplete or incorrect data reporting (or somebody posted on TikTok about how to steal Hyundais). There are going to be errors when publishing data from 18,000 agencies, but the advantage of having 18,000 agencies reporting is that any one agency is not hugely important for accurately portraying general trends.
Second, it highlights the importance of the individual officer in properly reporting crime data. Accurate crime data relies on individuals properly coding offenses into their Records Management System. This is even more critical with the NIBRS transition which requires substantially more information to be captured about a wider array of offenses. Improperly coding an aggravated assault/attempted murder as a murder offense is how we might get 58 instead of 6 murders in a city (though I certainly don’t know for sure that’s what happened here).
Finally, from an analytic standpoint this case showcases the need to use multiple data sources when evaluating whether a particularly noteworthy change is due to a data glitch or is a real trend. This is true when looking at FBI reporting and it’s true when looking at an individual city reporting a rise or decline in the midst of other changes — such as surging response times.
The crime data errors that usually get attention are those of underreporting, but hopefully the example of Greenwood, SC shows that the errors can go both ways.
I think your hypothesis that "attempted murder" could have been coded as murder makes sense. What are the codes? Could someone fat finger the mistake? Could someone have misremembered the correct code for murder, maybe because they use it so rarely? Or, worse, could someone have written a cheat sheet (taped to their monitor maybe) with the wrong code on it?
These kinds of coding errors happen in lots of contexts where complex systems have been set up to capture fine-grained data (Medical records come to mind in this context).
Thanks for working on this.
Thanks Jeff: Another insightful article.
Folks at the FBI fully understand that law enforcement agencies are not reporting crime correctly. With police staffing levels being low and with confusion over the NIBRS, there are going to be errors.
But misreporting has been going on long before the NIBRS. The FBI offers months and times of crime that are clearly inaccurate because agencies are not providing the information thus defaulting to inexact times and months.
On average, U.S. residents experienced approximately 246,900 hate crime victimizations each year between 2005 and 2019 per the National Crime Victimization Survey versus 11,634 reported by the FBI in 2022. Yet we are making policy and having a public discussion over the wrong numbers.
The rate of crime reporting for cyber crimes or juvenile crimes or retail theft is ridiculously low let alone property or violent crimes.
I'm beginning to think that, beyond homicides, "reported" crime statistics are wildly inaccurate for an endless number of reasons. I do not blame the FBI. I believe that the problem lies with individual agencies and state reporting systems yet no one wants to take this on because it's essentially a voluntary system as to participation.
I had a two hour conversation with a reporter from USA Today partially discussing FBI statistics. He was quite surprised by the realities of "reported" crime data.
There doesn't seem to be a solution but "reported" crimes need to come with a disclosure. I believe that anything over or under ten percent needs to be used cautiously.
Best, Len.