“You are being lied to,” Elon Musk posted on X yesterday. He’d know, because he’s doing the lying.
Musk was retweeting a wildly false post insisting that violent crime is on the rise, by an X user whose avatar is an imperial stormtrooper from Star Wars (red flag!). The account’s previous brush with infamy came when Donald Trump posted a screenshot of the account suggesting that Swifties supported him; Taylor Swift cited that in endorsing Kamala Harris this week. Despite beginning with the words “FACT CHECK” in bold—another red flag—the post is actually a vivid example of a new big lie driven by Donald Trump and his allies, full of easily debunked nonsense.
The user’s fundamental claim is that despite what the FBI’s data and all other legitimate statistical sources show, crime—especially violent crime—is actually rising, as Trump claimed in this week’s debate. The former president tried to say that crime was up, and when moderator David Muir corrected him, Trump replied, “The FBI—they were defrauding statements. They didn’t include the worst cities. They didn’t include the cities with the worst crime. It was a fraud.” This is false. Violent crime is down. Trump is wrong, as is Musk.
[David A. Graham: Trump’s illusory answers to imaginary crime problems]
The X user makes a slightly, though only slightly, more sophisticated version of Trump’s argument. Or rather, he or she throws more claims at the wall in the hopes they’ll stick, but they shouldn’t. I’ll take some of the big claims in order. As the tweet reads:
Less than a year after taking office, Biden-Harris’s administration had the FBI dismantle the long-standing crime reporting system, replacing it in 2021 with a new, ‘woke’ system that is optional for state and local law enforcement agencies to use.
Here’s what’s real: Starting in 2021, the FBI’s national crime estimates were based on reports to a system called the National Incident-Based Reporting System, moving from the old Summary Reporting System. NIBRS itself is not new; it dates to 1988. The Biden administration had nothing to do with the switch. The decision to move to NIBRS was made in 2015, and it was implemented in January 2021, before Trump left office.
The old Summary Reporting System gathered only limited data on a limited number of crimes. The switch was intended to improve the quality of America’s crime data. But the data remain plagued with troubles. For one thing, national crime rates are not available until late the following year: 2023’s numbers are currently expected from the FBI some time later this month. And because the country has an estimated 18,000 law-enforcement agencies—from the 36,000 officers of the NYPD to local constabularies with a single officer—collecting good data from all of them is hard.
NIBRS has never solved all of those problems, but it does provide more detailed data than SRS, tracking more types of crimes, for example. The reason the FBI kept using SRS was that not enough agencies had switched to NIBRS. To fix that, the FBI announced that, starting with 2021 numbers, it would collect data only from agencies that reported via NIBRS, and would stop using the old system.
Crime experts widely agree that, as a result of that transition, the numbers for 2021 are dubious. In the past, typically a small percentage of agencies had failed to report stats to the FBI—something like 5 or 6 percent. In 2021, a third of U.S. agencies failed to report. It’s important to remember that the FBI crime estimates are just that: estimates. Because the FBI had worse data, it had to make more assumptions in 2021
But by 2022, the most recently available year of FBI data, that problem was largely solved, partly because more agencies had shifted over to NIBRS. The X post says, “As a result, at least 6,000 law enforcement agencies aren’t providing data, meaning that 25% of the country’s crime data is not captured by the FBI.” That claim may be based on a July 2023 Marshall Project article saying that 6,000 agencies hadn’t submitted 2022 data. That was accurate at the time, but then the FBI decided to allow submissions via the old system, which meant that overall participation matched the historical average. I have no idea where the 25 percent number comes from, but all cities with more than 1 million people were included in the 2022 FBI data, while small towns and state police tend to have lower reporting rates. A greater number of crimes take place in larger cities, and no category of agency is at less than 77 percent, so that claim appears to be completely invented.
The rest of the post doesn’t stand up either. For example, it implies that liberal policies by prosecutors in New York City are falsely driving down crime rates in the data. But Charles Fain Lehman, a fellow at the conservative Manhattan Institute, notes that reported violent crime in New York has actually risen—hilariously, something that would presumably help the poster’s overall argument, if he or she weren’t so sloppy. Even so, the charge misses the point, because prosecutors don’t report these numbers—police departments do. These are statistics not about charges or convictions but about crime reports. Whatever the failings of progressive prosecutors, they don’t have anything to do with FBI crime estimates.
The X post also claims that NIBRS is “woke” and allows “agencies to record pronouns and gender identities, including transgender and nonbinary, as well as the sexual preferences of both criminals and victims.” As far as I can tell, this is invented out of whole cloth. The submission specifications include nothing like that.
“It’s not far-fetched to imagine that the Biden-Harris regime and the Democrats replaced the FBI’s universal crime data system with a new optional system to fabricate this massive decrease in ‘reported’ crime,” the post goes on.
But as we’ve seen, it’s not only far-fetched; all of the predicates are untrue. (The system has also always been voluntary.) The other problem is that although the FBI numbers are the acknowledged national standard, they’re not the only numbers available that show the same results. Many cities and states make their numbers publicly available online. Those numbers tell a consistent story: In most places, crime rose sharply in 2020 and has been receding ever since, though in general it’s still higher than in 2019. The Real-Time Crime Index, an invaluable new tool for tracking changes in crime made by the independent statistics firm AH Datalytics, shows the clear downward trend in violent crime and other offenses.
The X post is more or less totally false, but its goal is not to correct the record but to spread an atmosphere of fear and paranoia—to suggest to voters that they are not safe, and that the best way to guarantee their safety from the “American carnage” Trump has described is to vote him into office and abridge certain people’s rights. Trump has always seized on crime fears and lied about incidence of crime, but he’s working especially hard at it now. In addition to the inconvenience of his own 34 felony convictions, Trump has the problem that crime spiked in his last year in office and has been dropping since. Rather than change the subject, Trump wants to change perceptions of reality.
Crime data are not as reliable, or as timely, as would be ideal. Some crimes—especially those such as domestic violence and child abuse, whose victims feel shame—are thought to be drastically underreported. People who distrust police may also hesitate to report crimes. Given these difficulties, researchers tend to look carefully at the murder rate, because it is thought to be the most reliable statistic, as murders are almost always reported, and nearly impossible to hide. Today, murder statistics also point to a general downturn in crime. And that gets at the real lesson: No crime data should be taken in isolation. It’s essential to look at as many metrics as possible, understand their limitations, and emphasize trends over absolute numbers.
[David A. Graham: America’s peace wave]
But not all the statistics measure the same thing. Trump and his campaign yesterday cited the National Crime Victimization Survey to insist that crime really is up sharply. But as AH Datalytics’ Jeff Asher, the best guide to understanding crime statistics, has written, NCVS is less reliable than the FBI crime trends because it doesn’t include murder (homicide victims seldom respond to surveys), doesn’t specify the year crimes occurred (it asks about the past six months), and is subject to the same problems that have bedeviled other public-opinion polls in recent years. But, Asher contends, the trends in the two sources usually align anyway: “Both measures tell us that the nation’s violent crime rate in 2022 was substantially lower than it was in the 1990s, largely in line [with] where it was over most of the last 15 years, and likely slightly higher than where it stood in 2019.” The numbers for 2023, released yesterday, show a decline from the previous year. It’s also nonsensical for Trump to claim that the FBI is producing fraudulent numbers but then cite Justice Department figures as the gospel truth.
Mark Twain joked that there are lies, damned lies, and statistics. But some statistics are actually pretty reliable, which is why cynics turn to lies instead.
