The U.S. employs two distinct measures of crime. The Federal Bureau of Investigation’s Uniform Crime Reporting program counts the number of crimes reported to police annually. The Bureau of Justice Statistics, in its National Crime Victimization Survey (NCVS), asks 240,000 people a year whether they have been victims of a crime. The two measures have diverged since 2020: The FBI has reported less crime, while more people say they have been victims. For example, between 2021 and 2022, the FBI UCR showed reported violent crime fell by 2.1%, but the NCVS showed reported violent crime increased by 29.3%. The figure at the top of this post clearly shows how the two measures rarely move together.
As previously noted, the news media relies exclusively on FBI-reported violent crime data. We previously discussed how the FBI-reported crime rates are unrelated to the NCVS. The evidence shown here indicates real problems with the FBI-reported violent crime measure and that the FBI data are extremely misleading after 2020.
The next figure in this post shows the changes in reported violent crimes for these two measures over the fifteen years from 2008 through 2022. As you can tell from the figure, the correlation coefficient is -0.2389. From 2008 to 2019, the correlation is very slightly positive: 0.0473. But from 2020 to 2022, these two numbers are almost perfectly negatively correlated: -0.9597. Thus, the two measures are unrelated over the earlier period, and in the later period, they move in opposite directions. None of this generates confidence in the FBI UCR data.
By contrast, reported property crime rates for these two measures are closely related over the whole period (correlation coefficient 0.8919), and their correlation coefficients are very similar in the two periods.
The divergence is due to several reasons. In 2021, 37% of police departments nationwide, including Los Angeles and New York, stopped reporting crime data to the FBI. The more important point here is how this has changed over time. As the Washington Examiner notes: “In 2019, 89% of agencies covering 97% of the population submitted data, but by 2021, that coverage plummeted to less than 63% of departments overseeing just 65% of the population.“
In addition, in cities from Baltimore to Nashville, the FBI is undercounting crimes those jurisdictions reported.
And the figures the agencies do report to the FBI do not match the agencies’ publicly reported figures. For Baltimore, the FBI reported 225 murders in 2023, but the city reported 262 — which means the FBI left out 37 murders. In Milwaukee, the police department reported a 7% increase in robberies, but the FBI showed a 13% drop. Nashville’s own data tallied more than 6,900 aggravated assaults in 2023, but the FBI counted only 5,941, leaving almost 1,000 of those offenses “missing.” This trend is consistent across the board: While 2022’s FBI city-level figures track the police’s own data, the 2023 numbers consistently undercount offense totals. Any year-to-year comparison overstates decline.
Mark Morgan and Sean Kennedy, “Bad data from the FBI mislead about crime,” Washington Examiner, April 5, 2024.
Here are two simple regressions showing that the changes in reported crimes using the FBI and NCVS data were unrelated in the pre-2020 period and negatively related in the period from 2020 on.
| Year | Change in NCVS Reported Violent | Change in FBI Reported Violent |
| 2008 | -3.3% | -3.5% |
| 2009 | -16.9% | -5.6% |
| 2010 | 1.0% | -5.2% |
| 2011 | 12.1% | -5.4% |
| 2012 | 3.6% | -0.4% |
| 2013 | -7.8% | -4.5% |
| 2014 | -13.2% | -1.8% |
| 2015 | -6.5% | 2.4% |
| 2016 | 0.0% | 4.7% |
| 2017 | 7.0% | -3.1% |
| 2018 | 7.6% | -1.1% |
| 2019 | -13.1% | -2.5% |
| 2020 | -23.3% | 5.9% |
| 2021 | 13.6% | -2.0% |
| 2022 | 29.3% | -2.1% |
The NCVS data is available here.
https://bjs.ojp.gov/document/cv22.pdf
https://bjs.ojp.gov/content/pub/pdf/cv21.pdf
https://bjs.ojp.gov/sites/g/files/xyckuh236/files/media/document/cv20.pdf
https://bjs.ojp.gov/content/pub/pdf/cv18.pdf
I just listened to you on Relevant Radio (4/20/2024). I agree with what you said but
where are the 2023, and 2024 rates on crime?
If I share this info, all the democrats I know will say these are old stats and blame these numbers on Trump, 2020, Covid spilling over to 2024 … heaven help us!
I pray you have current stats and can share them.
Thank you,
The NCVS data for 2023 will be out in September.
Good Morning Doctor Lott, Huge fan of your work. I am by no means an expert, but a question is niggling me. Since the NCVS data is sampled and not a full recording of crime from all citizens, is it possible the sampling rates may be skewed by a significant amount? Im sure i am going to have this thrown at me when i have a discussion with folks about this data anomaly.
Sincerely,
Fitz
Thanks, Kelly. The NCVS is much more accurate than a political poll. Of course, the sample is weighted to make sure that it accurately reflects respondent’s share of the population, though they do a much more detailed weighting by everything that you see in your typical survey, but also be charateristics such as where they live. Not only do they survey 240,000 people, not your typical political survey where a very large survey is 12,000, but they have much better access to telephone numbers than private organizations do. The bottom line is that even if you still believe that there is some skewing, the issue is why the FBI and NCVS used to track each other and are now over just the last few years negatively related. I hope that this helps.
Underreporting of crime is not knew. In fact there were cases in the press that shed light on this.
https://youtu.be/OTW3RFNPEoM?si=1RG7i1PRDGHDu1b0
https://www.wusa9.com/article/news/local/dc/internal-documents-whistleblowers-show-alleged-underreporting-of-crime-by-dc-police-mpd/65-3ce1c9eb-ce1e-4823-a11c-66775c147b10
https://youtu.be/MUWpWJ7_o2g?si=KKuBssIbKqCzrUw2
It seems to me that the offense least likely to be inaccurately reported is homicide. People are not likely to fail to report a homicide or a missing loved one, even if there is a perception it won’t be investigated. There’s physical of homicide – dead bodies. People aren’t likely to fail to report a dead body in the street, or a missing loved one. To the extent that there is correllation between homicide and violent and property in general, the change in homicide rate would be a proxy for the violent crime rate, in general.
The FBI’s Crime Data Reporter shows that from 2018-2019, homicide was 5 – 5.1 per 100,000, rose to a peak in 2021 of about 6.8 per 100K, then dropped somewhat in 2022 to about 6.3 per 100,000. 2022 is last year available. If that downward trend continues, then one could reasonably say violent crime along with homicide has decreased since pandemic peak in 2021. Only time will tell.
https://cde.ucr.cjis.gov/LATEST/webapp/#/pages/explorer/crime/crime-trend
Correction to my comment above.
Correction: physical >>evidence<>crime<< in general
Correction to above:
physical EVIDENCE
violent and property CRIME in general
Mr Lott,
Thank you for your commitment to good quality research.
Does the FBI get its data from individual police departments, or is it collected from the State reports? And is there an estimate on how many departments actually report? I do find it hard to believe that every department in every state is focused on getting these numbers in, above other much more immediate priorities. Thank you for your time, and all your work.
Thank you very much. Your comment is greatly appreciated.