Crime Statistics 2022: Interpreting the Trends

Time for my annual update on the newly released annual national crime statistics. Last year, I reported on homicide statistics from the CDC and nonfatal violent crime statistics from the National Crime Victimization Survey. I wrote that the statistics for 2021 from the Uniform Crime Reports (UCR), the FBI’s statistics compiled from law enforcement agency submissions, were unreliable because of the huge amount of missing data. For 2022, however, the FBI allowed submissions under the previously used data collection system and reduced the missing data, so I am including statistics from the FBI in this year’s update.*

Homicide

First, let’s look at homicide statistics. As described in Lohr (2019), there are two national sources of information on homicide: statistics from the UCR (from law enforcement agencies) and statistics compiled from death certificate information by the Centers for Disease Control and Prevention (CDC). Figure 1 shows the homicide rates (number of homicides per 100,000 population) from the UCR and CDC (data sources are listed at the bottom of this post).

As you can see from Figure 1, for most time periods the two data sources give similar trends for homicide. The CDC line is usually higher than the line from the UCR because the CDC definition of homicide is slightly broader than that used in the UCR, including negligent as well as intentional homicides (Lohr, 2019). The CDC line has a spike for 2001 because it includes the victims of the 9/11 terrorist attacks (these were excluded from FBI statistics).

However, the difference between the two sets of homicide statistics appears to have widened slightly for 2021 and 2022. For most years between 1993 and 2017, the ratio of the UCR homicide rate to the CDC homicide rate exceeded 0.9. For 2022, the UCR homicide rate is only 85% of the CDC homicide rate. Both sets of statistics show a decrease in homicide from 2021 to 2022, but further investigation is needed of these widening discrepancies.

Violent Crime Reported to Police

There are two major national sources of data on violent crimes other than homicide. The UCR captures crimes that are known to law enforcement agencies. The National Crime Victimization Survey (NCVS) measures crime through asking persons in a large nationally representative survey sample about their experiences with crime — both crimes that were reported to the police and those that were not reported to the police. The crime types and definitions differ for the two sources, but there is a general similarity between rape, robbery, and aggravated assault in the NCVS (termed “serious violent crime”; the NCVS also measures simple assault) and the same crimes measured in the UCR.

Figure 2 displays the time trends for rape, robbery, and aggravated assault as measured by the NCVS and UCR between 1993 and 2022. The UCR rate for violent crime (rape, robbery, aggravated assault) reported to the police has been roughly constant since 2000. The NCVS rate for rape, robbery, and aggravated assault reported to the police shows more fluctuation. Some of that fluctuation occurs because the NCVS statistics are computed from a sample of households; the dotted lines in Figure 2 show the 95% confidence intervals that reflect the variability due to sampling. For most years since 2000, the rate of rape, robbery, and aggravated assault from the UCR has been close to the NCVS rate of serious violent crime reported to the police. Note that this similarity has occurred even though the UCR and NCVS measure different populations. The NCVS does not measure victimizations of children under age 12, people in prison or nursing homes, people experiencing homelessness, or U.S. nonresidents visiting the U.S. All of those groups are included in the UCR statistics.

Figure 2. Rape, robbery and aggravated assault as measured in the UCR and NCVS. The NCVS statistics for 2006 are deemed unreliable (Thompson and Tapp, 2023) and are not displayed.

What about changes for 2022? Rosenfeld and Lauritsen (2023) wrote that we don’t know whether crime went up or down in 2022 because the NCVS and UCR moved in different directions. However, they reported NCVS statistics from the top line in Figure 2, which includes all serious violent crimes including those not reported to the police. They mentioned that part of the reason for the discrepancy may be decreased reporting of crimes to the police for 2022 (as measured by the NCVS), and that explanation is consistent with the data. Although the decrease in the percentages of serious violent crimes reported to the police between 2021 (52.2%) and 2022 (48.0%) is not statistically significant, the slope from 2021 to 2022 in Figure 2 is greater for all serious violent crimes (black line) than for the set of violent crimes reported to the police (blue line). For 2022, the UCR rate (red line) is very close to the lower NCVS confidence limit for victimizations reported to the police (lower blue dotted line).

There are also other potential explanations for differences in trends. The 2020 and 2021 NCVS statistics were affected by changes in sampling spurred by the COVID-19 pandemic. In 2020, the Census Bureau, which collects the NCVS data, had to suspend face-to-face interviewing for several months and could not visit households that were scheduled to join the NCVS sample. As I discussed in my post about the 2020 crime statistics, people who are interviewed face-to-face, and those newly joining the sample, tend to exhibit higher victimization rates. Although the 2020 NCVS data were weighted to attempt to compensate for the missing higher-victimization people that year, it is possible that the weighting did not correct for all of the bias and that the 2020 NCVS statistics were biased downwards.

The NCVS response rate has also been dropping over the last 20 years. Krueger et al. (2023) wrote: “The overall response rate [the percentage of survey-eligible people selected for the sample who participate in the survey] of the 2022 NCVS was 53.9 percent, which causes concern of potential bias in the NCVS estimates.” Moreover, response rates increase by age, so that the persons most likely to be victims of violent crime (those in younger age groups) are least likely to provide data to the survey. Weights are used to compensate for the missing data but some bias may remain in NCVS statistics after the weighting.

It would be helpful if the Bureau of Justice Statistics, which administers the NCVS, performed an in-depth analysis to see what accounts for the increase in crimes not reported to the police. Is the increase concentrated in a specific set of survey respondents, or in particular demographic groups?

Is Crime Up, Down, or What?

The good news is that homicide rates from both sources appear to have decreased from 2021, although these are still higher than the levels in 2019. Kegler et al. (2023) reported that there were 5,223 more firearm homicides in 2022 than in 2019, which means that all of the increase in homicide between 2019 and 2022 can be attributed to additional deaths by firearms. More study is needed, however, on the increasing discrepancies between the CDC and FBI estimates of numbers of homicides. How much of that is explained by the more expansive definition used by the CDC, and how much is caused by homicides missing from the FBI data? One way to study the discrepancy would be to link records for the two datasets and identify which records are found in one source but not the other, as in Feldman et al. (2017).

The NCVS reports an increase in serious violent crime for 2022, both overall and for crimes reported to the police. These increases are statistically significant but it is unclear how much decreased response rates and pandemic-related changes in data collection may have contributed to them. It appears that the serious violent crimes reported to the police are roughly consistent with previous trends, but more analyses are needed to understand the large increase in crimes not reported to the police in the NCVS. Next year’s data should help clarify the patterns.

Footnotes and References

*In 2021, the FBI switched from collecting only counts of crimes to the National Incident-Based Reporting System (NIBRS), which collected details about individual crime incidents. The problem was that many law enforcement agencies did not submit NIBRS data in 2021, and only about 65% of the US population was represented in the data. Non-participating agencies included the New York City, Los Angeles, and Phoenix police departments, and the procedures used to estimate crime totals in 2021 did not account for the fact that large cities such as these have unique crime patterns — there really is no way to accurately estimate crime totals for major cities using data from smaller cities in the same state.

For 2022, the FBI allowed law enforcement agencies to submit count data under the old system; they also retrospectively collected 2021 data under the old system from a sample of agencies that had not submitted NIBRS data in 2021. For 2022, the law enforcement agencies submitting data represent about 94% of the US population and almost all major cities participated. The FBI weighted the data to account for the other 6%. The documentation gives no details on how the data were weighted for 2022 but the procedure is likely similar to that used in previous years and described in Lohr (2019).

Feldman, J.M., Gruskin, S., Coull, B.A., and Krieger, N. (2017). Quantifying underreporting of law-enforcement-related deaths in United States vital statistics and news-media-based data sources: A capture-recapture analysis. PLoS Medicine, 14(10), e1002399.

Kegler, S.R., Simon, T.R., and Sumner, S.A. (2023). Firearm homicide rates, by race and ethnicity — United States, 2019-2022. Morbidity and Mortality Weekly Report, 72(42), 1149–1150.

Lohr, S. (2019). Measuring Crime: Behind the Statistics. Boca Raton, FL: CRC Press.

Rosenfeld, R. and Lauritsen, J. (2023). Did violent crime go up or down last year? Yes, it did. https://counciloncj.org/did-violent-crime-go-up-or-down-last-year-yes-it-did/ Part of this article was reprinted in the Washington Post, November 10, 2023, p. A21.

Krueger, F., Scalyer, Z., Farber, J., and Hornick, D. V. (2023). Source and accuracy statement for the 2022 National Crime Victimization Survey. In National Crime Victimization Survey 2022 User Guide. Ann Arbor, MI: Inter-University Consortium for Political and Social Research.

Thompson, A. and Tapp, S. N. (2023). Criminal Victimization, 2022. NCJ Report 307089. Washington, DC: Bureau of Justice Statistics.

Data Sources

The CDC homicide rates for years through 2021 came from https://www.cdc.gov/injury/wisqars/fatal/index.html; the rate for 2022 is preliminary and was obtained from https://www.cdc.gov/nchs/nvss/vsrr/mortality-dashboard.htm.

The National Crime Victimization Statistics came from the annual issues of Criminal Victimization published by the Bureau of Justice Statistics (see, for example, Thompson and Tapp, 2023, for the 2022 statistics).

I downloaded UCR data from the FBI Crime Data Explorer on October 24, 2023. To download the tables of national estimates from the UCR, go to the main webpage of the Crime Data Explorer, then click on “Documents & Downloads.” Scroll down to the section “Crime in the United States Annual Reports” and download the CIUS Estimations data. Table 1 contains the crime statistics from 2003 to 2022. Note that the FBI updates statistics as more data come in, so graphs in this post may differ slightly from my previous writings. For example, the 2022 file lists a violent crime rate of 377.7 violent crimes per 100,000 population for 2017; Crime in the United States 2018 listed a rate of 383.6 for 2017.

The statistics I’ve reported in this post are from Table 1 of “Crime in the United States.” These are the FBI’s national estimates of crime rates and numbers, derived from submissions made by 15,726 law enforcement agencies and weighted to account for missing data. For 2022, Table 1 lists 1,232,428 violent crimes (including homicide) and 21,156 homicides, giving rates of 369.8 violent crimes and 6.3 homicides per 100,000 population. Tables 4 and 5 of “Crime in the United States” give the estimated total amounts of crime in each state. Arizona had an estimated 31,754 violent crimes including 500 homicides, giving rates of 431.5 violent crimes and 6.8 homicides per 100,000 population.

Note that the Crime Data Explorer has other sets of numbers for crime statistics in other locations of the website. Most of these do not account for missing data and thus should not be cited as statistics for the amount of crime. The main webpage has two sets of statistics. The top graph contains national and state estimates, and these statistics are consistent with those in Tables 1, 4, and 5 of “Crime in the United States.” The rest of the page contains statistics from NIBRS, and these are not nationally representative — they are based on the 13,293 law enforcement agencies that submitted data to NIBRS. The NIBRS-submitting agencies covered areas serving about 75% of the U.S. population, so statistics based only on the NIBRS agencies underestimate the amount of crime. The webpage says that there were 809,381 violent crime incidents reported by those 13,293 agencies for 2022, but that statistic does not include crimes from the law enforcement agencies that submitted data under the Summary Reporting System (the system used for data submission through 2020 and then resumed in 2022) and it does not estimate the amount of crime that occurred in the nation.

Yet another set of statistics is found if you download “Analytical Tables” from the “NIBRS Estimation Tables” section of “Documents & Downloads” (I have to give these long directions because most pages have no direct link). These are also supposed to be national estimates, but they are computed only from the law enforcement agencies that submitted NIBRS data. These estimates are not as reliable as those from Table 1 of “Crime in the United States” that I referenced in this post. The Table 1 estimates are based on data from agencies representing about 94% of the population; the “Analytical Tables” are based on data from agencies representing 75% of the population. In essence, the “Analytical Tables” estimates are not making use of all of the data that the FBI collected and these should only be cited for states where all law enforcement agencies submit NIBRS data.

There’s one more set of statistics, and this is under the “Expanded Homicide Data” tab on the left side of the Crime Data Explorer. These give the homicide counts from the 15,726 law enforcement agencies that submitted either NIBRS data or submitted Summary Reporting System data with Supplementary Homicide data. For 2022, there were 19,200 homicides from the law enforcement agencies that submitted data. This statistic, however, does not account for the law enforcement agencies that did not submit data for 2022 and is thus lower than the Table 1 estimate of 21,156 homicides.

Sharon LohrNovember 20, 2023