When Crime Statistics Diverge
Understanding the Two Major Sources of Crime Data in the U.S.
June 2025
By Janet Lauritsen, Ph.D., University of Missouri – St. Louis, and
Ernesto Lopez, Council on Criminal Justice
This report highlights key differences between the nation’s two measures of crime, the Uniform Crime Reports (UCR), which tracks crimes reported to law enforcement authorities, and the National Crime Victimization Survey (NCVS), an annual series of household interviews that captures offenses that are both reported and not reported to police. The report explores the strengths and limitations of the two sources, the similarities and differences in long-term and short-term trends they document, and the importance of considering both sources when assessing changes in crime over time.
Key Takeaways
- The nation’s two major tools for measuring crime, the UCR and the NCVS, show similar long-term trends for rates of serious violent crime, motor vehicle theft, and burglary.
- When the NCVS is adjusted to be comparable to the UCR by counting only incidents reported to police and matching the incident year of the offense, notable variations in short-term trends emerge between the two sources.
- Such short-term differences likely occur because the NCVS is a household survey of people 12 and older. In addition, some victims, such as businesses, are accounted for in UCR data but not included in the NCVS survey. Victimization rates are within a margin of error, meaning the true level of victimization falls within a small range of values.
- Much of the difference between the two sources stems from variations in methodology, rather than inaccuracies in the data.
- The NCVS functions as an important, useful, and independent measure of overall crime in the nation, providing valuable insights into crime types not reliably captured by the UCR, detailed information about victimization harm, and the reasons why people did not report offenses to police.
Glossary
- Bureau of Justice Statistics (BJS): A bureau within the Office of Justice Programs (a component of the U.S. Department of Justice) that serves as the nation’s principal source for criminal justice statistics. BJS is responsible for collecting, analyzing, publishing, and disseminating information on crime, criminal offenders, victims of crime, and the operation of the criminal justice system at all levels of the government.
- Federal Bureau of Investigation (FBI): The principal federal law enforcement agency of the U.S. tasked with investigating and enforcing federal laws, protecting the country against terrorism and cyber threats, and upholding national security. The FBI operates under the jurisdiction of the U.S. Department of Justice and conducts investigations into a wide range of criminal activities.
- National Crime Victimization Survey (NCVS): A survey developed by the Bureau of Justice Statistics and conducted annually by the Census Bureau that serves as the primary national data source on criminal victimization. The survey currently collects information from approximately 240,000 people in 150,000 U.S. households that are representative of the nation’s demographics. Participants, limited to those aged 12 and above, are interviewed about their experiences with criminal victimization, including frequency, types, and impacts. Data cover non-fatal personal crimes (e.g., assault, robbery, sexual assault) and household property crimes (e.g., burglary, motor vehicle theft) as well as reporting to law enforcement.
- National Incident-Based Reporting System (NIBRS): A comprehensive system for collecting and reporting incident-level data on crimes reported to law enforcement in the U.S. Part of the Federal Bureau of Investigation’s Uniform Crime Reporting Program, NIBRS is the successor to the Summary Reporting System, which provides limited information on a select number of crimes. NIBRS captures information on a wide range of offenses, including details such as victim demographics, offender characteristics, and the relationship between victims and offenders.
- Serious Violent Crime/Victimization Rate: This rate is the number of aggravated assaults, rape and sexual assaults, and robberies per 1,000 people. Homicide is excluded since the measure is not in the NCVS. The population denominator used here for the NCVS and UCR is the population reported by the UCR. This makes rates more comparable.
- Summary Reporting System (SRS): A long-standing component of the Federal Bureau of Investigation’s Uniform Crime Reporting Program that collects monthly summary information on eight major crime categories: homicide, rape, robbery, aggravated assault, burglary, larceny-theft, motor vehicle theft, and arson. While still in use by some agencies, the SRS has been retired and replaced by the National Incident-Based Reporting System, which captures more detailed information on each criminal incident. Historically, more agencies have submitted offense data using the SRS than NIBRS. SRS agencies can submit optional supplemental property crime details, including the specific larceny types (e.g., shoplifting, theft from motor vehicles, purse snatching, etc.).
- Uniform Crime Reporting (UCR) Program: A program operated by the Federal Bureau of Investigation that generates crime statistics using data on crimes reported to federal, state, local, and tribal law enforcement agencies. Law enforcement agencies voluntarily report their crime data to the FBI, which aggregates and analyzes the information to produce comprehensive crime reports, including the annual Crime in the U.S. publication.
Introduction
After an historic 30% single-year increase in homicides in 2020 and continuing higher rates of violent crime the following year, most Americans in 2022 believed reducing crime should be a top priority for the federal government.1 In 2023 and 2024, however, about half of the American public believed crime was elevated, though police-recorded crime statistics showed the opposite trend.2 While it’s long been true that public perceptions of crime and actual crime trends often conflict, this gap was exacerbated by the large difference between numbers released by the nation’s two principal sources of crime data: the FBI’s Uniform Crime Report (UCR) and the Bureau of Justice Statistics’ National Crime Victimization Survey (NCVS). The UCR program, which collects data from law enforcement agencies across the U.S., reported that violent crime experienced a 2% drop from 2021 to 2022. But the NCVS data, which surveys about 240,000 people a year, showed that violent victimization rose 75% across the same period.
These contradictory estimates made it difficult to assess whether violent crime had increased or fallen between 2021 and 2022. Criminologists Richard Rosenfeld and Janet Lauritsen summed up the conundrum this way in a brief for the Council on Criminal Justice in October 2023: “Did Violent Crime Go Up or Down? Yes, It Did.”
Understanding changes in crime during this period was particularly important for assessing the possible impacts of the easing of the COVID-19 pandemic. In addition, the large difference between the UCR and NCVS percentage change estimates became a heated point of political contention.3 The controversy intensified in the fall of 2024 when the FBI, without explanation, revised previously released figures, leading some to further question the integrity of the UCR data.4
Politics aside, conflicting signals from the two major data systems and revisions to data can produce confusion about whether crime is increasing or decreasing, and by how much, and also raise concerns about the validity of U.S. crime data more generally. This report offers a brief summary of how the UCR and NCVS differ, and why they may be expected to differ in their estimates of crime levels and the percentage change in crime from year to year. Additional details about crime data can be found in BJS reports on the topic.5
Important Differences in the UCR and NCVS
While both the UCR and NCVS are important for understanding the nation’s crime levels and trends, there are differences between these sources that may lead to discrepancies in the data they produce. Additionally, the NCVS provides critical information that is not captured by the UCR, including victim’s reasons for not reporting the offense, the use of victim services, detailed injury information, and the survey’s special focus on identity theft.
The fundamental distinction between the UCR, launched in 1930, and the NCVS is the source of the data upon which each system relies. The majority of UCR crime data are provided by state and local police departments to the FBI and are dependent on reports by individuals (including businesses and other organizations) to law enforcement. Most crimes in the U.S., however, are never reported to the police, and less severe crimes (e.g., theft and simple assault) are reported at even lower levels. To account for this omission, the NCVS household survey has been producing crime victimization rates since 1973. As of the 2023 release, the survey collects information from approximately 240,000 people in 150,000 U.S. households that are representative of the nation’s demographics. Participants, limited to those aged 12 and above, are interviewed about their experiences with criminal victimization, including frequency, types, and impacts. The survey also includes information about whether incidents were reported to police.
Other important differences between the UCR and the NCVS, such as the type of offenses covered, are discussed below and summarized in Table 1. It is important to note that the UCR program contains two systems for counting crime. The Summary Reporting System (SRS) is one format law enforcement agencies can use to submit monthly crime counts for eight offenses. The second system is the National Incident-Based Reporting System (NIBRS), through which agencies can submit detailed information on specific incidents covering 52 offense types.
Table 1. Overview of UCR and NCVS Characteristics
Reporting and Population Differences
The greatest source of disparity between the UCR and NCVS are different reporting levels for offenses. Moreover, reporting to law enforcement varies according to the severity of offenses. For instance, in 2023, 57% of aggravated assaults were reported to police, while 41% of simple assaults were reported to police (see Table A1 in the Appendix). This means that lower-level offenses (e.g., simple assault), which are the most common crimes, are also the least likely to be recorded by police, leaving the NCVS as the only reliable tool to capture these offenses. People who are interviewed for the survey and report a victimization are asked a series of questions about the incident and its outcomes, including whether it was reported to the police.
The Unique Contributions of the NCVS
These data have been used to provide additional details about crimes that have not been available in the FBI’s SRS, such as intimate partner violence, stalking, identity theft, and school crime. The NCVS also provides the public and policymakers with the reason why someone did not report their victimization to police and whether they sought help from a victim service agency. Another distinctive feature of the NCVS is that it captures unique sociopsychological harms, such as strained personal relationships, that may result from victimization. These measures are either so inconsistently reported to law enforcement, the trends are unreliable, or are not measured by any other national survey.
Another reason for potential discrepancies in the UCR and NCVS crime rates is that the two sources cover overlapping but distinct populations. As a household-based survey, the NCVS does not include people who are homeless or those who reside in institutions such as prisons, jails, and nursing homes; nor does it include crimes of violence against individuals under 12 years of age. Incidents experienced by people in these groups are included in UCR data if the incident is reported to the police.
In addition, while the NCVS excludes crimes against businesses or other organizations (offenses such as commercial burglary, shoplifting, and motor vehicle theft), these incidents are included in the UCR. If there are disproportionate changes in commercial- and household-based thefts, or in violence against persons who are included and excluded from the NCVS sample, then differences in the UCR and NCVS rates and changes in those rates would be expected.
Timing Differences
The time periods covered by the UCR and NCVS annual reports, which overlap but are not identical, may also contribute to differences in the crime rates and trends documented by the two sources. The most recent 2023 UCR report includes incidents that occurred during 2023, while the latest NCVS report is based on victimizations reported to interviewers during 2023. NCVS interviews take place during each month of the year and ask participants about offenses that took place in the previous six months; therefore, victimizations reported in the earlier months of 2023 may have occurred in the later months of 2022, and some incidents occurring in later 2023 will not be reported until interviews are conducted in 2024.
The use of interview year data for the annual NCVS reports is done so that crime information is released in a timely manner, and this decision does not often have large effects on annual rates. But if there are periods of rapid change in victimization, it can have an impact, depending on when during the year the change is concentrated. This dynamic occurred between 2020 to 2022.
Changes to Methodology and Collection
Since its first report in 1973, the NCVS (originally called the National Crime Survey, or NCS) has undergone several technical and methodological changes, including questionnaire changes in 1992-1993 that were designed to improve victimization recall and produced higher rates of some forms of violence. More recently, the administration of the NCVS was temporarily modified during the emergence of the COVID-19 pandemic. The changes included suspending in-person interviews for new sample participants and modifying the procedure for existing participants. While BJS notes that the 2020 NCVS is comparable to other years, it is possible that even a small change might appear large in the annual change estimates when compared to the UCR. For more detailed coverage of the NCVS methodology, please visit the Bureau of Justice Statistics page on the NCVS.6
Though not usual, administrative changes can also result in differing trends between the UCR and the NCVS. In 2021, the FBI replaced the SRS with the NIBRS. In 2020, approximately 95% of the U.S. population was covered by an agency reporting either SRS or NIBRS data. But in 2021, the share of the population covered dropped to about 65% because the FBI did not accept SRS-formatted data. As of this writing, the FBI had reverted to accepting both SRS and NIBRS data. This change, along with an increase in the number of agencies submitting NIBRS data, had improved population coverage to 94% by 2023.7 While the FBI attempts to adjust national crime estimates to account for missing or incomplete reporting from law enforcement agencies, large changes in the number of reporting agencies can affect yearly crime rates.
While both the UCR and NCVS are crucial to furthering our understanding of crime and victimization in the U.S., each data source has unique strengths and weaknesses. When the findings of the UCR and NCVS conflict, it does not automatically suggest that one data source or the other is presenting inaccurate information. Rather, the distinctive factors that characterize the sources likely contribute to the differences.
UCR and NCVS rates are expected to differ because the NCVS includes incidents both reported and unreported to police, covers a slightly different population as a household survey, and is based on incidents reported by interview year and not incident year.
UCR and NCVS Trends
Long-term Trends
This section examines three types of offenses: serious non-lethal violent crime, burglary, and motor vehicle theft. As noted above, the primary reason crime rates can differ across the two data sources is the inclusion in the NCVS of crimes that people did not report to law enforcement. When the NCVS rates are limited to victimizations that were reported to police, the differences between the NCVS and UCR rates narrow. Figure 1 shows the general decline in serious violent crime across the UCR and NCVS from 1993 to 2023.8 As expected, reported and unreported violent crime rates from the NCVS exceeded the rates reported to police in both the data sources during this period. Beginning in 2000, UCR rates were not significantly different from the adjusted NCVS victimization rate of those crimes that were reported to police for any sustained period.9
Figure 1. UCR and NCVS Serious Violent Crime Rates, 1993–2023
In addition to the differences in population and periods covered (i.e., report year compared to incident year), the NCVS is a household sample survey, and victimization rates are point estimates within a given margin of error. This means that while the victimization rate is a specific value, the true victimization rate likely falls within a range of values. For instance, the serious violence rate for victimizations reported to police in 2023 provided in Figure 1 is 3.8 per 1,000, but there is a 95% probability that the true victimization rate falls between 2.9 and 4.7. UCR annual rates are also technically estimates since the FBI imputes missing data when providing national crime estimates, but the range of possible values is generally not reported.10
Figure 2 shows that burglary rates generally declined from 1993 to 2023. In contrast to the trends for violent offenses described above, there was little difference between the UCR and reported-to-police NCVS rates for burglary. Both data sources show declines in burglary during the 1990s, relatively little change from about 2000 to 2012, and continued declines thereafter.11
Figure 3 shows the motor vehicle theft rates from 1993 to 2023. Long-term trends in motor vehicle theft from 1993 to 2023 are generally similar regardless of whether they were derived from UCR or NCVS data. Vehicle theft trends were flat or declining from approximately 2010 to 2015, and thereafter they began increasing, especially from 2021 to 2023. Also, the rates of motor vehicle theft are higher in the UCR data than the rates derived from NCVS data.
Figure 3. UCR and NCVS Motor Vehicle Theft Rates, 1993–2023
Some of these differences may be the result of thefts in which motor vehicles were owned by commercial or other organizational entities that were not surveyed by the NCVS. Even if that’s the case, however, victimization information from these sources is likely not high enough to fully explain the gap. For instance, according to 2023 NIBRS data, about 90% of motor vehicle theft victims were individuals. Additionally, like many other offenses, the definitions used by the FBI are similar and would not explain why the UCR rates are consistently higher than those derived from the NCVS.
These three trends highlight an important point about comparing the UCR and NCVS: long-term trends generally tell the same story, but short-term comparisons are more sensitive to methodological differences.
Short-term Trends
Turning to short-term trends, Figure 4 shows the impact of using NCVS 2018 to 2022 estimates based on interview year versus incident year while also considering reporting to police. The interview or incident year distinction is important when comparing periods of rapid social change (e.g., a global pandemic) because rates can shift quickly, and these shifts will be more evident in year-over-year comparisons. The data in this analysis end in 2022 because the full NCVS incident-year data for 2023 were not available at the time of publication.
Figure 4 compares serious violent crime rates using the UCR and different adjustments to the NCVS for 2018 to 2022. The first two column groups show the NCVS serious violence rate by interview year (i.e., as presented in BJS reports) and by incident year (i.e., when the incident occurred). In 2018 and 2019, there was very little difference between the rates based on these distinctions. The differences between the NCVS interview and incident year rates begin to widen, however, beginning in 2020. For example, in 2021, the interview year rate of 4.7 was approximately 30% higher than the incident year rate of 6.1. Further analysis (not shown) indicates that this larger gap was the result of notable increases in violent incidents during the second half of 2021, many of which would not have been captured until the 2022 interview cycle started. For an illustration of how monthly reported-to-police crime patterns have shifted in a sample of cities in recent years, please visit the Council’s crime trends reports.
As shown in the third through fifth column groups, the NCVS reported-to-polices rates are generally closer to UCR rates. Typically, the incident year rates of NCVS serious violence reported to police should most closely match the UCR, since the UCS is based on incident year. But this is not always true. For instance, in 2020, the serious violent crime rate for incident-year NCVS reported-to-police incidents was 2.4, while the UCR rate was 3.8. This shows that even when attempting to match the NCVS to the UCR, there are still individual years where the rates can be modestly different.
Figure 4. Serious Violent Crime Rates Using UCR and NCVS, 2018–2022
Figure 5 also shows the year over year percentage change in serious violent crime for the various measures. Apparent in this figure is the relative instability from one year to the next of the NCVS estimates compared to the UCR. Some of this is expected because the NCVS rates are based on sample survey data. But a notable part of this instability is due to the statistically low base rates of victimization in recent years. When victimization rates are low, a change in the annual rate can appear substantial even though the higher new rate remails low compared to the long-term trends. For example, an increase of 1 per 1,000 in violence would represent a 33% increase if the change is from 3 to 4 per 1,000. But a 1 per 1,000 increase in violence based on a change from 10 to 11 per 1,000 would result in a smaller percentage increase of 10%.
Lastly, it is possible that changes in UCR rates may also reflect law enforcement’s capacity to record crime. An analysis of UCR and NCVS rates from 1973 to 1992 suggests that the increase in UCR violent crimes observed during those years was attributable, in part, to increased police productivity.12 Thus, it is possible that a decrease in police productivity, such as a decline in staffing or proactive enforcement activities, may hinder law enforcement’s ability to record crime. Additional analysis is needed to determine if this dynamic applies to more recent years.
Figure 5. Serious Violent Crime Percentage Change Using UCR and NCVS, 2018–2022
These findings underscore that year-over-year comparisons between the UCR and NCVS should be made cautiously, but not ignored. Comparing the two sources may provide early evidence of emerging changes to crime and criminal victimization patterns. At the same time, however, some level of difference is expected given that the UCR and NCVS measure crime differently.
Conclusions and Implications
One benefit of having two separate systems to measure national crime levels is that each source can reveal information that the other cannot. UCR data, for example, are particularly valuable for providing jurisdictional level insights about crimes that come to the attention of local law enforcement. NCVS data cannot provide jurisdictional level data because of the costs associated with obtaining a sufficient sample size across the many jurisdictions in the U.S. In the future, however, it may be possible to develop local estimates of underreporting for some offenses, such as domestic violence, using the NCVS. Given its interview-based format, the NCVS also does not provide information on homicide, but comparisons to public health data sources suggest this crime is reliably measured in the UCR program. Overall, the NCVS does contribute unique and reliable information about crime that is independent of law enforcement agencies and provides information about victimization that is not collected by the UCR program. This independence is an important check on the accuracy of law enforcement data, especially given survey results showing that about 44% of Americans believe there is more crime than what is reported by the FBI, and 35% believe there is more crime than that reported by local police.13
UCR and NCVS crime and criminal victimization rates, as well as divergent estimates of annual change, are likely influenced by differences in their methods. Changes in any single year should be viewed in the context of long-term trends, where both data sources show an appreciable decline in violent crime since the early 1990s, as well as generally similar rates of key types of crimes that are known to the police. During years in which important events occur, either shifting victimization groups (e.g., an increase in commercial burglary) or causing variation in police reporting (e.g., drop in police trust), it is particularly helpful to use both data sources to understand whether changing conditions may or may not be influencing crime before discrediting either source.
As noted by the Council’s prior brief comparing the UCR and NCVS violent crime data for 2021 and 2022, the two sources told quite different stories about that period. During the 2024 presidential election campaign, the conflicting findings fueled continuous debate about whether crime was going up or down.14 Complicating matters, no office in the federal government investigated the competing UCR and NCVS year-to-year percentage change estimates in an attempt to explain the discrepancy. Further exacerbating the situation, in 2024 the FBI made significant revisions to UCR crime counts and rates for recent years without explaining why and how those revisions were made, a standard practice in statistical reporting. While the UCR and NCVS may not be perfectly aligned in year-to-year comparisons, and revisions can be performed for legitimate purposes, the absence of clear explanations can sow confusion and undermine public trust in crime data.
The findings of this brief support the recommendation of the CCJ Crime Trends Working Group to shift from the FBI to the Bureau of Justice Statistics the responsibility for analyzing and publicly reporting the crime data collected by the FBI from law enforcement agencies throughout the U.S. Requiring BJS to explore and explain notable discrepancies between the two sources each year would help illuminate the complexity of crime in the U.S., strengthen transparency, and enhance public trust.
Appendix
Table A1. Percentage of Victimizations Reported to Police
About the Authors
Janet Lauritsen is Curators’ Distinguished Professor Emerita at the University of Missouri-St. Louis, former chair of the Committee on National Statistics’ Panel on Modernizing the Nation’s Crime Statistics, and the 2022 president of the American Society of Criminology. She was a member of the Council on Criminal Justice Crime Trends Working Group.
Ernesto Lopez is a senior research specialist with the Council on Criminal Justice.
Acknowledgments
Stephanie Kennedy, policy director at the Council on Criminal Justice, provided editing and guidance to the authors for this report; other members of the Council on Criminal Justice team provided editing and additional support.
This paper was produced as part of the work of the CCJ Crime Trends Working Group with support from the Annie E. Casey Foundation, Arnold Ventures, the Harry Frank Guggenheim Foundation, Southern Company Foundation, Stand Together, and CCJ’s general operating contributors.
Suggested Citation
Lauritsen, J., & Lopez, E. (2025). When crime statistics diverge: Understanding the two major sources of crime data in the U.S. https://counciloncj.org/when-crime-statistics-diverge/
Endnotes
1 Pew Research Center. (2022). Public’s top priority for 2020: strengthening the nation’s economy. https://www.pewresearch.org/politics/2022/02/16/publics-top-priority-for-2022-strengthening-the-nations-economy/
2 Brenan, M. (2024, Oct 29). Smaller majorities say crime in the U.S. is serious, increasing. Gallup. https://news.gallup.com/poll/652763/smaller-majorities-say-crime-serious-increasing.aspx
3 Doan, L. (2024, Oct 16). Harris says violent crime is down. Trump says it’s up. Here’s a factcheck. CBS News. https://www.cbsnews.com/news/violent-crime-rate-trump-harris-fact-check/
4 See for example: Committee on Oversight and Government Reform. (2024). Comer demands transparency from FBI about quietly revised crime statistics. https://oversight.house.gov/release/comer-demands-transparency-from-fbi-about-quietly-revised-crime-statistics/
5 Morgan, R. & Thompson, A. (2022). The nation’s two crime measures, 2011-2020 (NCJ 303385). Bureau of Justice Statistics. For additional information see this presentation: Bureau of Justice Statistics. (2024). The Nation’s Two Crime Measures. https://bjs.ojp.gov/library/publications/nations-two-crime-measures-2011-2020; https://bjs.ojp.gov/media/video/71486
6 Bureau of Justice Statistics. (2023). National crime victimization survey (NCVS). https://bjs.ojp.gov/data-collection/ncvs
7 The 94% population coverage includes NIBRS (83% coverage) and SRS reporting agencies (11%).
8 The UCR data used in this report are from FBI Table 1 in the Crime in the United States by Volume and Rate, accessed Oct 3, 2024. NCVS data in this report are from National Crime Victimization Survey, Concatenated File, [United States], 1992-2023 (ICPSR 38963), Version Date Sept 11, 2024.
9 To make the UCR and NCVS estimates more directly comparable, the analyses in this section distinguish between total victimization in the NCVS and victimization that was reported to police. NCVS violence rates are also limited to the types of non-lethal violence measured in the UCR SRS data (i.e., rape and sexual assault, robbery, and aggravated assault). To further aid in the comparison, these analyses use the U.S. resident population provided in the UCR for the NCVS analysis. Since the NCVS does not include some populations (e.g., individuals under age 12), the population counts used by BJS to generate NCVS rates are lower than UCR population counts.
10 A recent exception was in 2021 when the FBI only accepted data from NIBRS-reporting agencies.
11 The similar rates of reported to police (NCVS) rate and the UCR correspond to an increase in the share of residential burglaries relative to non-residential burglaries. See CCJ’s fact sheet on burglary for more information.
12 O’Brien, R. M. (1996). Police productivity and crime rates: 1973‐1992. Criminology, 34(2), 183-207. https://doi.org/10.1111/j.1745-9125.1996.tb01202.x
13 Faucheux, R. (n.d.). National survey of U.S. Adults on the American criminal justice system September-October 2024. CrimeChannel.org. https://irp.cdn-website.com/47fdcd9f/files/uploaded/CRIME_CHANNEL_NATIONWIDE_SURVEY_REPORT_2024.pdf
14 For a detailed discussion see: Farley, R. (2024, May 3). Trump’s bogus attack on FBI crime statistics. FactCheck.org. https://www.factcheck.org/2024/05/trumps-bogus-attack-on-fbi-crime-statistics/
