The evidence on safety and justice in the United States comes from two federal crime measures that count different things: police-reported offenses compiled by the FBI, and the Bureau of Justice Statistics’ National Crime Victimization Survey, which also captures crimes never reported to police. Both show violence far below its early 1990s level; the FBI estimated a violent crime rate of 327.6 per 100,000 inhabitants in 2025, and the survey measured 23.3 violent victimizations per 1,000 persons age 12 or older in 2024, compared with 79.8 in 1993.1,2 The harder questions, such as how much prison, policing or a given gun law changes crime, rest on a smaller and more contested body of research.
In plain words
Safety and justice evidence starts with counting crime. There are two main ways the government counts crime. One counts crimes that police know about. The other asks a large sample of households whether they were victims, whether or not they called the police. The two give different numbers because about half of violent crimes are never reported.
Both counts say violence is much lower than in the early 1990s. Both also have weak spots. The police count went through a big change in 2021, and many departments did not report that year. The survey cannot count murders, because it interviews victims.
Counting crime is easier than explaining it. Researchers still disagree about how much prisons and police reduce crime, why some groups are imprisoned at much higher rates, and what many gun laws do.
The companion pages explain what the public prioritizes and who decides.
Defined terms
Safety and justice data use several technical terms.
Uniform Crime Reporting (UCR) Program. The FBI program that compiles crimes reported by law enforcement agencies. Its national estimates fill in for agencies that did not report complete data.1
Summary Reporting System (SRS). The older UCR format, which under the hierarchy rule required agencies “to report only the most serious offense if an incident included several crimes”.6
National Incident-Based Reporting System (NIBRS). The current UCR format, which “allows law enforcement to report up to 10 co-occurring offenses per single incident” with details on victims, offenders and circumstances.6
National Crime Victimization Survey (NCVS). A BJS household survey of persons age 12 or older that collects “information on nonfatal personal crimes” and household property crimes, “both crimes reported and not reported to police”.2
Clearance. A reported offense counted as solved “by arrest or exceptional means” in FBI data.1
Imprisonment rate. Sentenced prisoners (more than one year) per 100,000 residents.7 It excludes most people in local jails.
Crime levels: what the data show
Crime levels in the United States, by every federal measure, fell sharply from the early 1990s and fell again after a rise around 2020.
Police-reported crime. The FBI estimated 1,119,768 violent crimes in 2025, a rate of 327.6 per 100,000 inhabitants, down from 362.9 in 2024; violent crime fell an estimated 9.3% and murder and nonnegligent manslaughter 18.1%.1 The 2025 murder rate of 4.1 per 100,000 matched 1955 and 1956 as the lowest since FBI estimates began in 1936, after a 20-year high of 6.6 in 2020 and 2022.1 Property crime fell an estimated 12.4%, and motor vehicle theft about 23%.1
A second police-based estimate. BJS and the FBI also publish estimates built only from NIBRS data. That series put the 2025 violent crime rate at 335.2 per 100,000 persons, an 11% decrease from 376.1 in 2024, with homicide at 4.3 per 100,000.3 The two police-based series differ by several points because they use different methods; the FBI says they “should be viewed as analogous measures rather than as one being more accurate than the other”.1
Victimization survey. The NCVS measured 23.3 violent victimizations per 1,000 persons age 12 or older in 2024, not significantly different from 2023 and higher than in 2020 (16.4) and 2021 (16.5).2 In 1993 the rate was 79.8.2 About 1.45% of persons age 12 or older experienced at least one violent victimization in 2024.2
Death records. The CDC’s vital statistics, which come from death certificates rather than police reports, counted 20,162 homicide deaths in 2024, a rate of 5.9 per 100,000 population.4 That is higher than police-based estimates for the same year, 5.1 per 100,000 in the FBI series and 5.2 in the BJS NIBRS series,1,3 because the systems use different definitions and sources, a reminder that even the most serious crime is not counted one way.
Reported versus unreported crime: NCVS and UCR compared
The two main federal measures of crime and safety, the NCVS and the UCR, differ in what they count, which is why they cannot be read as two estimates of one number.
Coverage. The NCVS “excludes homicide because the National Crime Victimization Survey (NCVS) is based on interviews with victims,” and it includes simple assault, which the FBI’s violent crime total does not.2,1 It excludes people living in institutions such as prisons and hospitals.2
Reporting. About 48% of violent victimizations and 30.5% of property victimizations were reported to police in 2024.2 The share varies widely by offense: 73% of robberies but about 24% of rape or sexual assault victimizations were reported, down from 46% for rape or sexual assault in 2023.2 Victims cite reasons such as “fear of reprisal or getting the offender in trouble” and “believing that police would not or could not do anything to help”.2 A change in police-reported crime can therefore reflect a change in crime, a change in willingness to report, or both.
Precision. The NCVS is a sample. In 2024, half of households received a redesigned questionnaire, so the published legacy estimate rests on 69,396 household interviews, about 60% of eligible households.2 BJS later reported that the redesigned instrument produced a violent victimization rate 23% higher, 28.7 per 1,000, than the legacy instrument’s 23.3, and said that “As expected, the 2024 NCVS-R generally produced larger estimates”.5 Survey design can shift the level of measured crime even when crime itself has not changed.
Timing. NCVS crimes are classified by survey year, and the 2024 survey covered crimes from July 1, 2023 to November 30, 2024.2 Police data generally follow the incident date.3
The NIBRS transition and data gaps
The FBI’s move to incident-based reporting improved the detail of safety and justice data but created a period of weak national coverage. “On January 1, 2021, the FBI attempted to fully transition law enforcement agencies from SRS reporting to NIBRS”; because many agencies were not ready, the FBI later accepted either format for 2022 and 2023, partly because of “concerns about the completeness of the 2021 data”.6 In 2021, agencies reporting at least one month of NIBRS data covered 64.8% of the U.S. population, rising to 89.5% in 2025.3 The FBI now states that its 2021 estimates “are below a statistically acceptable level to be nationally representative and are not comparable to other yearly estimates”.1
Gaps persisted after 2021. The Congressional Research Service reported that 14.8% of agencies reported no data to the UCR program in 2023, and that as of May 2024 fewer than half the agencies in Florida (18%), New York (31%) and Pennsylvania (11%) reported NIBRS data.6 Participation is voluntary.6 In 2025, the FBI’s combined NIBRS and SRS data came from 17,075 agencies covering 96.2% of the population.1
Estimates are also revised. The FBI’s first 2024 release put the violent crime rate at 359.1 per 100,000;8 its 2025 release updated the 2024 rate to 362.9.1 Year-over-year percentages compare estimates, each with its own imputation for missing agencies, which is why small changes deserve caution.
Evidence on policing, prison, pretrial detention and reentry
The causal evidence on safety and justice policy is mixed in strength. The findings below are among the most cited; each has critics.
Prison and crime. The National Research Council’s 2014 review found that “These changes in punishment policy were the main and proximate drivers of the growth in incarceration,” that most studies find incarceration reduced crime though “the evidence does not clearly show by how much,” and that “the crime-reducing effects of incarceration become smaller as the incarceration rate grows”.9 At yearend 2023, the imprisonment rate was 360 sentenced prisoners per 100,000 U.S. residents, down 25% from 479 in 2013.7 Of sentenced state prisoners at yearend 2022, 62% were serving time for a violent offense, 13% for a drug offense, 13% for a property offense and 12% for a public order offense.7
Deterrence. The National Institute of Justice summarizes that “Police deter crime by increasing the perception that criminals will be caught and punished” and that “There is no proof that the death penalty deters criminals”.10 Police cleared 47.4% of reported violent crimes and 17.4% of reported property crimes in 2025,1 so a large share of offenders face no arrest for a given offense.
Police staffing. Chalfin and colleagues (2022) estimate that larger police forces reduce homicides, with larger per capita benefits for Black victims, while increasing low-level arrests that disproportionately affect Black Americans; they also find that larger forces make fewer arrests for serious crimes, “implying that police force growth does not increase racial disparities among the most serious charges”.11
Pretrial detention. Using quasi-random assignment of bail judges, Dobbie, Goldin and Yang (2018) find that detention raises conviction rates, mainly through guilty pleas, decreases formal sector employment and “has no net effect on future crime”.12 Evaluations of reforms are still accumulating. An MDRC evaluation of New Jersey’s 2017 reform found that “rates of initial booking into jail were lower than predicted given pre-CJR trends” and that the reform reduced time in jail in the month after arrest, with effects on court appearance and new arrests left to later reports.13
Reentry. Of state prisoners released in 24 states in 2008, about 66% were arrested within 3 years, 82% within 10 years, and about 61% returned to prison within 10 years for a new sentence or a supervision violation.14 Arrest is not conviction, and returns for technical violations reflect supervision rules as well as behavior.
Racial disparities: the data and competing explanations
Racial disparities in safety and justice are large on several measures, and researchers disagree about how much of each gap reflects differences in offending, in enforcement or in policy.
Imprisonment. The adult imprisonment rate at yearend 2023 was 1,218 per 100,000 for Black residents, 606 for Hispanic residents and 231 for White residents; the Black rate fell 33% from 2013 to 2023.7
Victimization. In police-reported data, the 2025 violent victimization rate was 769.7 per 100,000 Black persons and 209.9 per 100,000 White persons; homicide victimization rates were 14.6 and 2.0.3
Victim reports of offenders. In the 2024 NCVS, victims perceived the offender as Black in 27.7% of violent incidents where offender race was known, compared with a Black share of 12.2% of the population age 12 or older; offender race was unknown in 19% of incidents, and these perceptions include simple assaults.2
Sentencing. The U.S. Sentencing Commission reported that Black males received federal sentences 13.4% longer than White males in fiscal years 2017 to 2021, and that differences “largely can be attributed to the initial decision of whether the sentence should include incarceration at all”.15
How the explanations differ. One explanation emphasizes differences in involvement in serious violence, pointing to victimization data and victims’ descriptions of offenders, both of which are independent of police decisions to arrest. Another emphasizes enforcement and policy: the National Research Council found that “Intensified enforcement of drug laws subjected blacks, more than whites, to new mandatory minimum sentences” and that long-sentence laws “have likely increased sentences and time served for blacks more than whites,” while also noting that Black Americans “had long been more likely than whites to be arrested for violence”.9 A third explanation treats both as products of concentrated disadvantage, which the data on this page do not test. These explanations are not mutually exclusive, and their relative weight is contested.
Gun violence: the data and the research
Firearm deaths in the United States, the part of safety and justice where crime and public health overlap, are mostly suicides, a fact often lost in debates framed around crime. In 2024 the CDC counted 27,593 firearm suicides (8.1 per 100,000) and 15,364 firearm homicides (4.5 per 100,000).16,4
RAND’s ongoing review of gun policy research, updated January 29, 2026, identified 207 qualifying studies covering 68 of 144 policy and outcome combinations; it found some evidence of an effect for 18, inconclusive evidence for 50, and no qualifying studies for 76.17 Its stronger ratings include supportive evidence that child-access prevention laws reduce youth firearm suicides and unintentional child deaths, supportive evidence that shall-issue concealed-carry laws and stand-your-ground laws increase homicides, moderate evidence that background checks and waiting periods decrease homicides, and limited evidence that high-capacity magazine bans decrease mass shootings.17
Defensive gun use is among the least settled questions. RAND notes estimates ranging from 116,000 incidents a year using the NCVS to 2.2 to 2.5 million in a 1995 study by Kleck and Gertz based on a private survey; it calls the high end “not plausible given other information that is more trustworthy” and says the NCVS figure “almost certainly underestimates the true number”.18
Evidence
The safety and justice evidence on this page follows a hierarchy. Counts come from federal statistical programs: the FBI’s UCR,1,8 BJS’s NCVS and NIBRS estimates,2,3,5 BJS correctional and recidivism series,7,14 and CDC death records.4,16 Causal claims come from peer-reviewed quasi-experimental studies,11,12 consensus or agency reviews,9,10,17 and program evaluations.13 For comparison, the World Prison Brief puts the overall U.S. prison population rate, including pretrial detainees, at 542 per 100,000 in 2023, with pretrial detainees 25.5% of the total.19 For how to judge any measure, see Measurement.
What is unknown
Several questions about safety and justice remain open despite decades of research.
- Why crime fell in 2024 and 2025. The data show the decline; this page found no consensus causal account.
- The true level of unreported crime. The redesigned NCVS questionnaire produced a violent victimization rate 23% higher than the legacy questionnaire for the same year,5 so the measured level depends on survey design.
- Local trends where reporting is incomplete. National estimates impute missing agencies; local figures in places with partial reporting may be unreliable.6
- Marginal effects today. Most incarceration research covers the era of rapid growth; the effect of further changes at today’s levels is less certain.9
- Long-run effects of pretrial reforms on court appearance, new offenses and case outcomes.13
- The effects of most gun laws, where RAND found no qualifying studies for more than half of policy and outcome pairs.17
Competing interpretations
Safety and justice evidence supports at least three broad readings.
“The system works better than people think.” The murder rate is tied for its lowest level since 1936 by the FBI’s measure,1 the imprisonment rate has fallen for a decade,7 and public fear has lagged the data. On this reading, policy should consolidate gains rather than overhaul the system.
“The numbers hide the problem.” About half of violence goes unreported,2 reporting coverage was thin in key years,6 and victimization rates differ several-fold across groups.3 On this reading, national averages understate what high-crime neighborhoods experience.
“The costs are uncounted.” Crime statistics measure offending, not the effects of enforcement on families, employment and trust that the National Research Council documented.9 On this reading, a full accounting would weigh those harms alongside crime reduction.
Each reading uses real data. They differ in what they treat as the main cost.
Application to this framework
Safety and justice evidence shapes how the framework presents results.
- Measurement discipline. When a participant’s priority cites “rising crime,” the framework can show both federal measures and their limits rather than pick one. See Measurement.
- Priority versus fact. A high priority for safety and justice is valid as a statement of concern even when crime is falling. The framework reports it as a priority, not as evidence about crime. See Priorities.
- Opinions on contested evidence. On questions such as bail reform or gun laws, opinions are asked with exact wording and voluntary sharing, and the evidence quality is shown alongside. See Opinions and Question design.
- Tradeoffs. Evidence about effects informs, but does not decide, how much safety a community will trade for cost or liberty. See Tradeoffs.
Hypothetical example: a demo result shows Safety and Justice ranked first in a county whose police-reported violent crime fell. The framework would display the priority beside the county’s reporting coverage and a note that victimization surveys are national, not county level, without claiming either number is “the truth.”
What My Vote Forward can and cannot do here
My Vote Forward can collect and display primary data with dates, denominators and caveats, and link readers to the original sources. It can show where measures disagree.
My Vote Forward cannot settle empirical disputes about the effects of prison, policing, pretrial detention or gun laws, and it does not take a side on the causes of racial disparities. It produces no crime statistics of its own and does not verify local agency reports. Its websites conduct no official votes, and the demo uses simulated data. A priority is not a policy mandate and not a measurement of crime.
Unresolved questions
Safety and justice evidence leaves several design questions open for the framework.
- Should a priority result be shown beside local crime data at all, given the risk that participants read the data as a rebuttal?
- How should the framework handle years, like 2021, when national police data are not comparable?
- When two official estimates differ, as the FBI and BJS NIBRS estimates do for 2025, which should a summary lead with?
- How can evidence quality ratings, such as RAND’s, be shown without implying a policy recommendation?
- What would it take for local victimization surveys to exist, and who would fund them?
Sources
- Federal Bureau of Investigation, Uniform Crime Reporting Program, “UCR Summary of Reported Crimes in the Nation, 2025,” released August 2026. Jurisdiction: United States. cde.ucr.cjis.gov/… (accessed 2026-10-26).
- Bureau of Justice Statistics, Susannah N. Tapp and Emilie J. Coen, “Criminal Victimization, 2024,” NCJ 310547, September 2025, tables 1, 4 and 11, appendix table 1 and Methodology. Jurisdiction: United States. bjs.ojp.gov/document/cv24.pdf (accessed 2026-10-26).
- Bureau of Justice Statistics, Lizabeth Remrey, Andrea M. Gardner and Erica L. Smith, “Crime Known to Law Enforcement, 2025,” NCJ 312174, October 2026, tables 1 and 5 and appendix table 16. Jurisdiction: United States. bjs.ojp.gov/document/ckle25.pdf (accessed 2026-10-26).
- Centers for Disease Control and Prevention, National Center for Health Statistics, “FastStats: Assault or Homicide” (National Vital Statistics System mortality data, 2024). Jurisdiction: United States. cdc.gov/nchs/fastats/homicide.htm (accessed 2026-10-26).
- Bureau of Justice Statistics, “Victimization Estimates from the Redesigned NCVS Instrument, 2024,” NCJ 312200, October 9, 2026. Jurisdiction: United States. bjs.ojp.gov/… (accessed 2026-10-26).
- Congressional Research Service, “The National Incident-Based Reporting System (NIBRS): Benefits and Issues,” R46668, first published February 1, 2021, updated July 7, 2025, text read via the EveryCRSReport.com mirror because congress.gov returned an access error. Jurisdiction: United States. congress.gov/crs-product/R46668 and everycrsreport.com/reports/R46668.html (accessed 2026-10-26).
- Bureau of Justice Statistics, Derek Mueller and Rich Kluckow, “Prisoners in 2023: Statistical Tables,” NCJ 310197, September 2025, highlights, table 6 and tables 16 and 17. Jurisdiction: United States. bjs.ojp.gov/document/p23st.pdf (accessed 2026-10-26).
- Federal Bureau of Investigation, Uniform Crime Reporting Program, “UCR Summary of Reported Crimes in the Nation, 2024,” released August 2025. Jurisdiction: United States. cde.ucr.cjis.gov/… (accessed 2026-10-26).
- National Research Council, “The Growth of Incarceration in the United States: Exploring Causes and Consequences,” 2014, Summary. Jurisdiction: United States. nationalacademies.org/read/18613/chapter/2 (accessed 2026-10-26).
- National Institute of Justice, “Five Things About Deterrence,” June 5, 2016. Jurisdiction: United States. nij.ojp.gov/… (accessed 2026-10-26).
- Aaron Chalfin, Benjamin Hansen, Emily K. Weisburst and Morgan C. Williams Jr., “Police Force Size and Civilian Race,” American Economic Review: Insights 4(2): 139 to 158, 2022. Jurisdiction: United States. aeaweb.org/… (accessed 2026-10-26).
- Will Dobbie, Jacob Goldin and Crystal S. Yang, “The Effects of Pretrial Detention on Conviction, Future Crime, and Employment: Evidence from Randomly Assigned Judges,” American Economic Review 108(2): 201 to 240, 2018. Jurisdiction: United States. aeaweb.org/… (accessed 2026-10-26).
- MDRC, “Evaluation of Pretrial Justice System Reforms That Use the Public Safety Assessment” (New Jersey Criminal Justice Reform), November 2019. Jurisdiction: New Jersey. mdrc.org/… (accessed 2026-10-26).
- Bureau of Justice Statistics, Leonardo Antenangeli and Matthew R. Durose, “Recidivism of Prisoners Released in 24 States in 2008: A 10-Year Follow-Up Period (2008 to 2018),” NCJ 256094, September 2021. Jurisdiction: United States (24 states). bjs.ojp.gov/… (accessed 2026-10-26).
- United States Sentencing Commission, “2023 Demographic Differences in Federal Sentencing,” November 14, 2023. Jurisdiction: United States (federal courts). ussc.gov/… (accessed 2026-10-26).
- Centers for Disease Control and Prevention, National Center for Health Statistics, “FastStats: Suicide and Self-Harm Injury” (National Vital Statistics System mortality data, 2024). Jurisdiction: United States. cdc.gov/nchs/fastats/suicide.htm (accessed 2026-10-26).
- RAND Corporation, Gun Policy in America, “What Science Tells Us About the Effects of Gun Policies,” updated January 29, 2026. Jurisdiction: United States. rand.org/… (accessed 2026-10-26).
- RAND Corporation, Gun Policy in America, “The Challenges of Defining and Measuring Defensive Gun Use.” Jurisdiction: United States. rand.org/… (accessed 2026-10-26).
- World Prison Brief, Institute for Crime and Justice Policy Research, Birkbeck, University of London, “United States of America” (2023 data). Jurisdiction: United States, comparative. prisonstudies.org/country/united-states-america (accessed 2026-10-26).
