Does Gun Control Reduce Violence?

Gun control is not one policy, so the question has to be asked law by law. The RAND Corporation's review of the research rates the evidence on 18 kinds of state gun law, most of which restrict guns and a few of which loosen the rules.

RAND, a nonprofit research organization, ran its Gun Policy in America project for a decade, from 2016. The project kept only the studies that met fixed methodological criteria and rated the evidence they give. The fifth and final edition of its review, published in January 2026, covers research published through December 2024.

RAND set out to rate the 18 kinds of law against 8 outcomes: suicide, violent crime, unintentional injuries and deaths, mass shootings, police shootings, defensive gun use, hunting and recreation, and the gun industry. That makes 144 pairings of a law with an outcome.

After reviewing several thousand candidate studies, RAND found 207 that met its criteria. Those studies gave evidence on 68 of the 144 pairings. In 18 pairings they showed some evidence of an increase or decrease in an outcome, and in 50 they were inconclusive. For the remaining 76 pairings, no qualifying study existed at all.

An inconclusive rating is a statement about the studies, not about the law. The rest of this article depends on one sentence of RAND's: "where we conclude that evidence for a policy is weak or inconclusive, that does not mean that the policy is ineffective; the policy itself might well have important effects." No rating on RAND's scale is a finding that a law has no effect.

RAND's review says the same of policies with limited evidence: "They may or may not have the effects they were designed to produce; available scientific research cannot yet answer that question."

RAND gives several reasons a rating can stay weak. A law may have been adopted in too few states, or too recently, for outcome data to build up. An outcome may be too rare, as mass shootings are, for a study to detect a change. Some of the data needed are not collected reliably. Laws that govern who may buy a new gun also act on a small share of the roughly 400 million firearms Americans held in 2023. So their effects are hard to detect even if they work as intended.

Moderate or stronger evidence that a law reduces deaths or crime exists for 5 of the 18 kinds of law.

RAND calls this scale relative: it compares the evidence within gun policy research, not with other fields. All the studies RAND included were observational, and it says the study of gun policy is in its infancy compared with the study of smoking and cancer.

Six of the 144 pairings reach RAND's top tier. They belong to four kinds of law. Two of the six are findings that a law increases homicide.

Child-access prevention laws let prosecutors charge adults who intentionally or carelessly let children have unsupervised access to a gun.

RAND suggests why the evidence is clearer for three of the four kinds of law. Child-access prevention, concealed-carry and stand-your-ground laws govern how existing guns are stored, carried or used, rather than who may buy a new one. So they may have more chance of producing effects a study can detect.

The two increase findings come from laws whose case rests partly on deterrence. RAND writes that stand-your-ground laws are intended to reduce barriers to self-defense and deter crime, and that laws letting more people carry guns could deter attackers who fear armed victims.

RAND notes that concealed-carry permit holders are rarely convicted of violent crimes, so their own offending is unlikely to explain the rise in homicide. It offers candidate explanations, such as more carried guns being stolen and passing into criminal markets, and says these may or may not explain the effect.

In a 2005 report, a National Research Council panel reviewed John Lott's conclusion that right-to-carry laws, also called shall-issue laws, reduce violent crime. The Council is the principal operating agency of the National Academy of Sciences. The panel found that the evidence could not establish a causal link between those laws and crime rates. Its one dissenting member, James Q. Wilson, read the evidence as showing that the laws reduce murder. The other 14 members replied that "the scientific evidence does not support his position."

Below the top tier, RAND's ratings of an effect are moderate or limited, and each holds only for particular outcomes.

RAND finds moderate evidence that checks by licensed dealers reduce firearm homicides, by about 5% on average across the studies. For universal checks, which also cover private sales, it finds moderate evidence of fewer total homicides, about 3% fewer on average. It finds limited evidence that universal checks reduce firearm homicides.

RAND finds moderate evidence that barring people under domestic violence restraining orders from having guns reduces intimate partner homicides, both in total and by firearm.

The 1994 federal ban required a study of its effects. The research team that carried it out updated the study for the National Institute of Justice in 2004. If the ban were renewed, these evaluators wrote, its effects on gun violence would likely be small at best and perhaps too small for reliable measurement.

Assault weapons were used in about 2% of gun crimes before the ban, according to most studies, and in no more than 8%. The ban had also not yet reduced the criminal use of large-capacity magazines, likely because millions of pre-ban magazines were exempt.

RAND rates bans on assault weapons and high-capacity magazines together. It finds limited evidence that bans on high-capacity magazines reduce mass shootings and their fatalities, and inconclusive evidence on violent crime.

RAND finds moderate evidence that waiting periods reduce firearm suicide and moderate evidence that they reduce total homicide. A 2017 study in the Proceedings of the National Academy of Sciences used 45 years of state data, 1970 to 2014. It estimated that waiting periods cut gun homicides by roughly 17%. Its authors write that the size and significance of their suicide estimate vary across model specifications.

Red flag laws allow orders that remove guns from people judged to pose a risk. RAND finds limited evidence that they reduce firearm and total suicide. Its rating rests on three studies. A study across states from 1990 to 2018 found about 4% fewer suicides and 6% fewer firearm suicides after the laws passed. The remaining two, of Connecticut and Indiana and of San Diego County, found a suggestive fall in Indiana's firearm suicide rate and uncertain or imprecise results elsewhere.

The best-known estimates of red flag laws' effect on suicide come from separate studies of Connecticut and Indiana. The people whose guns were removed had suicide rates about 40 times the general adult population's in Connecticut and 31 times in Indiana. The Connecticut study estimated roughly one suicide averted for every ten to twenty removals, and the Indiana study one for every ten. A 2024 update of the Connecticut estimate, with different assumptions, put it at one for every 22.

RAND did not count the Connecticut estimates toward its rating, because they extrapolate suicide attempts rather than compare with a control group. Like the Connecticut estimates, case series of orders used against people who threatened mass shootings have no comparison group, so they cannot show that any shooting was prevented. RAND rates the evidence on mass shootings inconclusive.

If you are thinking about suicide, or worried about someone who is, this list of helplines gives numbers to call or text in several countries.

RAND upgraded permit-to-purchase laws in the fifth edition from inconclusive to limited evidence of reducing total and firearm homicide. Most of the studies follow one or two states, such as Connecticut, which adopted a permit-to-purchase law in 1995, and Missouri, which repealed its law in 2007. RAND calls the studies together compelling evidence that these laws are associated with fewer homicides. Its scoring rule still caps the rating at limited, because only one study that found a significant effect rested on more than three states changing their laws.

Estimates from so few states are unreliable guides to other states, and other changes can coincide with a law. For example, Missouri's repeal came at the start of the Great Recession. Two authors of the Missouri and Connecticut studies, Daniel Webster and Cassandra Crifasi, have also argued in print that states should prioritize purchaser licensing, as they call such laws.

RAND rates the evidence on gun-free zones inconclusive for violent crime, and finds no qualifying studies for anything else. For armed staff in schools and for lost and stolen firearm reporting, it finds no qualifying studies on any outcome.

Suicides were 62% of American gun deaths in 2024, though, as Pew Research Center notes, they tend to get less public attention than gun homicides.

In the CDC's figures as Pew presents them, which go back to 1968, gun suicides have outnumbered gun homicides in every year since 1981. So a figure for all gun deaths, presented as gun violence, is mostly about suicide, and that changes which laws could matter. The laws with top-rated evidence of reducing deaths, child-access prevention and minimum-age laws, act on young people, and two of their four top-rated findings concern suicide.

The suicide share also raises the objection that people who cannot reach a gun will use something else. For up to twelve years, a California cohort study followed 26.3 million adults who had not acquired a handgun. Those who acquired a handgun were about four times as likely to die by suicide as non-owners of the same age, sex and race living in the same neighbourhood.

All of the extra risk came from firearm suicide; the new owners' suicides by other methods were no higher. The risk of firearm suicide was highest just after the first handgun was acquired and stayed raised for years. The study is observational, so it cannot show by itself that removing access prevents suicides.

RAND's ratings add evidence on laws, with moderate evidence that child-access prevention laws reduce total suicides among young people, and limited evidence that waiting periods and red flag laws do. None of this shows that people never turn to another method.

Download this conclusion: suicide

This section as one image: the 2024 split of American gun deaths, 27,593 suicides and 15,364 homicides of 44,447. It also shows the cohort study of suicide risk after people acquired a handgun.

Download the card (PNG, 1,080 × 1,350)

Estimates of how often Americans use a gun defensively run from about 116,000 a year to about 2.5 million or more. The 116,000 comes from the National Crime Victimization Survey (NCVS) of 1992 and 1994, the 2.5 million from a 1993 telephone survey. The National Research Council found that the two estimates differ by a factor of nearly 22.

The gap comes from how the surveys define defensive gun use and how they ask about it. The NCVS is large and run for the government. It asks about self-protection only of people who first report being victims of certain crimes, and it never asks about gun use directly. RAND says the NCVS figure almost certainly undercounts. In the 1993 survey, 66 respondents out of 4,977 reported a defensive use in the past year. The 2.5 million figure extrapolates them to the national adult population.

In a survey of something rare, a small share of false reports can inflate the estimate a great deal. David Hemenway, a health policy professor at Harvard, calculated that the 1993 survey implies 845,000 defensive uses during burglaries. That is more than the estimated number of burglaries of gun owners who were at home and awake.

Gary Kleck, one of the survey's authors, replied in 1999, as RAND summarizes him, that official burglary counts are themselves too low. RAND treats the 2.5 million figure as not plausible against more trustworthy information and treats the frequency question as unresolved.

RAND found five qualifying studies of any law's effect on defensive gun use, and it rated the evidence inconclusive for every law they covered. On hunting and recreation it found no qualifying studies at all, and it notes that both outcomes are frequently raised in gun policy debates.

Download this conclusion: defensive gun use

This section as one image: the two survey estimates of defensive gun uses a year, 2.5 million and about 116,000, and why RAND treats the higher one as not plausible.

Download the card (PNG, 1,200 × 630)

The 40% traces to a telephone survey in November and December 1994 about guns acquired in 1993 and 1994, in which 251 acquisitions were reported. Its authors concluded that about 60% of acquisitions went through a licensed dealer, so about 40% did not. The Brady Act's dealer checks on handgun buyers began only in February 1994.

A 2015 survey of 1,613 gun owners, published in 2017, put the modern figure lower. Of those who had acquired a gun in the past two years, 22% did so without a background check. Among private purchases, the share without a check was 57% in states that do not regulate private sales and 26% in states that do. The two surveys asked different questions, and the 2015 survey's lead author said it is not clear how much of the fall from 40% to 22% that explains.

The 43 is a raw count ratio from a 1986 study of gun deaths in homes in King County, Washington, from 1978 to 1983, not an adjusted risk estimate. Most of the 43 were suicides, 37 for every self-protection homicide.

The same lead author's later case-control study of gun ownership reported an adjusted odds ratio of 2.7 for homicide in the home. The National Research Council noted that reading that ratio as a cause assumes that owning a gun is otherwise unrelated to the risk of homicide.

The 94% is from the Crime Prevention Research Center, founded by John Lott. It counts only shootings in public places, leaves out those tied to gang violence or other crimes, and uses a wide definition of a gun-free zone.

Everytown for Gun Safety, a gun-control advocacy group, found about 10%, counting shootings with four or more killed anywhere from 2009 to 2016. A Washington Post fact-check applied Lott's method to the same years and got about 86%. The Center's and Everytown's counts differ in what they call a mass shooting and a gun-free zone. So the two numbers do not describe the same set of events.

The 31% is from a 2016 study by Adam Lankford of public mass shooters in 171 countries from 1966 to 2012. A 2019 rebuttal by John Lott and Carlisle Moody counted many more foreign attacks, including attacks by groups, from 1998 to 2012. It put the United States at less than 3% of mass shooting incidents and deaths, and less than 1% of shooters.

Lankford replied that Lott and Moody's own data give the United States more than six times its share of the world's population in public mass shooters who attacked alone. The exchange continued into 2020 without agreement on which attacks count.

Mass shooting counts also differ because databases define the term differently. Databases reporting several hundred mass shootings a year define a mass shooting as four or more people shot, anywhere. Databases reporting a handful define it as four or more killed, in public. For 2024, RAND lists 502 from the Gun Violence Archive, which counts four or more shot. It lists 3 from the Violence Project, which counts four or more killed in public. So a figure from a database of people shot cannot be set beside a figure from a database of people killed.

Download this conclusion: four numbers

This section as one image: where the 40%, the 43 times, the 94% and the 31% come from, and what the counter-evidence on each found.

Download the card (PNG, 1,080 × 1,350)

Some laws do, and RAND's ratings say which. Six of its 144 pairings of a law with an outcome reach the top tier. Four show child-access prevention and minimum-age laws reducing gun deaths and injuries among children and young people. Two show stand-your-ground and shall-issue concealed-carry laws increasing homicide. Below that tier, background checks and waiting periods have moderate evidence of reducing homicide or suicide. For 126 of the 144 pairings the studies are inconclusive or missing, so on most questions the research has not yet given an answer either way.

Download the conclusion

The verdict as one image, with RAND’s 144 pairings as its chart: 18 with an effect found, 50 inconclusive and 76 with no qualifying study. It is 1,200 by 630 pixels, the size social networks use for link previews, and anyone may download it and post it.

Download the card (PNG, 1,200 × 630)