Are increases in gun sales to blame for the homicide increase?

New reports published by the Centers for Disease Control (CDC) and the Johns Hopkins Center for Gun Violence Solutions discuss the recent surge in gun violence, outlining trends from recently-released 2020 gun death data. The results are making headlines.

As many of us know, homicides surged by nearly 30% during the first year of the pandemic, and gun-related homicides increased by 35%. When it comes to overall gun deaths, including suicides, there was a 14% increase from 2019 to 2020. A wide number of factors influence state-level gun death rates including demographics, socioeconomic factors, and gun policies. For example, places in severe poverty also seemed to have a greater increase in gun-related homicides. Counties with a small proportion living below the poverty line experienced a 22% increase in firearm homicides, while counties with higher poverty levels experienced an increase of 40% or more. It also seems like the homicide increase may have impacted mid-sized cities the hardest. Gun homicide also tends to occur in highly concentrated areas. One study found that, in 2015, 26% of all firearm homicides in the United States occurred in census tracts that contained only 1.5% of the population.

The recent rise in gun violence also corresponded with a record-breaking increase in firearm sales, with 22.8 million guns being sold in the United States in 2020. In 2021, U.S. gun sales declined slightly — to 19.9 million — representing the second-busiest year on record. Pre-pandemic, the record for yearly firearm sales was 2016, with 16.7 million guns sold. As of today though, gun buying has largely returned to pre-pandemic levels. In January 2022, nationwide gun sales were down by 42.6% relative to January 2021.

The 2020 increase in gun sales was largely due to first-time buyers, who accounted for one-fifth of gun sales. Interestingly, about half of the first-time buyers were women, yet women are much less likely than men to commit homicide. Rather than buying a gun with the intention of committing homicide, many people likely bought them for purposes of protection. A 2020 study showed that nearly 60% of gun owners in California feared that gun violence would happen to them, which was their primary reason for owning a gun. Still, many are citing the increase in gun sales as the reason for the uptick in gun violence. But it’s more likely that gun homicides are related to many factors, such as the ongoing mental health crisis, reductions in proactive policing, and increases in protests and political polarization. In fact, gun sales often increase amidst political polarization and with more restrictive gun control policies.

Considering that many people purchase firearms for the purposes of protection in response to community upheaval and violence, it seems plausible that increases in violence could have led more people to purchase firearms, rather than the other way around. Despite the fact that firearm purchases and gun crime surged at the same time, in order to determine that one event caused the other, three criteria must be established: 1) association between the two factors (they either increase together, decrease together, or move in opposite directions together), 2) proper temporal ordering of variables (i.e., the causal factor comes before the supposed effect), and 3) elimination of alternative explanations for the relationship (i.e., “non-spuriousness”). When it comes to the argument that “increased firearm purchases led to increased gun deaths,” the first criteria is met, though the latter two are not.

This argument forms the basis of the CDC and Johns Hopkins reports, which both claim that gun prevalence is a large factor increasing gun violence. Rather than focusing on gun-related homicides specifically though, the reports both focus on all gun-related deaths. The latter measure includes suicides, which encompassed 56% of gun-related deaths in 2020. But looking at gun-related deaths rather than specifically focusing on homicides might lead to misinterpretations. Of course, instances and rates of gun violence are inflated when looking at gun-related deaths instead of gun-related homicides.

The studies relied on data from the National Vital Statistics System (NVSS), which was integrated with population data, urbanization measures, and poverty measures. Population-based firearm homicide and suicide rates were examined by age, sex, race and ethnicity, geographic area, level of urbanization, and level of poverty. The findings do not support causal inferences, though, as briefly stated above. While there was a correlation between increases in gun sales and increases in homicides, there is one point that we cannot stress enough: correlation does not equal causation.

Neither report did an in-depth examination of the factors contributing to the rate increases, though several explanations are possible. These include reductions in proactive policing and calls to de-fund the police, as well as increases in protests and political polarization. One possibility mentioned in the report is that increases in gun violence are attributable to increases in domestic violence. The CDC report states that “access to firearms is one of the primary predictors of lethality in abusive relationships.” They support this claim by citing research that found increases in domestic violence during the pandemic. However, other research on this shows that increases in domestic violence were short-lived and decreased after stay at home orders ended. The Johns Hopkins study makes a similar claim, citing a meta-analysis — but the meta-analysis didn’t look at the timing of the domestic violence increase. Failing to look at the entirety of the research on this topic is a critical oversight, especially considering the data from multiple cities (including DallasLos Angeles, Indianapolis, Chicago, and others) that found pandemic-related increases in domestic violence to be fairly short-lived. Because the increase in domestic violence was short-term, it seems misleading to say that this was a major reason for the major increase in homicides, which remained a prevalent issue throughout 2020 and 2021 (after stay at home orders ended).

Per the Bureau of Alcohol, Tobacco, Firearms and Explosives (ATF) data, tens of thousands of new guns purchased in 2020 turned up at crime scenes, which is about twice as many as in 2019. Because more newly-purchased guns were found at crime scenes in 2020 relative to 2019, the CDC claims that increased gun sales were the main reason for the homicide spike. This sounds plausible at first. However, determining how many crimes were committed with newly purchased guns is of course more nuanced than the authors make it sound. Law enforcement measure the “time to crime” (i.e., how much time passes between the purchase of a gun and the crime) for each recovered firearm, which they use as a proxy to identify how many newly-purchased guns are being found at crime scenes. If new guns are being used for crimes more often, then the average “time to crime” will be shorter. But decreases in the “time to crime” are also related to other factors, such as improvements in law enforcement technology (e.g., ballistics tracing and Crime Gun Intelligence Centers [CGICs]). In fact, the number of recently purchased guns being found at crime scenes has actually been increasing since 2013.

The Johns Hopkins study tried to make a convincing case that gun violence was most prevalent in states with less restrictive gun control policies. But when taking a closer look at the actual data used for the Johns Hopkins study, it seems misleading. First, these numbers are entirely based on total gun deaths, not gun-related homicides. Second, they selectively reported results that they knew would support their narrative, without the disclaimer that correlation does not equal causation. Importantly, there is some research from before the pandemic that shows some evidence for the effectiveness of background checks laws and requiring permits to purchase a firearm, but the effects of many of the other specific types of laws are uncertain (e.g., laws that restrict firearms in public places, “Stand your Ground” laws). Unfortunately, the Johns Hopkins and CDC reports use very simplistic analyses that don’t tell us much about this.

According to the Johns Hopkins report, more restrictive gun laws lead to lower gun death rates and less restrictive laws lead to higher rates. But, the analysis itself is quite limited, no tests are ran for statistical significance, and the results appear to be selectively reported. If you review the actual data and look at the different gun laws across all of the states (aside from the ten listed here), the same conclusions do not hold up. While it is possible that some gun laws are effective, the authors do not do a great job of analyzing this. For this post, I combined NVSS data with data on gun ownership and state gun laws to look at this a little more closely.

In the Johns Hopkins report, the authors list out ten states — the five states with the most gun deaths (Mississippi, Lousiana, Wyoming, Missouri, and Alabama) and the five states with the least gun deaths (Hawaii, Massachusetts, New Jersey, Rhode Island, and New York). Then, they tried to find laws that states had in common, and tried to attribute gun death numbers to these policies. Of the five states with the lowest gun death rates, all of them had some kind of Extreme Risk Protection Order (ERPO) law, and four out of five of them had purchaser licensing laws (Hawaii, Massachusetts, New Jersey, and New York). They falsely indicate that Rhode Island has a firearm purchaser law, but a footnote reveals that it is not the same as the laws in the other states and only requires a firearm purchaser waiting period. Of the five states with the highest gun death rates, all of them had “Stand your Ground” laws, while three (Mississippi, Wyoming, and Missouri) had permitless carry. The authors again falsely indicate that Alabama has permitless carry, but a footnote reveals that this law was passed in 2022, so it is misleading to include this as a “permitless carry state” when the law was not in effect at the time the data were collected.

The five states with the lowest overall gun death rates (Hawaii, Massachusetts, New Jersey, Rhode Island, and New York) all had ERPO laws and four out of the five (excluding Rhode Island) had purchaser licensing laws. However, when you look at the actual data and consider all of the states with firearm purchasing laws and ERPO laws, the link to gun-related homicide rates becomes less dramatic. For example, the District of Columbia had the highest homicide rate in 2020 (22 per 100,000), yet they had both a firearm purchaser law as well as an ERPO law. Similarly, Illinois ranked #9 for gun homicide rates and Maryland ranked #10, despite both of these states having firearm purchaser laws and ERPO laws. Additionally, both Illinois and Maryland ranked very low (#42 and #40, respectively) in terms of gun ownership, which contrasts with the claim that more gun ownership leads to more murders.

Similarly, the five states with the highest overall gun death rates (Mississippi, Lousiana, Wyoming, Missouri, and Alabama) all had “Stand Your Ground” laws, while three of the five (Mississippi, Wyoming, and Missouri) had active permitless carry laws. However, once again, when you look at all of the states with “Stand your Ground” or permitless carry laws, the link to gun-related homicide is not nearly as strong. For example, Idaho had both laws on the books as of 2020 and ranked #4 in gun ownership, yet they ranked #45 in gun homicide rates. Similarly, South Dakota ranked #34 in gun homicide rates despite having both of these laws on the books and ranking #9 in terms of gun ownership. Alaska, ranking #3 in gun owenership, also showed low gun homicide rates relative to other states (ranking #31 in gun homicide rates), despite having both of these laws on the books.

Gun ownership on its own doesn’t seem to correlate with gun homicide rates, either. Montana ranked #1 for gun ownership but #32 in gun homicides. Wyoming ranked #2 for gun ownership, but their gun homicide rate was so low that the calculated rate was unreliable and not reported by the CDC.

Another thing that is not mentioned in either report is that, in states with more restrictive gun control policies, ghost guns (untraceable, unserialized guns that can be assembled at home using parts purchased online) are becoming more of a problem. Considering the alarming rise over the past year in the number of ghost guns recovered during criminal investigations, especially in communities with high rates of gun violence, this is a major concern that may represent a serious backfire effect of restrictive gun control policies.

The main recommendations from the Johns Hopkins report are that 1) states should implement firearm purchasing laws (i.e., permit-to-purchase) and firearm removal laws (e.g., domestic violence protection orders, ERPOs), and 2) cities should proritize funding for Homicide Review Commissions  and community violence interventions. As mentioned though, it’s hard to know whether more restrictive gun control policies would lead to increased use of ghost guns, which may even exacerbate the problem and affect law enforcement’s ability to clear homicides. Secondly, firearm removal laws are difficult to implement successfully, which can render them ineffective. As for community outreach programs (e.g., Cure Violence), the results from these types of programs are not as promising as they are made to believe.

To really reduce gun crime, a better approach would be to improve policing. This could include increased use of proactive policing, and it also could include greater usage of technology to assist law enforcement in their investigations. For example, use of technology such as Crime Gun Intelligence Centers (CGICs) can dramatically enhance law enforcement’s ability to clear crimes, particularly murders. Other cities, such as Boston, MA and Rochester, NY have made major revamps to their homicide and forensic units that have increased homicide clearance rates. Drones are another tool that can also be used to monitor and identify prolific offenders. Because the most violent crimes are committed by such a small percentage of the population, identifying and cracking down on these individuals using policing and invesigative tactics would likely be helpful in reducing gun homicides. In terms of gun laws, perhaps there are some laws that are effective, but more research is needed to understand which laws are effective and why. It’s simply not accurate to say that “gun ownership increases crime” or that “more restrictive gun laws lead to crime”; the relationship is not that simple.

2 Responses

  1. Charles Andrews says:

    You suggest several times that “reductions in pro-active policing and calls to defund the police” might be related to the firearm homicide increase. And yet the states with the highest number of gun homicides are all republican controlled conservative states where neither of these factors is likely to be present. You would also expect this to be more of a factor in urban areas which tend to be more liberal, yet that also is not reflected in the data which shows that rates in rural and urban areas are not greatly different. Do you have any data that supports your speculation?

    • Elizabeth Berger says:

      Hi Charles,
      Thanks so much for reading and taking the time to comment! I appreciate you sharing your thoughts and adding to the discussion.

      So first, we need to think about gun homicides in terms of rates, not numbers. When you look at rates of gun homicide, there are some very liberal states/territories (e.g., the District of Columbia, Maryland) with very high homicide rates. On the flipside, there are some very conservative states that have lower homicide rates (e.g., Idaho, South Dakota). The data I used for this was the same data that CDC and Johns Hopkins used for their reports, accessed here: At the same time, I’m interested to see what data you are using to support your assertion, if you don’t mind sharing.

      Relatedly, I believe the data show that gun suicides are higher in rural areas, but that is not the same as homicides. When it comes to gun homicides, the rates are highest in highly-concentrated urban areas (which are typically more liberal, as you mention). The connection between high population density and homicide rates is nothing new — however, there is probably a “sweet spot” of population density and the impact on crime (in this study, crime rates increased with population density up until a sweet spot of 500 people/square mile: Of course, this relationship likely has myriad causes (e.g. gang violence, poverty, more anonymity, etc.) that all work together to create a “perfect storm.” I think the biggest point I would like to make here is that it’s not as simple to say that conservative places have higher homicide rates and liberal places have lower homicide rates. This issue is VERY complex, and making these types of blanket statements is harmful and makes us more polarized.

      There is some evidence that certain types of laws might impact violent crime (see…but the evidence is stronger regarding impacts on accidental deaths and suicides. And most of the research on this focuses on suicides, as suicides comprise the largest portion of gun deaths. There is less research on gun policies and homicides. Because the CDC and Johns Hopkins reports I mention above are focused on gun-related deaths (including suicides), it kind of presents a muddled picture, which is important to point out. The RAND resource explains why we need to think beyond the simple aspect of “conservative vs. liberal” OR “strict vs. loose gun control”; it’s more important to look at the impacts of different laws themselves, for which the findings are much more nuanced than people think.

      In particular, for central concerns in gun policy debates (e.g., mass shootings), there is inconclusive evidence on the effects of different policies (see RAND resource above, also As an (albeit anecdotal) example, New York has some very strict gun laws, but there was a terrible mass shooting in Buffalo recently. Of course, I will acknowledge that mass shootings are an entirely different animal than typical homicides. But they are one of the central concerns in the gun debate, so I would be remiss if I didn’t mention.

      Second, you say that Republican-controlled conservative states are less likely to see reductions in proactive policing — I’m not sure I would agree. Do you happen to have data on that? First, states across the board have limited proactive police stops due to COVID (see paragraph 5 of this post: Second, other reductions in proactive policing tend to be more informal and correlate with fear of becoming the “next viral incident” (see:

      Unfortunately, measuring reductions in proactive policing is very difficult to measure as it is often informal. Probably the best proxy for doing so is by measuring reductions in police-initiated calls for service, which have decreased alongside the homicide increase (see; also I stand by my assertion that “correlation does not equal causation,” so I do not attribute the homicide increase ONLY to reductions in proactive policing. However, proactive policing has been shown to decrease crime in numerous studies and meta-analyses, so I do think it is a promising approach (see:; also; I also cite additional research in this post here:

      Another thing to consider is that when police clear more crimes, they are able to get more of these people off the streets. This is true for both fatal shootings and non-fatal shootings (see more info here: There are technologies that can help law enforcement clear crime faster — see the last paragraph, the literature on Crime Gun Intelligence Centers is particularly interesting — but these cost money. In this way, de-funding the police would limit agencies’ abilities to acquire technology that can significantly help them be more effective in responding to crime AND clearing crimes. In addition, shrinking police budgets lead to understaffed departments, which means more officers are required to work overtime. This leads to more burnout, lack of sleep, and physiological changes that reduces one’s ability to cope with stress and respond to incidents (you can see more research on that in the “officer wellness” section of this post:

      One of the biggest problems in studying the reasons for the homicide spike is that it is very, very difficult to determine true causality. For that reason, it is misleading to say that any one factor led to the homicide spike. That is why I say that reductions in proactive policing might be one of the factors at play here — that claim is based on the breadth of the research, and not just one study. All too often, people (particularly laypersons) will see a study such as the Johns Hopkins one and conclude that “A-ha! That’s it! That’s the reason for the homicide increase!” And I think that is very problematic. When we look at decades of research, proactive policing has been shown to work, so I think we should be encouraging more of that.

      There is a lot more research on this topic that I could dive into and share, but this reply is already very long and nuanced. I hope this helps answer your question!