A new study in JAMA Pediatrics asserts that states with “permissive” gun control laws experienced higher pediatric firearm mortality rates following the 2010 Supreme Court decision in McDonald v. Chicago. The study analyzed data from 49 states spanning 1999 to 2023. This has to be one of the dumbest studies done in a long time. After Public Health officials fumbled the COVID response, you’d expect some hesitation before trusting their research on crime, a field far outside their expertise. The fact that they don’t even mention policing and law enforcement in discussing crime rates should provide some warning to the media.
We have previously written extensively on the false claim made in the first sentence of this study: “Firearm deaths are now the leading cause of death among US children and adolescents” (see here and here).
Unlike typical research, which compares crime or suicide rates before and after states change their laws and contrasts those changes with states that didn’t alter their laws, this study ignores how laws change over time. It takes what could be panel data which allows one to account for average differences across states and years (so-called fixed effects). The paper limits there discussion to a purely cross-sectional comparison. The purely cross sectional comparison cannot be used for any discussion of causation. They don’t even try to account for basic factors like law enforcement practices—such as arrest and conviction rates, imprisonment rates, or the death penalty—that influence crime. Nor do they account for any factors that might explain changes in suicides or accidental gun deaths in the 2011 to 2023 period.
The study categorizes the level of gun control laws in each state into one of three broad categories and assumes that there laws remain constant over time, and lumps many different laws together in an arbitrary manner. Are these additive? Do we simply add a concealed-carry law to a safe-storage law to universal background checks? For these gun control advocates, is a red flag law twice as important as having lots of gun-free zones? There is so much arbitrariness in how a measure that combines these different laws and even what laws to include. How did they decide to have eight strict gun-control law states, eleven permissive states, and 30 most permissive states? Why not 1/3rd in each of these different categories?
The methodology falters in several ways. It relies on the epidemiological concept of “excess deaths,” commonly used to gauge the impact of diseases like Covid-19. The authors applied Poisson regression, using only a time trend as a control variable, to estimate expected deaths for three state groups. They labeled the gap between predicted and actual deaths as “excess deaths,” attributing these to permissive state laws. This gap, however, could simply reflect error or residuals, encompassing random error, omitted variables, heteroskedasticity, serial correlation, measurement errors, and other statistical challenges inherent in such analyses.
Among these issues, omitted control variables stand out as the most glaring flaw. The authors failed to account for Covid-19, which spiked murders while reducing other street crimes. A basic dummy variable could have mitigated this, but a more refined approach would have incorporated variables for each state’s lockdown policies. Other neglected factors include standard crime equation variables: police presence, arrests, executions, population density, unemployment, income, alcohol consumption, poverty, and welfare. These omissions render the estimates biased and inconsistent, leading to unreliable forecasts.
Another critical issue is unobserved heterogeneity when grouping states. California differs vastly from Alaska, Massachusetts from Texas, and Hawaii from all others in geography, culture, climate, and other fixed characteristics, beyond just gun control laws. These fixed effects act as omitted variables in regressions that aggregate states. The only solution is a fixed-effects model using panel data, which this study clearly did not employ.
This study has the same problem as the synthetic control approach, and it has no control variables in forecast period. Does whatever deviations between the forecasted levels of deaths arise from changes in law enforcement or some other factors? They make no attempt at trying to account for any other factors that can explain changes in firearm deaths over time. They just assume that no factors are changing after 2010.
The study labels states as “most permissive” or “permissive” and see if there is a rise in “excess” juvenile firearm deaths (homicides, suicides, and accidental deaths) after the 2010 McDonald v. Chicago ruling.
To determine expected deaths, Poisson regression was used to fit the data in prepolicy period (1999-2010), adjusting for time trends and legal status groupings. After model fitting, expected deaths were predicted for each year of the post policy period and for the total period (2011-2023).
— There are other very basic statistical problems with the study. As mentioned earlier, the paper doesn’t account for unobserved heterogeneity. The authors also employ a Poisson model which has a particularly strict condition, namely that the variance of the dependent variable must be equal to its mean (called “equidispersion”). This is unlikely to be the case with the data used in this paper, where we expect the variance to be much larger than the mean (overdispersion). There are tests for equidispersion, but the authors make no attempt to justify their regression model. If the data are overdispersed, which is usually the case in crime models, then the appropriate model is the negative binomial which allows the variance to differ from the mean. Use of a Poisson model on over-dispersed data causes the width of the confidence intervals to be underestimated, leading to spurious inferences.
— While the paper assumes that all state laws remain unchanged, for example, a state like Colorado is listed as having permissive gun laws even though its laws got much stricter by the end of the period examined. In 2011, Colorado did not have universal background checks, a high-capacity magazine ban, or a red flag law. Universal background checks and a high-capacity magazines were effective July 1, 2013. Red Flag law went into effect in 2020. So if firearm deaths increased in Colorado, why should that be blamed on permissive gun control laws?
— The categorizing states into three different groups (most permissive, permissive, and strict) is arbitrary. All sorts of different gun control laws are lumped together and how they are weighted and where the cut off points to determine what group a state is put into is arbitrary. The correct way of dealing with this is to account for the different gun control laws separately as some may matter and others might not.
— Why look at homicides (which include murders and justifiable homicides) instead of just murders?
— Why look at firearm homicides instead of total homicides?
This JAMA study has gotten a lot of uncritical news coverage. Here is how the New York Times described the research.
Firearm deaths of children and teenagers rose significantly in states that enacted more permissive gun laws after the Supreme Court in 2010 limited local governments’ ability to restrict gun ownership, a new study has found.
In states that maintained stricter laws, firearm deaths were stable after the ruling, the researchers reported, and in some, they even declined. . . .
“It’s surprising how few of these are accidents,” Dr. Faust said. “I always thought that a lot of pediatric mortality from guns is that somebody got into the wrong place, and I still think safe storage is important, but it’s mostly homicides and suicides.” . . .
The study, published Monday in JAMA Pediatrics, examined the 13-year period after the June 2010 Supreme Court ruling that the Second Amendment, which protects an individual’s right to bear arms, applies to state and local gun-control laws. The decision effectively limited the ability of state and local governments to regulate firearms.
The researchers classified states into three categories based on their gun laws: most permissive, permissive and strict. They used a Centers for Disease Control and Prevention database to analyze firearm mortality trends from 1999 to 2010 — before the Supreme Court ruling — and compared them with the 13-year period afterward. . . .
Roni Caryn Rabin, “Gun Deaths of Children Rose in States That Loosened Gun Laws, Study Finds,” New York Times, June 9, 2025.
0 Comments