This article was originally sent to JAMA Pediatrics with the hope of allowing an alternative interpretation of its recently published paper, Firearm Laws and Pediatric Mortality in the US, to be presented to its readership. Because JAMA’s editors denied publication, K. Alexander Adams presents it here in the hope of furthering fair academic discussion of firearms laws and their impacts on public safety.
Firearm Laws and Pediatric Mortality in the US, an article by Faust et al. published online in June of 2025, contributes to the firearms literature by attempting to analyze the impact of firearms laws on pediatric firearms death. Unfortunately, the methods used by the article fail to present a convincing case that firearms restrictions reduce pediatric firearm mortality.
The way the study operationalizes firearms law is questionable. The article splits states into three groups: “most permissive,” “permissive,” and “strict.” The issue with broad categories is it masks the major differences between different gun laws. It could be possible that every gun law is ineffective, yet one single effective law explains why the “strict” group had less excess mortality. This is why most researchers study laws individually and do not arbitrarily lump them together. A three-tier system of restriction level, crafted through an opaque arbitrary system by the study’s authors, says nothing about which laws actually work, and it makes it even harder to determine the underlying theoretical reasons some laws work and others do not.
The study fails to adopt a panel data approach despite having the data with which to do so. Using panel data, one could control for unobserved heterogeneity between states and derive tentative causal conclusions from the data. Instead, the authors opted to use a purely cross-sectional approach, which does not allow one to draw causal conclusions. Further, the authors make no effort to control for factors, for which nearly every other research paper attempts to control, that may influence firearm homicide rates, firearm suicide rates, or firearm accidental death rates. There is no explanation as to why the study employs an inferior method when the data could have been easily set up as a panel study.
The paper’s use of “excess mortality” makes little sense in a criminology context. The gap between predicted and actual deaths could simply be a result of random error, omitted variable bias, or serial correlation, especially given the cross-sectional nature of the study. Again: why use an inferior cross-sectional methodology to calculate “excess mortality” when panel data was available?
Finally, the use of a Poisson model to calculate the counterfactual scenarios is almost certainly the incorrect methodological decision, rendering the results uninformative for policy formulation. Poisson models strictly assume equidispersion (that is, the variance has the same value as the mean) but the data used almost certainly is overdispersed (the variance is greater than the mean). One could run a likelihood-ratio test to see if the assumption holds, but the article does not mention whether this was done.2 If the equidispersion assumption fails, as it likely does, a negative binomial regression model should have been used instead. The result of using the wrong statistical model is that the results cannot be considered reliable.
Overall, the study fails to make a convincing case that “strict” firearms laws are causally related to lower pediatric firearms mortality, and the methods used are of little use to policymakers.