Tracing the Brady Act's Connection With Homicide and Suicide Trends

Given the ideological rhetoric that too often passes for "fact" in debates over the regulation of firearms in the United States, we should be encouraged by innovative evaluations of current or proposed policy interventions in this area. Although political considerations will always play a prominent role in policy development, politics that has to contend with the results of good science should produce better policy than politics based on poor science or none at all. Accordingly, the question is whether the article in this issue of THE JOURNAL by Ludwig and Cook¹ evaluating the 1994 Brady Handgun Violence Prevention Act—the most important national policy initiative related to firearms in over 2 decades—is good science. If so, what are its implications for current policy? If not, what lessons might it contain for improving scientific assessments of the Brady Act and other health-related public policies?

Ludwig and Cook assess the impact of the Brady Act on adult homicide and suicide rates in the United States. Their study exploits a natural discontinuity in the implementation of the law across the nation: 19 states had Brady Act–type background checks and waiting periods in effect before the law's implementation in 1994, and therefore were exempt from its requirements. The other states implemented checks and waiting periods in 1994. Ludwig and Cook reason that, if the law is effective, the Brady Act states should exhibit greater post-1994 reductions in gun homicide and suicide rates than the 19 non–Brady Act (or "pre-Act") states. Controlling for other influences, the authors found the expected significant reductions in Brady Act states compared with non–Brady Act states in firearm suicide rates of persons aged 55 years and older, but no significant difference in suicide rates for younger adults or in homicide rates for adults of any age. Their results are robust against differing model specifications, estimators, and outcome measures.

Under the circumstances, Ludwig and Cook have conducted a sound and thoughtful investigation, but the circumstances militate against the degree of investigator control normally expected of good science. For instance, the investigators could not assign states to treatment and control conditions, nor prevent the possible contamination of one group by the other. The very meanings of "treatment" and "control" are turned on their heads in this investigation, if treatment means the presence of the intervention and control means its absence. In this case, the control group was receiving the treatment (background checks and waiting periods) prior to the treatment group.

While hardly an ideal study design, it is just the sort that real-world circumstances commonly force on the public policy analyst. The choice for policy researchers is usually between designs that afford some comparison of outcomes across cases that differ in a policy's presence, magnitude, or timing and those permitting no comparison. As Ludwig and Cook observe, comparative designs are generally more powerful, but the investigator must be sensitive to the possibility that the presence of the policy in one group influences the outcome in the other group. That is both the key limitation and strength of the study by Ludwig and Cook. It is not only possible but likely that the preexisting background checks and waiting periods in the non–Brady Act control states influenced firearm availability in both groups. But Ludwig and Cook's recognition of this possible "contaminating" influence and how it might affect their results is itself an important contribution to understanding the complex relationship between point-of-sale restrictions on handgun purchases and firearm violence.

The Brady Act affects the acquisition of handguns from licensed gun dealers in the so-called primary firearms market. However, guns used in homicide and other crimes tend to come from the "secondary" firearms market, consisting of legal and illegal transfers from unlicensed sources.^{2 - 3} Although technically distinct, the 2 gun markets are closely interconnected: nearly all of the guns in the secondary market originate in the primary market. For example, a father purchases a gun from a licensed dealer as a gift for his son, or a "straw purchaser" acquires guns from a licensed dealer and sells them to gun traffickers. Secondary market transfers therefore depend to some extent on the ease with which guns may be acquired in the primary market. If primary market restrictions such as background checks and waiting periods reduce the supply of guns to the secondary market or otherwise drive up the price of secondary transfers, secondary market transactions should shift to locations with fewer or weaker primary market restrictions. In that way Brady Act–type restrictions in one state could alter the availability of guns through the secondary market in another state—and there is no way of knowing whether they are treatment or control states.

Trace analysis reports indicate that a significant fraction of guns used to commit crimes were originally purchased out of state.⁴ As Ludwig and Cook acknowledge, direct evidence of the impact of the Brady Act on interstate firearms trafficking does not exist. It is badly needed. Without such evidence, the magnitude or even the direction of the possible contamination bias in their study cannot be determined. But even if that bias is large, its effect is limited primarily to estimates of the Brady Act's effect on criminal violence. Suicide is much less likely to be affected by shifts in the secondary firearms market, in part because the Brady Act background checks appear to be more effective in prohibiting felons than those with a predisposition to suicide from purchasing firearms legally. Technically, mental illness prohibits the purchase of a handgun under the Brady Act; however, in practice very few applications are turned down on these grounds. The great majority are rejected because of felony convictions or indictments.⁵

Ludwig and Cook find a significant reduction in firearm-related suicide associated with the Brady Act. That the association is limited to this firearm category is additional support for the hypothesis that the reduction in suicide is attributable to the Brady Act's restrictions on handgun purchases. Importantly, the Brady Act's impact on suicide seems to result from the waiting period requirement that was phased out at the end of 1998. That result is consistent with the evidence that background checks screen out few handgun applicants for mental illness. Less obvious is why the association with suicide is limited to older adults. Ludwig and Cook anticipate an age-related influence by observing that older persons have higher suicide rates and lower gun ownership rates than younger adults; however, the assumption of lower gun ownership among older adults may be questionable (R.R., unpublished data, 2000). Nonetheless, the inclusion of suicide in their analysis of the Brady Act represents an important advance over previous assessments limited to criminal violence.⁶

Although the absence of evidence on secondary transfers of firearms precludes strong conclusions regarding the effect of the Brady Act on homicide rates, the article by Ludwig and Cook is a potent stimulus to the necessary research. Knowledge of how primary market regulations affect the secondary firearms market is the single most important next step in research on how the Brady Act and similar strategies affect levels of criminal violence in the United States. The accumulation of data on firearms traced by state and federal law enforcement agencies should facilitate studies of the relationship between the 2 gun markets.⁷ Until the results of that research are available, current knowledge does not warrant relaxing or abandoning any of the Brady Act–type restrictions on handgun purchases.