Council, 2003).¹⁰ Tracking firearms possession through arrest might also serve to detect emerging problems in high-risk populations. Quarterly data collection, such as that conducted by ADAM, permits monitoring of local trends over short time intervals.

Such data, however, may not be useful for answering many of the policy-related questions considered in this report. Although the ADAM samples are representative of the local arrestee populations for which the surveys were administered, the 35 data collection sites are not a representative sample of urban areas nationwide. Moreover, a survey of arrestees cannot be used to infer acquisition and use among criminals or the general population. The data are not representative of the relevant populations and can be influenced heavily by police priorities and procedures. Thus, these data alone cannot be used to infer the effects of guns on crime or the effects of interventions on gun use or the market for weaponry in general.

Suppose, for example, one found that the fraction of arrestees reported to possess firearms does not vary by the strength of local regulations. It may be, as suggested by the data, that regulation has no effect on the market. It may also be that regulation affects the crime and the ownership rates, but among the arrested populations the ownership and use rates are unchanged. And it may be that regulations influence policing and the accuracy of self-reporting in unknown ways. ADAM data do not reveal the association between regulation and the behavior of offenders, potential offenders, the crime rate, or policing. Thus, observing that the prevalence of gun ownership and use among arrestees changes after some interventions does not reveal how gun use or crime more generally changed in the population of interest.

Using proxy measures of ownership raises different issues and questions. In the proxy approach to measuring ownership (proxy approaches have not been developed as measures of firearms use) researchers have sought to find measures that would indicate whether firearms were available. A variety of these have been proposed, but it appears that the one the research community has settled on is the proportion of suicides committed with a firearm (Kleck, 1991; Cook, 1991). This measure has been found to