between gun ownership or availability and suicide? Building such a model would presumably involve estimating the probability that an individual commits suicide conditional on gun ownership (or availability in some sense). What data are needed to do this? What data are needed to estimate the effects of policy interventions on the probability of suicide or on the substitution of other means of suicide for guns? What other prior information is relevant? What covariates should be included? Are data on them currently available? Do data on covariates exist in a form that could be combined with gun ownership or availability data? Is it necessary to construct a new data set that includes both ownership or availability data and the covariates?
If one is interested in answering the question of whether adolescents with a gun in the home are more likely to successfully commit suicide than adolescents who do not have a gun in their home, then home-level data on gun possession and adolescent suicide are needed rather than aggregate data concerning the numbers of guns in circulation. This type of information could be used to address the basic question of what proportion of the adolescents with a gun in their home eventually commit suicide with a gun. Answering causal questions about firearms and suicide may require additional information.
The same questions can be asked about the probability of committing a violent crime with a gun conditional on ownership or availability. Similarly, what data are needed to support improved research on firearms markets and how criminals or suicide victims obtain firearms? How, if at all, would improvements in trace data be used in studies of the effects of policy interventions on firearms markets or any other policy issue? What would the desired improvements contribute to research on policy interventions for reducing firearms violence? How can trace data be used, considering the deficiencies of these data?
Ultimately, linking the research and data questions will help define the data that are needed. For example, attempting to answer the seemingly basic research question, “How many times each year do civilians use firearms defensively?” by using samples of data collected from crimes reported to the police is a mismatch between the data source and the research question. These surveys cannot reveal successful forms of resistance that are not reported to the police.
This effort to think carefully about the data needed to answer some of the basic research questions should take place in collaboration with survey statisticians, social scientists, public health researchers, and representatives from the Bureau of Justice Statistics, the National Institute of Justice, the Centers for Disease Control and Prevention, the Bureau of Alcohol, Tobacco, and Firearms, and others. The research program should assess data limitations of the existing and proposed data sets, regularly report the results of that research both in the scientific literature and in forums acces-