Counting Guns

Abstract

At the fall 2001 Social Science History Association convention in Chicago, the Crime and Justice network sponsored a forum on the history of gun ownership, gun use, and gun violence in the United States. Our purpose was to consider how social science historians might contribute nowand in the future to the public debate over gun control and gun rights. To date, we have had little impact on that debate. It has been dominated by mainstream social scientists and historians, especially scholars such as Gary Kleck, John Lott, and Michael Bellesiles, whose work, despite profound flaws, is politically congenial to either opponents or proponents of gun control. Kleck and Mark Gertz (1995), for instance, argue on the basis of their widely cited survey that gun owners prevent numerous crimes each year in theUnited States by using firearms to defend themselves and their property. If their survey respondents are to be believed, American gun owners shot 100,000 criminals in 1994 in selfdefense–a preposterous number (Cook and Ludwig 1996: 57–58; Cook and Moore 1999: 280–81). Lott (2000) claims on the basis of his statistical analysis of recent crime rates that laws allowing private individuals to carry concealed firearms deter murders, rapes, and robberies, because criminals are afraid to attack potentially armed victims. However, he biases his results by confining his analysis to the years between 1977 and 1992, when violent crime rates had peaked and varied little from year to year (ibid.: 44–45). He reports only regression models that support his thesis and neglects to mention that each of those models finds a positive relationship between violent crime and real income, and an inverse relationship between violent crime and unemployment (ibid.: 52–53)–implausible relationships that suggest the presence of multicollinearity, measurement error, or misspecification. Lott then misrepresents his results by claiming falsely that statistical methods can distinguish in a quasi-experimental way the impact of gun laws from the impact of other social, economic, and cultural forces (ibid.: 26, 34–35; Guterl 1996). Had Lott extended his study to the 1930s, the correlation between gun laws and declining homicide rates that dominates his statistical analysis would have disappeared. An unbiased study would include some consideration of alternative explanations and an acknowledgment of the explanatory limits of statistical methods.