A new analysis of the impact of right to carry (RTC) laws by Stanford University's Abhay Aneja and John Donohue and Johns Hopkins University's Alexandria Zhang debunks the "more guns, less crime" thesis. More importantly perhaps, the authors also sound a cautionary note about interpreting the findings of any one study--including their own.
Aneja, Donohue, and Zhang uncover four types of issues with Lott and Mustard's original study:
- "The comparison of crime between RTC and non-RTC states is inherently misleading because of factors such as deprivation, drugs, and gang activity, which vary significantly across gun-friendly and non-gun-friendly states (and are often difficult to quantify). To the extent that the relatively better crime performance seen in shall-issue states during the late 1980s and early 1990s was the product of these other factors, researchers may be obtaining biased estimates."
- Lott and Mustard's study covered the period 1977-1992. "Crime rates declined sharply across the board beginning in 1992. [...] Moreover, the average crime rates in non-RTC states seemed to have dropped even more drastically than those in RTC states, which suggests that crime-reducing factors other than RTC laws were at work."
- The data used by Lott and Mustard are problematic: Aneja, Donohue, and Zhang found multiple coding errors. In addition, county data are notoriously inaccurate, inconsistent, and incomplete. Finally, Lott and Mustard used a flawed contemporaneous arrest rate variable.
- The statistical model specification used by Lott and Mustard is problematic in a number of dimensions--omitted variable bias, endogeneity, collinearity, and treatment of standard errors.
The authors correct for some of these problems (fixing data inaccuracies, poorly constructed arrest ratios, incorrect standard errors, and using an extended 1977-2000 dataset). They find that "RTC laws increase crime--for rape, aggravated assault, robbery, auto theft, burglary, and larceny. There is not even a hint of any crime decline."
In an effort to probe the robustness of the results and improve the model, Aneja, Donohue, and Zhang then proceed to revise the model (using state rather than county data; extending the study to 2006, then 2010; removing collinear variables; including better controls; ...) and run a rigorous and extensive set of analyses with varying model specifications.
Their results make three important points:
- No matter the model specification, they can find no evidence to support the "more guns, less crime" claim. On the contrary, the evidence suggests that the effects of RTC laws on crime are positive, meaning that adopting RTC laws appears to result in crime increases. This effect is strongest and most consistent for aggravated assault.
- "Different models yield different estimated effects": while no model shows that RTC laws decrease crime, the impact of RTC laws varies from model to model. In some models, RTC laws are associated with substantial and statistically significant crime increases across multiple crime categories. In others, the extent of the crime increase is more modest, or observable only in one crime category. In other words, the results are not robust to model specification.
- "Our ability to ascertain the best model is imperfect." The authors conclude that "the 'best practices' in econometrics are evolving. Researchers and policymakers should keep an open mind about controversial policy topics in light of new and better empirical evidence or methodologies."
For an in-depth discussion of the very complex statistical issues involved in analyzing the relationship between RTC and crime, please refer to Aneja, Donohue, and Zhang's article.
Gun Violence in America (http://www.teachingwithdata.org/resource/3864)
Fear of Crime (http://www.teachingwithdata.org/resource/3155)
Crime Victimization in the US: A Data-Driven Learning Guide (http://www.teachingwithdata.org/resource/3437)
Generational Trends in Attitudes about Gun Ownership: A Data-Driven Learning Guide (http://www.teachingwithdata.org/resource/3448)
Crime and Victims Statistics (http://www.teachingwithdata.org/resource/3261)