In 1997, Lott and Mustard (1977) set off an ongoing controversy by famously contending that so-called shall-issue laws (SILs) -- state laws providing for the liberal issue of concealed gun permits -- deterred violent crime. They argued that these laws increase the probability that a would-be perpetrator's crime will fail because he is more likely to be threatened with a gun or shot by his intended victim. In this controversy the weapon of choice has been the diff-in-diff estimator applied to state and county panel data spanning various intervals of time and different investigators have come to different conclusions spanning from "more guns less crime" to "more guns more crime." This study brings to bear more general methods, a cohort panel data model (CPDM). Treating violent crime as a career, we interpret their deterrence hypothesis in the context of forward looking choices to enter and exit this career and estimate the corresponding cohort panel data model (CPDM). Our model distinguishes among three key parameters or effects of SILs, (i) a direct effect on entry decisions, (ii) a surprise effect on exit decisions by individuals who entered criminal careers prior to the passage of SILs, and (iiI) a selection effect on exit decisions on those how entered after the passage of SILs. Entry into a criminal career is confined to youth's "entry window." Thus, the evolution of crime rates depend on the evolution of the shares of the population in entry and exit windows and when these shares are covered by SILs. Applying generalized least squares with autocorrelated errors to state-panel data on changes in violent crime rates, our preliminary results strongly reject the deterrence hypotheses. Moreover, as the CPDM nests the diff-in-diff model, we reject the restrictions that reduce the CPDM to the diff-in-diff. Thirdly, we expect that future draft of this paper will shed light on the way the reliance on diff-in-diff estimator estimates has fueled the controversy. We can show that the diff-in-diff estimator is a weighted sum of the three key CPDM parameters and the weights depend on the span of years covered by the sample. Thus, for example, early investigators used samples that ended shortly after the initial wave of the adoption of SILs. Correspondingly, it seems likely that their estimated diff-in-diff parameters put heavy weight on (ii), the surprise effect. In contrast later investigators used longer sample periods covering more years since SILs were adopted and, we expect to show that their estimated diff-in-diff parameters were more a mix of the three effects and averaged out to no effect.
Read full abstract