More Guns Less Crime by John Jr (best free e book reader .txt) 📕

generalized the Lott and Mustard model to include state-specific trends in an effort to control for these unobserved factors. ... we report the results for models with a quadratic time trend. The only significant impact estimate is for assaults, and its sign is positive, not negative. (Dan Black and Dan Nagin, "Do Right-to-Carry Laws Deter Violent Crime?" Journal ofLegal Studies, January 1998, p. 218)

Much more was controlled for than "a single national trend" in my study (e.g., as just mentioned above, state and county trends as well as other crime rates). While it is reasonable to include individual linear state trends or nonlinear trends for regions, including nonlinear trends for in-

dividual states makes no sense. The approach by Black and Nagin is particularly noteworthy because it is the one case in which an academic study has claimed that a statistically significant, even if small, increase in any type of violent crime (aggravated assault) occurs after the law.

Consider a hypothetical case in which the crime rate for each and every state follows the pattern that Black and Nagin found in their earlier paper and that I showed in this book (discussed in chapter 7, pp. 136-37): crime rates were rising up until the law went into effect and falling thereafter. Allowing a separate quadratic time trend for each state results in the time trend picking up both the upward path before the law and the downward path thereafter. If the different state crime patterns all peaked in the year in which their state law went into effect, the state-specific quadratic trends would account for all the impact of the law. A variable measuring the average crime rates before and after the law would then no longer reflect whether the law raised or lowered the crime rate. 79 This is analogous to the "dubious variable" problem discussed earlier. If enough state-specific trends are included, there will be nothing left for the other variables to explain.

If shall-issue laws deter crime, we would expect crime rates to rise until the law was passed and then to rise more slowly or to fall. The effect should increase over time as more permits are issued and more criminals adjust to the increased risks that they face. But the quadratic specification used by Black and Nagin replicates that pattern, state by state. Their results show not that the effect from the quadratic curve is insignificant, but that the deviation of the law's effect from a quadratic curve over time is generally insignificant.

To see this more clearly, take the hypothetical case illustrated in figure 9.15, in which a state faced rising crime rates. 80 The figure shows imaginary data for crime in a state that passed its shall-issue law in 1991. (The dots in the figure display what the crime rate was in different years.) The pattern would clearly support the hypothesis that concealed-handgun laws deter violent crime, but the pattern can easily be fitted with a quadratic curve, as demonstrated with the curved line. There is no systematic drop left over for any measure of the right-to-carry law to detect— in terms of the figure, the difference between the dots and the curved line shows no particular pattern.

Phrased differently, the deterrence hypothesis implies a state-specific time pattern in crime rates (because different states did or did not pass shall-issue laws, or passed them at different dates). All Black and Nagin have shown is that they can fit such a state-specific pattern with a state-specific quadratic time trend, and do this well enough that the residuals no longer show a pattern.

EPILOGUE/ 211

Year

8385878991 939597 Right-to-carry law passes in 1991

Figure 9.15. Fitting a nonlinear trend to individual states

3Should one expect an immediate and constant effect from right-to-carry laws with the same effect everywhere?

While he includes a chapter that contains replies to his critics, unfortunately he doesn't directly respond to the key Black and Nagin finding that formal statistical tests reject his methods. The closest he gets to addressing this point is to acknowledge "the more serious possibility is that some other factor may have caused both the reduction in crime rates and the passage of the law to occur at the same time," but then goes on to say that he has "presented over a thousand [statistical model] specifications" that reveal "an extremely consistent pattern" that right-to-carry laws reduce crime. Another view would be that a thousand versions of a demonstrably invalid analytical approach produce boxes full of invalid results. (Jens Lud-wig, "Guns and Numbers," Washington Monthly, June 1998, p. 51) 81

We applied a number of specification tests suggested by James J. Heckman and V. Joseph Hotz. The results are available from us on request. The specifics of the findings, however, are less important than the overall conclusion that is implied. The results show that commonly the model either overestimates or underestimates the crime rate of adopting states in the years prior to adoption. (Dan Black and Dan Nagin, "Do Right-to-Carry Laws Deter Violent Crimel" Journal of Legal Studies, January 1998, p. 218)

Black and Nagin actually spent only a few brief sentences on this issue at the very end of their paper. Nevertheless, I did respond to this general point in the original book. Their test is based upon the claim that I believe "that [right-to-carry] laws have an impact on crime rates that is constant over time." 82 True, when one looks at the simple before-and-after average crime rates, as in the first test presented in table 4.1 and

Crime rate before law

Crime rate after law

-5-4

-1012345

Years before and after implementation of the law

Figure 9.16. What was the crime pattern being assumed in the simple test provided in table 4.1?

a corresponding table in my original work with Mustard, this was the assumption that was being made. 83 Figure 9.16 illustrates the crime pattern assumed by