Last week I commented on Lott's LA Times editorial where he claimed that
Examining all the multiple-victim public shootings in the United States from 1977 to 1999 shows that on average, states that adopt right-to-carry laws experience a 60% drop in the rates at which the attacks occur, and a 78% drop in the rates at which people are killed or injured from such attacks.
I pointed out that he once again failed to mention the results in Duwe, Kovandzic and Moody. Homicide Studies 2002 6:4. Here is the abstract:
Right-to-carry (RTC) laws mandate that concealed weapon permits be granted to qualified applicants. Such laws could reduce the number of mass public shootings as prospective shooters consider the possibility of encountering armed civilians. However, these laws might increase the number of shootings by making it easier for prospective shooters to acquire guns. We evaluate 25 RTC laws using state panel data for 1977 through 1999. We estimate numerous Poisson and negative binomial models and find virtually no support for the hypothesis that the laws increase or reduce the number of mass public shootings.
They did find some evidence that carry laws increase the number of people killed and wounded in mass public shootings, but this was in only one of the models.
Well, in his reply to comments on the LA Times article, Lott finally mentions the Duwe article. He writes:
for a recent paper that gets the same results I do when this more narrow definition is used see Duwe, Kovandzic and Moody, "The Impact of Right-to-Carry Concealed Firearm Laws on Mass Public Shootings," Homicide Studies, November 2002
That's right, Lott is claiming that Duwe got "the same results" as Lott. Most people would, I think, reckon that with Lott saying "78% drop" and Duwe saying "virtually no support" for an increase or a decrease, that their results were not the same.
So, what tortured reasoning does Lott use to claim that Duwe got "the same results"? Here is what Lott says:
Whether one looks at two or more people killed or injured as well as three or more people killed, the drops are huge and quite statistically significant. Only when you examine multiple victim public shootings involving four or more people killed is the number of deaths so small that the drops are no longer statistically significant (for a recent paper that gets the same results I do when this more narrow definition is used see Duwe, Kovandzic and Moody, "The Impact of Right-to-Carry Concealed Firearm Laws on Mass Public Shootings," Homicide Studies, November 2002, for a similar result when examining the few cases where four or more people are killed).
This misrepresents both Lott's findings and Duwe's paper to make it look like Duwe's work supports Lott. First, in the Lott and Landes paper, they write that when they examined shootings with four or more people killed:
The results are similar to our earlier ones. We find that right-to-carry laws reduce the number of deaths, and that these deaths were increasing before passage of the law and falling thereafter.
Duwe et al could not reproduce this result. When they used a Poisson model as Lott did here, they found statistically significant increases. Buried in an endnote on page 302 of The Bias Against Guns Lott admits that the results are not significant if he uses a negative binomial model (instead of Poisson) but insists that the results still show decreases. And while Duwe et al also found that the results were not significant, in some specifications there were increases and some decreases.
Lott's description of the differences between looking at shootings with two injuries and those with four killed is also misleading. He makes it seem that the only difference is that the sample size is smaller, causing the "huge" drops to stop being statistically significant, but if there were huge drops with the smaller sample, then they would still have been statistically significant. In fact, the drops are smaller and are actually increases in some specifications. Duwe et al also point out that the smaller sample has a big advantage---it is possible to collect ever single such shooting, while the larger sample could be biased by the omission of some cases.
Lott's description of Duwe as getting "the same results" is reminiscent of the way he has described Ayres and Donohue's work. Most observers would say that their results contradict his, but Lott insisted that:
Their own county level data that did the year by year breakdown actually showed that Lott and Mustard were correct, but they weren't smart enough to know it.