Deltoid

Jon Buck said:

NCS didn't do a very good job of asking; they only asked about defensive weapon use after the respondent answered positively to having been a victim of a crime.

Right, so cases where someone whipped out a gun without being threatened with violence weren't counted.

Kleck makes this criticism of the NCS (in "Point Blank", if I remember correctly). This leaves out the case of successful self-defense, in which the intended victim was not victimized because they used a gun in self-defense;

No it doesn't. Such a person is still a victim of the crime of attempted robbery, or assault, or attempted whatever.

obviously, most of these people would have answered negatively to the question "Have you been a victim of a crime?"

Since NCS does not ask this question, this point is irrelevant.

You have misunderstood Kleck's criticism. He is arguing that NCS undercounts crimes such as domestic assaults and therefore undercounts defensive gun uses in such situations. A valid criticism, but the undercount would have to be truly vast to account for the discrepancy between the 80,000 NCS derived count of defensive gun uses and Kleck's 2.5M. Nor is there any evidence that NCS significantly undercounts burglaries, so it does not account for the factor of 28 discrepancy above.

Read the article again; Kleck gives 2,500,000 as the upper bound for self-defense uses, and 800,000 as the lower bound. Obviously, the real number is probably somewhere in between. The same criticism of the NCS figure applies here, too; robberies prevented by with-gun self-defense are likely to go undiscovered using their questioning methodology.

NCS asks: "Did anyone try to rob you?" Are successful defenders going to say "no"?

Point the first: I don't think Kleck's 15% number is all that accurate, since it is derived by dividing gun deaths into gun woundings, and gun woundings are likely to be underreported, especially minor ones received by criminals shot at by their victims, which is what we are talking about.

Sure, it is possible that the correct figure is, say, 10%. It would have to be 1% to explain the discrepancy, though, and this is just not believable.

Point the second: even police officers involved in shootings tend to over-report woundings of the people they shot; I see no reason why a civilian poll would be any different. Note that some of Kleck's other data suggests that only 1-2% of self-defense uses involve killing or wounding the attacker.

This was computed by taking Kleck's estimate of 1500-2800 justifiable with-gun homicides, and applying the 15% fatality rate to get 9000-17000 woundings and dividing by his estimate of 1M self-defense uses. Since his estimate of self-defense uses was an input to the calculation, we cannot use this figure to cross check the validity of his estimate of defense uses.

Running your same calculations with the two percent number yields 7500 justifiable with-gun homicides, which is still high but closer to a realistic number. Using Kleck's old figure of 1.2 million self-defenses (still within the range he is claiming), and the two percent figure, we get 3600 justifiable homicides, which is much more reasonable than 30,000, and closer to the 1,500 to 3,000 number claimed by Kleck and others.

Not surprising: This calculation boils down to:

   3000*.15*1.2M
   -------------
        .15*1M

It would have come out exactly if you had used Kleck's original 1M estimate.