Eugene Volokh writes:
(Incidentally, am I mistaken in thinking that it's the NCVS numbers
which are usually cited to show that self-defense with a firearm
decreases the likelihood of injury, compared to no self-defense?)
No, you are not mistaken. In "Point Blank" Kleck dismisses the NCVS as
not adequate for measuring DGUs (because the NCVS undercounts some
crimes) and then a few pages later uses the NCVS to measure the
effectiveness of defensive gun use. I don't see how he can have his
cake and eat it.
Another disturbing thing about his treatment of the NCVS in "Point
Blank" is that he suppresses the NCVS estimate. Readers are not told
what it is, merely that it is "not adequate". What is "not adequate"
here is Kleck's reasons for rejecting the NCVS. The reasons given by
Kleck could not possible account for an NCVS undercount of more than a
factor of two. Elsewhere in the book, Kleck is quite happy to
multiply five correction factors together to get an estimate of DGUs
from one survey. Why doesn't he apply one correction factor to the
NCVS estimate to get an estimate corrected for the undercount?
The answer is that the estimate obtained differs by an order of
magnitude from the estimate he gets from the Hart Poll and would
reveal that he offers no adequate explanation for the discrepancy.
Kleck has wrestled with the NCVS estimate for almost ten years now and
still has not come up with an adequate explanation.
(I recognize that certain numbers, for instance the wounding
numbers, are based on samples too small to be significant. But I
assume the burglary and robbery numbers aren't.)
The wounding numbers appear to correspond to 24 actual reported
woundings. This implies a 95% confidence interval for the Kleck survey
derived estimate of DGU woundings of 100,000-300,000. The interval is
wide, but even the lower end is 5-10 times Kleck's earlier estimates
and exceeds the NEISS estimate for ALL firearm woundings. The
wounding numbers cannot be dismissed as some sort of statistical
fluke.