The survey is confidential. The person surveyed cannot get into
trouble for mentioning a defensive use with an illegal weapon.
Nonetheless someone is less likely to report such a use, so the NCVS
will likely underestimate such uses. However, this hardly supports
your claim that it is a "gross underestimate" unless almost all
defensive gun uses are conducted with illegal weapons.
David Veal writes:
Considering the problems occasionally (and well publicized)
encountered by people using armed resistance to crime, it is possible
that people would be hesitant to talk to a government official even
if they were fairly certain they were in the right. Many jurisdictions
(especially those which are both firearms hostile and have a lot of crime)
have policies which effectively mean any self-defense, no matter how
justified, goes to the prosocuter's rubber stam .... er, I mean
Grand Jury. :-)And while we may be talking about a confidential survey, and the
wrong level of government, a lot of people may not be capable of
either making that distinction, or interested in doing so.
If someone really does think that the Census Bureau will breach
confidentiality and pass things to the local prosecutor the smart
thing to do is to refuse to participate in the survey (rather than
participate and then lie). After all, if you participate you may
accidently let something slip. The NCVS gets a very high
participation rate, so this suggests that the number who lie in this
way is very small.
It is also possible that Kleck's high estimate was partly caused by
people telling stories about gun defences that happened to someone
else, or completely making it up. For example, his study implies that
citizens shoot 200,000 criminals each year, which is ten times as high
as the estimate he obtained by other means. This suggests that at the
very least, some exaggeration is going on.
We can reconcile the Kleck estimate and the NCVS one by assuming that
some people falsely claimed to Kleck that they used a gun (false
positive) and that some people falsely claimed to the NCVS that they
had not (false negatives). If this is so, the true estimate lies
somewhere in between 2.5M (Kleck) and 80,000 (NCVS).
To determine where it lies we need to know the relative likelihoods of
false positives and false negatives. If they are equal (that is, the
percentage of non-(gun defenders) who claim to use a gun to Kleck is
the same as the percentage of gun-defenders who claim not to have used
one to the NCVS) we can solve some equations to find the true
estimate: Letting l be the fraction who lie, and d be the true
estimate we have:
d*l = d - 80,000 (The fraction of gun defenders who lie is the
discrepancy between the NCVS estimate and the true figure.)
(250M-d)*l = 2.5M - d (The fraction of non-(gun defenders) who lie is
the discrepancy between the Kleck estimate and the true figure.)
Solving for d:
d = 80,000 /(1 - (2.5M - 80,000)/250M)
= 81,000
Under this assumption, respondents to the NCVS and to Kleck are both
99% truthful.
Readers may be puzzled as to why, if both groups are equally truthful,
the true estimate is not halfway in between the NCVS and Kleck
estimates. The reason is the number of people who did not use a gun
for self defence is much much more then the number who did. A very
small percentage of non-(gun defenders) corresponds to a very large
percentage of gun defenders.
Even if NCVS respondents are 20 TIMES as likely to lie as Kleck's
respondents (surely a vast overestimate), the true estimate is still
only 99,000.