I asked the question and have at least a partial answer after reading some of the published research.
First of all, the 65,000 number of defensive gun uses per year is based on data from the National Crime Victimization Survey (NCVS) and I was unable to download the actual study - my computer complained the PDF file was corrupted. So I cannot explain the factor of 2000 discrepancy between the 65,000 number and the documented 34 incidents in 11 months of 2012.
However, there is reason to be skeptical of the far higher reported rates of defensive gun use based on telephone surveys. In particular, there are two studies reporting roughly 2 million defensive gun uses per year: the Kleck and Gertz study that reported the 2.5 million use estimate and the Department of Justice study that reported a 1.5 million use estimate. Both of these studies were based on national random telephone surveys of about 4500 and 2500 households respectively. The reasons to be skeptical about these survey results are stated by the authors of the Department of Justice survey as shown in all the quotes below:
First, of all, using a survey to estimate the rate of occurrence of a rare event will suffer from the false positive effect. The study authors state (where DGU = defensive gun use and my emphasis is added in bold):
Any estimate of the incidence of a rare event based on screening the
general population is likely to have a positive bias. The reason can
best be explained by use of an epidemiological framework.[15]
Screening tests are always subject to error, whether the "test" is a
medical examination for cancer or an interview question for DGUs. The
errors are either "false negatives" or "false positives." If the
latter tend to outnumber the former, the population prevalence will be
exaggerated.
The reason this sort of bias can be expected in the case of rare
events boils down to a matter of arithmetic. Suppose the true
prevalence is 1 in 1,000. Then out of every 1,000 respondents, only 1
can possibly supply a "false negative," whereas any of the 999 may
provide a "false positive." If even 2 of the 999 provide a false
positive, the result will be a positive bias--regardless of whether
the one true positive tells the truth.
Respondents might falsely provide a positive response to the DGU question for any of a number of reasons:
o They may want to impress the interviewer by their heroism and hence
exaggerate a trivial event.
o They may be genuinely confused due to substance abuse, mental
illness, or simply less-than-accurate memories.
o They may actually have used a gun defensively within the last couple
of years but falsely report it as occurring in the previous year--a
phenomenon known as "telescoping."
Of course, it is easy to imagine the reasons why that rare respondent
who actually did use a gun defensively within the time frame may have
decided not to report it to the interviewer. But again, the arithmetic
dictates that the false positives will likely predominate.
The study authors then go on to state (here NSPOF = National Survey of Private
Ownership of Firearms - this survey):
In line with the theory that many DGU reports are exaggerated or
falsified, we note that in some of these reports, the respondents'
answers to the followup items are not consistent with respondents'
reported DGUs. For example, of the 19 NSPOF respondents meeting the
more restrictive Kleck and Gertz DGU criteria (exhibit 7), 6 indicated
that the circumstance of the DGU was rape, robbery, or attack--but
then responded "no" to a subsequent question: "Did the perpetrator
threaten, attack, or injure you?"
So note that these reported statistics rely on only 19 respondents to generate the 1.5 million estimated rate of defensive gun uses and that some of these 19 reports were inconsistent. Another possible unreliable report was:
one woman reported 52 (defensive gun uses in a year)!
The survey authors also state:
Some troubling comparisons. If the DGU numbers are in the right
ballpark, millions of attempted assaults, thefts, and break-ins were
foiled by armed citizens during the 12-month period. According to
these results, guns are used far more often to defend against crime
than to perpetrate crime. (Firearms were used by perpetrators in 1.07
million incidents of violent crime in 1994, according to NCVS data.)
Thus, it is of considerable interest and importance to check the
reasonableness of the NSPOF estimates before embracing them. Because
respondents were asked to describe only their most recent defensive
gun use, our comparisons are conservative, as they assume only one
defensive gun use per defender. The results still suggest that DGU
estimates are far too high.
For example, in only a small fraction of rape and robbery attempts do
victims use guns in self-defense. It does not make sense, then, that
the NSPOF estimate of the number of rapes in which a woman defended
herself with a gun was more than the total number of rapes estimated
from NCVS (exhibit 8). For other crimes listed in exhibit 8, the
results are almost as absurd: the NSPOF estimate of DGU robberies is
36 percent of all NCVS-estimated robberies, while the NSPOF estimate
of DGU assaults is 19 percent of all aggravated assaults. If those
percentages were close to accurate, crime would be a risky business
indeed!
In addition the authors note:
Should the number of DGUs serve as a measure of the public benefit of
private gun possession, even in principle? When it comes to DGUs, is
more better? That is doubtful, for two kinds of reasons:
o First, people who draw their guns to defend themselves against
perceived threats are not necessarily innocent victims; they may have
started fights themselves or they may simply be mistaken about whether
the other persons really intended to harm them. Survey interviewers
must take the respondent's word for what happened and why; a competent
police investigation of the same incident would interview all parties
before reaching a conclusion
o Second and more generally, the number of DGUs tells us little about
the most important effects on crime of widespread gun ownership. When
a high percentage of homes, vehicles, and even purses contain guns,
that presumably has an important effect on the behavior of predatory
criminals. Some may be deterred or diverted to other types of crime.
Others may change tactics, acquiring a gun themselves or in some other
way seeking to preempt gun use by the intended victim.[16] Such
consequences presumably have an important effect on criminal
victimization rates but are in no way reflected in the DGU count
So, my conclusion is that these kinds of random telephone surveys are not an accurate way to estimate the annual rate of defensive gun use in the US.