In Point Blank Gary Kleck writes:
The aggressor's possession of a handgun in a violent incident
apparently exerts a very slight net positive effect on the
likelihood of the victim's death. The linear probability
interpretation of the OLS coefficient implies that the presence of a
handgun increases the probability of the victim's death by 1.4%.
thus the violence-increasing and violence-suppressing effects of
gun possession and use almost exactly cancel each other out. This
small association is statistically significant, however, because of
the very large (n=14,922) sample size.the effects of aggressor weaponry are quite substantial when taken
stage by stage, i.e., when separately examining attack, injury, and
death. This is why impressive-appearing results can be obtained
when researchers examine, for example, only the last stage, looking
solely at the impact of guns on the likelihood of the victim's
death, among those wounded...The findings also imply that if gun possession were reduced among
aggressors in violent situations, total assault injuries would
increase, the fraction of injuries resulting in death would
decrease, and the total number of homicides would remain about the
same....
This is perhaps the most bone-headed claim Kleck makes in his book.
Kleck's data implies that a 10% reduction
in gun possession by aggressors would result in a 2% decrease in
injuries and a 6% decrease in homicides. This is definitely not
"remain about the same."