Kevin Drum tries to understand John Lott’s weighting of responses in the 1997 survey. Kevin is not trusting his internal bullshit detectors–actually they may have malfunctioned from an overload. Overall, the survey is large enough to produce decent results, but one must remember that focusing on respondents within the sample reduces the sample size one is actually addressing.
In Lott’s work that means if 25 people reported defensive gun use, 25 is your sample size. Given that, the margin of error is +/- 20%.
At that point there is no reason to weight because you don’t have a sample size appropriate to make any reasonable conclusions. His new study will not have sufficient sample size of those using defensive gun use to make any reasonable conclusion either.
Lott is a hack. The only real problem with that is people listen to this particular hack. From what I’ve read of his stuff Gary Kleck is far more reliable and trustworthy in his research, but doesn’t get the press coverage because he isn’t a hack.