|
The initial problem with Kellermann's NEJM 1993 study, and the forthcoming UPenn study, is that they start at the wrong end by taking as their study group people who have already been shot, and then comparing them to people who haven't. That makes it almost unavoidable that you're going to be stuck with a ton of variables that you have to control for; better to avoid that right off the bat.
What strikes me as a more useful starting point is to assemble a study group of private citizens who keep a firearm for home defense, or have a CCW permit and carry for self-protection, depending on what you specifically want to measure. Then, once you've established who your test subjects are, you assemble your controls by selecting ones who most closely resemble each individual test subject in every socio-economic and behavioral aspect except firearms ownership, so as to isolate that one variable to the greatest extent possible. I don't think there's any ethical issue when you're only asking people to do what they would have done anyway.
Then, over, say, a five-year period, you require the members of both the study and control group to report becoming the victim of any of a number of types of crimes, attempted or completed, such as assault (physical or sexual), robbery, "hot" burglary (while occupant is home), non-negligent manslaughter, and murder (obviously, in the case of the last two, if completed, somebody else will need to inform the researchers). At the end of the study, you tally up how many times members of either group have become victims of one of the crimes, break that down by foiled attempts vs. completed crimes, and crunch the numbers.
It's going to be comparatively expensive and time-consuming at the front end, and it may not even yield any statistically significant results, but any results you do get are going to be a damn sight more reliable than the econometric modeling garbage that the lines of Kellermann and Branas have engaged in.
|