« Responses to "protests" shutting down Trump presentations | Main | Expansion of right to carry laws »
John Lott demolishes gun study in The Lancet
Debate here. It underscores what criminologist Gary Kleck has said--all the antigun studies are being published in medical journals because the editors, and peer reviewers, don't know anything about criminology. Here they initially looked at several years, then decided to focus on just one because the data on laws was supposedly more comprehensive (??? Ascertaining State laws in 2009 should be no more easy than ascertaining them in 2010 or 2008. I suspect they ran several years, and picked the one that gave the desired result). And, as Lott points out, there are so many variables State-to-State, that it's better to use time-series analysis... here are the years when various States adopted various laws, did their crime rates before and after vary in ways that those of other States did not? In the end, the study's spokesman is reduced to "the study shows we need more studies," hardly a headlines-drawing conclusion.
The moment I heard of it -- universal background checks will reduce homicides by 90% -- I knew either the authors or the headline-writers knew nothing about the issue. Imposition of a totalitarian regime with random searches and a presumption of guilt couldn't reduce homicide rates by 90%.
And to think they managed to get this into The Lancet...
4 Comments | Leave a comment
There are liars, damn liars, and statisticians. For a study to be truly valid, the statistical model and data collection process must be designed BEFORE the experiment. So one could develop an hypothesis, select a statistical method, set up the data collection procedure, and wait for some future year to obtain data in order to have any validity in the study. Using statistics on past, uncontrolled actions simply produces lies.
For every American murdered by a firearm, 25 Americans are murdered by incompetent, lazy and/or careless doctors and other healthcare professionals.
THAT would seem to be a much more pressing issue than worrying about firearms.
Thanks, David. A minor quibble. You want to use a "panel" data set where you can follow lots of places over time. The most accurate option is to see how a place’s crime rates change after new laws go into effect and then compare that change with the changes that occurred in places where the laws didn’t change. What makes this different than pure time series is the part that comes after the word "and" in the preceding sentence.
Thanks again.
The Lancet isn't what it used to be, and hasn't been for nearly a generation now....
Unfortunately, as I try and teach medical students and residents, the journal crap gets published in doesn't change the essential fact that it is CRAP.