EVEN MORE ON LOTT….Here are the demographic questions that Lott used for a survey he conducted recently to reproduce the results of his mysterious 1997 survey. He claims that it is very similar to the one he did in 1997:
(We obviously have the area code, write down sex from the voice if possible.)
I have two demographic questions for the survey.
What is your race? black, white, Hispanic, Asian, Other.
What is your age by decade? 20s, 30s, 40s, so on.
That’s it? And from that he was able to weight responses up or down by as much as 4:1, magically morphing 2 out of 25 respondents from 8% down to 2%?
Look: the purpose of weightings is to ensure that your sample is similar to the demographics of the United States as a whole. So if your sample contained, say, 11% blacks, but blacks are actually 12% of U.S. population, then you would weight the black responses a little more heavily so that they constitute 12% of the total survey responses.
But weights like that are very small, only a few percentage points. Lott’s response, basically, is that “black Vermonters,” for example, are very uncommon. So if such a person ended up in his survey, he would represent .04% of the survey sample (1 out of 2,424), while he represents only .01% of the general population. Thus, that one response would need to be downweighted 4:1.
This is simply ridiculous. Even the most partisan hack wouldn’t try to get away with something like that.
Why does anybody believe this stuff?