Comment: More Flipping Or Something Else?

(See in situ)


More Flipping Or Something Else?

Since I have a strong interest in this kind of analysis (computer scientist by trade), I decided to reproduce some of the author's results and then investigate some new data.

Among the new data were NH 2008 Primary results (both Republican and Democrat) and NJ 2008 General Election results. The NJ results are not per-precinct but rather per-municipality, which I think is a suitable substitute. As you can see from the images linked to below, the same "flipping" trends are present. I also ran some computer simulations and found that votes "lost" or "gained" are very statistically significant in all three cases.

http://tinypic.com/r/1t27tg/6
http://tinypic.com/r/wthf9v/6
http://tinypic.com/r/9ggit4/6

Contrary to one of the assertions in the Vote-Flipping Paper, this is not a Republican-only phenomenon, and does not "benefit" only Republicans in the General election matchups.

I think these new charts suggest that there is some other underlying explanation for this apparent "flipping" phenomenon. I have not yet tried matching up demographic data here, but one obvious thing that comes to mind is that more populous precincts/municipalities will tend to be in urban areas. Regardless of Party affiliation, I think it may be reasonable to consider whether city voters tend to be more liberal/moderate, especially relative to others in their own Party.

I don't pretend to have solved any mystery here, but I do think that these charts cast some doubt on the "flipping" hypothesis (unless flipping is rampant by both sides on a huge scale, which I suppose is possible). I plan to continue studying data from other election years and non-Presidential elections in particular to hopefully find out more.