I wanted to see if there were any statistical basis for examining electability. So I decided to look at the Bush vs. X hypothetical matchups that some pollsters do. While I think these are meaningless in terms of predicting a winner in Nov. 2004 (every single poll, even those in Dem strongholds like Rhode Island, predict a Bush win), I decided to look at the comparative
margins each candidate gets. The results show that there isn't much of a different between the candidates, at least, based on the small sample size I used.
Methodology: I look at polls over at DC's Political Report. Only seven states have reasonably fresh polls showing matchups between Bush and the top five candidates. (For some reason, Edwards is consistently ignored by pollsters.) I ignored any state which had very old polls or which did not poll for all or 4/5 of the top five.
The states: CA, CT, NH, NY, OH, PA, RI
Data adjustments: Clark wasn't polled in the Ohio or RI polls, so I gave him the average of the other four candidates' margins for those two states.
Lieberman's margin in CT was, unsurprisingly, outsized - 11% vs. an average of -0.5%. For the sake of this experiment, I adjusted his margin to make it the same as the next-highest person's, Kerry's, which was 4%.
The results:
Lieberman: -4.67%
Kerry: -5%
Clark: -5.94%
Gephardt: -7.5%
Dean: -8%
So from top to bottom, there is little more than a 3% difference in average margin. That strikes me as rather small, especially since most polls' margins of error are seldom as good as 3%. (Though feel free to quibble with my adjustments. Those were just judgment calls.)
And if you look at just the three swing states included (NH, OH, PA), the top-to-bottom difference shrinks to less than 2%. I know, we're stretching the sample size very thin here. I'd love to be able to run these #'s with more states, but I just don't have any more data.
