FWIW, I have been told that members of the Unofficial Phatbox BBS line up very neatly along a 215/135 degree axis, so, as always, the sample is they key, eh?

Did you mean 215/35? If so, that's consistent with here--I computed the angle as 36 degrees.

OK, more seriously, what kind of model do you have in your political skunkworks? How many dimensions *will* do the trick?

I'm not really sure. I've long wanted to get my hands on raw data from one of the broad issue polls so I could do a principal components analysis and see what comes out.

I think two dimensions, one for economic and one for social issues is a good start. But I wouldn't be surprised to see some other major factors pop up.

I also read an article a while back where the authors had studied the voting record in Congress from roughly 1946-1996 and tried to determine the dimensionality there. Their conclusion was that it was mainly one-dimensional except during the 60's when a second dimension came into play on civil rights. They also noted that the explanatory power of the primary dimension (i.e. the degree of partisanship) was highest at the end of the survey period.

--John