Hacker News Reader

Democrat versus Republican makes more sense as urban versus nonurban than anything else.

Very true (and parent makes a really good point, also).

I do wonder: rather than a single axis, how many would be appropriate? If you did PCA on... well, some crazy mishmash of factors -- demographic, political, socioeconomic, geographic -- what would be the eigenvalues of today's US politics? (And what's the minimum sensible number of values to capture the range of experience?)

Man, I can just hear my cofounder exhaling in frustration, at such an ill-posed problem, and abuse of terminology... and no doubt there are real, quantitative answers to some of this (and PCA is probably naive compared to... latest whiz-bang etc)... But still.

I imagine at that point the cause and effect would become mightily, mightily mixed up. Maybe race is a cause of political affiliation (it certainly isn't an effect), but surely certain factors, like donations, would be an effect of political affiliation. As it already stands, it seems reasonable to me that the urban/suburban/rural split might be equally as much an effect as a cause.

As it stands, that sort of analysis might be interesting, but is of little value without separating cause and effect (and surely those will be different for everyone).

Reply on news.ycombinator.com