I can't find any direct links, but I'd suggest Robin Hanson's blog (overcomingbias.com). His hypothesis is that most political/voting activity is done primarily for the purposes of a sense of identity and signalling. He discusses this in a number of posts.

His basic thesis: I have no incentive to vote "correctly", since the odds of my vote affecting the result are nil. Therefore, it costs me almost nothing to vote D solely to show my friends what a caring intellectual I am, or to feel good about myself. My hypothesis is that this is the main reason scientists vote D (the caring/intellectual image).

One hilarious example of this effect at work can be found at http://www.pipa.org/OnlineReports/Pres_Election_04/html/new_... -- the gist of it is that polling indicates that a decent number of Bush voters in 2004 believed that Bush held policy positions that he never held and in fact never claimed to have held. It is not that Bush lied to these voters about his issue preferences: they just liked him for other reasons and ascribed their own issue preferences to him.

John Sides explains a related phenomena here: http://www.themonkeycage.org/2010/08/me_talking_to_politico_... whereby voters don't necessarily know much about most issues and end up adopting issue positions held by parties they trust. See the link towards the end of that piece.

In general, there's been a ton of political science work on how election outcomes are largely determined by non-issue factors, like economic conditions. An electorate that voted based primarily on issues should have stable preferences that don't swing wildly based on the state of the economy.