This paper focuses specifically on US presidential elections, though the broad principles are applicable to other elections. One of the authors is Nate Silver

This paper focuses specifically on US presidential elections, though the broad principles are applicable to other elections. One of the authors is Nate Silver (of fivethirtyeight.com), who has probably spent as much time doing statistics on US presidential election results as just about anyone else in the world. Another is Andrew Gelman, who's no slouch either. Academically he's got way more depth than Nate Silver, but he isn't as focused on elections. Authors Gelman and Edlin wrote a paper about why it might be rational (in utility-maximizing terms) to vote that's famous enough that I've read it (which isn't necessarily very famous). One-sentence summary (though the whole thing is only 9 pages long and not terribly scary): in a US election, your chance of deciding the election if you vote ranges from about 1 in 10^7 in swing states to about 1 in 10^10 or less in very one-sided states. The average vote, in some sense, has about one chance in 60 million of being decisive.