It seems to be common wisdom among economists that voting is irrational, on the grounds that you are extremely unlikely to be the deciding vote. I don’t think this is the right way of looking at it. Obviously, if you knew how everyone else was voting, then you could decide whether or not your vote was needed beforehand. But us mere mortals will, I think, have to stick to Bayesian analysis. Before the votes are counted, you have some estimate of the probability that your side will win. Every vote you know about updates this probability – even if there are millions of votes. If you friend tells you he is voting Obama, and you didn’t already know that, then you must consider an Obama victory to be ever-so-slightly more likely than before. And the same analysis applies to your own vote. If the probability of an Obama victory is, in your estimate, X without your vote, then it must be X+epsilon after you factor in your vote. A small epsilon, perhaps, but not zero! If you think that increasing your side’s chance of victory by epsilon is worth standing in line for two hours, then it is quite rational to vote. This binary stuff is for computers; even voters – notoriously a bunch of sheep – can do better than that.