In theory, people vote for a candidate because they prefer that candidate. Recent research from a pair of economists indicates that in practice people may prefer a candidate because they voted for him.
Harvard's Sendhil Mullainathan and Yale's Ebonya Washington, in a paper for the National Bureau of Economic Research, examined surveys conducted from 1976 to 1996 that asked young adults about their attitudes toward a candidate two years after the candidate's election. They discovered that those who were eligible to vote two years earlier were "twice as polarized as ineligible ones" in their opinion of the candidate. The ones who got to vote showed more approval for "their" candidate, and more disapproval for the one they didn't vote for, than those who started with the same opinions but couldn't vote to express them.
The effect held up even after Mullainathan and Washington took into account possible confounding variables such as age and level of information. "Two years after an election," they conclude, "a citizen…may hold a favorable opinion of [a] politician in part to avoid the internal discomfort of having voted for a person for whom the individual has a poor opinion."
One of the conclusions these findings suggest is an efficiency argument for term limits, since they limit the number of elections in which our judgment of candidates is skewed by our previous votes for them. Another conclusion is that high voter turnout is not an unalloyed good, since it guarantees that more voters next time will have irrational biases in favor of incumbents.