This can't be good news:
In a series of novel experiments, researchers Gregory A. Huber (Yale), Seth J. Hill (University of California, San Diego), and Gabriel S. Lenz (University of California, Berkeley) found that voters are susceptible to these biases even when given financial incentives to behave otherwise and when the information necessary to avoid these biases was readily available.
In their study, "Sources of Bias in Retrospective Decision Making: Experimental Evidence on Voters' Limitation in Controlling Incumbents," reported in the latest edition of the American Political Science Review (APSR) published by Cambridge University Press, Huber, Hill and Lenz asked around 4,000 citizens to play a series of games assessing the performance of fictional politicians. As they played, they experienced changes in their earnings. The experiments also manipulated whether they won or lost in a lottery, when they learned about an upcoming election, and the rhetoric surrounding the election.
The findings suggest that incumbents who associate themselves with good news for which they bear no responsibility, implement policies that generate good news close to elections at the expense of overall voter welfare, and use rhetoric that encourages people to focus on how they feel in the here and now, ignoring the long-term, could benefit at the ballot box.
The results of the games showed that even though the players' earnings were the only accurate information they had about their incumbent politician's performance, the other manipulations affected the players. They tended to punish or reward the incumbent based on whether or not they had won or lost a lottery, and gave greater weight to earnings closer to the election when they learned about the election closer to it or after certain rhetorical statements. They persisted in this irrational behavior even when it was made clear to them that their fictional incumbent had had nothing to do with the lottery and that events closer to the election were no more informative of the incumbent's true performance than events further from the election.