Freedom, Markets, and Well-being Fall 2017

Sharks Eat the Retrospective Voting Theory


Achen and Bartels compare three theories of democracy:

  1. The Populist Theory: the government in a democracy reflects the will of the people.
  2. The Retrospective Voting Theory: democracy promotes good government because government officials who perform badly are voted out of office while those who perform well are kept in.
  3. The Group Theory: people vote for parties that they think of as representing the groups with which they identify. That is all that happens in voting; elections do not turn on questions of policy or even the competence of the incumbents.

Achen and Bartels believe the first theory is false on the grounds that voters do not have specific enough preferences about what the government should do. They claim that this is commonly accepted among political scientists.

The chapter we discussed today assembled evidence against the second theory, the Retrospective Voting Theory.

Our next class will concern their case for the Group Theory.

Retrospective Voting Is Blind

Achen and Bartels say that retrospective voting is “blind.” It seems to me that I was operating with at least three different assumptions about what this means.

  1. Voters always vote against incumbents when things go badly.

  2. Voters act irrationally when they punish incumbents for events that are not their fault and favor replacements who are likely to do no better or even worse.

  3. Retrospective voting does not reliably produce good government.

The first point is false if taken literally. Achen and Bartels themselves note this, so they presumably favor something weaker, such as that voters “usually” (or “often” or maybe “randomly”) punish incumbents when things go badly. Something like that is probably true, but it does not really offer much of a prediction about how democracies work.

The second point is debatable. As Professor Brown pointed out, when a principal has very little control over an agent, the sensible thing to do might well be to punish the agent whenever things go wrong. This might be the only way to get at an agent who performs badly, even if it also punishes agents who do not perform badly.

The third point is probably the closest to what they mean. One thing to note about it is that it implicitly comparative. Retrospective voting is not reliable compared to what? If there is no more reliable way of producing good government, retrospective voting might well be the best thing on offer.

Our Discussion

Achen and Bartels take pride in the shark case because they have enough data to eliminate all the competing possible explanations of the vote in that year. Etelle granted the point but noted, rightly enough, that one case hardly proves that all votes work this way. I think that they believe they have a lot of suggestive examples and that the shark case is supposed to be the one they can rigorously document. But it’s still possible that there are hidden differences between the shark case and the others.

We talked about whether people now are more informed than they were in 1918. There have to be some differences. Everyone would know about a plague of locusts now, but the people of Nebraska did not appreciate its scale then. At the same time, I think Aaron is right to say that the people in our seminar room are really quite unusual in how much attention they pay to information that is relevant to politics. And we may be overly confident in the extent of our knowledge.

Professor Brown made a point about the economics of information and public goods problems. It’s really costly to acquire information about politics and public policy. But the vote is going to go one way or the other regardless of how informed any particular individual is. So there is very little incentive to be informed.

Tristan noted that there are interesting questions about which officials get blamed. Do federal officials get booted when there is a natural disaster while state and local ones do not? If so, why?

I noticed that in the discussion of the influenza case, Achen and Bartels said that the voters did not punish the party in power because no other party made an issue of it. I wonder how big an effect that has. If it is a large one, then you can imagine a democracy functioning better if the leaders of the major parties agreed to keep some things off the table. In particular, I am thinking of Senator Goldwater’s asking to meet with President Johnson in order to discuss the 1964 election. Specifically, Goldwater wanted to make sure that race would not be an issue in the campaign, despite the fact that it would have surely helped him.

If only everyone were as wise.


Achen, Christopher H., and Larry M. Bartels. 2016. Democracy for Realists: Why Elections Do Not Produce Responsive Government. Princeton: Princeton University Press.