Towards smarter democracy
I'd add that people use their political opinions to signal things they'd like their acquaintances to believe about their personality. For example, people support aid to the poor to signal to their friends and co-workers that they are generous people, and thus good people to be friends or co-workers with. The actual impact of any resulting policy on poor people or taxpayers or others generally remains unobserved by friends and co-workers and thus is irrelevant. Even if a friend believes the policy is mistaken, but that the opiner's heart is in the right place, this is sufficient for successful signaling. It's coming across as having a genuine desire to help the poor that counts. Similarly, people are motivated to vote in order to signal their altruistic participation in the community.
The one area Caplan fails to explore is very small-scale elections. I've long thought that juries, for example, are a much better example of democracy than large-scale elections. Jurors are forced to take the time to learn the facts (thus overcoming rational ignorance) and have a real case at hand (thus overcoming much of the mismatch between social signaling to ignorant friends and reality). They can clearly foresee the impact their verdict will have on the defendant. While Caplan extols the virtues of judges and other experts like himself (he'd like to have a Supreme Court of Economists to overturn bad economic legislation), such a Platonic oligarchy of experts needs to be checked and balanced by democracy. Juries, and the right of juries to judge law as well as fact (and for lawyers to argue law as well as fact in front of the jury), along with an unimpeachable right to trial by jury, are a much better way than large-scale elections to check oligarchy.