I sort of wonder how this book would sound if written this year instead of two (or three) years ago. While I'm a fan of free markets and market mechanisms, it is harder to dismiss "antimarket bias" as irrational. Right now we are confronted with rather substantial evidence that the financial industry's efforts did not contribute to the common good, but rather enriched themselves at the expense of everybody else. If humans are not capable of evaluating long-term risk correctly, or if influential agents suffer irrational biases in their economic decision-making, then perhaps the naive view is more correct than economists would like.
Still, Caplan's thesis seems to make sense: voters obtain the policies they want, even if these policies make no sense.
Unfortunately, there is plenty of evidence on the other side of this! Lawrence Lessig's current work provides "easy examples" where sound policy and even voter preference lines up on side A, yet the government sticks with side B. Marijuana policy is an easy example--- although most people favor restricting it, solid majorities favor fines only (no jail time) and allowances for medical use. But this doesn't translate into law.
That is, voters can't obtain the policies they want, even if these policies make sense. (FWIW, I think Lessig's ideas about why this occurs are too simplistic by far.)
That brings us to Diane Feinstein, long-term Bush II enabler, and her latest snit-fit about not being consulted on the appointment of Leon Panetta to head the CIA. (Don't blame me, I voted for Tom Campbell--- though I can't think of a Republican since then.) Here, I think, is somebody who fits both models! Feinstein lines up on the liberal side of economic policy, giving her supporters the "irrational" policies they demand. But she utterly fails to be responsive to her base's agenda on a host of other issues. I don't think this is because of fundraising--- it's more likely that on civil liberties, war, etc., there is little reason for her to fear electoral defeat.
Feinstein needs only provide a clear distinction on a sufficient subset of voter preferences in order to carry the day. The remainder she can deal with based on less rational (or at least less vote-seeking) considerations, following the model Caplan lays out of "rational irrationality". When the cost (probability of electoral defeat) is low, Feinstein can consume more "irrationality" (voting against her supporter's desires). Caplan talks about "slack" within which politicians can vote for rational policies but this doesn't cover the whole story. A market with (usually) only two contenders, and high transaction costs, is not a great one for voters.