As Stigler and others have suspected, if people are irrational for centuries, maybe the definition is wrong. As they said in A Princess Bride:

Fezzini: Inconceivable!
Inigo: You keep using that word. I do not think it means what you think it means.

Bryan Caplan's Myth of the Rational Voter is a very good book (which, btw, is not at any of my local bookstores, supporting his thesis). It is interesting, with fun facts about stupid voters and smart economists (OK, a little self-serving, but so is Oprah). Most importantly, he took the time to make it brief, something all too few authors are willing to do these days (see Bugliosi's latest book on the Kennedy assassination: 1612 pages!).

But as to his main point that voters are irrational, he mentions the masses perpetual belief in protectionism, suspicion of market solutions, support for make-work, all stuff economists know to be inefficient, part of a patronage system that delivers fewer benefits than costs [there's a problem here in that Keynes--quite the popular economist--helped rationalize almost every antigrowth proposition at some point in his career (including killing Kulaks, see Euthanasia of the Rentier), just as Krugman can be counted on to support anything Kim Il-Jung does in practice--increasing returns to scale, etc. Let's ignore that for now]. But, and this is a Jennifer Lopez sized butt, what if people are maximizing relative status? That is, what if my utility function weren't U(x) but U(x-mean(x))? Wouldn't then I still spend less on cabbage when cabbage prices rise in either case, but at the policy level, I don't give a crap about things that are broadly affecting aggregate income--I just want to earn $100 more than my wife's sister's husband (to steal from H.L.Mencken). In experiments, because any single purchase doesn't affect the mean, we see evidence of maximizing U(x), but at the policy level, we see an indifference to mean(x). Solution?!

I argue in Why Risk is Not Related to Return that a relative status utility function can explain the failure of the risk-return empirical relationship. Perhaps it can explain persistent irrationality on policy matters?

The question then becomes, why do so many economists think it's obvious that people should maximize their absolute wealth (in a Von Neumann-Morgenstern way) while the average guy cares about his relative status? Perhaps because economists are elitists who would easily work in any decent University worldwide, so his 'relative status' is very peculiar. Perhaps it's because while economists enjoy the cynical assumption of 'self interested' motivation, they can't believe that self-interest could be that petty (more baubles are ok, but relative wealth?).

In any case, one may learn about 'better' utility functions, just as you can learn to be more considerate, patient, etc., but fundamentally, humans evolved their preferences in a zero-sum world, growth really starting in the mid 18th century, so human nature (preferences) evolved before then.

The bottom line is that if the average person is looking at the world as zero-sum game, it makes sense he will favor policies that he can understand (eg, we give a monopoly to mailmen) over one's he can't, for targeted benefits (likes) and diffuse costs (indifferent), and there is nothing in the system that will make him want to be more enlightened.