I went to an interesting talk today on risk pricing. The equity premium puzzle is an issue that no one has really resolved yet. The idea is that the long-run return to holding "risky" equities remains excessive to the present day. One answer that the behavioralists give is that humans are simply risk averse, but many people find the degree of risk aversion necessary to explain the phenomenon inplausible.
The speaker's idea was to combine rational agents (with potential differing apetites for risk) with Bayesian agents who lean about the economy. The Bayesians are stupid, but through their learning converge the economy over time to the rational model. There are a few rational agents in this economy--the Warren Buffets--who know what's going on. Over time, these guys get richer and richer. The main conclusion of the paper is that the market can put a high price on risk, one that converges over time to a reasonable price, in an economy populated by smart people and stupid people who are quick learners.
There is a distributional aspect to this model as well, however. The smart people raise their share of income over time, due to their superior knowledge. Of course, in economic shock periods, as now, they do not do so well, but in the long run, under normal economic conditions they make out like bandits. I've discussed before the education model of inequality, which explains how the educated do better than the non-educated. This model, in which there are super-educated people, plausibly explains the movement at the very top of the distribution. This is happening in an institutional setting where it's harder and harder to deny the smartest and most productive people the proper value of their labor. You can wonder what is the optimal tax to apply to this group, but you should also wonder how to join it.