David Friedman on Behavioral Economics: Intriguing Research Project, with Reservations
I have long argued that the economic assumption of rationality is useful not because it is a complete and correct description of real world behavior but because it describes that part of behavior that is predictable. If half the time an individual takes the actions that best achieve his goals and half the time he acts at random, then modeling his behavior as rational with some random error probably does as good a job of predicting it as we can do. When dealing not with a single individual but with the aggregated affects of many individual actions, random actions will tend to cancel out, making the predictions of a rational actor model more accurate than they would be for a single actor. Further, in some but not all cases, the actors that most matter are precisely those selected for successful decisions in the past—successful speculators, for example, have more money to speculate with and so a larger effect on markets, than unsuccessful ones.