Sunday, July 5, 2009

Behavioral economics doomed?

Rather interesting lecture claiming that behavioral economics shouldn't be as hot as it is. I don't buy it.

First of all, the claim isn't really that strong: that behavioral economics has plenty of hope of improving on classical economics, but little hope of replacing it. Well of course. We know classical economic theory works amazingly well in the laboratory and in the real world in almost all situations, in the aggregate. We're certainly not going to throw out that success just because psychology leads to some quirky behaviors sometimes.

Secondly, I firmly put learning in the category of behavioral economics. When Levine says several times that a phenomenon isn't a result of irrationality but of imperfect learning, this only says to me, as a behavioral economist, that there is hope to model the process of learning so that these quirky situations can be understood more generally and improved upon, not that behavioral economics has nothing to say on the matter.

Thirdly, I don't throw out all hope of behavioral economics "replacing" classical economics just because classical economics is in fact so useful. Relativity didn't "replace" Newtonian physics; it kept the results of Newtonian physics in the domain in which they are accurate and refined the theory to be applicable in more contexts. This is a perfect analogy for where I see behavioral economics heading.

Classical economics makes predictions without any mechanisms for individual behavior. It just calculates aggregate outcomes which would result if agents were perfectly rational. This "as if" approach does not actually claim that agents are perfectly rational. It says that the aggregate quantity that results is as if the agents were perfectly rational and individual level behavior were in fact not the noisy misunderstood mess that it is (and Newtonian physics makes predictions as if mass did not bend spacetime.)

Behavioral economics should retain the aggregate predictions of classical economics but actually try to model the individually noisy, dynamic process that leads to it. In that sense, I think there is plenty of hope for behavioral theory replacing classical theory, even if classical theory remains incredibly useful in most situations and continues to be the shortcut method of calculation (just as I do not take time dilation into account when calculating the velocity of a projectile on impact).