It’s a rare privilege to be able to choose which one (or two) of 17 Nobel economics laureates I can have dinner with. The 4th Lindau Meeting on Economic Sciences achieved exactly what it sets out to do, at least for the invited young economists: educate, inspire and connect.
Despite being inspired by Diamond’s brilliant intellectual attacks on Prescott, and by Mirrlees’ astute analysis of economic policy, I was a little surprised by the issues raised during ‘Panel on Behavioural Economics’. For a start, the only economist to win a Nobel for what we now call ‘behavioural economics’ – Daniel Kahneman – wasn’t there. It was left to Maskin, Akerlof, Selten, Aumann, Phelps and McFadden (and Martin Wolf) to fill the gap.
The following issues were raised:
1) Criticism: Behavioural economics is essentially a patchwork of unrelated models, which explain specific deviations from classical economic theory. There is no single ‘behavioural economic theory’. My answer: The whole of economics is a patchwork of models, which explain specific phenomena. You shouldn’t apply moral hazard to a lemons market. A good economist uses the right model to explain the right phenomenon, but the best (i.e. curious) economist comes up with the best model when all other models seem to fail.
2) Criticism: Experiments put people into artificial situations, which bear no resemblance to reality. My answer: The only reason why we set up experiments is because we observe particular deviations outside the lab. In the lab, we can measure (very) precisely the effects of particular treatments and be more accurate about cause and effect. Experiments may not prescribe perfect policy, but they can push it in the right direction.
3) Criticism: People can learn in any any environment, so eventually most deviations from classical models disappear anyway. Real-life mistakes are costly so people learn fast. My answer: firstly, mistakes don’t always disappear in the long-run. Secondly, these statements are almost contradictory: we do not always have the opportunity to learn over long periods of time in real life because mistakes are costly. Experiments allows us to see what happens when we can’t learn.
Experiments must be done carefully. Plott and Zeiler showed recently that Kahneman’s endowment effect can be undone by extensive subject training. David Eil, whom I met at Lindau, also showed that he could reverse hyperbolic discounting by allowing subject to choose the time delay between two fixed payments and produce what he, with Vincent Crawford’s encouragement, called ‘hypobolic discounting’.
Doing good experiments and creating good models to explain the data they generate is simply good science. Ignoring the data is not.