On Noah Smith on weather chaos and economics

On weather and economics. As you have mentioned we disagree on the promise of using more realistic micro foundations. I am not optimistic. The reason is thateople are much more complicated than molecules. The ideal gas law plus something on evaporation and condensation are extremely simple and fit the data very well. Meteorologists have solid micro foundations based on elementary chemistry and a bit of thermodynamics -- all, I think, stuff done pretty early in the 19th century. Psychology is not at that stage yet and might never be. Micro foundations for the social science can't be approximately accurate -- people are too complicated.

Ironic that so much of general equilibrium theory is due to Arrow, a guy trained in meteorology.

On chaos: it was huuge in the late 80s. Larry Summers asked me to look at chaos and see if it could appear in asset prices. Total failure. Might be me, but the models gave very simple predictable dynamics. Weather can be forecast a bit ahead (not where tornadoes touch down or even when it srarts to pour but you know the sattelite stuff is useful). The long run is very hard.

Asset prices are unpredictable in the short run, but some can predict in the long run (say Shiller). The difference is that there are not a lot of molecules looking for quick profits looking for short run patterns and eliminating them.

For other prices not updated continuously on double auction markets and for real flows, things are better, but people looking for patterns and making them go away matter there too. People aren't rational, but they aren't stubbornly dumb enough for me to model them.

Back to Summers, this was the context in which he said " You know my opinion. In the history of economic theory, the day in which they came up with the utility function wasn't an especially good day. The day in which they came up with the idea of the supply curve was an especially good day". Playing with models of chaos is the only thing that made me suspect that he wasn't right.