Thursday, April 26, 2012


I have highlighted in the past some exceptionally bad examples of forays of physicists into Economics (search for "Econophysics"). Not all are that bad, but they generally have in common that they portray the economy just as exogenous stochastic processes where no economic agents take decisions. That can make sense in some contexts, but these are rare cases. Now it seems chemists are venturing into Economics, what good could that bring?

Yochanan Shachmurove and Reuel Shinnar are an economist and a chemist who managed to get a paper into the working paper series of the department of Economics at the University of Pennsylvania. So it must be a serious piece. Their point is that is Chemistry, they have to deal with chemical reactors that depend on many variables and are very difficult to predict. This is not unlike an economy, where a multitude of factors may matter in ways so complex with some randomness thrown in that forecasting is very difficult as well. The authors suggest to use partial control, which involves identifying a few crucial variables and monitor those for forecasting. That does not look like much of an innovation to economists, as we are used to abstract modeling, factor analysis, econometrics and simple rules like the Phillips Curve or the Taylor Rule.

The methodology of partial control they are trying to push, though, hits a few roadblocks when applied to Economics. The first is that it needs an objective, which is easy to set in a chemical plant (it is the choice of the plant manager) but not so easy for an economy as a whole. The second is that it needs to separate the problem into independent units. They suggest, for example, to treat the United States as independent from the rest of the world. That may work for some questions, but many it does not, especially when the point is to summarize complex interactions. Third, the procedure requires a hierarchy of reactions. Much of our understanding of general equilibrium would not be captured by such constraints. Fourth, the method relies a lot on the ability to manipulate controls. That is easy in a chemical plant, but an entirely different problem in an economy. The authors take the example of the Fed and interest rates. Well, the Fed has a target on one very special interest rate, all others are market driven.

While Shachmurove and Shinnar offer scattered examples of how to apply partial control, there is no sense of how a complete model would look like. I would wait to see a model in operation before calling this an interesting modeling strategy for Economics.


Anonymous said...

I anticipate this was written by a retired chemist, much like the previous pieces were written by retired physicists.

Vilfredo said...

Reuel Shinnar is not only retired, he is deceased, according to the link.