The title of this (relatively) recent article in Philosophy of Science caught my eye, and the content is quite interesting indeed. The authors, two Finnish philosophers, examine why economists have been so reluctant to embrace computer simulation. They consider simulation to be, roughly, a mimetic model the actions of economic agents that generates some epistemically important result reliant on the interdependence of the actions of these agents. That is, simulation is different from the use of computational techniques to, for instance, calibrate a DSGE model.
They claim that reluctance to use simulation is due to economists’ reliance on the “perfect model”, a construct which captures economically relevant relations in a simple and tractable way, and in such a way that logical arguments may be ported from one problem to another. That is, economists follow Kitcher’s “explanatory unification,” by which more and more economic relations are explained with fewer and fewer explanatory arguments. Learning the argumentative patterns is more important to economists than learning the results, and therefore simulations which focus on results, and contain argumentation within the black box of a simulation are seen as unsatisfying by many economists. To a lesser extent, analytic proof focuses on necessary conditions and numerical exercises give sufficient conditions for some economic event to come about. Sufficiency is in many ways less satisfying for explanation than necessity. I completely agree with all of this.
Much of this article could, I think, be summarized by focusing on the “inexactness” – Dan Hausman’s phrase – of economics. For all practical purposes, Newtonian gravity equations are completely and perfectly able to describe how gravity will act on objects we interact with. To the extent that we can improve our measurement of gravity (by more accurate constants, or by using relativity to deal with large and small objects), we are making progress toward some sort of ideal science. Economics is not like this. It is utterly Sisyphean to try to move toward true results/parameters/etc. in economics. We are nowhere close to being able to accurately model interaction at the scale of an economy. Therefore, economics is the art of developing patterns of argumentation – metaphors, for instance – that offer some insight into how the social world functions, and using these patterns for explanation and prediction. The nature of analytic proof is that it offers intuition as to why “if x then y” works, by mapping the mathematical objects back into their real world counterparts. Such a program will only work if mathematical constructs are used in a consistent manner, and this is why I think you see economists insisting on a far tighter paradigm than other social sciences: we need to know exactly what each proof means by “power” or “decision” or “social welfare” in order to perform the mapping.
This understanding of the method of economics is useful for reading economics papers, actually. No one cares about the mathematical detail of a proof, save to the extent that something new is introduced (like MCS twenty years ago) which allows us to derive new results. No one cares about the exact numbers derived by a model. In this way, the highly mathematical field of economics is actually just as qualitative as the other social sciences. What the mathematization gave us a set of easily portable and commonly known argumentative structures, which allows progress within economics to move faster than in our sister fields.
http://www.econ.iastate.edu/tesfatsi/WhyEconomistsShunSimulation.Lehtinen2008.pdf (Final published version)