Ep. 459 Chicago-Trained Alberto Bisin Defends the Use of Mathematics in Economics
Bob has a delightful discussion with one of his NYU professors on the recent controversy regarding the use of math in economics.
Mentioned in the Episode and Other Links of Interest:
- The YouTube version of this interview.
- Paul Romer’s critique of “mathiness” in the theory of economic growth.
- Bob’s doctoral dissertation, with mathematical appendix showing the problems with assuming the real rate of interest equals the marginal product of capital. (The punchline is equation [4] on page 190.)
- The link for this episode’s sponsor, Monetary Metals.
- Help support the Bob Murphy Show.
About the author, Robert
Christian and economist, Chief Economist at infineo, and Senior Fellow with the Mises Institute.
Some interesting highlights for me (correct me If I misrepresent):
– Dr. Bisin explains that Krugman’s view of models is that they serve to illustrate some economic principle or relationship, not accurately predict. If Krugscum is really saying that consistently, it seems more defensible to me than Friedman’s acausal ‘anything that predicts’ standard.
– Dr. Bisin explains that the common expectation that Economists ought to be able to predict crashes if their models are correct — is akin to expecting geologists to predict the timing of the next earthquake. Great analogy. In the language of chaos theory this is trying to predict a phase transition in a complex system; the feedback mechanisms combined with multiple variables makes for unsolvable problem.
Now I would like to say there was more here to comment-on but for me there wasn’t.
It’s perhaps an interesting episode for people who already know Dr. Bisin.
I mean, for 95% of the time I was wondering when we’d get around to discussing the ways math are used in economics
— Economic Laws — invariant properties of networks of transacting humans: Logical but almost no math
— Market and Market-Intervention Analysis
— Methodology: What mathematical methods can be applied to what problems?
— Butterfly Collecting: How did these actors affect those outcomes?
— Economic Forecasting
— Business Analytics
All these are in the broad sense, ‘economics’, and very different legit applications of math.
Starting out with some structure like this would then naturally lead to the different degrees in which mathematics and modeling apply to each but also the larger issue of what is legimitately called ‘Economics’. Are we really engaging the science of economics by just applying analytical methods to historical datasets — or is that just stamp collecting?
Is it economics to focus on some particular business or process and try to forecast — or is that just economizing — and thus outside the legimitate bounds of the study of economic laws and analytical methodology.
So what are you doing when you’re using math ‘in economics’? Are you just butterfly collecting and writing a little story about the shape of the wings? Or are you using math to derive new analytical tools that will be later used to economize? There are very different things happening here and describing them in some kind of overview would have benefited most of the audience.
There was no construction for this podcast; instead we got the fireside chat about furniture and the he-said-she-said of personalities the world won’t care about in a few short years.
The death of listener comments makes me wonder if we’re already just trying to scream the truth into an eventual AI training dataset.
I think you should just start your own podcast?
I am an Operations Researcher who uses Mixed Integer Programming to model and solve Less than Truck (LTL) tractor and freight routing problems. Our models are extremely large and precise. We solve the models multiple times per day to determine how freight should be loaded and moved by the company I work for. One can think of this model as a very specialized micro level Econ model. These models are checked every day by people in the field using the model results, and we get feedback nearly every day on small issues here and there. We work to close the gaps between reality and model behavior constantly. These math based models deliver extremely specialized and practical results. To do this, though, means that they can’t be used to make sweeping generalizations outside of their very particular application. I think these are the sorts of applications of math modeling that make the most sense. Hopelessly inaccurate models of enormous systems like national economies are by and large useless. This is one of the reasons I feel so uncomfortable about the Fed reserve making important decisions like setting national interest rates. The models they use are aggregated in the extreme. There’s plenty of space for math. But I’m not sure if it’s really all that useful at the macro economic level.
One additional comment on this. It seems like a lot of macro Econ modeling dependents on model smoothness and continuity so that shadow prices and other marginal metrics can be derived to make claims about prices/etc. So much of life, however, is discontinuous and combinatorial. When things get highly combinatorial and discontinuous, most of the marginal results break down. That’s certainly true in my line of work. One of the reasons we use MIP rather than LP and NLP.