I often ask macroeconomists why their models are so simple, far simpler than the real world. I asked one macroeconomist what “constraint prevents you from examining less simplified models?” His answer was clear and direct: “In my case it’s IQ” but he goes on to claim that this is true of “mainstream macro people.”
He goes on to explain his “problem”. He says it would be “very hard to understand the economics of a two hundred sector model”. The real world has (at least) tens of thousands of sectors and actors. If he finds it difficult to understand a two hundred sector model, then why should we have any confidence in his attempts to understand the real world with far simpler models?
Macroeconomics cannot be an experimental science, but neither can problems in stellar physics. Astrophysicists instead use computer models which are simplifications of real stars but can be solved numerically. By solving many instances of those models they can determine if the numerical models are consistent with observations, and discover how various parameters affect the results. In this way, they can gain some understanding of what makes the model “tick”. A prime example of this is the Flash project created by Bob Rosner at the University of Chicago.
Macroeconomists (or, at least those meeting a minimal IQ level) could do the same thing: build a computational version of a macro model, solve it for a wide range of empirically plausible values for the key parameters, and explore the results to learn which parameters produce which outcomes. My collaborators and I have been solving 200-sector models for over a decade. The 2019 JPE paper I wrote (which represents work done by 2014) showed that this is quite feasible even for a demanding stochastic model of climate and the economy. The problem we had was Editor Piazessi’s opposition to doing sensitivity analysis, Editor Heckman’s threat to reject the revision if I would dare to criticize Piazessi in any document, and Lars Hansen’s refusal to give Cai and me any access to UC computers even though UC was getting >30K in overhead every year from NSF for this work. IQ is not the only problem getting in the way of macroeconomics research making use of modern computational science.
Readers of my blog will not be surprised by this after seeing Chari’s 2010 Congressional comment on the mental weaknesses of macro minds. Readers should also note that I am not the one asserting that macroeconomists have insufficient IQ, but that I am just reporting what macroeconomists tell me and Congress.