Some Intellectual History for Matt Yglesias

I was taught that, at its base, the emphasis on microfoundations grew from the notion that macroeconomics should not posit individual behaviour or responses (such as, for example, money illusion) that would be considered irrational or otherwise at odds with the implications of microeconomics. It seems like a self-evidently good idea that the underlying models or assumptions about decision-making and market behaviour should be consistent between the two disciplines since, if they disagree, they can't both be right.

This tension arose mostly with the primacy of Keynesian economics and its emphasis on the study of the behaviour of aggregates rather than of individuals or firms.

Without the discipline of microfoundations, it becomes logically impossible to distinguish between a) observed relationships that are highly dependent on past circumstances, and b) true structural or behavioural relationships.

The Lucas Critique, to my understanding, played a huge role in this as well; your macro data are only relevant to the regime in which they're observed unless they have microfoundations underpinning them.

4. Lucas read Hayek.

5. Leijonhufvud read Hayek.

6. Hicks rejected Keynes -- and rediscovered his inner Hayek.

It was NOT a secondary fact in this history that the founder of modern "Keynesian" economics abandoned the theory, and immediately began writing papers and books on equilibrium "lags", "Austrian" capital theory, and other rediscoveries of his "inner Hayek".

Hicks = founder of ISLM = modern "Keynesian economics".

It probably didn't help that one of Hayek's earliest criticisms of Keynes was his utter and total lack of micro reality in his theory. Hayek was proud of the micro foundations of Austrian "macro" theory. Austrian macro was specifically derived from micro because before Keynes there was no distinction.

Maybe someone can help me understand the practice of modeling in macro. The standard in business economics is to validate a model by having it predict out of model data. Often, the data will be divided into a training and testing set. It doesn't seem that macro modelers follow this approach. They simply fit the model to data and accept it. But their is no trick to fitting a model to data. It's fairly easy to achieve a good fit. The acid test of a model is its forecasting ability. That's how business economists compare various models.

It seems to me that it would be fairly trivial to have all the modelers step forward with their favorite models and try to predict a standard data set, such as GDP for the 1990's, based on data before the 1990's. Why is that such a difficult thing to do?

Of course, someone would have to step in and write a model for the Austrians since most Austrians oppose any such effort. But the results would be very enlightening.

Microfoundations have also, it turns out, added discipline to modeling. Because assumptions are less arbitrary, they're easier to test.

You write "the obsession with microfoundations looks pretty silly, and it is."

Almost anything carried to the point of true obsession can be made silly. But I don't think it's silly to try pretty hard to connect to microfoundations.

When it is known how a theory maps onto microfoundations, it can give more ways to check it and/or to influence the outcome, and so can easily be very useful. This is not just some mindless cargo cult extrapolation from the practice of those subfields of physics which rest directly on exact physical laws. It shows up naturally in other fields like chemistry, biology, medicine, and various engineering-ish disciplines (some with names involving "engineering," some with names like "metallurgy" and "aerodynamics").

On the other hand, my impression of anthropology, sociology, and much of psychology is that they share your lofty disdain for microfoundations. If so, it is clearly possible to explore quite complicated fields without worrying much about microfoundations.