In the spirit of xkcd.com/793/ I will now offer uninformed opinions about what is wrong with macro-econ out of the blue.
The author of Unlearning Economics (whom, since I don’t know who they are, I will henceforth call UE) thinks most macro-econ is useless because it comes with a set of poorly defined and unrealistic assumptions. In particular, they argue against the counterargument to such attacks, namely Milton Friedman’s assertion that the asssumptions don’t matter as long as the predictions are correct. UE has several objections to this: that theories that don’t give micro-understanding aren’t enlightening even if correct, that there is often insufficient evidence to discriminate between theories, that experimental evidence in economics is too coarse-grained, that the assumptions are often not clearly specified, and that relaxing assumptions is always a positive good.
To me, UE’s argument goes like this: macro theories aren’t testable, so make them be things that sound reasonable. My problem with this argument is that just because a theory sounds reasonable isn’t good evidence that it is correct. If macro theories are all “explanations” rather than things that offer testable hypotheses, then that is terrible, but making them “explanations” that sound better doesn’t make the situation any less terrible. The only way to improve the situation is to offer testable hypotheses and then test them.
Consider the DSGE v. Minsky Model that UE brings up. They claim that without looking at internal mechanics, there was nothing to figure out which was correct until one predicted a crash and the other one did not, and then a crash happened. Well, this is how we found out which one is more correct! Looking at internal mechanics didn’t do it – both models have a lot of unreasonable assumptions. We know which of the two is more correct because and only because we were able to discriminate between them in terms of their predictions.
Or let’s take UE’s example of different kinds of models: water boiling and freezing. The two models they contrast are “water demons which make it boil and freeze at specific temperatures” and “water is made of two hydrogens and an oxygen”. One of these is certainly more true than another. But the reason the first model is terrible is not because it’s ridiculous, but because it’s untestable. I can make it testable by saying “oh, and these water demons are distracted by annoying music, so if you play LMFAO loud, water won’t boil”. Then it’s fine, we go and test it and figure out it’s false. The second model is even worse because not only is it not by itself testable, it also isn’t connected to the question. That is, it fails at being an explanation, too. I guess they’re trying to say that microfoundations are important to macro theories but this is a very poor way of doing so. And I don’t buy it anyway. If you had models with various assumptions and a way to discriminate between them with experiment, then you’d know which assumptions were tenable and which were not. Then, you would have a better understanding of what’s important in the leap from micro to macro than you would by simply taking all of micro wholesale.
Inherent in my criticism above is a disagreement with the notion that relaxing assumptions is always a good idea. It’s not. The whole point of modelling is to get away with treating things as simply as we can. The famous Einstein quote goes “Make things as simple as possible, but not simpler.” That’s how modelling works. Otherwise, you need giant supercomputers to figure out anything about anything, and way more giant supercomputers than we have to figure out anything remotely interesting about anything remotely interesting. Figuring out what is important and what is not should not be seen as a barrier to understanding simply because it stops us from considering unimportant things. Instead, it is itself a large part of understanding.
So now where are we? We claim that if there is a problem in macro, it’s not the lack of microfoundations, but instead that macro theories don’t offer testable hypotheses. So why don’t they? Here is where I speculate: it must be either because macro experiment is not progressing rapidly enough, or because macro theorists don’t care if their theories are correct. How physics advances is like this: people put forward hypotheses which agree with available data and then they have to say what not yet available data would differentiate between their theory and other ones. (This is why experimentalists tend to distrust anthropic principlists and string theorists.) And then experimentalists go and collect that data. So the problem in economics is either that there’s no new data that somehow expands on data which already exists, or that theorists don’t care if it’s there.