Thread by @JWMason1

Just this monring I've been told, on two different threads by two different economists, that a permanently inverted yield curve "must" imply expectations of ever-increasing deflation, and that higher savings by rich than by poor "must" mean wealth inequality rises without limit.

Neither seemed to even notice that these conclusions only hold only if you formalzie the question in a specific way and under specific assumptions.

A big part of learning to "thinking like an economist" is learning to substitute a statement within some standard model for any claim about real economies, ideally so seamlessly that you don't even realize you're doing it.

To be clear, there's nothing wrong with formal models. They are often quite useful! Even highly mathematical ones, sometimes.

The problem is when you lose the ability to distinguish the map and the territory - when you come to think of the terms of the model not as tools for answering specific questions but as real objects existing in the world.

This is obvious in the rational expectations literature, where the existence of a "true model" is a foundation of the theory. It comes through in terms like "structural" vs "reduced form", which suggest that some models have a different ontological status than others.

It comes through in the way econometrics is taught and discussed, where the premise is always that there is a true, underlying data generating process out there to be discovered, as opposed to just more or less useful ways of describing certain observable patterns.

It's all very strange.

For anyone interested, I've struggled with these questions in this old post on Haavelmo jwmason.org/slackwire/the-…, and in a talk I gave last year - adudio here: jwmason.org/slackwire/lect…

I think one thing that unites this is the desire for a kind of knowledge that simply isn't available. For instance, two common defenses of the insistence on microfoundations is that it allows welfare analysis, and that only microfounded models are useful for policy.

The first suggests that we can have positive, objective knowledge about which outcomes are morally better; the second suggests that we can identify a set of parameters that are truly fixed, invariant to all policy chocies. Obviously neither is possible.

Rather than accept that economic knowledge is always going to be local, historically contingent and dependent on judgements from outside, we construct an imaginary world as object of analysis, in which the kind of knowledge we want would actually be possible.

You can see this clearly in Lucas' classic 1976 essay on rational expectations. He starts with some very reasonable points about the limits of economic forecasting based on historical relationships between aggregates - that the parameters are gong to drift over time and so on.

And then there's this sort of turn where he says, one reason the paramters change is because people change their behavior based on policy chocies. But if you could do economics based on people's true underlying desires and choices, instead of observed behavior, you'd avoid that.

In effect, he says, since theory based on economic observables can only give us limited, approximate knowledge based on particular time, place, context, let's instead imagine a world where we could have more general and exact knowledge, and see what econ would look like there.

As @arpitrage points out, another piece of this is weak standards of correspondence required between original question and equivalent model one. You just have to "match the moments" of your stylized facts, with little/no concern for underlying mechanisms. ineteconomics.org/perspectives/b…