The somewhat suspect epistemology of economics
“Don’t think, but look!” – Wittgenstein, Philosophical Investigations (66)
I’ve been wanting to write something about the epistemology of economics for a while, and have been spurred on by Noah Smith’s post yesterday, linking back to an older piece from Frances Woolley at Worthwhile Canadian Initiative. I recommend you go and read them yourselves if you like, and here I take a very similar line to Noah.
As anyone who did the equivalent of Logic 101 will know, deduction is essentially a schema for truth preservation. A logically valid inference is one where it is not possible for the premises and the negation of the conclusion to be true at the same time. If you put in truth and do it right you will get more truth out at the end – but if you start with garbage, then all bets are off.
People often talk about the idea of there being a conflict or dichotomy between inductive and deductive styles of reasoning. I don’t think that’s the right way to think about it. Sure, if you deduct from premises that are self-evidently false (read: most premises in economic models) then you have no reason to think the conclusion will be true. But if you go about inducting will-nilly without some kind of vaguely deductive model about how the world works, then you will end up believing all kinds of crazy things. What I think we do most of the time is create models from simplified premises that may not be ‘true’, but are a starting point from which we can see whether the simplification still gets us to the right answer. But how can we know whether it gets us to the right answer?
By testing it. By making predictions and seeing if they come true.
Deductive models, when they begin with premises that are either obviously false or gross simplifications, are only ever worth anything as a preliminary to inductive testing. Furthermore, if you are beginning with premises that you think are self-evidently true, then you need to seriously question whether you are right*, or even saying anything at all**.
In Frances’ post, she describes a genuine distinction between the ways in which traditional economist and behavioural psychologists in fact go about trying to understand the world, and ends on a wistful note:
Dabbling in economic psychology or behavioural economics is a little like taking the red pill – you go down the rabbit hole, and wake up realizing that the entire world is an illusion…
I want a purple pill – a merging of the red and the blue – that would allow me to merge behavioural insights into a coherent model of economic behaviour…
But I don’t know if such a thing is even possible.
I can see there is a genuine issue with the mathematical complexity of a more realistic model of decision making. The model may become too complex to manipulate in order to generate novel results. But I don’t think this is the biggest problem. I’m pretty sanguine about the fact that straightforwardly false assumptions are knowingly put into models (it works in physics!). The economist’s biggest problem is that even with really simple assumptions, the phenomena he/she is attempting to model are often too complicated to be amenable to robust inductive testing. And this can be used as an excuse not to try***.
My belief that recessions are caused by monetary disequilibrium is basically the product of a model where you simplify the economy to assume there is only money, generic units of ‘output’ and sticky prices. It generates the prediction that reducing the value of currency in such circumstances will help achieve full employment of resources and an economic recovery. But who says this is the ‘right’ simplification to make? Well, this chart from a 1992 book by Barry Eichengreen, is a good place to start (h/t Brad DeLong)
Obviously, this graph by itself doesn’t constitute proof. Some people who would disagree with me also cite this chart in support of their view, or disagree that it provides much by the way of evidence. But if it wasn’t for the fact that the graph looked like this rather than the other way around, I’d think that maybe the money-output simplification is not a good one to make. The point being, the model generates predictions about what we should expect to happen in recessions when either the money supply is increased, or it is signalled that the central bank will do so if necessary. The model gives you an idea for where to look. It is not a substitute for looking.
People trained in economics are often very good at thinking at the margin. And at the margin, in macroeconomics I think we**** need a bit less abstract thinking and a lot more casual empiricism. It doesn’t take a genius to point out that some very popular explanations for the housing boom don’t even pass the laugh test once you take a cursory glance at the relevant data. It shouldn’t be a point of dispute, pace John Cochrane, that nominal changes obviously have real effects. People with some economics training ought to be very good at quickly correcting such things. But quite frankly, we could do a lot better – especially when it comes to correcting ourselves. Too often we don’t even bother to look; easily satisfied by a little just-so story about incentives, or an elegant but unrealistic model of a messy reality.
(File under self-admonishment.)
*I’m not going to argue now about what may or may not count as ‘self-evidently true’, but suffice to say that the set of things that are self-evidently true is a somewhat miniscule subset of the things people believe to be self-evidently true
**I’ve heard people defend the economist’s model of ‘rationality’ on the basis that you can take pretty much any action by any person and identify a set of preferences that makes that action ‘rational’. But if you say that, you haven’t created a ‘theory of action’ with any useful content, but rather created a schema by which we can model people’s actions if we can identify the terms of the schema empirically. If you have no way of telling what a person’s preferences are except insofar as they are revealed by their actions, and you stipulate the preferences are always consistent, then you don’t have a useful theory of anything. You can ‘model’ ex-post facto anything I do by attributing some set of beliefs and desires to me, but that doesn’t mean you remotely understand or predict what I will do or have done
***If you think that it is not in the business of economics to generate empirical predictions, then I have the right to ask why I should give a rat’s ass about economists’ views on policy
****I did IB economics and intro micro/macro at Oxford. It counts, OK?