Saturday, 31 October 2009

Falsifiable economics

So, let's say I claim that all swans are white.

Prove it, you insist. Well, say I, here's a white swan - here's another one - and here's another. Therefore it's quite likely that the rest of them are white too.

You probably wouldn't consider that to be much of a proof. Quite rightly, you'd insist on the scientific standard. Make clear what a counterexample would look like (a black swan) and show that a reasonably comprehensive effort to find black swans has failed. Second best might be a detailed biological analysis of swan DNA and physiology to convince you that they simply don't have the capability to produce black pigment - but even that is not very reliable. After all, by definition I am only analysing the white ones.

So why in economics do we accept the equivalent of white swans as "evidence" for macroeconomic theories?

If a Keynesian wants me to believe her theory about government borrowing resolving the paradox of thrift, she should tell me what evidence would disprove her story. For instance, if we could see two different countries, one with an substantial change in government borrowing which did not bring about the expected change in aggregate demand.

If an Austrian wants me to believe his story about credit booms, business cycles and reallocation of resources, he should similarly explain what evidence he would accept as a counterexample.

Don't worry about who's going to go out and collect the data - I'll send the Keynesian to disprove the Austrian theory and vice versa. It's called "gains from trade".

Mostly in economics we see people invent stories and then come up with reasons why their story could be true. It's time for economists to be willing to explain how they could be wrong, and let us go look for evidence that they are.

2 comments:

PunditusMaximus said...

Re: Keynesians: Don't we have huge piles of this sort of information based on cross-country analysis of various nations' flailing during the Great Depression?

Lucas said...

Interesting story from Walter Block that relates to this:

Block was doing empirical work on the connection between rent control and housing quality for his dissertation. Basic theory is pretty obvious: rent control creates a shortage of housing, so landlords don't need to compete for tenants. So, you have bad housing quality in areas that are rent controlled (controlling for other things, like income).

But, he ran some versions of his regressions and got a positive relationship between rent control and housing quality. He took this result to his dissertation advisor (who certainly claimed to be an empiricist) - who declared that Block obviously made a mistake. As Block says "What was testing what? It certainly seems that, in this case, this empiricist was using the theory to test the data!"

Just a curiosity question:

What evidence would be sufficient to show you that falsifiability is not universally appropriate standard for evaluating the truth of a proposition regarding economics?