Saturday, June 16, 2012

Two out of three macro tests ain't bad

Krugman says Keynesianism passes three tests:

"The policy response to financial crisis has, in effect, given us a great natural experiment in macroeconomics — an experiment that can and should be viewed as a test of two views of the economy. One view — which includes both freshwater macro and much of what Austrians say — is in effect classical macro as Keynes described it, in which the economy is always constrained by supply. The other is a more or less Keynesian view in which a depressed economy is constrained by demand, not supply.

These two views had strong implications on three fronts. One was interest rates: would large budget deficits drive rates up, as a classical view implied, or would they do no such thing under depression conditions? A second was the effects of austerity (which has been much larger than the weak efforts at stimulus, and therefore provides the real test); would austerity policies release resources to the private sector, as per the classical view, or lead to economic contraction? Finally, a third implication involved inflation: would large increases in the monetary base produce soaring inflation, again as classicists of all kinds claimed, or do no such thing under depression conditions?"

Brad DeLong agrees and Gene Callahan (who apparently doesn't think he's unpopular enough already in Auburn, Alabama) also agrees.

I think I two-thirds agree on the tests and disagree strongly on one other point. There is a very good case for aggregate demand side issues (and especially liquidity preference issues) given the situation with crowding out and with inflation.

I don't think the interest rate test is a good test of Keynesianism. It's actually worse than that. What's been going on with interest rates is simply a test of whether you understand economics or not. The thing is, zero lower bounds and liquidity traps really aren't a "Keynesian" thing at all. Or at least you don't have to be a Keynesian to understand that bonds and cash becom interchangable (or much closer to interchangable) at the zero lower bound. You also don't have to be a Keynesian to understand that the zero lower bound acts like a price floor, and so moving the supply and demand for loanable funds around is not going to change the price of loanable funds so long as the equilibrium interest rate remains below that price floor (think about it - does changing labor supply or demand change the price of labor for workers with reservation wages below and marginal productivities above the minimum wage?).

Now, there can be disagreement over when we'll get out of this liquidity trap. But the point is, the only thing you need to be able to say "we won't see increasing interest rates until well after we see aggregate demand pick up" is a good sense of economics, not Keynesianism per se. It's obvious why Keynesians would be more focused on this than other people, I guess. But the fact that Keynesians are often the only ones saying this to me says more about gut reactions by non-Keynesians to the phrase "liquidity trap" than anything else. A price floor is a price floor - this is micro 101, not Keynesianism.


The other thing I disagree with here is that this is a "natural experiment". No way no how. You start telling people that they think the logic of random assignment applies. And when you think the logic of random assignement applies you do dumb things like scatterplots with GDP growth against government spending. And that's wrong. This is not a "natural experiment" as the word "natural experiment" is traditionally understood. The downturn has been big enough that I think you can safely confirm a few stylized facts about liquidity traps and aggregate demand depressions, but I don't think you can do more than that. It is certainly going to be as tricky to pull a fiscal multiplier estimate out of this mess as it has been for every other recession.

Don't go to a trade theorist to explain what a natural experiment is (even if he has a Nobel). You're better off asking a labor empiricist instead.


  1. I might have left too much between the lines for readers on that last point... might be tempted, if they thought this was a "natural experiment" as it is traditionally defined, to think that we could just compare the impact of different spending levels on GDP growth to get a multiplier. In other words, one might be tempted by claims about "natural experiments" to read the slope of that scatterplot as a multiplier. But that's wrong and Krugman is wrong - this is not a natural experiment and that slope would be a biased estimate of the multiplier.

  2. Daniel Kuehn: Speaking of multiplier estimates, have you yourself conducted any research papers on multiplier estimates? Are any of them ready to be published, or will they take a while?


All anonymous comments will be deleted. Consistent pseudonyms are fine.