Friday, September 03, 2010

David Friedman On Macro

I am teaching an introductory Macro class this semester, and it always makes me feel like a snake-oil salesman. David Friedman has a great new blog entry on Macro. I have two additional thoughts on the subject.

1) I have another objection to Macro: we cannot see the counterfactuals; there are no good natural experiments to test hypotheses. When some states do one thing, and others do a different thing, we end up with lots of different observations that can be used to test a hypothesis. We can see how two similar states that do two different things differ from each other, and this tells us something about the effects of policy. With the U.S., however, there are no other data points--different countries are too dissimilar to make direct comparisons. As a result we cannot know what would have happened if macro policy had been different. What would have happened without TARP, or the stimulus bill? We don't know. There's no way to know. Combine this with the poor predictive power of macro theory in its current state and you get something that is of dubious scientific value. Aside from "Inflation is always and everywhere a monetary phenomenon," there's not a lot of Macro that I would be willing to stand behind firmly.

2) Friedman suggests that insufficient attention has been paid to regime uncertainty--the reluctance of individuals and firms to make costly economic decisions when policy is uncertain. Bob Higgs has been writing about this topic for some time. I blogged about it once. Here's Jerry Jordan, former Fed bank president, making the same point. Here's Don Boudreaux reading James Madison on the subject. Here's Russ Roberts (after an interview with Higgs) on the subject, and again here. Here's Scott Sumnerdismissing the idea. Finally, here's Tyler Cowen on regime uncertainty (he calls it "policy uncertainty"), and not atypically, he takes a muddy position. My point here is that I'm not sure if insufficient attention has been paid to this topic. Lots of economists are aware of it. How much attention should the topic receive? I don't know. Given that it's so very difficult to test macro hypotheses, would paying attention to the topic of regime uncertainty make any difference at all?

1 comment:

The Motsinger said...

Per the first point, Jim Manzi has a similar take: http://www.city-journal.org/2010/20_3_social-science.html

I am sympathetic, but then I really wonder what econometrics is for. Angrist & Pischke have a recent article in JEP arguing that, contra Ed Leamer, the 'con' has been taken out of econometrics. They argue, as far as I can tell, that instead of simply building a model with (quasi)reasonable assumptions, specify what would go into an experiment and then look for a historical instance where it looks like that single independent variable changed and nothing else.

Still seems like a con to me, though, since an actual experiment presumably breaks the causal chain when manipulating the independent variable, and a 'natural experiment' can do no such thing.