The only thing missing from this "major, nationwide economic experiment" is a control group (which is half the experiment, right?)

Yesterday there was a lot of blog talk about Mike Konczal's article on the "major, nationwide economic experiment" testing monetary and fiscal policy - the term that he used for the fiscal austerity and monetary expansion we've been having since last fall.

I didn't weigh in at the time for the reason laid out in the title: it's really too early to say anything much less have this expansive claim about "experiments" that makes no sense. I thought Krugman's post on Konczal's article was pretty good at laying out the argument but only suggesting that "the results aren't looking good for the monetarists". I'll take "aren't looking good" over "nationwide economic experiment" - it claims far less. And of course I suspect both Krugman and Konczal are ultimately right, but you can't claim experimental validation at this point.

We're going to need lots of retrospective studies on this many years from now when all the data are in and we can have several thoughtful years grinding through it rather than putting something together for a Washington Post deadline. These studies are probably going to take multiplier estimates, apply them to the policy variables, and determine what contributions monetary and fiscal policy likely made to the recovery (the approach of Vernon (1994) for the Great Depression) or determination of multipliers from the episode itself (the approach of Romer (1992) or Gordon and Krenn (2010) for the Great Depression), or what is effectively the same thing - estimating the output gap and determining the contributions of policies to the closure of that gap based on the sequence of their appearance (the approach of DeLong and Summers (2009) to the Great Depression).

I think we can make Krugman's suggestive statements about how we think things are falling out (if we care at all about the current unemployed we ought to make these suggestive statements), but I don't think we're ready for the full empirical exercise generating or applying multipliers or counterfactuals just yet.

Which brings me to some accusations of "begging the question". A good example is David Henderson's post on the Konczal article. He writes: "But Irwin's last sentence begs the question. It amounts to saying that the Keynesian model is right about fiscal policy's potence. Certainly Irwin believes that and has a right to believe and write it. But Konzcal can't rely on that if he wants to claim that fiscal policy is more potent than monetary policy. Konzcal is using what I call "proof by assumption."."

This is unfair to Konczal I think and not what we should expect of people in newspaper or even journal articles. You take prior empirical evidence and you use it as the best guide to the world that we have. That's what Konczal and Irwin are doing. They can't relitigate every single scientific question in a newspaper article - they have to take the science as given. It's unfair to say they're committing "proof by assumption" when they do that - what they're committing is "proof by prior evidence". The same is true, for example, of Vernon's (1994) work cited above. He takes prior multiplier estimates and uses them in an analysis of the Great Depression. For a lot of reasons I consider this more accurate than what Romer (1992) does, generating a multiplier from scratch from a fairly crude difference-in-differences estimator. Now could you come along and accuse Vernon (1994) of doing "proof by assumption". What is his only assumption? That we have decades of good empirical work on multipliers and he can rely on that literature. If you don't like the scientific literature of course you might want to tar that as "proof by assumption", but I think that's unfair.