Paul Krugman and I occasionally disagree on the fine points of some macro issue, but almost never do I see him making a basic logical error. Until today:
Dean Baker points us to Feyrer and Sacerdote, who use cross-state variation in stimulus spending per capita to estimate the employment effects of the stimulus. They find a clear positive effect: states that got more money per person did better on jobs. And as Baker points out, the national effects must have been larger, since some money spent in New Jersey presumably creates jobs in New York and vice versa.
One thing I might point out, by the way, is that this is something of a “Well, duh” result. Of course more federal spending in a given state or county creates more jobs. And the burden of proof should always have been on stimulus critics to explain why this doesn’t mean that stimulus spending creates jobs at the national level too. In normal times you can argue that the positive job effect of higher spending is washed out by higher interest rates — that fiscal expansion will be offset by contraction on the part of the Fed. But with interest rates up against the zero lower bound, that argument doesn’t apply.
The argument about Fed policy in the second paragraph is flat out wrong, as I’ve pointed out 100s of times. But even if it is right, this study tells us nothing about the macro effects of fiscal stimulus. It’s a near perfect example of the fallacy of composition. Every single anti-stimulus model would predict exactly the same finding at the micro level. If the federal government builds a billion dollar military base in Fargo, North Dakota, I think all economists agree that the number of jobs increases—in Fargo, North Dakota. Does the number of jobs increase at the national level? Very possibly yes, but nothing in the Feyrer and Sacerdote study addresses that question. Krugman, Dean Baker, Brad DeLong, etc, are very smart guys, I don’t know why they keep hyping this study.
Even worse, Krugman suggests that their instrumental variable analysis (which used state population) tells us that stimulus mattered. Small states got more aid, and small states did better job-wise. My preceding argument suggests that study is also irrelevant, but there’s another problem here. Smaller states may have a different industry mix (say more commodity-oriented) that allowed them to better ride out the recession. So it’s doubly irrelevant.
By the way, just a few weeks ago I criticized a Brad DeLong post on similar grounds. If Paul Krugman would have read that post he would have avoided falling into the fallacy of composition.
Micro studies can’t tell us whether fiscal stimulus works. Micro studies can’t tell us whether monetary stimulus works. Micro studies can’t tell us whether RBC or sticky price models are better. Micro studies can’t tell us anything about macro. That’s why macro is a different field.