So there’s a new paper [.pdf], full of regressions, confidence intervals and the like, purporting to show that the Obama stimulus package destroyed jobs on net.
Full disclosure of a busy consultant: I didn’t actually read the paper. From reading this critique by Noah Smith of “Noahpinion” (whom I am going to add to my blogroll, since I like him in the same way that El Guapo was amused by the Three Amigos), it looks like the study may indeed have been fishy.
But my point today is neither to praise nor condemn the paper. Instead, it is to point out a Klassic Krugman Kontradiction. Remember, in a Krugman Kontradiction, he doesn’t actually contradict himself, the way a lesser man might. Instead, Krugman uses true data in order to frame the debate this way or that, always in order to ridicule his opponent, even if he himself is taking the opposite side of an issue from one episode to the next.
So in Krugman’s discussion of this latest stimulus paper, he writes (and I’m reproducing it in full):
STUPID STIMULUS TRICKS
So there’s another the-stimulus-didn’t-work paper (pdf) making the rounds, and as usual being seized on by people who have no idea what the issues are with this kind of estimation.
Basically I’m with both Dean Baker and Noah Smith here, but I thought I might add some more general discussion.
What this study claims to do is estimate the effect of the stimulus by looking at cross-state comparisons. So the first thing we should understand is just how difficult it is to do that.
Remember, the stimulus was not big compared with the economic downturn. The original Romer-Bernstein estimate was that it would, at peak, reduce unemployment by about 2 percentage points relative to what it would otherwise have been. And most of that effect was supposed to come through measures that would have been common to all states: tax cuts, transfer payments, etc.. At most, differences between predicted effects among states should have come to no more than a fraction of a percentage point off the unemployment rate.
Meanwhile, there were large differences in actual unemployment changes by state. Here’s the change in the unemployment rate from 2007 to 2010:
Obviously there were factors other than the stimulus driving the great bulk of these differences. At the top are the “sand states” that had the biggest housing bubbles; at the bottom, cold places where nobody lives.
To tease any effect of the stimulus out of these interstate differences, if it’s possible at all, would require very careful and scrupulous statistical work — and we’d like to see some elaborate robustness checks before buying into any results thereby found.
The latest anti-stimulus paper shows no sign of that kind of care. It makes no effort to control for the differential effects of bubble and bust. It uses odd variables on both the left and the right side of its equations. The instruments — variables used to correct for possible two-way causation — are weak and dubious. Dean Baker suspects data-mining, with reason; the best interpretation is that the authors tried something that happened to give the results they wanted, then stopped looking.
Really, this isn’t the sort of thing worth wasting time over.
So if you go and re-read Krugman’s post, you’ll note that the one actual argument he spells out–as opposed to mere assertions of dubiousness that may be valid, for all I know–is that these guys didn’t adjust for the housing bubble’s obvious effects on unemployment among the states.
Dean Baker was particularly aghast that these Ohio State researchers could have omitted such an analysis. He went so far as to say:
It also would have been nice to see [in this paper purporting to show that the stimulus destroyed jobs] a variable for the drop in house prices by state. The economics profession as a whole was too thick to notice the $8 trillion housing bubble on the way up, or to realize that its collapse would have any impact on the economy. Now that the collapse of this bubble has led to the worst downturn since the Great Depression, one might think that economists would finally start paying attention to it….At this point, it should be economic malpractice to run state employment regressions without including a housing price variable.
OK, so it’s pretty clear that the pro-stimulus Keynesians can’t believe these moronic researchers didn’t model the obvious fact that unemployment rates across the states were affected by how bad the housing bubble was in each state. I mean, who could possibly have missed that?
In this context, Krugman’s post from December 2008 strikes me as ironic. At the time, he was ridiculing any “hangover theory” of recessions, which blamed out current malaise on the malinvestments of the housing bubble years. Krugman wrote:
So the hangover theory, which I wrote about a decade ago, is still out there.
The basic idea is that a recession, even a depression, is somehow a necessary thing, part of the process of “adapting the structure of production.” We have to get those people who were pounding nails in Nevada into other places and occupation, which is why unemployment has to be high in the housing bubble states for a while….
One striking fact, which I’ve already written about, is that the current slump is affecting some non-housing-bubble states as or more severely as the epicenters of the bubble. Here’s a convenient table from the BLS, ranking states by the rise in unemployment over the past year. Unemployment is up everywhere. And while the centers of the bubble, Florida and California, are high in the rankings, so are Georgia, Alabama, and the Carolinas.
Incidentally, Brad DeLong liked Krugman’s empirical evidence so much, that he incorporated it into one of his speeches around that time.
Now the problem with the above (note that that BLS link is dated at this point) is that Krugman was looking at the year-over-year change in unemployment across states as of December 2008, whereas the housing bust had been well underway by then. As I pointed out in my reply, if you started your comparison from when the housing bubble burst, then there was a very tight fit in the rankings of states with housing declines and unemployment increases. So the “hangover theory” made a lot of sense, if you did the timing properly.
I guess I should be grateful that Krugman apparently got the lesson. When he wanted to use the obvious connection between the housing bust and unemployment to suit his purposes–as when he ripped the Ohio guys who were attacking the stimulus–Krugman looked at the change in unemployment from 2007-2010, as you can see when Krugman introduces his chart above.
By all means, Daniel Kuehn and others, point out in the comments that strictly speaking, Krugman hasn’t contradicted himself in these two posts. But if you don’t see why he’s being slippery with his opponents, I guess we’ll have to agree to disagree.