John Taylor has what I consider to be a pretty good post about the stimulus. He points out that if we evaluate the stimulus' impact using the same models that we used to predict its impact, we don't gain a lot of added value from the backward-looking part of the analysis. That is very true. He then points out the extreme difficulty of evaluating the stimulus (or other policy interventions) without basically assuming your conclusion through the choice of the model used to construct the counterfactual. This is also true.
He then presents a paper that he has written that attempts to evaluate the ARRA. According to Taylor, he has clear evidence that the stimulus didn't work.
This, of course, sent alarm bells ringing through my head. According to my Efficient Marketplace-of-Ideas Hypothesis, if there were an easy way to show convincingly that the stimulus failed, somebody would already have shown it. Previous studies purporting to show the failure of stimulus have been less than convincing. So I went into this paper expecting to find serious problems.
But instead I found Taylor pretty much agreeing with the Keynesians...
Taylor evaluates the ARRA by looking at changes in specific components of output - federal government purchases, state government purchases, and private consumption. According to Keynesian theory, if stimulus works it should work by raising either G (in the case of a rise in government purchases) or C (in the case of tax refunds or increased transfers). So Taylor just looks to see the degree to which this actually happened.
First, he shows that personal consumption doesn't appear to have been much affected by the tax rebates and transfer payments that made up a big chunk of the stimulus:
This graph looks convincing; you see bumps in stimulus income, but no bumps in consumption. Of course, Taylor is assuming a particular lag structure for his model; it's quite possible that the rise in spending due to tax rebates and transfers didn't occur immediately, but were spread out over the next few quarters, so that you wouldn't expect to see contemporaneous bumps. Similarly, when Taylor generates counterfactuals to show that consumption didn't rise, he picks a specific lag structure. Maybe he tried alternate specifications and didn't report them because they showed the same thing? Or maybe not.
Anyway, the arbitrary lag specification means that this model is not in any way conclusive. But it is still suggestive. And what it suggests is that tax rebates and transfer payments don't make for particularly good stimulus, because in a balance-sheet recession people will just use the money to pay down their debts. Helping people pay down debts may be good in its own right; but this is a reason why Keynesians often argue that government spending is a better approach to stimulus than tax rebates.
On that note, Taylor notes that federal expenditures didn't rise by very much due to the ARRA:
The most striking finding in Figure 3 is that only a small part of ARRA went to purchases of goods and services by the federal government. Measured as a percentage of GDP the amounts were immaterial: At the maximum effect, which occurred in the third quarter of 2010, federal government purchases due to ARRA reached only 0.21 percent of GDP and federal infrastructure only 0.05 percent of GDP.
These amounts are too small for the stimulus package to have had a significant effect on the overall economy. In this case the debate over the size of the government purchases multiplier is largely moot because the government purchases multiplier had virtually nothing to multiply at the federal level.
This precisely echoes the complaints that Keynesians had about the ARRA: not enough federal government purchases, not enough infrastructure spending.
Finally, Taylor looks at state-level spending, and concludes that the ARRA grants mainly replaced state borrowing with federal borrowing:
Figure 6 shows that in the absence of the 2009 stimulus grants, net borrowing by state and local governments would have been greater than it was with the grants. This is consistent with the view that state and local governments tried to smooth their expenditures in the face of temporary changes in income, much as households without borrowing constraints did...State and local governments used the stimulus grants to reduce their net borrowing (largely by acquiring more financial assets) rather than to increase expenditures[.]
This analysis is subject to the same caveat about lag structure as the earlier consumption analysis. But it also seems basically plausible to me, and highly suggestive that states mainly pocketed their ARRA grants instead of spending them on building roads and such.
Finally, Taylor presents a hypothesis that ARRA grants actually made states shift their spending from purchases to transfer payments, which (if this indeed happened) would have actually reduced GDP.
So to sum up, Taylor's paper says that ARRA didn't increase GDP because it didn't increase government purchases, and even hurt GDP to the extent that it reduced government purchases. This sounds exactly like a Keynesian critique of the stimulus bill - government purchases are good medicine for recessions, and we didn't do nearly enough of them! But somehow Taylor interprets his pro-Keynesian assumptions as an anti-Keynesian conclusion:
More generally, the results from the 2000s experience raise considerable doubts about the efficacy of temporary discretionary countercyclical fiscal policy in practice. In this regard the experience with the stimulus packages of the 2000s adds more weight to the position reached more than 30 years ago by Lucas and Sargent (1978) and Gramlich (1978, 1979).
What? But Lucas, Sargent, etc. thought that government purchases wouldn't raise GDP. That runs counter to Taylor's entire critique, which is that ARRA didn't raise government purchases enough!
Readers may go through this paper with a finer-toothed comb than I used, and may find significant methodological flaws that I overlooked. Please be my guest. But that won't change the larger point that Taylor is being a lot more Keynesian than he seems to think he's being.
So this makes two prominent conservative economists who have criticized Democrats for not being Keynesian enough (the other being Martin Feldstein). Is it possible that what these guys don't like is really just...well...Democrats?
Update: Mark Thoma points out that this is not the first time John Taylor has criticized the ARRA for not including enough government purchases.
Update 2: Paul Krugman concurs. John Taylor responds:
Now, I know that Krugman...would like a stimulus package with higher proportions going to...government purchases...But experiences from the 1970s raise serious doubts about the political and operational feasibility of such discretionary fiscal policy. So do recent experiences in many other countries...In a simple Keynesian model, all the government has to do to combat a recession is quickly increase government purchases, but the difficulty with doing so in practice is one of the classic arguments against discretionary fiscal policy.So Taylor is arguing that the type of Keynesian policy that Keynes himself suggested, and which modern-day Keynesians want, is simply politically infeasible. Fine. I myself am quite receptive to that argument, actually. But then there's the China example, which Taylor himself cites as an example of a rapid increase in government purchases in response to a recession. How does Taylor explain that? Is it autocracy, or maybe some kind of central planning, that is supposed to make real stimulus politically feasible in China but not here? And then there's another question: even if (real) stimulus turns out not to be politically feasible, does that mean that economists shouldn't recommend it, if it would be the best policy? I mean, the fact that Medicare privatization is politically infeasible hasn't stopped John Taylor from urging us to do it.