John Cochrane cites the work of John Taylor in yet another rant about how complete crowding out, or at least near complete crowding out, renders government spending or tax policy multipliers small or non-existent.
But what did Taylor actually find? First, as he noted just the other day, he has no objection to the theoretical arguments concerning fiscal stimulus (this is from the debate that Cochrane cites to support his claims that fiscal policy cannot work):
I want to emphasize that I am not saying that permanent or long-lasting changes in fiscal policy, such as tax reforms that lower marginal tax rates, cannot help the economy, or that the automatic stabilizers are ineffective.
…I am not objecting to certain features of macroeconomic theory that are sometimes labeled Keynesian, such as that wages and prices are sticky, or that aggregate demand has a role in short run fluctuations, or even that people’s expectations matter, though I would emphasize that expectations have important rational characteristics. For me such complications—which I have been researching for many years—suggest the need for fewer discretionary policy actions and more systematic rule-like ones. …
The basic idea behind Keynesian stimulus packages is presented in basic college courses: A shift down in aggregate expenditures can be countered by increasing government purchases—augmented by possible multiplier effects—which shift up aggregate expenditures and fill the “gap.” Temporary changes in tax payments and transfers work the same way except that increased consumption is supposed to fill the gap.
Estimated macro models used for policy evaluation—whether old Keynesian or new Keynesian—have this basic mechanism built into them. However, they differ greatly in their predictions of the policy impact because of different assumptions about expectations, the marginal propensity to consume, the speed of price adjustment, and crowding out of other spending. For example, Christina Romer and Jared Bernstein used old Keynesian models to predict the effect of the stimulus package of 2009 before it was implemented. They predicted large effects of the package with multipliers around 1.5. In contrast, in research with John Cogan, Volker Wieland and Tobias Cwik, I used a new Keynesian model to predict the effects of the 2009 stimulus. We predicted a much smaller effect, with multipliers averaging 0.5, even less when you include transfer payments.
The problem with using these existing macro models to answer the question of this debate “Did fiscal stimulus help the economy?” is that they will simply repeat the same prediction story over and over again. You learn virtually nothing if you use the same models to evaluate the impact that you used to predict the impact.
So it is necessary to look at what actually happened…
Thus this is an empirical question, not a theoretical one, and Taylor draws on his own empirical work with John Cogan to argue that the stimulus was ineffective.
But let’s look at those results a bit closer. First, in testimony before Congress about his findings Taylor notes there were three categories of stimulus, and that it’s important to distinguish among them:
Three components of ARRA are…: (1) Federal government purchases of goods and services (both government consumption and government investment), (2) Federal grants to states and local governments, and (3) Temporary transfers and credits which increase the disposable personal income of individuals and families.
For the purposes of assessing the impact of ARRA on the economy it is very important to distinguish between these three categories and consider each in turn. …
What does he find for government spending?:
Federal Government Purchases Perhaps the most striking finding in the data … is that only a tiny slice of ARRA has gone to purchases of goods and services by the federal government. Of the total $862 billion in the ARRA stimulus package, the amount allocated to federal government consumption summed to only $24.2 billion in the two years 2009 and 2010. The amount allocated to infrastructure investment at the federal level was $5.6 billion in 2009-10, or only 0.6 percent of the total ARRA.
Measured as a percentage of GDP the amounts are even smaller. At the maximum level, reached in the third quarter of 2010, federal government purchases were only 0.2 percent of GDP and federal infrastructure was only 0.04 percent of GDP.
Clearly these amounts are too small to be a factor in the economic recovery. The debate over the size of the government purchases multiplier does not matter here because the multiplier has virtually nothing to multiply at the federal level. On this account ARRA has not been effective in stimulating economic growth and job creation.
Taylor is not claiming that government spending didn’t work, or can’t work theoretically. He is saying there is no way to tell.
The second category is where Taylor puts most of his analytical effort, but before turning to that, what about the third category, temporary transfers and credits? Does that work? Here Taylor says:
These statistical techniques show that the effect of the temporary stimulus payments on personal consumption expenditures is much smaller than the effect of more permanent income changes and statistically insignificant from zero.
In this case he is arguing the stimulus didn’t work, but this is an argument about how the stimulus should be designed, not an objection to stimulus per se. He thinks the tax changes should be permanent, not temporary (however, permanent changes in taxes are a poor tool to apply against temporary cyclical problems, and I find the statistical evidence to be far more cloudy than does Taylor).
He also finds that the second type of stimulus, Federal grants to states and local governments, is ineffective, but it’s important to understand why. As he notes here:
The large federal stimulus grants sent to state and local governments for infrastructure spending were mainly used to reduce borrowing and thus did not result in an increase in purchases. … The explanation is that local governments in effect acted as many American households did: when they received the stimulus money, they saved it rather than purchased goods and services. …
To better understand this explanation one can look at other countries, and in particular at China’s recent stimulus package. This week I went to China and explored the question.
Local governments in China apparently did increase infrastructure spending in 2009 following the stimulus package. Why didn’t these governments simply reduce borrowing as did U.S local governments? Professor Chong-en Bai of Tsinghua University gave me the best answer using simple economic reasoning: the local governments appeared to behave more like liquidity constrained households than permanent income households. In China, local governments do not have much access to capital markets. They get their funding mainly from the central government, including loans from the central bank, and of course only for projects that are approved by the central government. At any point in time local governments are submitting new infrastructure projects for approval; some are being rejected and some are being accepted. If the central government wants to increase infrastructure spending by the local governments all it has to do is lower the acceptance criterion, instruct the central bank to provide the funds, and the volume of projects increases. This is apparently how the stimulus worked in China.
Note that the mechanism is essentially built into the structure of the economy, with characteristics similar to an automatic stabilizer in which the criteria are raised and lowered administratively according to the state of the business cycle.
Thus, the finding here is really about the institutional design of the stimulus. He is arguing that transfers to state and local governments can work, and work well. Just look at China. If the stimulus package is designed so that state and local governments cannot use the money to finance existing projects, i.e. substitute federal money for state money rather than increasing overall spending, then it will work just as well as it worked in China.
Thus, the lesson here isn’t that stimulus won’t work, not at all. The lesson is that the stimulus package must be better designed, e.g. so that state and local governments cannot substitute for other types of spending (though I should note, again, that I don’t think the empirical evidence is anywhere near as clear as Taylor implies). I think this is an important lesson. One of the biggest problems in our response to the Great Recession, something Paul Krugman has been highlighting recently, was the decline in state and local spending and employment. Getting the design of the stimulus right could have prevented that, and that rather than the more simplistic “stimulus doesn’t work” is the lesson that we need to take away from this experience.
Update: Paul Krugman notes that Christina Romer has rebutted the Cogan and Taylor results “in devastating fashion”:
…I thought I should point out something else about Taylor’s claim (pdf) that aid to state and local governments had no effect, because they would simply have borrowed the money. Christina Romer (pdf) has already answered this, in devastating fashion:
Cogan and Taylor (2011) present a different view. They show that states had been borrowing heavily before the Recovery Act, and then borrowed less after the receipt of the state fiscal relief. From this, they conclude that the state fiscal relief in the Recovery Act had no net benefit—it just replaced state spending financed by borrowing with state spending financed by Federal aid.
Cogan and Taylor’s analysis shows the importance of specifying the counterfactual. Most states have balanced budget requirements. The requirements leave some room for deficit financing of current spending for a year or two, by running down rainy-day funds or the use of various accounting devices, especially if the deficit is the result of a downturn that was not expected when the budget was passed. But states didn’t have the option of continuing the pace of borrowing they had done in the 2008 and 2009 fiscal years. Absent the Recovery Act, states would have been forced to contract spending greatly. Therefore, relative to the plausible baseline, state spending was substantially higher following the receipt of the Recovery Act funds.
This is just one of those things where you have to ask, what is Taylor thinking? Does he really believe that states and localities could have borrowed all the money that they received from the ARRA, and thus spent at the same rate? Because if he doesn’t believe that — and he shouldn’t — his whole case falls apart.
And if states and localities can borrow freely, how do you explain the drastic fall in their spending I have been documenting?
Update: See Brad DeLong as well:
…I had thought that we had gotten John Cochrane to admit that the national income identity was not a behavioral relationship–that when the government spent money there was no necessity (or even likelihood) that the people who would eventually pay the taxes to ultimately redeem the debt issued by the government to fund its spending program immediately cut back on their spending by as much as the government increased its. But here we have Cochrane again expressing what sounds a lot like the totally-wrong British Treasury view of the 1920s, again expounding what I call Eugene Fama’s fallacy:
The Grumpy Economist: Manna from Heaven: the Harvard Stimulus Debate: Suppose the government pays contractors to… dig a ditch from Fresno to Bakersfield (high speed rail.) Is anyone surprised that GDP goes up in those areas?… When a donut shop relocates from LA, and people spend their salaries on donuts, that counts for more multiplier…. [T]o build the base or rail line, the government had to tax or borrow the money. Cross-sectional studies do not measure the loss of demand in (say) Chicago from the money that got spent in Bakersfield…. Stimulus has to be paid for. In evaluating stimulus for the whole economy, you have to count the loss of demand from the paying-for-it side equally with the raise in demand or employment from the spending-it side…. [Y]ou can’t even rely on the magic of borrowed money — you have to defend the idea that taxing Chicago to dig a ditch in Bakersfield raises output on both places by one and a half times the tax. Not impossible ([Steinsson and Nakamura] try), but not as easy as it seems…
I would say that it is very easy to get to a multiplier of 1.5 in a liquidity trap:
- Government spending goes up by $1 this year.
- Private future tax liabilities go up by $1 this year and private investment spending goes down by about 5¢ this year.
- Higher government spending raises the current price level, and reduces the burden of debt, freeing up risk-bearing capacity to finance investment and leading indebted businesses to spend more increasing their capital stock and indebted households to spend more on consumption.
- The expectation that some of the increased government purchases will ultimately be funded by monetization raises expected inflation and, as long as the economy’s interest rates are at their zero nominal lower bound, reduces real interest rates and raises the incentive to spend more now on investment and consumption goods and repay it later.
At the ZNLB–in a liquidity trap–it is very hard to see how the policy-relevant multiplier could possibly be less than 1.5.
And Cochrane’s overall rhetorical line–that “stimulus must be paid for” so the fall in spending in Chicago completely (or largely) offsets the rise in spending in Bakersfield–is a complete and false red herring: true in a rigid cash-in-advance economy with a technologically-fixed velocity of circulation, a constant stock of outside money, and an inelastic supply of inside money, but not true in our economy.
No, I have no idea what Cochrane thinks he is doing. It’s a red herring. He really ought to stop. …
This post originally appeared at Economist’s View and is posted with permission.