Marcus Nunes sent me a new Christina Romer paper, which claims that fiscal stimulus is effective. I’ll argue that she has some good evidence, but also that there are weaknesses in her argument.
Her best evidence is a study that she did with David Romer, which examined two types of tax cuts; those done to boost the economy, and those done for other reasons, which can be viewed as “exogenous.”
What David and I did was to bring in information on the motivation for tax changes. For every legislative tax change, up or down, there is a huge narrative record about why it was passed. This narrative record is contained in Congressional reports, presidential speeches, the Economic Report of the President put out by the Council of Economic Advisers each year, and other documents.
We read all of those documents and classified tax changes into those taken in response to other factors affecting output and those taken for more independent reasons. We identified a number of tax cuts taken because the economy was slipping into a recession. We also found a number of tax increases taken because government spending was rising; for example, policymakers raised taxes dramatically during the Korean War. This is important because spending increases will tend to increase output, while tax increases will tend to reduce it. So in cases where the tax increase is caused by the spending increase, there are systematically factors going in opposite directions.
At the same time, we also found a number of tax changes taken not in response to current or forecasted economic conditions, but for more ideological or long-term reasons. For example, Ronald Reagan cut taxes in the early 1980s because he believed lower tax rates were good for long-term growth. Bill Clinton raised taxes in 1993 because he thought dealing with the deficit would be good for the long-term health of the economy.
We argued that to estimate the impact of tax changes, we should look at the behavior of output following these tax changes made for more ideological reasons. In other words, we dealt with some of the omitted variable bias problem by excluding from the empirical analysis the tax changes taken in response to economic conditions.
They found that the endogenous tax changes had a modest (but positive) effect on output, while the exogenous changes had a large impact on output.
I favor a pragmatic approach to research, so I applaud the Romers for using the narrative approach. However what’s being tested here isn’t really “stimulus,” it’s tax cuts. The traditional Keynesian model says fiscal stimulus will boost NGDP, and will also boost RGDP if there is slack in the economy. Otherwise you get higher prices. So there are actually two interesting questions worth testing; does Keynesian stimulus boost NGDP (i.e. spending) and does the higher spending lead to more real output.
To make things even more complicated, supply-siders have suggested an alternative channel through which tax cuts might boost RGDP; increasing the incentive to work, save and invest. The basic supply-side model doesn’t predict claim any impact of tax cuts on NGDP, indeed it was sold in the late 1970s as a tool for boosting output without boosting inflation. Even so, if the central bank is targeting interest rates, then supply-side effects could easily lead to more NGDP as well.
So let’s accept Romer’s argument that tax cuts boost RGDP. Is it a problem that we don’t know exactly how or why? It could be, because if it is due to supply-side effects then lump sum tax rebates and government spending increases wouldn’t necessarily work.
Romer discusses one such event that occurred in the spring of 2008, when the Bush administration issued tax rebates to boost consumption. John Taylor later pointed out that consumption did not seem to respond to these tax rebates, despite the temporary spike in disposable income. Romer replied:
The trouble with this analysis is, Professor Taylor wasn’t thinking about what else was going on at the time. Democrats and Republicans didn’t come together to pass the tax rebate for no reason. This was the heart of the subprime mortgage crisis. House prices were tumbling. Mortgage lenders like Countrywide Financial were in deep trouble.
Economists were worried that consumption was about to plummet. For most families, their home is their main asset. When house prices fall, people are poorer, and so tend to cut back on their spending.
Against that background, the fact that consumption held steady around the time of the tax rebate may in fact be a sign of just how well it was working. It kept consumption up for a while, despite the strong downdraft of falling house prices.
Romer’s right that without the rebates aggregate spending and output might have been somewhat lower during mid-2008. But Romer doesn’t consider whether that might have led the Fed to move much more aggressively in September 2008. As it is the Fed met two days after Lehman failed and did nothing (which effectively tightened policy sharply as the Wicksellian natural rate was plunging rapidly during this period.)
The Fed cited an equal risk of recession and inflation (i.e. economic overheating) when it left interest rates unchanged at 2% in September. It seems highly unlikely that the Fed would have been so passive if Romer’s counterfactual had come to pass. If so, then John Taylor might be right, but for the wrong reason. The real problem is that the Fed sabotaged Bush’s tax rebate. The real problem is that in new Keynesian models the fiscal multiplier is precisely zero if the central bank targets either inflation or NGDP.
This is the biggest weakness in Romer’s paper. When discussing Taylor’s critique she rightfully talks about the omitted variable problem. But then she basically ignores the problem of monetary policy counterfactuals when considering what would have happened without the Obama stimulus. The basic problem here is that she seems to hold three contradictory views:
1. She agrees with Bernanke that the Fed is not out of ammunition.
2. Elsewhere she praises Bernanke for acting aggressively in 2008-09, making the recession less severe.
3. Her three million “jobs created or saved” estimate for the Obama stimulus implicitly assumes that if Obama’s stimulus had not passed, then the Fed would have responded to the deeper downturn with almost criminal negligence.
Now I’m perfectly willing to concede that it’s unlikely a counterfactual monetary policy would have exactly offset any fiscal stimulus reductions. But I would also insist that monetary policy counterfactuals must be addressed in any multiplier estimates. And I see one Keynesian study after another completely ignoring this problem.
She also cites the study by Nakamura and Steinsson that looked at the cross-sectional effects of defense spending in various regions. But as I’ve pointed out many times, these multiplier estimates are completely consistent with the national multiplier being zero. In other words, models that assume no aggregate multiplier effect (such as the monetary policy offset model) would nonetheless predict that regional defense expenditures would impact regional GDP.
She also cites a study of how individual reactions to rebate checks depend on the date they were received (Jonathan Parker, et al.) This approach is somewhat better, but still fails to fully address the monetary offset problem. Stimulus might boost NGDP in Q2, and the Fed might take it all back in Q3.
On a more positive note let me acknowledge that the Romers’ tax cut study is very important. It suggests that we do know, at a minimum, that cuts in marginal tax rates boost RGDP. I’m willing to support that sort of fiscal stimulus. But at this point I’d have to say that’s all we know. In a world where central banks are targeting inflation (and by implication AD), then all multiplier estimates for demand-side stimulus will be highly uncertain, little more that estimates of monetary policy incompetence in the specific period being examined.