As much as I admire the Romers' earlier work, I have to say that I am rather skeptical of their methodology in this case. They attempt to distinguish tax cuts and tax increases that are "exogenous" - which basically boils down to those motivated by philosophical reasons - from those tax changes that were simply the result of the business cycle or concerns about the deficit. They then find that "exogenous" tax increases tend to slow down the economy. By a lot.
By an implausibly huge amount, in fact. According to their estimate, the 2001 tax cut boosted GDP growth over the subsequent 2 years by about 3%. Actual GDP growth between 2001:Q2 and 2003:Q2 was about 3%, so their conclusions suggest that without the 2001 tax cut, there would have been zero GDP growth for the entire two-year period of 2001-2003.
That would have made the 2001 recession - despite an incredibly expansionary monetary policy - as severe as the 1981-82 recession, which itself was the most severe economic contraction in the US since the Great Depression. Do we really think that the 2001 recession was going to be of that historical magnitude if it hadn't been for the 2001 tax cut?
I think the reason for this highly improbable result is that Romer & Romer use a very arguable classification system for their analysis of tax changes - one that seems to have some flaws in it. For example, they classify most of the 2001 tax cut and all of the 2003 tax cut as being done for "exogenous" (i.e. "philosophical") reasons, rather than for counter-cyclical reasons. Yet in both cases, despite the fact that I'm sure those tax cuts were not originally conceived in order to jumpstart the US economy, that is how they were sold. There would have been little chance for tax cuts the magnitude of the 2001 and 2003 tax cuts passing Congress if they had not been portrayed as medicine for the sluggish economy.
Let me be more specific. According to Romer & Romer, the 2003 tax cuts had no counter-cyclical component to them. Yet the 2003 ERP (caution: large pdf file) reads as follows, on pages 54-55:
[T]he recovery in investment could be delayed by weaker-than-expected profit growth, higher required rates of return arising from geopolitical and other risks, or a prolonged period during which companies focus on repairing their balance sheets. More general risks to recovery in 2003 include an increased sense of caution, which could lead households to pull back on their spending plans, and the potential for further terrorist attacks. To insure against these near-term risks while boosting long-term growth, the President has proposed a focused set of initiatives. Specifically, the President’s plan would [cut taxes in various ways.]That sure sounds like selling the tax cut as a fiscal stimulus to me. This is just one example of the problems that I have with the Romers' classification scheme, and one of the reasons why I am skeptical of their conclusions. (See Jim Hamilton's post for more notes of caution about their results.)
...Accelerating the marginal tax rate reductions would insure against a softening of consumption by putting more money in consumers’ pockets through long-term tax cuts, which have been shown to be more effective than temporary cuts in boosting near-term spending.
Estimating the effect of changes in tax laws on economic growth is a hugely important topic, but I don't think that this is the way to go about doing it. The Romers had great success identifying and classifying changes in monetary policy by looking at the minutes of the FOMC, but the primary conclusion I take from this paper is that the same technique has major flaws when applied to fiscal policy.
Monetary policy is determined by one group of about a dozen people, with two policy objectives - inflation and economic growth - and one policy variable - interest rates. Fiscal policy, on the other hand, is determined by a complex interplay between executive and legislative branches, involves changes to hundreds of spending and tax provisions, and is the result of hundreds of individuals pursuing scores of very different policy objectives, from macroeconomic performance to deficit management to philosophy to satisfying lobbying pressure to pure self-interest. So it is not surprising to me that something that works to analyze monetary policy does not work to analyze fiscal policy.
So as interesting as I found this paper, I didn't find it at all convincing. This is simply not a good way to get at the question of how tax policy changes impact the economy, I fear.
UPDATE: I did a bit of editing to the text for clarity.