There appears to be an error in N. Gregory Mankiw's "Economic View" column of January 11, 2009. The error is the association of Christina Romer with the proposition that the tax multiplier--the effect on GDP of a tax cut--is twice the spending multiplier. The Romers' article does not distinguish between the two, referring only to the "substantial multiplier... due to the procyclical behavior of investment" (p. 37 of the working paper version, at http://tinyurl.com/dl20090111). David Romer in conversation two years ago characterized the paper to me as "hyper-Keynesian... suggesting very large multipliers..." The Romers believe in a tax multiplier no larger than the spending multiplier, and they certainly do not believe that a balanced-budget equivalent reduction in taxes and spending provide any Keynesian stimulus at all.
Mankiw's comparison of the 1.4 estimated spending multiplier from Valerie Ramey's study with the 3.0 estimated tax multiplier from the Romers' study is inappropriate. The two studies use very different methodologies. They are not comparable. For example, the Ramey study on the effects of government spending--while a superb contribution to the literature, and one that I have assigned to my graduate students--does not fully control for the tax increases that often accompany spending increases. Thus it is very likely to understate the effects of spending increases alone: her study assesses the impact of the Korean-War military spending increase without taking account of the fact that it was accompanied by a large tax increase.
What Romer and Romer's study (and their earlier work on monetary policy) shows is not that tax cuts are uniquely effective, but rather that failing to consider the reasons for policy changes leads to underestimates of the effects of all types of stimulus. Because these issues of omitted variable bias are likely to be as strong for spending as for tax changes, the most reasonable interpretation of their paper is that all types of fiscal stimulus are more potent than conventional estimates would lead us to believe.
It is somewhat puzzling that Mankiw appears to believe that the Romers do think that tax multipliers are larger than spending multipliers, as they do not, and this is something that he could have very easily checked.