CRGreathouse said:
The CBO table suggests a multiplier of 0.1 to 2.5 for the various aspects of the ARRA (0.2 to 2.5 if the upper-end tax cut is excluded). But Romer & Romer measure a multiplier of 3.0 for tax cuts. So taxing money out of the system to stimulate it doesn't seem workable: a tax of T reduces the GDP by 3T, then increases it by kT where k is in [0.1, 2.5]. So at best, taxing T decreases the economy by 0.5T.
Let me rephrase, using general terms, what I gather of how your argument works, and you can let me know if I have it wrong.
The CBO has estimated multipliers (x_1,x_2,...,x_n} respectively for govt actions {g_1,g_2,..., g_n} using some methodology M1. Separately R&R estimate multiplier x'_1 for g'_1 using methodology M2.
Your argument is that:
(A) x'_1 is a better estimate than x_1 and g'_1 is essentially identical to or more appropriate than g_1,
(B) Instead of comparing x_1 with {x_2, ..., x_n}, it is more sensible to compare x'_1 with {x_2,...,x_n}.
If that is indeed what you are saying, then the following are my objections:
1. The methodologies are different!
R & R admit that their estimates for multipliers are larger than those derived from conventional estimates. They argue that their estimate is better in that it eliminates some systematic errors that exist in the conventional estimates. But they specifically caution against doing exactly what it seems you've done.
Quoting the 2010 AER paper, Conclusions, pg 799 (4th new paragraph):
" Similarly, our results do not speak to the issue of whether taxes are a more powerful tool of fiscal policy than government purchases. The fact that our estimates of the effects of tax changes are larger than conventional estimates of the effects of changes in purchases is of little relevance: conventional estimates of the effects of purchases, like conventional estimates of the effects of taxes, almost surely suffer from omitted variable bias."
2. g'_1 (from R&R) and g_1 (from CBO) are significantly different, in several ways:
Since the CBO action g_1 is a one year tax cut for higher income groups (i.e., a pulse of width one year), it does not accurately apply to the pertinent question: "how is there any stimulative effect to letting the Bush cuts lapse on the higher income groups?" (i.e., what is the short term effect on GDP of a step function-like tax increase on the high incomers). My argument assumes that (i) for small changes in tax rates, the response can be linearized, and hence, flipping the sign does not cause a huge error, (ii) we care about short term effects, which I shall define, for convenience, as the immediate year following the action (see also the introductory section of the CBO paper), and I therefore neglect effects of changes to tax policy beyond 1 year as well as any second order anticipatory effects resulting from the knowledge of cuts beyond 1 year. I think these are not completely terrible approximations for the situation.
On the other hand, g'_1 is a "generic" tax hike on all income groups. And while it has the right temporal profile, I believe it introduces a rather significant error in that it is not a high-income tax change, which, according to the CBO report, has a multiplier that is about 4 times smaller than one on the lower income groups. One might argue that (modulo methodology) these are essentially the same, if the contribution of lower income households to GDP were negligible compared to that of the higher income group. I do not know how true (or untrue) that is.
3. The multipliers calculated by both papers are time dependent (i.e., they vary with time measured from the implementation of the action). The time dependences for different multipliers have different forms. For instance, the CBO report discusses multipliers for purchasing of goods and services:
"For example, a one-time increase in federal purchases of goods and services of $1.00 in the second quarter of this year would raise GDP by $1.00 to $2.50 in total over several quarters, with most of that effect in the first two quarters and little effect beyond a year."
So the effect of government purchases is mostly short term, their multiplier of 1-to-2.5 being realized over a couple of quarters. On the other hand the 3X multiplier calculated by R&R is much slower coming into effect, and takes as long as 10 quarters to mature to that value. In the first few quarters (what I consider the short run), the multiplier stays below 1.
That's all for now.