Back when the stimulus was being considered, there was a lot of talk about what the multiplier of stimulus spending is. Basically, the multiplier is the effect of government spending permutating through the economy: I buy a wrench for my fighter jet from you, you hire workers, they feel richer and start shopping big screen televisions, and so forth . . .
If the multiplier is greater than one, stimulus spending has a big effect; you cause the economy to grow by more than the amount of the government spending. Theoretically, if the multiplier is large enough, and the growth occurs in the right places, you can get enough tax revenue to pay for the stimulus spending. Sadly, empirical estimates are much smaller than the 3x or 4x multiplier you would need for this to be true.
If the multiplier is one, the stimulus basically raises GDP by exactly the amount of government spending.
If the multiplier is less than one, the stimulus raises GDP by less than the amount of government spending. Since you have to pay back the full amount later, and the stimulative effects are only temporary, this is probably not a good deal. Not that this ever stops the government from doing other sorts of spending--F22, I'm looking at you . . .
Obviously, the size of the multiplier is a matter of hot political debate, with the estimates roughly breaking down along ideological lines: conservatives think it is small, liberals believe it is larger. These beliefs are obviously colored by your other opinions on the general wisdom of having the government take taxpayer dollars and spend it on stuff.
In the Wall Street Journal, Robert Barro and Charles Redlick discuss the findings of their new paper, which suggests that the multiplier for defense spending is about .6 or .7 at trend unemployment, rising towards one around an unemployment rate of 12%. They think that the multiplier for non-defense spending is smaller--though this is colored by the fact that it's very difficult to extract the stimulative effects of non-defense spending. As they note in the article, defense spending tends to happen mostly independent of the economy, which makes it relatively easy to see the effect. Non-defense stimulus tends to occur when the economy is already tanking.
Barro and Redlick also suggest that tax cuts are preferable to spending, a finding broadly consistent with the Romer and Romer paper that found substantial multipliers for tax cuts/increases. However, they note that this effect is harder to pin down over long time horizons.
I've discussed the underlying paper with Barro, and it seems pretty compelling; they've got a hell of a time series. On the other hand, I know that this work fits both his and my political convictions, so there's a good chance we're both missing something. No doubt liberals will jump on the paper with both feet, and we'll get to here about what that missing something might be.