It struck me reading DeLong and Summers that Keynesians like myself often inadvertently provoke opposition. When we discuss temporary increases in government spending, we typically assume it is debt financed. Furthermore, as the interest on higher debt has to be paid for, we generally assume taxes increase to do that. So even though our increase in government spending is temporary, both taxes and debt end up being permanently higher. At a rather basic and non-intellectual level, I think this puts many people off from the start.
Instead we could start with a temporary balanced budget increase in spending. That way neither taxes nor debt are higher in the long run. In addition, for anyone who has done their post graduate training in the last twenty years that is the natural way to start thinking about what is going on. Or if we want to avoid tax increases altogether, why not finance any increase in debt by reducing government spending rather than raising taxes? If raising debt in the long run is a problem (and I think there are good reasons why it might be), then why not use lower government spending after the stimulus not just to pay the interest on debt, but to pay off all the additional debt incurred by the stimulus. When you think about all these possibilities, the standard choice of debt finance paid for by permanently higher taxes is really the least likely to win friends.
Now you could say that Keynesians do this because this policy choice is the most effective form of stimulus. In that case, why do we nearly always choose an increase in government consumption, rather than public investment? Additional public investment will have some positive impact on the supply of output, which will raise future taxes in much the same way as the hysteresis effects that DeLong and Summers analyse.
So if I was trying to convince John Cochrane (or less ambitiously Tyler Cowen) of the efficacy of fiscal stimulus, how would I do it? I think I would use a two period model. The first period is Keynesian with interest rates stuck at the zero lower bound, and is the period in which we undertake a debt financed increase in government spending. The second period is classical, but where supply is influenced either by the additional infrastructure investment in period 1, or by hysteresis effects. All debt is paid off by the end of period 2 through lower government spending, but lower government spending does not influence output because this period is classical. (We could add a final third period which is the steady state and is uninfluenced by anything that happens in periods 1 and 2, just to show that we do not believe hysteresis or infrastructure effects last forever.)So what is there not to like about this policy? Output is higher in period 1 because demand is higher, and is higher in period 2 because supply is higher on average. There is no long run increase in debt. There is no increase in tax rates at any point. As period 2 is probably longer than period 1, we even have more time in which government spending is reduced rather than increased relative to base. However because output and therefore tax receipts are higher in both periods, the reduction in government spending required to pay off the debt might not need to be that large: this is how the DeLong and Summers argument would be translated in this set-up. The framework focuses on the essential reason why stimulus works. We shift demand into a period in which it matters - because monetary policy is ineffective at the zero lower bound in period 1 - and out of a period in which it does not - because monetary policy works in period 2.