I don't think Simon Wren-Lewis will find this an annoying argument against fiscal policy. I hope not anyway.
Let's assume that New Keynesian macroeconomics is 100% correct. What role is there for fiscal policy?
Simon agrees that in normal times there is no role for fiscal policy as a tool to control Aggregate Demand. Monetary policy can handle AD all by itself. Fiscal policy has a lot of micro jobs to do: making sure the right number of bridges get built at the right time in the right places; and making sure that the right people are taxed the right amount at the right time for allocative efficiency and equity. It would be suboptimal to distract fiscal policy from its microeconomic job and leave monetary policy less than fully employed at doing its macroeconomic job.
The New Keynesian case for fiscal policy as a tool to influence AD only applies at the Zero Lower Bound, when monetary policy is supposedly incapable of doing the job by itself. Let's concentrate on that ZLB case.
Even New Keynesian macroeconomists recognise that a commitment to future monetary policy can still be effective at the ZLB. The central bank promises that in future it will leave interest rates "too low for too long". So, why can't we say that monetary policy still works, even at the ZLB? Their reply is that such a promise might not be credible. OK. Let's grant them that assumption. Promises about future monetary policy are assumed not credible.
New Keynesian macroeconomists say that increases in government spending will increase AD when the economy is at the ZLB. That isn't quite right. In fact, it's wrong. That is not what their models actually say.
Let's take this bit slowly. First consider a permanent increase in government spending in a New Keynesian model. The result is no change in the natural rate of interest. The New Keynesian IS curve does not shift vertically up when there is a permanent increase in G. For a given time-path of Y (remember, we are asking if the IS curve shifts vertically up, not horizontally right, so this is a legitimate supposition) a permanent increase of G and a permanent decrease of C by an equal amount will leave the consumption-Euler equation still satisfied at the original time-path of real interest rates. Intuitively, a permanent increase in G means an equivalent reduction in permanent disposable income and an equivalent reduction in desired consumption. (New Keynesian macroeconomists may miss seeing this if they ask whether an increase in G shifts the IS curve right, as opposed to shifting it up. But if you put permanent income on the horizontal axis, the New Keynesian IS curve is horizontal, so it makes no sense to ask whether a permanent increase in G will shift that horizontal IS curve rightwards.)
(The only effect of a permanent increase in G in a New Keynesian model comes through the supply-side. If the government takes part of national income and spends it on (say) pyramids (which is what New Keynesian models typically assume, since G does not appear in the representative agent's utility function), then the representative agent may choose to work more hours, because he is now poorer. But this supply-side effect is not what New Keynesians need if they want to prevent deflation at the ZLB. It only makes things worse.)
OK, so what about a temporary increase in G? Won't that shift the New Keynesian IS curve upwards, and cause the natural rate of interest to increase temporarily? Yes and no. If you increase G from 200 to 300 today, and promise to reduce it from 300 to 200 tomorrow, that will increase AD today. But if you leave G at 200 today, and promise to reduce it from 200 to 100 tomorrow, that will have exactly the same effect on AD today. Today's increase in G is not what causes AD to increase. It's tomorrow's decrease in G that causes today's AD to increase.
Here's the intuition. A temporary increase in G from 200 to 300 and back to 200 is identical to a permanent increase of G from 200 to 300 plus a promise to cut G by 100 tomorrow. We have already established that a permanent increase in G does nothing. Therefore the effect of a temporary increase in G is identical to the effect of a promise to cut G in future.
A decrease in future G (once the economy has escaped the ZLB) causes an equivalent increase in future C. By consumption-smoothing, that increase in future C causes an equivalent increase in current C, for a given real interest rate between today and the future.
OK. Now let's compare monetary policy to fiscal policy at the ZLB. We can either promise to loosen future monetary policy. Or we can promise to cut future government spending. Which promise is more credible?
Hmmmmm. The answer's not so obvious, is it?
Notice something weird? I have made absolutely no change in the standard New Keynesian model. The only thing I have done is to re-frame the question. Instead of talking about the effects of a temporary increase in government spending I am talking about the effects of a promise to cut future government spending.
I have changed the definition of "doing nothing". Under the original definition, we do something now, when we increase government spending, and we do nothing in the future, when government spending just falls again, all by its little self. Under my re-definition of "doing nothing", we do nothing now, and we promise to do something in the future, when we will actively cut government spending.
Acts of commission and ommission. Trolley problems. Stuff like that. That's what we are talking about.
By changing the definition of "doing nothing" I have suddenly made both fical and monetary policy at the ZLB rely on the credibility of promises of future actions. The two policies are now on a par.
Now I'm going to go further, and change the definition of "doing nothing" with monetary policy.
Take the standard New Keynesian macro model, and bolt on a bog-standard money demand function. You can even make that money demand function perfectly interest-elastic at the ZLB, if you like.
It is well-understood by New Keynesian macroeconomists that if you add a money demand function it does nothing whatsoever to the model. The two models are observationally equivalent. For every interest rate reaction function there exists a money supply reaction function, and vice versa. But it lets me change the definition of "doing nothing".
Again, a credible promise to keep future interest rates too low for too long will increase the expected future price level. That means the future nominal demand for money will be higher too. In equilibrium, money supply equals money demand, so the expected future money supply will be higher too.
A permanent increase in the money supply today is equivalent to a promise to keep future interest rates too low for too long.
Let's see how the question looks now. The New Keynesians are asking us to believe that a promise to cut future government spending would be more credible than an actual increase in the money supply today. Really?
See how I have turned the tables, just by redefining what "doing nothing" means? All of a sudden it is fiscal policy that requires credibility of a promise of future action. Monetary policy simply requires a belief that central bank will "do nothing" in future. It won't decrease the money supply back down again.
Roosevelt was at the ZLB. He didn't promise to keep interest rates too low for too long. Roosevelt simply raised the price of gold. And it worked. Because people naturally assumed he would "do nothing" in future. By "doing nothing" they understood "not lowering the price of gold back down again". It worked because people thought of "monetary policy" as "setting the price of gold".
New Keynesian macroeconomists will be tempted to insist that monetary policy really is, is, IS, IS setting interest rates. Any monetarist could insist right back that monetary policy really is, is, IS, IS setting the money supply. And a gold bug could in turn insist that monetary policy really is, is, IS, IS setting the price of gold. That argument will get us nowhere. Nor will arguing over whether fiscal policy really is setting the level of government spending or setting the change in government spending.
It's all in the framing. There is no reality in these matters. These are all social constructions of reality. We theorists shouldn't be suckered into believing that our conceptual schemes are out there in the real world. Except, the conceptual schemes of real people out there in the real world are part of the reality of that world, for a social scientist. And the Neo-Wicksellian social construction of reality, in which "monetary policy" is defined as "setting interest rates", is a damned bad reality to construct. Especially when we hit the ZLB.