I've been having some interesting email discussions with Scott Sumner, and thus it is time for a macro post.
Scott Sumner and David Beckworth have recently been arguing that, once you take the Fed into account, the Keynesian multiplier (the effect of fiscal spending on GDP) is zero. This is true, they say, because the Fed acts to counteract any unusually rapid rise in nominal GDP. The idea is basically this: after a recession ( a fall in nominal GDP), in order to return to its previous trend, NGDP must grow faster than its trend rate of growth. But since NGDP = real GDP + inflation, and since a stimulus increases aggregate demand, this will involve higher inflation as well as faster growth. The Fed, Sumner and Beckworth hypothesize, will not allow higher inflation, and so will raise interest rates, thus canceling out the effect of the stimulus on aggregate demand. The multiplier, when measured in NGDP terms, and when defined to take the Fed's "reaction function" into account, will be zero.
Now, in the past, I've argued that the Fed would have to be crazy to do this. It would have to refuse to allow NGDP to grow fast in a recovery even after NGDP plummets in a recession. But maybe the Fed is crazy! If the Fed actually does have an emotional bias against inflation - which some people argue is a desirable quality in Fed governors, because it helps anchor low inflation expectations - then it just might do something very much like what Sumner and Beckworth imagine.
However, does this mean that fiscal stimulus is ineffective? I say no. Why? Because that would only be true if fiscal and monetary policy had identical and opposite effects on inflation expectations.
Realize that when we talk about the fiscal "multiplier," we're taking about a real multiplier - the effect of spending on real gdp growth. The growth rate of NGDP is equal to the real growth rate plus the rate of inflation. So it's possible for the real growth rate to rise in response to a stimulus while the nominal growth rate stays the same. This will happen if inflation falls. For example, you could go from having a 2% RGDP growth rate with 3% inflation before the recession, to a 4% RGDP growth rate with 1% inflation after the stimulus. NGDP growth is 5% before the recession and 5% during the recovery, but RGDP growth is different.
This is called a disinflationary boom. In the case of a stimulus-induced disinflationary boom, the real multiplier is nonzero even after taking the Fed's inefficiently hawkish reaction function into account.
How can a disinflationary boom happen? Well, inflation is determined in large part by expectations of future inflation. If the combination of fiscal stimulus and higher interest rates caused inflation expectations to be lowered (perhaps by giving the Fed an opportunity to prove its hawkishness), the Phillips Curve would shift downward, meaning you could get more real growth for any given rate of inflation. That would allow a disinflationary boom. The stimulus would push RGDP back to trend after a recession, while a permanent change would remain in the NGDP time series. In fact, if the episode increased the Fed's credibility in the long term, the combination of policies would have benefits beyond the effect of the stimulus.
Some people might argue that this describes the 1980s. Reagan's tax cuts, this story would say, acted as a Keynesian demand-boosting stimulus that raised RGDP growth, but that the accompanying inflation was tamed by the hawkish Volcker Fed. I'm not sure I believe that, but it at least sounds plausible.
Anyway, the upshot is that a nominal multiplier of zero, caused by a pathologically hawkish Fed, does not necessarily mean that the real multiplier - the multiplier we care about - must be zero.