Ryan Avant at the Economist has a good discussion of why quantitative easing has an impact. It is all about expectations. It isn't about how many and what type of securities the Fed purchases--how big its balance sheet will be. It is rather hints about the outcomes for the economy that the Fed wants to acheive. Market Monetarists, and Avant should be counted as one, want clear communication--a target for the growth path of nominal GDP. Where does the Fed want spending on output to be, which is, necessarily, how much money income does the Fed want people to earn? (Unfortunately, Avant, as does Sumner too often, plays to those tied to the status quo, talking about a commitment to full employment and higher inflation, with a mere passing reference to nominal income.)
Nick Rowe has another in his series of posts about why interest rates are not fundamental to monetary policy. Rowe has been arguing that New Keynesian models assume full employment in the long run, without there being any mechanism to get there. In their models, the central bank sets an interest rate, which determines spending now relative to later. In the model, real spending later is equal to potential output--the full employment level of output. A lower interest rate now then results in more spending now. However, if spending later is not tied down, then a lower interest rate now could just mean less spending later. That a lower interest rate now means more spending now is based upon an assumption rather than any implication of the model.
While all of this makes sense to me, it seemed to me that it was just about imperfections in New Keynesian modeling strategy. In the real world, central banks create and destroy money. Why can't the implications of the quantity of money the central bank creates be included as an assumption in the model, even if that part of the economy is not part of the model? Sure, calling it "general equilibrium," is a misnomer, but can't partial analysis provide insight?
In this post, Rowe connects his critique to Leijonhufvud's theory of the corridor. In Rowe's view, as long as most people believe that the economy will bounce back soon, then an interest rate policy will work well enough. But if people do not believe that the economy will bounce back soon, then a policy based on interest rates can fail. Leijonhufvud's theory was that the macroeconomy is reasonably stable within a corridor of small fluctuations, tending to revert to full employment. On the other hand, if the economy is somehow pushed well outside the corridor, it won't return.
Rowe's argument suggests that the problem Leijonhufvud described as the corridor is with the traditional focus of central banks--interest rates. The New Keynesian insistence that adjustmetns in short term interest rates, and promises about where their time path in the future, are all that is necessary to provide for macroeconomic stability, is mistaken. It follows from this flaw in their model. Instead, Rowe suggests that central banks need to emphasize the quantity of money. It is the quantity of money, and not interest rates, that determines the nominal economy in a monetary economy. When people sell goods for money and buy goods with money, the quantity of money matters.
I think Rowe is correct, however, I would go one step further. It is the nominal anchor that ties down the nominal economy. A monetary regime that fixes the nominal quantity of money has a nominal anchor, but a poor one. Of course, Rowe doesn't advocate a fixed quantity of money. My point is that to provide for macroeconomic stability, a monetary regime must tie down the level of some nominal magnitude. The price level or spending on output are much better than the quantity of money.
Do these nominal magnitudes require some commitment about the quantity of money? Not really. Changes in the quantity of money that are expected to be temporary have little effect on other nominal magnitudes. Changes in the quantity of money that are perceived as inconsistent with the central bank's goals for other nominal magnitudes will be seen as temporary and so have little effect.
And this brings us back to Avant's point. How many bonds or what type of bonds the central bank purchases is not important. The quantity of money is not important. What is important is the nominal magnitude the central bank is aiming to achieve.