Money is quite the slippery liquid asset.
How does money work, and how does this relate to NGDP targeting? Jim Hamilton recently criticized NGDP targeting as a regime without concrete foundations. What's special about the asset purchases that would, in and of themselves, lead to an increase in NGDP expectations? Brad DeLong got back on the issue, and commented that the key reason for why the Fed's purchases can have an effect is that the Fed can purchase assets that will not be as liquid outside of the zero lower bound. In effect, the key transmission mechanism is the provision of liquidity for non-liquid assets, which then commits the central bank to a higher money supply in the future.
Liquidity also played a key role in a recent paper by Kiyotaki and Moore on how monetary policy affects an economy. In their model, agents in an economy hold two assets: money and equity. In each period, one can expect to be able to dump one's entire stock of money. However, this does not hold true for equity. Money is money because it is liquid. Is currency money? Yes, because I can spend it. Are demand deposits money? Yes, because I can withdraw it and go spend it. Are stocks money? Yes, as long as the market is stable and I can sell them.
This framework leads to a simple conclusion: what qualifies as money is regime dependent. M1, M2, M3...M9001...M∞ have no real meaning without knowledge of the monetary regime. Are money market funds the equivalent of deposits at a commercial bank? Well, for most of history, yes. But when the market goes down and there is a flight to safety, money market funds lose the liquidity that makes money money. An old physicist joked "Imagine how difficult physics would be if electrons could think?" Under different monetary regimes, money can "think" in different ways! Depending on the expectations of the agents in the market, what is considered money can quickly change. As a result, what used to be money begins to behave in a very sporadic manner.
If money can change, old monetarist definitions of money supply make no sense. Because what is considered "money" changes with the monetary policy regime, trying to use monetary policy to control the path of the money supply is circular! Because there exists a continuum of assets with differing liquidity, there's no logical spot to use to mark the division between money and asset. Even houses can be money if the market is deep and liquid enough. If houses become important as a source of collateral, a collapse in the housing market can have effects similar to that of a monetary contraction.
In this light, Scott Sumner's argument that the tightness or looseness of monetary policy should be determined by macroeconomic conditions becomes much more concrete. If money is now a continuum of assets with different levels of liquidity, then the stance of M1, M2, or the interest rate on any given asset become useless for determining the stance of policy. When people point at the stock of M1 and M2 and say that monetary policy has been very loose, they miss the bigger picture that includes debt rehypothecation and the lack of safe assets. Because there's no way to comprehensively evaluate all of these interactions on the asset level, the only way to judge monetary policy is by the final macroeconomic outcome we care about: NGDP. Additionally, because the regime is so important for the behavior of money, more focus needs to be placed on finding the right monetary regime and less on the specific day-to-day actions.
Monetary policy must stabilize NGDP. One way it does so is by affecting the continuum of liquidity and the assets that can serve as money. Markets matter. Money matters. And most importantly, Regimes matter: for both monetary policy and the money it manipulates.