Blake Johnson replied in a comment on the Market Monetarist blog to my previous rubber band post. I realize that my previous post wasn't very clear, so I'll try to be more specific this time.
Let's do this with the MV = NGDP equation (where M is base money). If the central bank is initially credible in its level targeting (it doesn't matter whether it's price or NGDP targeting) and market expectations make sure that NGDP stays on the path no matter what the central bank does with M, they wont get any feedback at all as to what is a good size for M. If they keep M constant, V will grow 5% each year. Since the market expects NGDP to stay on the path, there wont be any signs in asset prices or TIPS spreads that the central bank is making M grow too fast or too slow. This is what I was referring to with the circularity problem. Perhaps I'm using that concept the wrong way.
Let's say the central bank keeps M constant. How much can V continue to increase? Surely there must be some limit. One USD can't give you 100 trillion USD of NGDP. So eventually the market will have to protest to this abuse and the central bank will have to increase M.
If they increase M enough, NGDP will return to the target path. There's no doubt about that. But the central bankers might not be 100% sure that NGDP level targeting works. This is the first time that they are in trouble. Everything was so easy previously, when they didn't have to do anything at all. So perhaps there is some limit to how much they are willing to increase M before they give up and switch back to the good old inflation target.
What if the market fears that this is the case? What if V has increased so much that the central bank wont be willing to increase M enough? What if NGDP keeps on falling due to this fear? And then, when the central bank has reached its increase-M-threshold, they abandon the level target.
Well, there sure seems to exist an easy fix to this problem. Just increase M at the same rate as NGDP. Problem solved?