I see a three step process. A monetary shock (money supply or demand) causes flexible asset prices to change immediately (stocks, commodities, exchange rates, etc.) This causes output to rise, and consumer prices and wages respond with a lag.
An expansionary monetary policy increases long run NGDP for reasons that have nothing to do with interest rates. (Recall that interest rates are only affected in the short run.) Instead, more money increases expected long run NGDP via the excess cash balance mechanism. This increase in expected future NGDP raises current asset prices (stocks, commodities, real estate, etc), which then leads to more current AD.
For some reason, I had missed this aspect of Sumner's thinking before. Some reactions:
1. Commodity prices and stock prices have been rising for quite some time. They spiked more in recent days, but they were already quite high. Why is the economy still weak? Perhaps falling real estate prices were offsetting rising commodity and stock prices?
2. I find this monetary transmission mechanism pretty scary. You are trying to drive up prices in bubble-prone asset markets in order to increase output. It seems to me that the risks of overshooting must outweigh whatever benefits you obtain. Sure, nominal GDP moves slowly enough so that you do not have to worry about overshooting that target. But meanwhile, you move from one bubble (Internet stocks) to another (real estate) to another (commodities). That can't be healthy. I felt much more favorably disposed toward monetary expansion when I thought that the transmission mechanism was through general prices rising faster than wages.