a bit harder than Scott thinks

I like reading this sort of paper, because it reminds me that history is moving in the direction of market monetarism.  There is no stable equilibrium where the monetary policy regime does not result in the financial markets expecting steady growth in NGDP.  

NGDP targeting of the forecast is easy, setting the forecast is a bit harder.

The currency banker executes, by contract, bets that NGDP grows at a particular rate referenced to time. But then the ledger reports NGDP at a regular interval and the bet becomes a sure thing, bets pile up so the currency banker needs to charge congestion fees. It is still doable as long as the natural NGDP growth remains stable at 3%, or it holds to something close.

But there is no need to be synchronous because there is no need to collect data, expected NGDP growth is right there, at the currency issuer, in loans to deposits, the active queues.    The currency issuing function gets first peek at these two stacks, and it can be asynchronous, settling the bets at any instance, with an interest charge, when it can do so with minimal currency risk.  In other words, asynchronously, there is no advantage in insider information from the data collector,  we are skipping the data collection loop. No one trader has a better look at the tradebook, and the pit boss is a half step ahead. 

The advantage of a time synchronous data collector is that we get an extra layer, central bank term insurance.  Not necessarily a bad idea,but I always had that function up in the smart layer. Remember, insuring time generally falls on government as part of currency insurance.  A bad but workable idea.