First, here is Scott Sumner's ideal world:
In an ideal world, we’d remove all discretion from central bankers. The Fed would simply define the dollar as a given fraction of 12- or 24-month forward nominal GDP, and make dollars convertible into futures contracts at the target price. If the public expected NGDP to veer off target, purchases and sales of these contracts would automatically adjust the money supply and interest rates in such a way as to move expected NGDP back on target. It would be something like the classical gold standard, but with the dollar defined in terms of a specific NGDP futures contract, instead of a given weight of gold. The public, not policymakers in Washington, would determine the level of the money supply and interest rates most consistent with a stable economy.
To proceed, not everyone will understand this post, which is DeLong on Scott Sumner:
As I understand Scott's proposal, it is this: Nominal GDP in the fourth quarter of 2007 was $14.291 trillion. A 5% growth rate from that base would give us a value of $17.455 trillion for the fourth quarter of 2011. Add on another 3% for the average short-term nominal interest rate we would like to see, and we have $18.153 trillion. Therefore the Federal Reserve would, today, announce that it stands ready to buy and sell dollar deposits to qualified customers at a price of $1 = 1/18,155,000,000,000 of 2011Q4 GDP.
If investors thought that nominal GDP in the fourth quarter of 2011 was likely to be lower than $18.15 trillion, they would take the Fed up on its offer: demand the cash now, pay off the contract in a year by then paying 1/18,155,000,000,000 of 2011Q4 GDP, and (hopefully, if they were right) make money–thus the money stock would increase. If investors thought that nominal GDP in the fourth quarter of 2011 was likely to be greater than $18.155 trillion, they would take the Fed up on its offer: give cash to the Fed now, collect the contract in a year by receiving 1/18,155,000,000,000 of 2011Q4 GDP, and (hopefully, if they were right) make money–thus the money stock would fall.
If nominal GDP were expected to fall, the Federal Reserve would be shoveling money out the door at negative expected nominal interest rates. If his scheme were applied today it would be quantitative easing on a pan-galactic scale, as everybody would run to the Fed with bonds to use as collateral for their promises to pay the expected futures contract in a year in exchange for the cash now.
The Federal Reserve would then become truly the lender of not just last but first resort. Why would anybody borrow on the private market even at 0% per year when they could borrow from the Fed at -3%/year? Savers would simply hold cash rather than try to match the terms that the Fed was offering borrowers. Borrowing firms would borrow from the Fed exclusively. The Fed would thus create a wedge between the minimum nominal interest rate that savers would accept (zero, determined by the alternative of stuffing cash in your mattress) and the nominal interest rate open to borrowers.
I expressed related reservations about a related version of the idea in the 1997 JMCB. I am all for (rough) nominal GDP targeting, and considering the forecast, and for Scott's work in general, but I don't think the "automaticity" versions of it work. NGDP targeting does best as a general guideline for the central bank, which the central bank follows to make the world a better place, but without renouncing some ultimate degree of discretion with regard to timing and targeting and how good a deal they offer everyone at this new and somewhat unusual version of the discount window.
It's a general problem with strict pegging schemes that some prices (or pxq variables) adjust more quickly than others, or are better and more quickly forecast than others, and that means arbitrage opportunities against the pegger and/or very dramatic swings in nominal interest rates.
So on this question I agree with Brad and not with Scott.
Still, there is a general rule: when Scott Sumner says you are wrong, you are wrong (this is somewhat distinct from the claim that "Scott Sumner is always right," though if he worded all his pronouncements in a particular way I suppose it would not be).
So perhaps Scott will say that Brad and I are wrong. Or perhaps he will say that I am wrong about the general rule in the first place. Or perhaps he will say that we have misunderstood him.
The broader underlying question is how strict a nominal GDP target or NGDP forecast target can be and that question is not very well understood.