Zachary David is wrong about futures targeting

Last week Zachary David wrote a critique of nominal GDP futures targeting that was favourably shared by a few people I respect, but whose arguments I don't find persuasive. In it, he argues that though targeting NGDP is a fairly persuasive argument in general, specifically Scott Sumner's idea of targeting the price of contracts on an NGDP futures market would not work firstly because of revisions, and secondly because unlimited liquidity at a fixed price breaks market.

I think both objections are wrong, and indeed are wrong in a way that actually illustrates very nicely why targeting expected NGDP growth through futures markets is an attractive monetary policy regime. (Incidentally, I forgot to post this when I wrote it, last week, and in the meantime Scott has written a reply himself.)

Prof. Sumner says the government should, as well as issuing inflation-linked bonds, issue NGDP-linked bonds. Just as a regular bond might pay at a specific interest rate and return a particular principal on maturity, and an inflation-linked bond does the same but with extra to neutralise the effects of inflation, an NGDP-linked bond would add money onto the final principal to reflect changes in NGDP. So if a regular £100 bond pays back £105 on maturity, and inflation was 2% over the period, an inflation-linked bond would pay back £107; if NGDP growth was 5%, an NGDP-linked bond would pay £110.

This means that the current spread in price between a regular and inflation-linked bond reflects market expectations of inflation. If you expect 2% inflation, you're willing to pay exactly 2% more for an inflation-linked bond—no more or no less. So one way to target 5% growth in NGDP would be for the central bank to buy and sell the NGDP-linked bonds until the difference is 5%. When investors stop buying or selling from the bank then they must expect NGDP to grow 5%.

But David says this market is weird: "in general, a well-functioning market shouldn't have an infinitely liquid counter-party triggering automatic causal effects against every position." He also says it "defeats the purpose of price discovery". But there are a plethora of similar examples where this sort of behaviour isn't weird at all.

Currently, central banks consider and cite market forecasts of inflation when they decide whether to increase or reduce the monetary base—exactly the same calculation as in Sumner's scenario, except more slowly and lumpily. Markets only fail to expect the inflation target to come up when they think the Bank isn't credible or is deliberately going to miss it (because, for example, there is a big supply shock, something that is not a problem for NGDP targeting).

In this real live market "expressing your view immediately causes your view to be less right". Indeed, in this futures market, a perfectly credible bank would not need to buy and sell much on this market at all, since investors would always expect it to. This is exactly the same as in regular inflation targeting: demand shocks to countries with credible central banks do not usually change inflation forecasts—people know they will balance the shock out.

Consider also Switzerland's currency peg, or the UK's entry into the ERM. The bets market actors made were not reacting to external shocks, they were questioning the willingness of the UK authorities to do what was necessary to maintain the peg—they were testing credibility. In Sumner's world, bets on the NGDP futures market are bets about credibility. The central bank is not shouting like Canute at the price to make it go where they want, they're changing the real world so it's the right price.*

David's other objection also seems quite clearly falsified by mundane features of normal markets. He says that revisions to data mean that while firms are betting based on the true value of NGDP at a given point, they are rewarded based on what the statistical authorities report, which involves ~1% of error. He notes Sumner's point that if these aren't systematically too high or too low, no one would be expected to lose out over time by betting consistently on the true value, but he still objects.

But just consider regular markets. Every single market involves trading on data that includes error. Inflation-linked bonds involve the error in the retail prices index—not to mention the subjective decisions about quality and basket inclusion the stats office must make. And other markets invest based on considerably worse data like Chinese household income or GDP or industrial production statistics. These are simply not big problems with markets; markets are made to aggregate the info that we have, generating the best possible "answer" from that. If you disagree, you can make money saying so.

Finally David says people will not trade on the market if it exists. It's not clear that that's a bad thing. The reason they won't trade—if they don't—is that there is no doubt what's going to happen: they trust the central bank to deliver 5% NGDP growth. There is no risk, in this world, that it will fail and that they should hedge against this risk, since their customers buy their products with nominal dollars. Bear in mind that if people overall really did think NGDP would overshoot they could buy infinite bonds off the bank and make incredible sums if they were right. The reason David expects little trading is that he thinks the Bank will be credible.

So David is wrong. His examples are falsified by existing markets: if his claims were true, then it would have been impossible or crazy for the Swiss National Bank to peg the Franc, and it would have made no sense for George Soros to short the pound in 1992. If he was right about revisions then practically no markets would work, since practically all data is revised or measured with substantial error, and yet markets based on that data function well. NGDP futures targeting is a policy that could work—or if not, then for reasons not highlighted by David.

*David's sketched out scenario seems to me a partial model of the type which illustrates why formal modelling can be useful. If someone "just sells" NGDP contracts then it's true that the central bank will buy them at the price peg. But the key thing that extra money base would be expected to change the price of is NGDP contracts themselves. So any increase in the money base would be expected to trigger people into buying NGDP contracts. The individual move would sow the seeds of its own erasure and would have no net effect, even in terms of a yo-yo: markets would know in advance. It is not a reductio ad absurdum of the system at all.