I’ve had a lot of questions about NGDP futures targeting, and so I thought I should do a post answering some of those questions. My preferred target is a 5% NGDP growth trajectory (aka level targeting) but to keep the discussion manageable I’d like to put off practical questions about which target, which time period, levels vs growth rates, etc. Let’s just focus on the core concept. Will it work? By “work,” I mean would it have prevented the crash of 2008? I say yes.
For simplicity, let’s assume that the central bank decides on a 3.65% NGDP growth path. To respond to David Glasner’s concerns that we first need to catch up to trend, we will set the first NGDP target 6% above current estimated NGDP. Each day, the target for NGDP is raised one basis point. Assume the policy begins on May 16, on which day the Fed estimates NGDP to be $15 trillion. Then they would set the NGDP target for May 16, 2010 at $15.9 trillion, 6% above current estimated levels. Each day they raise their NGDP target by one basis point. From now until forever. Thus the target NGDP for May 17, 2010 is $15, 901,590,000 and zero cents.
At this point you may be scratching your head and wondering what I mean by daily NGDP estimates. Right now U.S. GDP is calculated quarterly. I assume that the government could come up with monthly NGDP estimates (doesn’t Canada do this?) And it’s obvious they could use monthly estimates for other potential targets such as the CPI or a nominal wage index. However, nothing really crucial hinges on that assumption; it just makes the plan easier to explain.
But how do we get daily estimates? Simply take a weighted average of two consecutive monthly announcements. Thus the May 31, 2010 NGDP estimate is assumed to be an average of the May and June 2010 NGDP announcements. For the May 21, 2010 estimate, it would be about 5/6 times the May 2010 NGDP announcement, plus 1/6 times the June 2010 announcement. I don’t know if these numbers are exactly right, but I think you get the idea. Note that it is not particularly important that the actual future GDP is equal to these futures estimates, just that we have a reasonable way of determining the payoff for future NGDP contracts. Also assume that at maturity the contracts are worth 1/1,000,000,000,000th of nominal GDP, roughly $15.90.
Now let’s assume that it’s May 16, 2009 and the Fed begins targeting the price of futures contracts due on May 16, 2010, at a price of $15.90. The Fed agrees to buy and sell unlimited quantities of these contracts at the target price. This guarantees that the market price of the contract is always precisely equal to the Fed’s NGDP target. Each buyer must put 10% into some sort of margin account. The Fed pays interest on these margin accounts. How much interest? Enough to assure that these policy futures markets are highly liquid. So liquidity is also not a problem. The contracts are settled about 14 months later, after all the relevant GDP data has been announced.
Before the start of trading, the Fed sets the monetary base at the level that they believe will lead to an expected future NGDP of $15.9 trillion. Once they open the futures market for trading, they announce that each $100 purchase of NGDP futures will trigger a Fed open market sale of $500, presumably of short term T-securities. I don’t believe the Fed would need to buy any other assets, but if they did, then they could buy TIPS, longer maturity T-notes and bonds, high quality foreign government debt, U.S. agency debt, foreign agency debt, high quality corporate bonds, in that order (although I have an open mind on this question.)
Trading would continue until the market reached equilibrium. This would occur when the monetary base had settled at a level where there was no longer any demand for trades. I think it is important to distinguish between how this market would operate during ongoing operation, and when it is initially set up. Under normal conditions the demand for base money is relatively predictable from day to day. There are trend, seasonal, and day of the week fluctuations in the demand for base money, all relatively predictable. Thus as a practical matter the Fed would rarely take a significant net long or short position on NGDP futures. Even more importantly, as long as their forecasts of base demand were unbiased, the net short and long positions from one day to the next would be entirely uncorrelated. Because actual future daily NGDP estimates would be (by construction) highly serially correlated, this means the Fed would take on relatively little risk over time.
On the day the program was first launched there might be somewhat greater uncertainty regarding the demand for base money. That is because the demand would be contingent on the program being in operation. And this problem could be especially severe in a liquidity trap. In my view, this problem could be mostly eliminated through a penalty rate on excess reserves. Such a policy would make the demand for base money much more predictable, as currency would then make up about 90% of base demand, and currency demand changes relatively slowly
Once the program was in operation, there would be several potential weaknesses:
1. Even if the program stabilized 12 month forward NGDP expectations, it might not stabilize longer term NGDP expectations if the public expected the Fed to abandon the policy at some point in the future. However, I don’t see this as a major drawback, as I believe stabilizing 12 month forward NGDP expectations would keep nominal wage rates well behaved, and (like Earl Thompson) I regard aggregate nominal wage instability as the key factor behind macroeconomic instability. Furthermore, the Fed could also trade longer term NGDP contracts, not as a way of affecting the base, but rather of overcoming the time inconsistency problem by setting itself up for future losses should it abandon the policy.
2. The price of the contract may not be equal to the optimal forecast of next year’s NGDP. This is actually two separate problems:
a. The price of NGDP contracts may not equal the market forecast of next year’s NGDP.
b. The market forecast of next year’s NGDP may not be the optimal forecast.
Of course the first discrepancy is simply the risk premium, while the second is the deviation from the rational expectation forecast.
Hartzmark (1987, JPE) found no evidence of risk premia in commodity futures markets. But even the finding of no risk premium in commodity futures markets grossly overstates the size of the risk premia problem in an NGDP futures market, for three reasons:
1. Risk premia often exist because traders want to hedge against risk of price changes. However, unlike for commodities, there is no significant NGDP futures market today. This suggests relatively little demand to hedge NGDP risk.
2. NGDP is much less volatile than most commodity prices.
3. Only the time-varying part of NGDP risk premia would cause macroeconomic problems, except at the date of adoption.
Some people draw false analogies with the TIPS market. The TIPS/conventional bond yield spread may sometimes be distorted by shifts in liquidity preference. This is a problem because TIPS are not pure inflation plays, but rather close substitutes for conventional bonds which trade in a somewhat more liquid market. NGDP futures should not be attached to any sort of bond.
I won’t spend much time on the rational expectations debate. You probably know I am a big believer in Ratex, and the “wisdom of crowds” in general. When I began this project in the 1980s I used an “archipelago model,” where each person observed local shocks before the Fed. And all of those individual observations were aggregated by the futures market. But as we will see, the real advantages of futures targeting lie elsewhere.
In a more recent paper with Aaron Jackson in Economic Inquiry (2006), we showed that futures targeting would eliminate the need for the central bank to estimate a structural model of the economy. This is a big advantage, as there is no general agreement among macroeconomists as to which structural model is best. In a BEJ Contributions to Macroeconomics paper (2006), I argued that futures targeting would also eliminate the need to agree on a single monetary policy instrument (or indicator if you prefer.) This is a huge advantage, as the Japanese discovered that the short term interest rate instrument was ineffective once rates hit zero. The market would choose whatever indicator they thought best. Paul might look at the fed funds rate as an indicator of Fed policy, Nick might prefer M1 or M2, and Arthur might get inspiration from the price of gold. A populist’s dream: an FOMC with 300,000,000 million potential members. And Ben could retire to one of those newly affordable condos on Miami Beach. (Affordable on a government pension thanks to . . . oh nevermind.)
The recent crisis convinced me that there is one more advantage, which may be greater than all of the others. NGDP futures targeting would eliminate uncertainty about how much credibility the Fed’s policy had in the markets, at least for as long as investors expected the policy to persist. As a practical matter, central banks are very reluctant to shift course after committing to a new policy, so I don’t think credibility would be a big problem, unless they foolishly picked a target variable that did not mesh with their true social welfare function. (This is one reason I don’t like inflation targeting, an issue I will discuss in another post.) As long as the policy was credible, the market could estimate the demand for base money under the assumption of precisely known NGDP growth expectations.
In contrast, consider the fiasco of last October. The basic problem had nothing to do with whether the market could forecast NGDP better than the Fed (although I think it did forecast better last October) but rather that the Fed wasn’t even targeting it’s own forecast. Why did the Fed not set the base at a level expected to produce on-target NGDP growth? Probably because there was enormous uncertainty about how credible such a policy would be. The easiest way to understand the Fed’s dilemma last October is to consider the widespread and simultaneous fears of deflation and hyperinflation.
People feared deflation for good reason; it was quickly becoming a reality. But they also feared hyperinflation for good reason; the base was soaring at unprecedented rates. I don’t doubt that there was some rate of base increase that would have generated the desired inflation expectations, but the Fed probably feared that a base increase that large, if successful, would raise velocity and thus we would rapidly overshoot in the other direction—toward high inflation. The root of this problem is credibility; the Fed doesn’t know how credible its inflation promises will be, and hence errs on the side of mild deflation, rather than hyperinflation. Of course the markets understood this fear, and quickly figured out that disinflation, not hyperinflation, was the real risk. Markets aren’t dumb. They sniff out Fed indecision like a schoolyard bully sensing the timidity of his next victim.
So the central banks face a particularly difficult problem when then run out of conventional ammunition. They know that some amount of QE will create inflation expectations, but they don’t know exactly how much. Even worse, they don’t know how credibly the markets would view their attempts at reflation. And the velocity of any injected base money would be highly sensitive to the expected rate of NGDP, and thus highly sensitive to the credibility of the Fed’s promise to target its forecast.
With NGDP futures targeting you don’t have that problem. The same entity that sets the monetary base (i.e. the market); also forecasts the velocity at which the base will circulate. NGDP growth expectations would have never fallen below target if the Fed had already been employing a NGDP futures targeting regime in 2008. I believe this is the great advantage of futures targeting, not the other three advantages discussed in my earlier papers. This argument has never been published, or perhaps I should say never until now—because Nick Rowe tells me to forget about journals, blogs are the new forum for research. (I suppose I could consider this blog as being a sort of journal, where I serve as author and editor, and you guys are referees.)
1. What about the circularity problem discussed in 1997 by Bernanke and Woodford? That problem only applies to targeting regimes where the Fed observes a private futures market; and adjust its policy instrument each time the futures price veers off target. Under this proposal the price would always be exactly on target; the information being provided isn’t so much the expected future NGDP, but rather the monetary base setting expected to produce on-target growth. Bernanke and Woodford said a system that forecast the instrument setting is not susceptible to the circularity problem.
2. If the price of NGDP contracts was always exactly equal to market expectations, why would anyone trade? To answer this question it will be helpful to describe a concept Aaron and I called “quasi-velocity,” the ratio of next year’s NGDP to this year’s base. NGDP futures traders each have to estimate quasi-velocity. Suppose in equilibrium the monetary base is $1 trillion and the target NGDP is $15.9 trillion. Then the market forecast of q-velocity would be 15.9. But only a few traders would estimate q-velocity to be precisely 15.9, which is merely the (dollar-weighted) median forecast. Those who expect actual realized velocity to be higher will buy NGDP futures contracts, and vice versa.
3. Why trade before you can observe where the market thinks the monetary base will be, as that information would be helpful to traders? I would expect some traders to wait until the last minute (as on ebay), but remember that each contract is only traded for one day. After that a new contract will be adopted. Traders could roughly estimate the monetary base by looking at the equilibrium monetary base from the previous day’s trades.
4. Isn’t this just turning monetary policy into a casino? Look how irrational investors were during the tech and housing booms! There are several responses I could make here. First, I think these bubbles were famous because they were out of the ordinary; markets are usually the best way we have of aggregating information. But let’s say I am wrong; then we could give the Fed research department some money to play with, and tell them to invest whenever the price seemed irrational, or bubble-like. If the Shillers of the world are correct this should make the futures targeting regime even more efficient, and should also provide a source of tax revenue for the Federal government.
More seriously, I don’t think market irrationality would be as worrisome as many might imagine, because I believe the biggest welfare cost of monetary instability comes not from fluctuations actual NGDP, but rather fluctuations in the expected NGDP growth path (the 1929, 1937, 2008 examples that I keep coming back to.) As long as NGDP growth expectations stay on target, then nominal wages will also be well-behaved, and any temporary fluctuations in actual AD will quickly dissipate, without producing any long lasting business cycles.
5. In another post I will address tangential issues. Economists disagree as to whether NGDP growth should be equal to total factor growth (Selgin), 3% (Woolsey) or 5% (as some wimps prefer.) There are others like Earl Thompson who insist (correctly in my view) that a nominal wage target is actually best. Most of the profession seems to prefer some sort of flexible inflation target. There are disputes about whether to target levels or rates of change. And should one target 6, 12, or 24 months out into the future? Of course any monetary regime, futures targeting or otherwise, faces all of these questions.
So here’s my claim. NGDP futures targeting would have maintained NGDP growth at the target rate right through the financial crisis. Even if the financial crisis had been every bit as bad, real GDP would have done far better had NGDP continued growing at a brisk pace. But the financial crisis would have been nowhere near as bad. Bank balance sheets all over the world were absolutely devastated by going from 5% positive to negative 2% expected NGDP growth. So although the lingering effects of the sub-prime fiasco would have continued in some housing markets, the housing slump would not have spread nationwide, and to Canada as well, the commodity and stock market declines would have been far less dramatic, the dollar would not have appreciated as dramatically in late 2008, and world trade would not have collapsed.
I’d like to thank Bill Woolsey for giving me some very useful advice, although he should not be blamed for any linger problems. (Bill often seems to understand the logic of my plan better than I do.) I should also mention that my colleague Aaron Jackson has worked with me on this idea. And finally that people like Earl Thompson, Robert Hall, David Glasner, Bill Woolsey, Kevin Dowd, Robert Hetzel and others pioneered the idea of using market expectations of economic aggregates to guide monetary policy.