Friday, July 6, 2012

Noah Smith on Futures Targeting

Noah Smith wrote a critique of Scott Sumner's proposal to use futures contracts on nominal GDP to target the growth path of nominal GDP.   Unfortunately, his understanding of the proposal came from Williamson's brief description.   Smith discusses a proposal to control monetary policy according to changes in the price of a futures contract.   The central bank would observe the price at which private investors trade a futures contract, and then use that to gauge appropriate changes in monetary policy.

Smith argues that monetary policy would be very volatile--the central bank would have to tighten or loosen  large amounts.   He makes this judgement based upon the volatility of stock prices relative to the discounted value of dividends.   Since stock prices vary much more than discounted dividends, then futures contracts on nominal GDP would vary more than nominal GDP.    He then notes that nominal GDP has been very volatile (though using Williamson's peculiar measure.)     And so, the futures prices would be extremely volatile.

It is certainly good to see more outside criticism of futures targeting.   Unfortunately, it is important to review what the proponents have said rather than just riffing off what other critics have said.   Sumner proposes having the central bank buy and sell the futures contract at a target price.   There would be zero volatility in the price of the futures contract.

I believe that Sumner has in mind a system where the target price increases at a 5% annual rate.  However, it would also be possible to define an index by dividing the actual value of nominal GDP by the target, which would mean that the price of the future contract would never change.

Of course, what this means is that the central bank always takes a position opposite to the net position of market traders.   If there are more bulls than bears on the market, the Fed must be a bear too.   If there are more bears than bulls, the Fed must be a bull.     Smith's argument would then be that because stock prices vary more than actual dividends, the Fed's position on the futures contracts would vary more than actual fluctuations of nominal GDP.   Since nominal GDP varies so much, then the Fed's position on the contracts would vary a tremendous amount.

There are several problems with his argument.

Estimates of nominal GDP volatility from periods when the central bank was targeting some combination of inflation and the output gap, targeting the unemployment rate and allowing accelerating nominal GDP growth, and  targeting some combination of inflation and interest rates (explicitly to help the Treasury fund deficits, no less,)  are not likely to help predict actual nominal GDP volatility under a monetary regime that solely targets the growth path of nominal GDP.

To the degree that stock volatility is generated by momentum trading (foolish investors buying stock because it provided capital gains in the past,) or some kind of "greater fool" process, (clever investors buying over-valued stock planning to unload it to really foolish investors,) or even clever people playing some zero sum "musical chairs" game, none of that applies to an index futures contract with a fixed price.   The contracts only provide a payoff if nominal GDP actually deviates from target in the direction predicted by the trader.  If it goes the other way, there are losses and with no change, there is no gain, just wasted transactions costs.

I would think that comparing the volatility of stocks to the volatility of dividends is not just looking at excessive volatility in stocks, but also at the theory that stock only has value if profits are paid out to the stockholders in the form of dividends.   The notion that stockholders accept that profits are retained forever and any one stockholder can only cash in by selling shares, is also being tested.   Suppose that certain blue chip companies pay out regular dividends, hoping to attract investors who see the stock as providing stable income, almost like a bond.   Other companies keep just about all profits and reinvest them.    Dividends are remarkably stable (perhaps more so than earnings.)   Stock values  vary depend ing on expected earnings.   (Of course, maybe Smith misspoke and he meant that stock prices are remarkably more volatile than earnings .rather than dividends.)

Also, discounting observed dividends by a variety of discount rates, and comparing that to stock price volatility seems to assume that discount rates are stable.   Isn't the hypthosesis that there is a discount rate being tested?  I would think that one of the things that impact stock price volatility is changes in expected future discount rates.

Most importantly, stocks form an important store of wealth for many people.   They have a long record of providing positive returns one way or another.   People storing wealth in this way have any number of personal reasons to buy or sell.  Further, the prices do change, and trying the beat the market is always a temptation.   Nominal GDP futures are a bet on whether nominal GDP will be on target or not.   If the system works perfectly, the expected return would be zero.   It is hard to see why these contracts would form an important store of wealth.

All of the proposals for futures targeting involve having the central bank change current monetary policy according to changes in its position on the futures contract.   If the market is long, and the central bank is short, then it should tighten.   If the market is short, and the central bank is long, then it should loosen.   The problem that concerns Smith is that the market position on the contract can be expected to vary more than the true expected value of nominal GDP away from target, (which means more than the expected payoffs) and so the central bank will be loosening and tightening more than necessary.    And even if nominal GDP stays on target, this offsetting loosening and tightening seems scary.   Well, let's change "scary" to "might be disruptive to real economic activity," or "somewhat or substantially reduce  human welfare."

Smith suggests that this volatility in central bank policy would be quarter by quarter.   There would be a quarter of tight monetary policy followed by a quarter of loose monetary policy.       Most proposals for futures targeting involve continuous adjustment.   The central bank trades the future all during the quarter and makes open market operations according to its position on the contract.   In other words, rather than the trades last quarter determining whether policy is tight this quarter, trades early today are determining monetary conditions later in the day.   Trades late in the day determine monetary conditions tomorrow morning.

Sumner has proposed discrete trading, but on a daily basis.    The central bank announces a tentative daily target for base money, the contracts are traded, and then the Fed adjusts the tentative target by the net position on the contract.   Perhaps the futures contract trades in the morning, and then open market operations in bonds occur in the afternoon.

This doesn't mean that there would be no volatility.   On the contrary, my point is that all of these proposals would allow for volatility to occur during each quarter, each month, and each week.   And with the more continuous proposals--during each day.

To some degree, this high frequency volatility is even more "scary," but not necessarily.   Wouldn't such fluctuations would be over before there was much chance for it to do harm?   Still, I believe that the quantity of base money should adjust to the demand to hold it, conditional on the nominal anchor.   And the least bad nominal anchor is slow steady growth of spending on output--a nominal GDP target.   To the degree trades of the future contract causes an excess supply of base money for the first part of the week, even if there is an offsetting excess demand at the end of the week, it would be undesirable.

These shortages and surpluses of base money would certainly cause liquidity effects on interest rates, and so there would be undesirable volatility in short term interest rates.   Now, I don't think that stable short term interest rates are necessarily desirable.   I think that short term interest rates should fluctuate due to short term changes in the supply and demand for credit.   If many people want to borrow this week, and few want to lend, having interest rates a bit higher this week, so some of those wanting to borrow wait a bit, and who were going to lend next week, bring forward their lending to this week, is desirable.

Still, it is hard to see how significant volatility would be introduced into short term interest rates, much less long term rates, by high frequency changes in the monetary base.   It would rather be expectations of persistent changes in the monetary base due to persistent shifts in the market position on the contract that would have have persistent effects on financial markets and then spending on output.

Of course, most criticisms of the proposal have been that no one will be motivated to trade the contract, not that they will do a lot of trading for reasons that have nothing to do with expectations of nominal GDP relative to target.

Anyway, my own view is that the central bank should be free to adjust monetary conditions as it sees fit, subject to the constraint that it trade the futures contract at the target price.   If it turned out that a short position by the market today was likely to be followed by a long position by the market tomorrow, then making  open market purchases today just to sell them tomorrow would be pointless.    (Part of my mindset continues to be consideration of a purely privatized version of index futures convertibility, where all the competing issuers of money are free to choose their positions on the contract just like anyone else.)

Finally, if futures targeting worked, wouldn't the variation of  the central bank's position on the contract vary more than nominal GDP?    It is supposed to provide a negative feedback mechanism.   Monetary conditions today are adjusted according to market expectations of future nominal GDP as revealed by the market position on the futures contract.  If it works, nominal GDP must actually have less volatility than the the market position on the contract.    Suppose nominal GDP varied, but the central bank's position on the contract always stayed zero.   Would that be a good thing?

Noahpinion: Excess volatility and NGDP futures targeting


  1. Bill,
    I disagree with this part:

    "Smith's argument would then be that because stock prices vary more than actual dividends, the Fed's position on the futures contracts would vary more than actual fluctuations of nominal GDP. "

    It is very useful to look at the chart of the Chicago VIX. When stockmarket options are expensive, hedging stockmarket positions with NGDP futures becomes very attractive, even though the expected return of these futures is zero. It is because even though the expected return of the portfolio does not change when the futures are added, portfolio volatility is reduced. So we should expect to see wild swings in the size of Fed's balance sheet that are correlated with the VIX. Of course, this is a good thing.

  2. It might be socially optimal to allow some slight volatility of NGDP along the NGDP target path in response to changing risk premia. In this case the solution is to target a narrow range of NGDP futures prices.

  3. I doubt that having spending on output vary according to "risk" premia would ever be a good thing. Unfortunately, I don't understand this idea to know whether it "might" be better to have a positive or negative correlation.

    I think the market interest rate should stay equal to the natural interest rate, the quantity of money equal to the demand to hold money, and spending on output grow at a slow, stable rate. To me, risk premia impact the natural interest rate and perhaps the demand for money. They may impact the proper allocation of resources between consumer and capital goods (and between different people's consumption and different firm's investment.) Why should spending on output change in aggregate?

  4. You should explain this better.

    I have call options on stocks (or say just an S&P) index. I will lose my premium if stock prices fall. One reason they might fall is that nominal GDP falls below target So, I hedge by selling nominal GDP futures.

    I have put options on the stock index. I will lose my premium if stock prices rise. One reason stock prices might rise is that nominal GDP rises above target. So I hedge by buying nominal GDP futures.

    Whoever is writing the options has the opposite incentive. So, there is trading with the Fed even though nominal GDP is expected to remain on target, though with purchases and sales roughly matched.

    It is these kind of transactions, however, that create worries about perverse manipulation.

  5. 123:

    My second comment replied to your first.

    My first comment to your second.

    Maybe I need to figure out how to make these comments nested.

  6. Bill, the problem with VIX is the following. To the extent VIX reflects real changes in risk aversion (it could simply reflect the changing risk of monetary policy errors), VIX affects the transmission of monetary policy. It is important to realise that futures peg is pegging the risk-neutral expected NGDP, and not the actual expected NGDP. When VIX is high, and NGDP futures trade at the pegged levels, markets are indifferent between the smaller probability of NGDP undershooting and the larger probability of NGDP overshooting. Because actual NGDP being below target is more costly to markets, the actual statistical NGDP forecast is above target.
    Recent speech called "Optimal outlooks" by Kocherlakota from the Minneapolis Fed contains a good discussion of risk neutral probabilities, even though he reaches the opposite conclusion, namely that the Fed should target TIPS spreads at 2%, even though it means that statistically inflation is expected to be 1.8%.
    So suppose VIX is high, NGDP futures indicate 5% NGDP growth, and this means that markets expect NGDP will grow by 5.5%. How can NGDP temporarily grow at the pace above the target? Well, there was a temporary explosion in money supply that will be reversed that was needed to make sure that NGDP futures remain pegged.
    Williamson has a chart that proves that for supply-side reasons it is optimal to have lower NGDP at night. So it also might be optimal to have a temporarily lower NGDP during a period of heightened risk aversion, if such risk aversion causes a temporary supply-side shock. For example, if there is a financial crisis during year one, we might achieve a higher total two-year output if we get 4.5% NGDP growth during year one combined with a catch up with the NGDP target path during year two. Tyler Cowen has made a related point in his
    "Williamson has doubts about ngdp targeting" post, he called the resulting increase in currency/credit ratio dangerous.
    NGDP level targeting is essential especially during the financial crises, however for supply-side reasons it might be better to target a narrow corridor along the NGDP target path.

  7. Thank you. I am not sure I understand all of this. I am concerned about the argument that speculators will want to be sell the contract to hedge against a recession induced reduction in their stock portfolios even when they expect nominal GDP to be on target. It seems to me that this would keep nominal GDP above target enough for them to pay for this insurance. If the central bank stays hedged, then speculators who buy the contract would get a positive expected payoff so that they cover the losses for stockholders when nominal GDP is below target. This would vary with risk aversion.

    I don't see why spending on output "should" be high when people are more risk adverse and vice versa.

    To me, it looks like overshooting. People are more worried about the future, and this causes them to spend less on output. Monetary conditions need to become more expansionary to offset this and keep spending on target. But those worried about the situation sell futures even when monetary conditions have expanded enough to keep nominal GDP on target, because they are willing to pay more to insure against these greater worries. Expected nominal
    GDP is even future above target than normal.

    It seems like overshooting.

    Is that it?

  8. "It seems to me that this would keep nominal GDP above target enough for them to pay for this insurance."

    That's right, on average NGDP would be above target in such situations.

    "It seems like overshooting."

    Yes, we should expect overshooting in such cases. There are points of view possible:

    1. Following Kocherlakota, we could say that such overshooting would be welfare-enhancing.

    2. Central banks could target actual expected NGDP, by allowing some fluctuation in NGDP future prices

    3. Cowen and Williamson would argue that it is optimal to have a temporarily lower NGDP growth when risk premia are elevated. If there is a level targeting, there would be catch-up growth later.

    There is a related problem if the Fed invests in Treasuries. It is possible that the Fed would systematically buy Treasuries when they are expensive and sell them when they are cheap. So the excess volatility would generate losses for the Fed on the asset side of Fed's balance sheet.