Tuesday, July 19, 2016

Criticisms of NGDP futures targeting

Zak David, a quant trader, recently wrote a post criticizing Scott Sumner's idea of NGDP futures market targeting. Sumner fired back with a defense of the idea, and Zak responded with an update to his post.

I want to see if I can explain Zak's ideas in a little greater detail. Basically, he's right.

To recap, the original NGDP futures targeting idea goes something like this:

1. The Fed sets an NGDP target (say, 5%).

2. The Fed then offers to enter into any number of NGDP futures contract with anyone who wants, at a price equal to the target. So if I take a $1000 long position in these futures, and NGDP comes in at 10% (double the target), I get $2000 back. If I take a $1000 short position, and NGDP comes in at 2.5% (half the target), I get $2000 back. And so on. The Fed is always on the other side of the deal, and I can make as many of these deals as I want (assuming I can post sufficient margin).

3. The Fed then makes monetary policy automatically in response to people entering into these contracts with it. If a person takes a long position in NGDP futures, the Fed tightens a bit to make sure NGDP doesn't actually come in above target.

Zak had three main criticisms of this idea:
A) Informed traders will not trade in this market,
B) Manipulators will trade in the market, and
C) Data revisions will introduce noise into monetary policy.

I'll ignore (C) and try to explain (A) and (B). Keep in mind that I'm not saying anything new in this post; just restating Zak's argument in my own words.

First, let's talk about why informed traders - the people we want to trade these contracts - won't even show up. Suppose I have some knowledge that the Fed doesn't, about macroeconomic forces. For example, suppose I see a big inflationary shock coming that, if the Fed doesn't counteract it, will raise NGDP to 10%. Will I take a long position in an NGDP futures contract in the market described above, thus revealing my private information to the Fed and helping it make better policy?

It depends. IF the Fed hits its NGDP target on average, then I will not. Because in that case, on average, I would lose money betting on the Fed not hitting its target. Why the heck would I bet good money on 10% - or 2%, or 6%, or 5.001% -- when the probability distribution of NGDP is distributed symmetrically around 5%? A negative expected return with positive risk? No thanks!

If the Fed DOESN'T hit its target on average, then I might be able to make some money entering into this futures contract. But if the NGDP futures targeting mechanism doesn't lead to the Fed hitting its target on average, why the heck would we want to use that mechanism to make monetary policy in the first place??

So IF the mechanism works, no informed traders would use it. Hence, whatever information they have about macroeconomic shocks will NOT reach the Fed.

That is criticism (A).

OK, so who definitely would trade in that futures market? Manipulators.

Suppose I'm just some deep-pocketed jerk with zero knowledge of the macroeconomy, and I want to make some free money at the expense of the country and the bond markets. Here's what I do. First, I sell TIPS and buy Treasuries - in other words, I bet against inflation. Then I take a huge amount of long positions in NGDP futures. The Fed tightens monetary policy. My TIPS go down relative to my Treasuries, and I pocket the spread. Then I take a bunch of short positions in NGDP futures -- at the exact same price as before -- to net out my previous long position.

I have manipulated the TIPS and Treasury bond markets. I have caused monetary policy to change. And I have made arbitrage profits.

This is bad, because it introduces noise into monetary policy, and it also causes the bond markets to be less efficient.

That is criticism (B).

Both of these criticisms are valid. In other words, an NGDP futures targeting policy as stated above would introduce zero information to the making of monetary policy, while introducing a nonzero amount of noise. You'd be better off setting monetary policy according to  a random number generator, because at least then you wouldn't be letting crooks rig the bond markets.

So Zak David is correct. The idea is not sound. If you want to use NGDP futures targeting to set monetary policy, you're going to have to think of a much better system than the one described above.

(Fun bonus point: Why doesn't criticism (A) apply to all financial markets? That's probably the biggest question in the field of market microstructure. Check out the Glosten-Milgrom model and the Kyle model for two classic answers to that question.)


1. Everybody is asking me to link to these old posts (post 1, post 2) by Mike Sankowski criticizing Sumner's idea. Check them out if you're interested in this debate (and if you're not, why have you read this far?).

2. As expected, Scott Sumner fires back. He accuses Zak and me of disbelieving the EMH. Well, does anyone think the EMH holds when the government pegs the price of assets? If the government offered to buy and sell infinite amounts of Tesla stock futures at the price of $200 a share, would that market be efficient? No, it would not. Would $200 be the best estimate of Tesla's future earnings? No, it would not. Scott, just because you call something a "market" doesn't mean it's efficient.

Sumner seems to have thought very little about how markets actually become efficient. Scott, you need price discovery. Which means you need informed traders to trade. Informed traders' trades are the mechanism by which information gets from the world into the price. If informed traders don't trade, the price is based on some combination of A) noise, and B) manipulation. Informed traders only show up if they can profit by trading on their information. For informed traders to profit, someone else must lose money. In most models of market microstructure, the money is lost by liquidity traders (sometimes called "noise traders", though that term is now usually used to mean fools who trade overconfidently on false information).

In Sumner's "market", the only liquidity trader is the Fed itself. If the Fed sets policy such that A) informed traders can make money trading against the Fed, and B) the Fed offers infinite liquidity, that means that informed traders can make infinite money at the Fed's expense. But the Fed controls policy. It can make very, very, very sure it does NOT lose infinite money to informed traders. If the Fed provides infinite liquidity, the only way for the Fed to make sure that it doesn't lose infinite money to informed traders is for the Fed to make sure it doesn't lose any money to informed traders. (Note that this would most likely involve setting monetary policy to overreact to any trade.)

Hence, an informed trader would have to be dumb to trade in Scott's market, because trading in Scott's market means fighting the Fed and losing. Hence, no informed traders will show up. So the only traders who do show up will be A) liquidity/noise traders, B) fools, and C) manipulators. Hence, if the Fed sets monetary policy in reaction to trades in Sumner's market, it will set monetary policy purely in reaction to noise and/or manipulation.


  1. Dumb question: Wouldn't it receive some liquidity from people using it to hedge?

    1. That's actually an incredibly *smart* question. If they do, you can use hedging demand to get a sense of how much people really care about inflation vs. recessions, and set policy accordingly. The problem is, people's hedging depends on how they think their trading decisions will affect Fed policy, so it's hard to actually get a good measure of people's preferences this way. I first met Narayana Kocherlakota at a macro conference when I sort of hijacked the Q&A of his keynote speech to debate this idea... :D

    2. I was going to ask isn't A wrong because informed traders would hedge. In fact, worse than wrong. The only information you'd get out of informed traders is hedge, but with B you wouldn't be able to identify them.

    3. Anonymous10:14 AM

      C is quite serious. GDP data are revised and quite noisy and essentially you give the government an incentive to falsify their production so as induce a particular Fed policy. Moreover, the only way to overcome the issue of short-term noise is to produce a longer term NGDP future, which will necessarily be less liquid and less tolerance of short-term mark-to-market problems. Look at long-term swap spreads. You can get 45 "free" basis points by buying the 30y swap spread and holding to maturity. Nobody's doing this in size sufficient to restore a positive spread. There's little reason to believe NGDP markets would turn out differently.

    4. I think you're ignoring another facet to Sumner's argument;

      '...there is a relatively simple way to reward monetary-policy decision makers. For example, assume that the Fed has a 3.65 percent nominal GDP growth target and that the committee sets the fed funds target at 2.25 percent, based on the preferences of the median voter on the Federal Open Market Committee (FOMC). Then the six hawkish FOMC members who advocated a fed funds target above 2.25 percent will presumably be concerned that the lower actual instrument setting will be too expansionary and will push the nominal GDP growth rate above 3.65 percent. The six dovish FOMC members would have expected below-target nominal GDP growth when the fed funds target was set at 2.25 percent.'

      Which brings the part I like best;

      'Next comes the first step toward a market-driven monetary policy regime. Assume that the salary of each voting member of the FOMC is tied to the accuracy of his or her NGDP forecasts. Thus, if actual NGDP growth turned out to be “too high”—that is, above 3.65 percent—then all those FOMC members who preferred a more contractionary policy stance (a higher fed funds stance) would receive a pay bonus, and those who voted for an even more expansionary policy would see their pay reduced.'

      Their money is where their votes are. Now, those incentives need to be linked to the monetary base (ultimately what we, the people, use to buy stuff);

      'The Fed would peg the price of NGDP futures at $1.0365, but only during the period where it was the target of monetary policy. During this period, changes in investor sentiment would affect the quantity of money, not the price of NGDP futures. For market expectations to determine monetary policy, there must be a link between NGDP futures purchases and sales, and the quantity of money. This link can be achieved by requiring parallel open-market operations for each NGDP contract purchase or sale. Because investors buying NGDP futures are expecting above-target growth in NGDP, the Fed should automatically reduce the monetary base each time an investor buys an NGDP futures contract, and it should automatically expand the base each time an investor sells an NGDP contract short. For instance, each $1 purchase of a long position in an NGDP futures contract might trigger a $1,000 open-market sale by the Fed. A purchase of a $1 short position would trigger a $1,000 open-market purchase by the Fed. In that case, investors would be effectively determining the size of the monetary base.
      The wisdom of the crowd would not be in guessing the number of jelly beans in the jar, but determining the number. Which would tend to push nominal GDP to the announced target. Sumner argues that had such a regime been in place in 2008, we would have avoided the worst of the financial crisis/Great Recession.'

      By harnessing incentives, as economists of all political stripes profess to believe. As for 'informed traders' not showing up...well, informed gamblers show in Las Vegas, last I looked.

  2. Anonymous5:24 AM

    I think there's an error in point two of your explanation of the NGDP futures. Consider what would happen if NGDP were zero, or negative.

    1. That's no problem, because futures contract profits can go infinitely negative.

    2. Anonymous3:58 AM

      For reference it would probably be implemented as
      (# percentage points above / below target) x ($X per point)

  3. Anonymous8:49 AM

    Noah, your 'fun bonus point' was the first thing that came to mind as I was reading critique A. I think it is only fair you explain why we actually see other financial markets when the critique A above could be leveled at them.

    1. The key difference is that in normal markets, the Fed doesn't make the markets. That's done by market makers, whom informed traders consistently make money off of (and who themselves make up the difference by making money off of liquidity/noise traders).

  4. If the manipulator succeeds in benefiting through his side investments, then the market is forecasting off-target NGDP, which would make the NGDP contract more attractive. Criticisms A and B don't explain why repriced side assets wouldn't imply repriced NGDP contracts, except to say that informed traders wouldn't want their informational advantage to be subsequently weakened. But that is circularly self-defeating, because the change in side asset values indicates changes in NGDP contract values. How do the side assets increase in value if contrary interests are so powerful that they hesitate to buy/sell NGDP futures for fear of shifting monetary policy?

    Maybe the A + B critique assumes that the manipulator has fooled the market as well? But then the A critique seems to imply that the market is aware of manipulation, but withholds arbitrage for fear of being too effective. That stretches the definition of market manipulation because it implies market cooperation.

  5. Anonymous9:42 AM

    Your (A) seems to miss the point. If, in a free market for NGDP futures, the Fed's target price prevails, then the expectation is that NGDP will come in on target. The Fed's promise to issue as many or as few of these contracts as the market demands at the target price *is* the beginning and the end of the policy, since it mechanically ensures the price will stay where it wants it to stay (without resorting to a price control).

    This issue ties into Grossman and Stiglitz, who pointed out that in a fully efficient market, there would be no incentive to acquire information, and so efficiency could not arise in the first place. This places a conceptual limit on efficiency (there has to be enough inefficiency to incentivize information acquisition), but near-efficiency is still possible.

    If the traders think NGDP will come in above (below) target, they will go long (short) on these futures, thereby triggering automatic adjustment of monetary policy. Knowing this, traders will know that their expectation that the Fed will miss it's target will always be wrong on average, and hence there will be no trade. But this is precisely the point! It fixes expected NGDP at the Fed's target, which means monetary policy is optimal (if you don't like NGDP, you can replace it with whatever other target you like). FWIW, John Cochrane favors the same idea, except he prefers CPI futures with a fixed target level.

  6. Think of an NGPD target as similar to an inflation target.

    The target is set by policy makers using a variety of criteria. The futures market would help inform the decision, not directly calculate what the target is.

    For instance right now Yellen and John Taylor have different ideas about where the natural rate is.

    The futures market would be a way for the market to signal the Fed just as now the market is commenting on Fed policy by having increasingly lower yield on safe assets. Lately the Fed has been saying the market is wrong, but as time goes by looks like the market is right again. Fed policy has been too tight given its inflation target.

    Inflation targeting has clearly failed. Kocherlakota wrote in Bloomberg about economic policy during the Obama years:

    "Obama's Economic Disappointment by Narayana Kocherlakota

    In January 2009, at the beginning of Obama’s first term, the nonpartisan Congressional Budget Office issued a 10-year forecast for the U.S. economy, including such indicators as unemployment, gross domestic product, the budget deficit, government debt and interest rates. Here’s a table comparing the CBO’s expectations for the year 2015 to what has actually happened:

    NGDP forecast to grow 33 percent, actually grew 22 percent.

    Real GDP, forecast 20 percent, actual 10."

    So inflation targeting has failed. So has the strategy of backing Hillary and calling for more government spending and investment. She's only promising $180 billion a year over 10 years, without raising the deficit.

  7. A much more sensible NGDP future market would be one where it actually directly drove the monetary base.

    For example: The govt could underwrite to buy or sell a fixed percentage of future GDP measured in dollars for gold.

    Assume (to keep it simple) that 0% NGDP growth is targeted. The govt commits to buy or sell 0.0001% of GDP for a gram of gold.

    If people's expectations of future NGDP falls the value of 0.0001% of GDP also falls and it becomes a good deal to sell gold at the fixed rate. This would increase the monetary base and will raise NGDP expectations back to the point where it meets the target and selling gold is no longer a good deal. And vice versa for too high expected.

    It should not matter if the value of gold in money terms varies as what is being held constant here is the dollar value of NGDP.

    1. 'A much more sensible NGDP future market would be one where it actually directly drove the monetary base.'

      Did you read Sumner's paper?

  8. Thanks for the pointer to Dave Zachary's blog - looks good. Might be a good Twitter follow too.

    I think it's worth expanding a bit on the observation that there is no existing NGDP futures contract, because there is a connection to the hedging question. Sumner requires the existance of speculators who in aggregate can estimate a more accurate probablity distribution for future NDP (conditioned on current CB policy) than the CB can estimate for itself. It doesn't matter whether there are heterogeneous agents who are right in aggregate or just one agent who is always right, because the CB will make an unlimited market. But suppose that in addition to speculators, there are also hedgers. Then we would have the preconditions for an NGDP futures market even if it had no effect at all on the policy decisions of the CB. But if the CB bases effective policy on the market, then it discourages both speculators, as shown by Zachary, and hedgers, because the volatility of NGDP will also be reduced (because targeting NGDP level will transform an unbounded process to a mean-reverting one.) So whatever the interest in an NGDP futures market would be in Sumner's world, it is strictly less than the interest in the actually existing world.

    Zachary mentions failed futures contracts linked to employment, but there is an example that has an even more direct impact on financial contracts than either NGDP or employment: inflation. We have markets in the cash instruments (inflation-linked bonds) and in the OTC derivatives (inflation swaps, caps, and floors) but repeated attempts to float an exchange traded inflation contract have all failed. So this question of the viability of a futures contract is not a theoretical one.

    1. Thanks Phil! I wasn't aware of the history of those OTC products. Very interesting.

    2. Yes, I literally talked with the inflation guy who is still a friend about listing inflation contracts. The size of this markets is small - too small to be listed.

      Very few institutions have a direct AND measurable AND highly correlated economic exposure to inflation. Not zero, but not many either.

  9. It's also worth pointing out that the existance of hedgers would be a serious problem for Sumner's plan because hedgers are willing to pay a risk premium and it is unlikely that hedging interest would be symmetrical. Therefore the market-implied expectations of future NGDP would not be the same as the market's actual expectation of futre NGDP and the CB would have to employ a model to estimate the risk premium (just as, e.g., the Cleveland Fed does for inflation.) So we would be thrown back onto the judgment of bureaucrats that Sumner is so anxious to avoid.

    1. You could accomodate that by having a sufficiently large NGDP target corridor. Changes in volatility premiums also communicates information about monetary credibility.

  10. I'm confused by point A: Even if a trader expects that the Fed will hit its target on average (that's her prior), why wouldn't the acquisition of new knowledge (a 10% spike in NGDP) lead to a an upward revision of the likelihood of the Fed missing its target? Why wouldn't this updating induce a trader to bet? Or, maybe more formally, is the argument that no possible expected deviation from the target could be large enough to make a bet rational, if we start with the prior that the Fed hits its target on average?

    1. You can think of it a little like the Grossman paradox. If the futures market is informationally efficient - that is, if my trading conveys my information to the Fed, allowing them to set monetary policy optimally - the Fed will observe my trade and readjust monetary policy so my trade doesn't make money.

    2. How does the manipulator make money assuming high efficiency? The same anticipation of Fed reaction in the NGDP market should also prevent the manipulator's target assets from being manipulated.

    3. Think of the Fed as a bookie and the trader as a bettor. The bookie wants a balanced book, so that she doesn't have to pay out no matter who wins. Actual existing bookies do this by adjusting the odds (i.e. the price) until their book is balanced. Under Sumner's proposal, the bookie does this by publicly changing the real odds of the contest, say by breaking the kneecap of one quarterback. The betting odds stay fixed at 1:1. So the trader sees 1:1 odds on the Titans to beat the Panthers, and takes the Panthers. But the act of making the bet causes the bookie to fix the game for the Titans.

      The basic problem is that Sumner wants to outsource the forecasting of NGDP to the private sector, but he doesn't want to pay them for doing the work - in fact, he wants to charge them! For a self-proclaimed libertarian, this is kind of a bizarre misapprehension of how markets work. He needs to figure out a scheme whereby Fed policy achieves its goal AND the traders providing the information get paid, but instead he has NAND.

    4. Phil, that is a nice example of the A + B argument. However, it doesn't explain how you simultaneously have a market that rewards a manipulator, AND credits the NGDP market + Fed response with such efficacy.

      The problem is that changes in asset prices to the benefit of the manipulator implies that the market's NGDP expectations have veered from the target. That creates exclusive events: 1) Manipulator benefits from off-target NGDP expectations and 2) Positional trades in NGDP futures bring NGDP expectations to target.

  11. Seems to me you are throwing out the NGDP targeting mechanism with Sumner's flawed implementation proposal bathwater.

    First you say "So Zak David is correct. The idea is not sound." The very next sentence is "If you want to use NGDP futures targeting to set monetary policy, you're going to have to think of a much better system than the one described above."

    This begs the question, and is worth a new post. If some better proposal for NGDP targeting were put in place by an economist with stronger understanding of betting/prediction markets, let's say Leighton Vaughan Williams created one. And this modified proposal was not gamable, or easily gamable at least. Then...is NGDP targeting a good idea or not? I suspect you are undecided. But hard to tell. Either way I think addressing this more fundamental question is more interesting than the tactical question of a particular proposal not being workable.

    1. NGDP targeting is different than NGDP futures targeting! That's the key.

  12. Let's see what Scott Sumner actually wrote in his 2013 paper;

    In most futures markets, a change in investor sentiment affects the price of the
    futures contract. This proposed market would be very different. The Fed would
    peg the price of NGDP futures at $1.0365, but only during the period where it was
    the target of monetary policy. During this period, changes in investor sentiment
    would affect the quantity of money, not the price of NGDP futures. For market
    expectations to determine monetary policy, there must be a link between NGDP
    futures purchases and sales, and the quantity of money. This link can be achieved
    by requiring parallel open-market operations for each NGDP contract purchase or
    sale. Because investors buying NGDP futures are expecting above-target growth
    in NGDP, the Fed should automatically reduce the monetary base each time an
    investor buys an NGDP futures contract, and it should automatically expand the
    base each time an investor sells an NGDP contract short. For instance, each $1
    purchase of a long position in an NGDP futures contract might trigger a $1,000
    open-market sale by the Fed. A purchase of a $1 short position would trigger a
    $1,000 open-market purchase by the Fed. In that case, investors would be effectively
    determining the size of the monetary base.

    Traders would continue buying and selling NGDP futures until the money supply
    had adjusted to the point where the market expected NGDP growth to be right on
    target. Reaching a market equilibrium does not mean that each market participant
    expects on-target growth. As in any market, there is a diversity of opinion. If there
    were not, there would be no reason for trading to occur. One common misconception
    is that no one would trade NGDP futures because they would know that NGDP
    futures prices would be set at the market expectation for future NGDP. But equilibrium
    prices reflect market expectations in all asset markets. The price of copper
    futures reflects the market consensus of where spot copper prices will be at the
    contract’s maturity. The existence of market equilibrium does not stop the trading
    of copper futures, as there are differences of opinion among traders.

    Recall that after the central bank stops pegging a particular contract (and starts pegging the
    next contract), its price can rise or fall. Thus, people would trade NGDP contracts
    for exactly the same reason they trade any other asset: in the hope that the price will
    move in the direction they anticipate (after the Fed is no longer targeting the price).

    Even an absence of trading of NGDP futures contracts might not cause problems
    for monetary policy. If no NGDP contracts were traded, it would mean that the
    market consensus expected the current setting of the money supply and/or interest
    rates to produce on-target NGDP growth. At worst, it would mean that any expected
    deviations from the NGDP target would be so small that investors did not think it
    was worth the time and effort to trade NGDP futures.

    Which would mean; 'Mission accomplished.'

  13. Anonymous2:26 PM

    I don't get why the Fed would have to make the market or why anyone would be able to get out of their position at any time by placing an equal and opposite bet. That makes very little sense.

  14. Two basic problems with inflation targeting:

    1. Central banks have inflation targets, and there are market-based measures of future inflation. But most central banks are currently systematically missing their inflation targets on the low side. So what makes you think these same central banks can hit NGDP targets?
    2. Even if you can hit NGDP targets, why is that a good thing? Central bankers seem to like to talk in terms of Taylor rules with output gaps and inflation targets in them. Isn't NGDP targeting just a special case of that?

    1. "Even if you can hit NGDP targets, why is that a good thing?"

      There are economists who believe that inflation has some magical power to increase real gdp. The mechanism is said to be either: (1) it overcomes downward sticky wages (which assumes that there is a materially large group of people with materially excessive real wages and no bargaining leverage) or (2) implicit debt relief for some at the expense of others.

  15. Reading this post with fresh eyes, I wonder if you are assuming that the manipulator possesses some structural advantage over the rest of the market? The manipulator's arbitrage creates opposing incentives to reverse the arb. In a literal sense, under an NGDP targeting system, EVERYONE is a manipulator, and the bet is that the market is sufficiently efficient so that future pricing approaches an unbiased estimator.

  16. How about this:

    You buy NGDP futures betting that the FED won't hit their target. You also short treasuries. If the FED tightens in response to the NGDP futures trade, you lose on the NGDP futures but gain on the treasuries.

    This would depend on the pay-offs to treasuries when the FED tightens, but it's not hard to find financial instruments that would have a pay-off that would exceed the losses from NGDP futures.

  17. Anonymous5:10 PM

    Guys, the Fed isn't setting the *settlement* price of the contract. It's *pegging* it. Big difference. It would not be futile to trade this contract against the Fed.

    Yes, the Fed can adjust policy to keep from losing huge amounts of money, but it can't control *actual* NGDP perfectly.

    I must be making an error becaus this seems obvious to me.

    1. You're right, so if the Fed is risk-averse it'll set policy so as to *overreact* to each trade against it, just to make damn sure no one can take it for billions in a day. That overreaction will A) be sure to drive away informed traders and leave the market to noise traders and manipulators, and B) introduce noise into policymaking.

    2. The Fed is not really free to overreact or underreact. It has to react by just the amount required to equalize the number of buyers and sellers. It's really acting as a Walrasian auctioneer who clears the orange futures market, not by adjusting the price (that is pegged), but by planting and tearing up orange trees.

  18. Anonymous5:41 PM

    Whoa, whoa, whoa. This is a different objection that we can get to in a moment. I want to get clear on the first objection first. There *is* a reason for an informed trader to trade this, and there is a clear way for that trader to make money. Can we agree on that much?

    1. Oh no no no. This is the same objection, not a different one. And no, there *isn't* a reason for an informed trader to make this, and *no* way for him to make money.

    2. Anonymous6:19 PM

      (Sorry for not replying to the original comment. I misclicked.)

      OK, let's break it down.

      Simple Scenario
      Peg is at 5%. I believe that if the current state of affairs continues NGDP will come in at 4.9%, so I sell some NGDP futures to the Fed. NGDP comes in at 4.9%. I make money. Yay.

      Overreaction Scenario
      Peg is at 5%. I believe that if the current state of affairs continues NGDP will come in at 4.9%, so I sell some NGDP futures to the Fed. Let's say I do this 2 days from settlement. Now, the Fed wants to make sure it doesn't lose any money to me, so it starts blowing up the base, overreacting, as you say. I see this, and I think they have succeeded such that if the new conditions continue until settlement, NGDP will come in at 5.1%. Since there is no bid-ask spread, I can come back and reverse my position without pain. So, one hour before settlement, I get net long NGDP futures. Fed doesn't have time to materially impact the NGDP numbers for the period in question. I make money. Yay.

      Where am I going wrong?

    3. The key is the second scenario, since your "simple" scenario assumes the Fed doesn't react to the "market".

      Now, if the Fed believes that NGDP reacts to policy with a lag, it won't allow short-maturity futures trading for exactly this reason. If it thinks the policy lag is 4 months, for example, it won't make a market at 3 months out, since that would simply be giving away free money without getting it any information that would help with policymaking.


    4. @Noah Smith, you are drawing a false distinction between "manipulators" and "informed investors". If the Fed is responding to futures pricing, then every trader is a manipulator, and every manipulator is an informed investor. The updates are even more question begging than the original post. What obstacles prevent "manipulation" above and below the target? If Fed policy is tighter than target, and I can "manipulate" policy towards the target, then I can gear my portfolio towards loosening.

    5. Anonymous4:27 PM

      Ok, let's go with the four-month lag. Are you saying the June contract will expire in February, or that the Fed will simply quit making markets in it until February? (Scenario to follow once I know what you have in mind.)

    6. If the "market" allows people to trade infinite amounts of futures 10 seconds before maturity, when NGDP is known (and different, at least infinitessimally, from the price), that allows someone to make infinite money with zero risk at the Fed's expense.

      You figure out the rest. ;-)

  19. Does not this whole thing make no sense in the first place?

    Sumner says that he prefers a scheme in which money is "convertible" into futures. Convertibility makes sense if you get a real asset, like gold. However, why would you pay money to get a futures contract, when you can enter into a futures contract with no money down?

  20. Anonymous11:22 AM

    "If Noah Smith wants to seriously challenge the policy he needs to provide a plausible argument for large and time varying risk premia in the NGDP futures markets. So far, no one’s been able to do that. But that’s the sine qua non of any criticism. Otherwise, I simply don’t care. Manipulation? Who are the victims? And did this occur under Bretton Woods?"


    this is getting good.

  21. I have a much simpler idea for the Fed to keep ngdp on track. A non-basic basic income. Whenever there is a shortfall and it can't loosen, it distributes a bonus to everyone, and whenever there is a surplus, it tightens monetary policy. It would be directly under their control and much more effective, albeit still subject to corrections and noise.

  22. Much of the objection seems to center on facing the Fed as a counterparty who is trying to make money off you. OK, let me suggest a change to Sumner's scheme (it's really more a change in framing than in substance). Prohibit the Fed from taking any sizable position in the futures market. Oblige it by law to adjust policy until there are zero net bets. No one fears losing to the Fed because the Fed has no position in the futures market. Participants in the futures market are betting purely against each other: those who think the Fed is too tight against those who think it is too loose.

  23. I have nothing better to do with my life than to try to carry the Walrasian auction analogy further, so here is another proposed market setup. In this setup, the Fed never acts as a counterparty in the futures market at all. It never takes any position, not a single contract for a single second.

    The Fed does NOT offer to act as a counterparty in the futures market. It never takes any position, not even temporarily. Instead it runs a daily auction where it invites traders to submit schedules of their demands as a function of open-market operations. A typical schedule will say “I want to: buy 10 contracts if you expand base money by $1 billion, sell 5 contracts if you do nothing, sell 20 contracts if you contract base money by $1 billion” and so on. From these schedules the Fed computes the open-market operations that result in zero net demand, carries out the operations, and records the trades. The Fed itself never has to take any position.

  24. Why not just a regular prediction market? Why can't the Fed just observe the prices in this market instead of having to get involved?

    1. I think this is reasonable. At the end of my post I noted that the Fed could observe the Fed Funds Futures market (and that they probably already do). If Sumner wanted to subsidize an academic NGDP futures market for "research purposes" like he's been doing, then it might be somewhat useful.

  25. Are comments closed? First, NGDP cannot be smoothly and linearly manipulated by the Fed, so the real danger of Sumner's proposal is that it might lead to drastic changes in the money supply. Second, you must have "noise" for markets to work; this is textbook 101 (noise traders create markets according to EMH). Sumner's idea is no worse than traders in Fx who bet a currency won't break a certain band (Soros/Bank of England).