« Did War Based Fiscal Stimulus End the Great Depression? | Main | Why readers believe this blog is too left wing -- or too right wing or too... »


Feed You can follow this conversation by subscribing to the comment feed for this post.

I don't think it's as paradoxical as you suggest.  The front end will rally, the back end will widen.  So the task of the market is figure out around what point the steepening is going to occur.  Somewhere in the two to five year range, maybe?

OT: Your power steering metaphor is highly evocative for me.  I accidentally once reversed power steering hydraulics.  Almost broke my arm when I tried to drive it.

Also... Though I agree that buying stocks might be more effective than buying bonds, I think there are far better ways to deal with disinflation that don't require changes in legislation.  E.g.:

1) Fed prints money
2) Fed gives money to treasury (or buys debt from treasury with it - whatever)
3) Treasury sends cheques to all citizens
4) Disinflation? Repeat.

This will eventually cause inflation, which is the communication strategy you are looking for.

Actually it would, of course, require budget legislation. But still... Wouldn't have to change the fed charter.

K: Nice save with the second comment.

I can't speak on what the Federal Reserve/Treasury can and can't do, but I'm pretty sure the functions of the Federal Reserve and the Treasury are roughly combined in the Bank of Canada, and I don't see anything in the BOC legislation that allows them to cut a cheque to a Canadian Citizen. Therefore, to do #3 would require a bill be passed in Parliment. Besides, I would imagine that cheques cut to citizens would come of the Consolidated Revenue Fund, which is not BoC's responsibility.

You can check yourself, starting at Section 18 and on:


Paul:"the functions of the Federal Reserve and the Treasury are roughly combined in the Bank of Canada"

In Canada it might be the department of finance sending out the cheques.  And yes, it would require a budget law. But, anyways, I assumed we were talking about the US since Canada is not (apparently) in a liquidity trap.

The action is a success if the real yield declines (e.g. TIPS having a negative yield) while the inflation expectation increases. It is true the two can offset to show no change in the nominal yield. The most recent Debt issuance with a negative real yield highlights the success of the communication strategy - the real cost of borrowing is lower while the inflation expectation is higher.

MTJ: But if the economy is expected to recover, real rates should rise. Maybe the policy was a total failure, which is why real yields have fallen? Who knows?

K: "OT: Your power steering metaphor is highly evocative for me. I accidentally once reversed power steering hydraulics. Almost broke my arm when I tried to drive it."

OMG! I didn't even know it could be done!

"The front end will rally, the back end will widen. So the task of the market is figure out around what point the steepening is going to occur."

That's normally how it is supposed to work, when central banks operate on the short end. But if the central bank says it is operating on the long end, isn't it trying to flatten the yield curve?

"1) Fed prints money
2) Fed gives money to treasury (or buys debt from treasury with it - whatever)
3) Treasury sends cheques to all citizens"

Yep. That's the exact case of helicopter money. It's what Australia did. But you can't subsequently reverse it, if you need to, without tax increases.


"K", above, describes the money-financed deficit form of a "helicopter drop". My question is, doesn't this describe the current (expected) situation? Of course the Fed is not allowed to buy Treasury bonds at auction. No matter -- it just buys the bonds that the banks bought at auction. The anticipated level of QE2 (around $100b/mo.) just about equals net monthly Treasury issuance to finance the deficit. My observation is that there seems to be a lot of talk about how QE gets "money into circulation", while the fact that the Fed will be financing Treasury spending seldom gets mentioned.

BTW, should the GOP carry the House in the mid-terms, Congressman Ron Paul is expected to be the Chairman of the House Subcommittee charged with Fed oversight. Yes, laws can be changed, but it seems unrealistic to think that the President would have the votes for such an action in the coming Congress. Further, the Fed faces record-low popularity and more aggressive oversight from Congress as a result of its quasi-fiscal actions in 2008-2009. To ask Congress for the authority to purchase stocks in this environment is to invite all sorts of Fed-threatening amendments to the enabling legislation.

Finally, if your proposal has merit, why not argue that the fiscal authorities should undertake it? It is completely within their purview, and certainly the Fed has no more expertise in buying stocks than Treasury does. If Congress could authorize the Fed to buy stocks (which, again, is highly unlikely), they could just as well authorize Treasury to do so, and the Fed could finance it. This would at least leave intact our democracy's system of oversight and accountability over fiscal actions without threatening, in the long term, the Fed's independence. Surely, you can imagine that, following a Fed purchase of stocks, any future President or Congress, in an effort to maintain popularity, would urge the Fed to prop up or "juice" the stock market. Were Treasury to have to do so instead, it would be held accountable for such a blatant manipulation of prices for political benefit.

Nick:"But if the central bank says it is operating on the long end, isn't it trying to flatten the yield curve?"

I am not aware that they are intending to flatten the whole curve, all the way out 30 years. If it were me, I would try to shape the curve into what I would consider to be a credible path for the short rate.  I.e. If we keep the short rate at zero for the next 3 years, there's going to be so much inflation/growth by then that rates will have to rise fast.  So I'm going to buy 3 yr yields down to zero and let subsequent ones equilibrate to a natural level.  Its a way of clearly communicating the path of front end rates.  I like the idea of the Fed announcing for how long they plan to keep rates at zero, along with this strategy.  "A trillion dollars of QE" is useless as a communication tool, since quantity of money has so little relationship to inflation, so it would be much better if they talked about the path of rates.  I don't know what they plan to do or say, but at the very least, I don't think there is any inherent communication paradox in QE.

As far as helicopter money goes, reversal does require taxation.  But that's OK.  Helicopter money is negative taxation.  It's only natural to reverse it with taxation.  And it is a far more effective stimulus than bond/stock purchases since most of the recipients would be debtors rather than savers.

David Pearson:  You're absolutely right.  Advocating a helicopter drop is the same thing as advocating fiscal stimulus.  The only real economic difference between fiscal/monetary stimulus is who gets the money in the end.

OT: The threads on the hydraulic input and output lines were the same size on that car.  It was like a Stephen King novel.  Pure demonic possession.

"But you can't subsequently reverse it, if you need to, without tax increases."

Is that a bug or a feature? Stock purchases are easy to reverse, and if they're expected to be reversed, they may not have much impact on stock prices. (In particular, investors may have a model of the economy in which the inflation rate consistent with high-enough-to-produce-recovery stock prices is one that is too high for the Fed to tolerate.) Of course, the Fed can always overwhelm the market and force stock prices up, but that might require a huge increase in its balance sheet, which it might find unpalatable. The fact that the Fed can make more of a commitment in the helicopter drop case (because tax increases are more difficult than stock sales) is a way to overcome the "expectations trap."

Negative TIPS yields make my brain hurt. Why would anyone in sign-up for that?

The Fed has really made a dogs breakfast out of this. How many people will say to themselves: "Hey honey! The Fed is sorta maybe kinda trying to tinker with the long end of the yield curve ... let's go buy a washing machine!". It strikes me as ridiculous. They're living in their own little echo chamber and mistaking it for reality.

David: I tend to agree. I find it hard to see the difference between:
1. Fed helicopter drop
2. Fed lends money to Treasury, which drops it.
3. Fed lends money to individuals, who lend it to Treasury, which drops it.

It may make a difference to expectations of future policy. Is it Fed or Treasury which gives future guidance? And we do live in a world in which central banks are supposed to be in charge of AD, rather than departments of Finance. And this is the same answer I would give if you suggest Treasury buys stocks, instead of the Fed. Whose communications strategy should we be listening to? A communications strategy is more credible if there is one voice, and that one voice has both current and future responsibility for the policy in question. That means, given history, an independent central bank.

K: Actually, I'm not exactly sure just which bit of the term structure the Fed wants to operate on. And if it were successful, and expected to be successful, that would change the whole term structure again, in ways that are hard to foresee. Just precisely *when* will real income and inflation recover? Again, at a minimum, the whole communication strategy gets very muddled.

Andy: That's a topic for another post, that has been rattling around in the back of my mind for some time. The Fed wants a strong response to its entrance, but an easy exit. But the easier its exit is seen to be, the weaker will be the response to its entrance. I need to draw some supply and demand curves for the Fed's entrance and exit.

Patrick: Negative TIPS? So invest in canned food instead! But if the Fed announced a clear inflation target, I think people could understand that. "Look Honey, the Fed says stuff will cost more next year. Let's go buy a washing machine now".

Andy: Really good point.  Helicopter drops are extremely convincing.

Patrick:"Negative TIPS yields make my brain hurt."

Me too.  Especially in the absence of NGDP futures.  Low rates may still cause some people to invest, but consume?  I doubt it, until we see some serious real inflation - not just expectations thereof.  And that isn't likely to happen absent consumption.  I think it's a big mistake to confuse financial market inflation expectations with consumer expectations. And in this case, it may be the markets that are wrong, and consumer "real rates" may be much higher than the TIPS yield suggests.  


As a limiting strategy, the Fed could buy treasuries in order starting from the shortest maturity and working their way out the curve.  They announce that they are going to do this, and that when they are done, they are going to keep the short rate at zero out until the maturity at which they stopped.  Then, at every increment they buy that yield down to zero.  Each incremental purchase is stimulative so presumably it would cause steepening of the rest of the curve.  They stop when they like the shape.  There is a very coherent communication strategy:  The Fed chooses the first part of the curve.  The market chooses the rest.  Instead of setting the short rate, the Fed sets the end time of zero part of the curve.  I think the market could understand this very well: as usual they are watching one degree of freedom determined by the Fed.

I am not saying that the Fed wont muddle it.  I am just saying that there is no inherent inconsistency in communication of QE.  You do however, need to leave some part of the curve in control of the market, in order to prove the effectiveness of the stimulus.

Nick:"Look Honey, the Fed says stuff will cost more next year. Let's go buy a washing machine now".

I wonder.  I suspect consumers will react after they get burnt.  Not before.

Yes, we need helicopter drops. No, stock purchases are not the answer. What if we relax the assumption of EMH, and say that stocks are already very expensive, like in 2000 (or perhaps in 2010). This is what happens:
"The Fed will loosen monetary policy by buying stocks, this will raise money expenditures and real profits. Expectations of higher real profits will raise stock prices. Our presence as a significant buyer will depress the required return on stocks, even though it is already too low. Our success will increase the probability that our exit strategy will start soon, this will increase the required return on stocks." The result - stock prices are a very poor indicator of policy success.

This is what happened in Japan:
"The Ministry of Silly Walks will loosen fiscal policy by building bridges to nowhere, this will raise money expenditures and prices of construction materials. Our success will increase the probability that our exit strategy will start soon, this will depress the prices of construction materials." The prices of construction materials are a very poor indicator of policy success.

In all these four cases the real indicator of policy success is M3. In case of helicopter drops, M3 will increase automatically. In case of QE2, M3 will increase a little bit, and velocity of M3 will increase a little bit, but it is hard to make quantitative predictions. The same applies to stock based QE and infrastructure investments.

K, good points. This is what basically happens in Sweden. Swedish central bank publishes the forecast of short term interest path with every policy decision - they are doing permanent QE2 without any need to buy actual government bonds. And differences between the central bank interest rate forecast and actual market data are very informative - they give us a constantly updated market opinion about which parts of central banks forecast are suboptimal.

Nick, buying stocks has one very significant advantage vs. QE2: bond bubbles are much more dangerous than stock bubbles.

Although the Fed can't legally purchase stock directly, it could target stock prices, using bond purchases as a vehicle by which to manipulate them. (In fact, the Fed is normally in a position of using one market to manipulate another: it doesn't operate in the interbank lending market, but it uses the Treasury market to target the federal funds rate.) Stock-price targeting has interesting implications. If the Fed followed such a policy consistently in pursuit of its mandate, then private sector stock purchases (in aggregate) would represent bets on a weaker economy, as investors anticipated an increase in the Fed's stock price target. That should give us some degree of automatic stabilization.

I don't find the washing machine example very compelling: it's hard enough to get adults to believe in Tinkerbell over longer horizons, but at a one-year horizon, depending on the characteristics of the Phillips curve, she may fail to exist even if people do believe in her -- which makes it that much harder to get them to believe. If you're on a flat section of the curve, you have to move along it by convincing people that it's less flat than it actually is. Over longer horizons, it becomes more and more plausible that we will eventually reach a steeper section of the curve, and the Fed can promise more inflation. It won't make people buy washing machines sooner, but it could make them buy stock, and it could make businesses buy durable productive assets that otherwise would not have been profitable to buy.

Nick:"And we do live in a world in which central banks are supposed to be in charge of AD, rather than departments of Finance."

I agree.  So right now the Fed can't do helicopter drops. But this is not an argument against helicopter drops.  It's an argument for giving the Fed authority to do helicopter drops.  

What if the Fed just purchased 3-month T-Bills? Those rates are already at zero so nobody would judge the policy by whether or not yields went down. They could combine these T-Bills purchases with a commitment to move the TIPS breakeven inflation curve back to its 2004-2008 range (2.3%-2.5%). This would have the positive feedback effect without the legal constraints of stock index purchases.

Nick, As usual, very good post. There are lots of assets the Fed (in theory) could buy; stocks, real estate, foreign exchange NGDP futures. In each case the price would move in the "right" direction.

Andy Harless makes a very good point. As long as they can target an asset price that is observable in real time, they don't actually have to buy that asset. Indeed in my view they could buy zero interest T-bills (although that might not be enough.) Right now stock and foreign exchange prices are observable in real time. Inflation expectations are as well, although there are some questions about the accuracy of TIPS spreads, and I don't know much about the liquidity of the CPI futures market. There is no NGDP futures market. So it's one of those "lesser of evils" choices.

Regarding expectations (mentioned by several commenters), in my view it's not about the average guy or gal. The Fed needs to convince markets that it is aiming for a higher NGDP three years out (compared to what is currently expected.) If they can do that, stock, commodity and commercial real estate prices will rise right now. That will start the ball rolling, and the average guy will respond to higher asset prices, not to the Fed. (Or at the very low end, rising employment opportunities as higher asset prices lead to more capital formation.)

BTW, what's your take on the recent posts by Krugman? Doesn't he seem a bit unimaginative? To me it seems obvious that the markets are reading the whole QE discussion as an implied increase in the Fed's inflation target---or perhaps a commitment to hit the target, where previously that had been lacking, or perhaps a movement in the direction of level targeting. Krugman's usually got a pretty sophisticated take on expectations. I can't imagine why he keeps insisting it's all about maturity transformation. He's right that by itself that wouldn't do much.

"As long as they can target an asset price that is observable in real time, they don't actually have to buy that asset."

If we added one clause to that statement, "as long as they can target an asset price and have the power to change it ...", we would have the "bazooka theory." That theory has failed dismally in practice. Targeting something without the power to affect it amounts to the "air bazooka theory": pretending to have a bazooka when everyone can see that you do not. I think that Nick will like this view, because the air bazooka theory is a common thread in many of his posts. But advocating this theory amounts to claiming that pretending to have a bazooka will work better than actually having one. That is a doubtful claim. Why should manifestly empty threats have the power to change expectations?

Scott: "Indeed in my view they could buy zero interest T-bills"

This the same as doing repos, only the money comes back after a month instead of a day (repo is just QE on overnight government debt).  Even the Fed doesn't believe there is any point doing repos at the zero bound.  I can't see how buying T-Bills would be any different.

I just realized the real reason that we need NGDP futures.  It's not so that we and the Fed can observe the market expectation of NGDP.  It's so that the Fed can manipulate those futures in a big way.  Lets say they bought next years future up to 5%.  Then anyone with a project that depends on 5% NGDP can invest in that project and short the future.  In effect they can hedge out their macro exposure.  They don't have to take the Feds word that NGDP will grow.  It's more like arbitrage.  And if the economy does not produce 5%, investors will receive a direct stimulus from the Fed.  It would work in the other direction too.  The Fed could manipulate futures down and people would invest in futures rather than real projects.  Arbitrage ought to be a far more powerful force than manipulating expectations and gets us around the problem of Fed credibility.

And, Nick, yes it's a bit like buying stocks.  But it's much better because 1) it doesn't transfer wealth disproportionately to savers and big business and 2) it is way more transparent and direct both as a communication tool and as a hedging tool.

I agree with Phil. But even if they had a bazooka, and everyone saw they had a bazooka, it wouldn't do any good if everyone knew they wouldn't ever use it.

The Fed needs to take a lesson from Tony Soprano on credibility and managing expectations. Sometimes you need to back-up your threats with real muscle. Most of the time you don't have to resort to muscle if you have a reputation, but every once in a while somebody tries to call your bluff. That's when you need to remind everyone that you really are the baddest dude in town.

The specific machinery doesn't matter all that much so long you have the ability to force people to do what you want them to do. I was only 6 years old at the time, but didn't Volcker teach us that? The Volcker Fed said "we're going to stamp out inflation". Everyone said: "That's nice. I'll believe it when wages and prices stop rising". Then the CB applied muscle; they raised interest rates to 18% (or whatever it was) and everyone started believing them. To this day, does anyone doubt a competent CB's ability to stamp out inflation at will?

At ZIRP with high unemployment and idle capacity the CB can't call our bluff with interest rates, and all the wishy-washy talk of bending the yield curve and keeping rates low for a long time is not going to convince anyone of anything (at least not in any reasonable time frame). The Fed really needs to call our bluff and apply some muscle.

The problem with "targeting and not buying" is that markets front-run.

Let's say the 10yr is at a 2.6% yield, and the Fed announces that it will buy bonds to hit a target of a 2.2% yield. Traders will quickly vacuum up the bonds, so that the 10yr quickly hits 2.2%. At that point, Treasuries are "trading inventory", waiting to be unloaded on the Fed. Along comes the Fed and says, "well, the yield is at our target, so we won't actually have to buy many of those bonds." Traders, realizing that there is no big buyer for their inventory, dump the bonds, and the yield quickly rises back to 2.6%.

Something like the above has occurred in 10 and 30yr Treasuries in the past few months. This is an indication that the bond market is not giving QE much credibility, particularly since Dudley and other Fed officials have stressed that an important stimulative impact of QE is lower long term rates. What lower rates? They have just round-tripped, and they are at levels they traded at before the Bernanke Jackson Hole speech.

David Pearson: Yes. If the Fed promises to do QE, they may actually have to do it...

David Pearson makes a good point. It is a contradiction to hope to influence expectations and yet assume that people will not trade on those expectations.

But the Fed *wants* people to trade on those expectations.  It's not a problem!  And if they trade against the Fed's desires (i.e. rates go back to 2.6%), the Fed should punish them, by actually doing the QE they were threatening.

David and Phil: OK, I get your point.  You are saying that the QE yield target is inconsistent with resulting probable path of the short rate and that therefore the market wont respect the QE target.  There are two answers to this:

2) The Fed can overwhelm the market whenever it wants.  If it wants 2.2% yield, it gets 2.2% yield.

1) The Fed should never set curve shape targets that are inconsistent with the path of short rates.  That's why they should always only determine the initial part of the curve and let the market do the rest.

No, when the Fed wants a 2% yield, and that isn't a plausible yield, then what the Fed gets is not a 2% yield but no yield as it owns all the securities outstanding. Of course in reality the markets would stop being liquid long before then, and therefore they would stop transmitting expectations.

The index aggregates expectations, but changing the index will not change the expectation.

When you push too hard on the speedometer to speed up the car, you get a broken speedometer, not a faster car.

RSJ:  OK, but I said probable, not plausible. The distribution of plausible yields is very wide.  3yX3y swaption vols are around 30% - thats 1 standard deviation log-normal.  Two standard deviations is forward yield range of 1.6% to 5.4%.  And that doesn't even account for the smile, which is huge.  And if the Fed says the yield is 2.2% because 1) we're going to force it there and 2) that's going to be the path of the short rate, then it suddenly becomes a lot more plausible.  So no, the Fed can't put the yield anywhere.  But it can put it anywhere it *wants*, which is to say that it can use the curve as a very convincing way of asserting its intended path of rates.

The Fed can overwhelm the market whenever it wants. If it wants 2.2% yield, it gets 2.2% yield.

It is highly unlikely that the Feds has this kind of market power.

At the long end of the yield curve the Feds "power" is close to non-existent as opposed to the shell game on the interbank market.

Any possible change in the long term T-notes can be explained by psychological herd behavior rather than the Feds actual power.


Of course the Fed is not allowed to buy Treasury bonds at auction. No matter -- it just buys the bonds that the banks bought at auction.

The banks are not the major buyers at T-bond auctions, non-banks are.

My observation is that there seems to be a lot of talk about how QE gets "money into circulation", while the fact that the Fed will be financing Treasury spending seldom gets mentioned.

Your remark is based on the assumption that the same group of non-bank entities that have recently bought government securities from the Treasury would be willing to turn around and re-sell those securities to the Feds, not to someone else, on the open market.

That's a rather far-fetched proposition .

vjk:"It is highly unlikely that the Feds has this kind of market power."

Based on what?  What's going to stop it? Restrictive balance sheet limits I'm not aware of?  The most charitable interpretation that I can make of your comment is "The Fed doesn't currently exercise that kind of market power." Is that what you meant?

"At the long end of the yield curve the Feds "power" is close to non-existent as opposed to the shell game on the interbank market."

There's only $846Bn of outstanding treasury bonds (everything from 11 to 30 years). Do you really think the Fed couldn't control the long end of the curve? (Not that I think they want to.  It's the short end they want to control.)


The total bond market size is about $32 trillion. The global bond market size is about $100 trillion.

There is no special yield curve for Treasuries only.

Not that I think they want to. It's the short end they want to control

If they want to control the short end, why would they buy long term ? Does not make sense.

Besides, they can control the short end only upwards, its already being at 0-25 bps.

K, the Fed can offer to buy them at 2%, and they will buy all of them at 2%.

In the money markets, there is a certain demand for actual cash, both to settle transactions as well to prevent bank overdrafts or to meet bank reserve requirements. That demand is inelastic over the short run. There is no substitute for cash when cash is needed to settle a payment.

But in the capital markets, there are plenty of substitutes for long bonds, or agencies, or what have you. There is no requirement for the private sector to try to outbid the CB, whereas a bank short of reserves must try to outbid the CB. The private sector can just sell the asset to the CB. Sure, there will be front running and noise traders, but these are short term and unreliable as policy transmission mechanisms.

So the private sector does sell 800B in treasuries to the CB, but assuming that they were indifferent in holding those bonds, then 1/2 of the proceeds of the sale will be used to purchase put options and the other half will be used to purchase call options, and no new investment occurs, neither do asset prices rise. Broker Dealers end up sitting on a lot of cash that is deposited with banks, creating 800B in excess reserves.

In the capital markets, you are not going to get asset prices to rise by retiring assets -- you need to change the expectations of the market participants, so the 3/4 of them go long and only 1/4 go short. But trying to change the index that measures those expectations is not going to work.

"The total bond market size is about $32 trillion". And almost none of it beyond 10 yr.

"If they want to control the short end, why would they buy long term ? Does not make sense."

Do you know what they are going to buy? Last time they bought a lot of agencies. Duration generally around five years or less. And no, I wasn't proposing they buy t-bills at 25 bps. I think they are going to buy mostly less than 3-4 yr debt, as I've discussed above, in some of the first comments. If they start buying the long end they get into the conflict that Nick discussed in the post. It's highly counterproductive.

RSJ: "K, the Fed can offer to buy them at 2%, and they will buy all of them at 2%."

No, they won't. Only if no one believes that 2% is a reasonable average path of the short rate. In reality the range of opinions is extremely broad. There are plenty of people who expect us to fall into an extended liquidity trap, QE or not. And lots who think we'll get rampant inflation. And if the Fed sticks to targeting yields that are consistent with its intended path of rates, it's extremely unlikely that an overwhelming majority of the market will dissent. And if they do, the Fed has enormous power to punish them, either by manipulating yields or through the ultimate path of the short rate.


Do you know what they are going to buy?

I vaguely recall there was some talk about 10 years and above.

My understanding was they intended to lower interest rates at the long end of the yield curve.

That's futile for two reasons: the Feds cannot influence much rates beyond the interbank market rate; the outcome of the exercise is going to be even more bloated bank cash accounts at the Feds (which already stand at $1tril in "excess") and not much more.

So, it is unimportant whatever maturity they are going to buy.

A simple answer to the Feds inability to influence long terms is sheer capital market size (bonds combined with other substitutes as RSJ said). There other theoretical factors of arguably less importance predicting the Feds impotence to influence long term.

All of the above may be bunk so one may be wrong.

But, on the other hand, judging by the Japanese QE experience re. long term and the QE1 voyage re. the same, one may be right.


What is "the ultimate path of the short rate." ?

By the "ultimate path" I just meant the path that is ultimately followed (out of the multitude of possible paths).

vjk:"I vaguely recall there was some talk about 10 years and above"

So you had proof they're idiots, but you lost it...

"the Feds cannot influence much rates beyond the interbank market rate"

You repeating this won't make it true. I see no reason why the Fed can overwhelm the overnight market, but not the market for term debt. Especially since they have the power to discipline those who disagree with them through the setting of the path of the short rate (for the last time). You need to acknowledge this point.

"the outcome of the exercise is going to be even more bloated bank cash accounts at the Feds..."

I never said it was going to work. My contention is that there exists a QE strategy with a coherent communication strategy. I don't know if the Fed is going to follow such a strategy, but I'm not going to a priori assume they're stupid. But even if they do follow the best possible strategy, it may not be sufficient to pull us out of a this liquidity trap (assuming that's what it is).

"A simple answer to the Feds inability to influence long terms is sheer capital market size"

How about the potential sheer Fed balance sheet size?

The expectation hypothesis has been shown to be inconsistent with the data for about two decades now. From countless studies. It's hard to think of a statement that has more solid econometric backing than "EH is false".

Here is a reasonable survey: http://research.stlouisfed.org/wp/2003/2003-022.pdf

The fact is, you have _no_ _idea_ about what the actual relationship is between the short and long term rates. You have a theory that is contradicted by the data, but these contradictions are ignored for theoretical convenience. Actual markets are not disciplined by theoretical convenience.

Institutionally, the CB can control the zero rate because banks *must* outbid the Fed when they are short of reserves, and by draining, the CB has the power to place them in a net deficit position with respect to reserves. Therefore they will pay whatever the CB decides.

That is a very special case.

But no one is required to get into a bidding war with the CB for assets of maturity > 0. No one is forced to hold these assets, or can be required to have a short position in these assets. If the CB offers to overpay, then that is not discipline, it is free money, and those who hold the asset will sell as much to the CB as it is willing to buy. When it stops buying, then the CB loses control of the rates. If it buys all the assets, then still there will be an implied risk-free rate at that maturity, and you have no plausible mechanism to control this rate.

In the interest of full disclosure, I am long bonds, and I think they will fall, Japan style. The fed doesn't need to try to make them fall.

I am in the RSJ camp. His point is exactly why Sumner's idea of manipulating NGDP futures prices won't work. If the fed is prepared to make an unlimited market, it can indeed make the futures price whatever it wants, but there is no reason for the underlying NGDP to fall into line with the futures price. This is the general case for asset prices, not an exception.

RSJ, isn't there a substitution effect there? When the Fed buys treasuries and drives down yields, the spreads on risky bonds grows and eventually money flows there as well.

RSJ: I'm not familiar with the literature on the EH, but the EH as stated in the paper you cite, is a hypothesis that forward rates are literally expectations of future short rates under the historical measure.  They aren't.  This isn't even consistent with an arbitrage free rate model.  In an arbitrage free world, discount factors are the reciprocal of the expected value of a bank account under the risk neutral measure.  And the historical measure is related to the risk neutral one by an arbitrary conditional time varying drift.  This drift (as well as convexity effects resulting from a literal expectation of rates) can likely explain the rejection of a simplistic EH under an empirical study.  I don't think this is very interesting, and it certainly doesn't prove the existence of arbitrage in fixed income markets.

Anyways, when I use the word "expectation" I usually mean "in an arbitrage free model under the risk neutral measure".  So when I very loosely say "the yield is the expectation of the path of the short rate" I don't mean in a trivial econometric way.  It should really be read as "the price of bonds is determined by an expectation of a function of the path of the short rate".  And if the Fed specifically announces the path of the rate, then the market will arbitrage the yield to a level that is consistent with that path. Yes, the path has to be credible and it has to be respected, which means the Fed shouldn't try to lay out a 10 year curve.  It should start from the front and work its way out the curve as necessary.  

As far as your argument that the Fed has special power over the short rate, I need to think more about the mechanism.  But here is why I'm initially skeptical:  If the Fed's control over the short rate was just a technical trick and not a result of the unlimited balance sheet of the Fed, then the Fed funds rate would become disjoint from the real short term funding costs of arbitrage players.  This is not the case: arbitrage players have near unlimited access to funding at fed funds plus a small fixed spread.  This is because the banks are willing and able to source those funds in the fed funds market.  If equilibrium were achieved though an non market mechanism the banks would be either unwilling or unable to do this. Furthermore, yields and the short rate path are related by "expectations". This means that moves in one will cause significant flows/pressure on the other through carry trades. Over the long run, the average pressures in one ought to be about the same as the average pressures in the other. If the Fed has the power to directly effect the real funding costs of market participants then I think I think it ought to have the power to directly effect the market for expectations of functions of the future path of those funding costs. That's my gut feeling, but no, I don't have empirical evidence.  

Finally, I agree with Nick and you and most of the rest of the people here, that indiscriminate buying across the curve by the Fed (if that's what they choose to do) likely does not have a coherent associated communication strategy, and is therefore likely to be very ineffective.

Phil Koop:"Sumner's idea of manipulating NGDP futures prices won't work"

I said that, above.  But Sumner may have said it before me, somewhere else.

"there is no reason for the underlying NGDP to fall into line with the futures price"

Is too.  The reason is arbitrage.  People can invest in projects even if they 1) depend on a high level of NGDP, and 2) don't believe that level will be achieved.  They can do this by investing in the project and short selling the future at the high level of NGDP futures price caused by the Fed.  The Fed effectively guarantees the level of NGDP.  So arbitrage will lift NGDP to the futures level.  At least in principle.  I'm not so sure if it would work very well in practice.  But you can't say that there is no reason.


I think in the case of a Fed S&P target, Nick would argue that Fed buying creates private demand for the asset. In other words, the Fed will not end up owning all the stocks, because by buying stocks, it will make private buyers also want to buy stocks. The link between the S&P level and private demand for stocks is the "wealth effect" on private spending that raises expected corporate dividends.

So what's wrong with the above premise? By buying stocks, doesn't the Fed make private holders "richer"? Isn't there a "wealth effect" of rising equity prices on private consumption spending?

Of course, Fed-induced capital gains on stocks, all else equal, would support more current consumption by equity holders. The problem is that this increased spending is offset by the lower spending of savers that now face lower future returns from too-high equity prices. In effect, the Fed's purchases may merely be viewed as a transfer of wealth to current holders from future holders forced to accept lower returns. The wealth effect benefit to current holders is only half of the equation needed to calculate the impact on total spending.

Of course, the above all holds true for permanent Fed bond purchase programs. In fact, a Fed bond price target would be no different than a stock one. This is because, using a dividend discount model, it makes no difference whether you raise the price of stocks through asset purchases, or raise the "value" of the same by targeting a lower l.t. bond yield and thereby reducing the discount factor. The problem in each case is that future savers face lower returns, and therefore must save more to support the same future consumption.

There is an exception to the above: the special case where rising prices beget expectations of HIGHER future returns. This is what we had during the housing bubble. Actors saw rising house prices as evidence that housing capital gains would continue well into the future, so they needed to save less to support the same future income.


Agree with your comment on the real rate, however the inflation expectation will only increase if the policy is a success at reflating. i.e. Fed is pushing down real rate, if a success the real rate and the inflation expectation will increase. The real rate impact is messy but the inflation component will only go up if successful at reflating.

I'm losing track of the comments here. Just want a add a few points:

Whether or not the Fed actually has or needs a bazooka, that can do real things regardless of expectations, the communications strategy is a very important part of its arsenal. And a communications strategy that has that internal inconsistency can't work as well.

A promise, or conditional promise, to use a bazooka in future, if credible, can be as effective or more effective than using a bazooka today.

There are games, games of pure coordination, where a credible communications strategy is all you need. Cheap talk does the job. We all want to meet somewhere. Where shall we meet? A single voice saying "Grand Central Station" is all that's needed.

If people did expect recovery, then nominal and real interest rates would rise, and the equilibrium monetary base would fall. With a credible communications strategy, the Fed would actually need to point the bazooka in the exact opposite direction. In a sense, it's not 50% communications, nor 100% communications, it's 150% communications (if what you are communicating is interest rates and the monetary base).

The main effect of stock prices on AD is not via wealth effects. It's Tobin's Q. Higher real stock prices reduce the cost to firms of funding new investment via stock issuance. The P/E ratio on stocks is just as or perhaps more important than the dividend/price ratio on bonds, especially government bonds.


The effect of a lower cost of capital on investment is questionable in the current environment. We already have record low long-term real corporate borrowing costs. Assuming indifference between debt and equity, large corporations have every incentive to invest in new projects; and yet the three month average of core durables goods orders has been stagnant in 3q, and the ISM and regional survey new orders sub-components have been downright weak. Small businesses might benefit from a lower cost of debt, but their problem is likely to be more a lack of collateral than a high interest rate on loans.

Similarly, in the housing sector, we have record low nominal mortgage rates, and perhaps real as well. And yet housing starts are in the tank.

The two sectors of the economy most sensitive to capital costs are not reacting to lower rates/credit spreads. Clearly, other factors are at work besides the cost of capital for new projects. Would a lower equity cost of capital succeed where the lower cost of debt did not?

BTW, just to be clear -- corporate borrowing costs are low because credit spreads are at record lows, not just because Treasury real yields are low. The Fed has already had a major impact on the corporate cost of capital.

Could the Fed buy State Govt. bonds at negative interest to finance a stimulus?

Yes, if they want to subvert their mandate.  Eventually they'll lose money on the trade, and the treasury will have to make up the shortfall.  So it's just fiscal stimulus.


My contention is that there exists a QE strategy with a coherent communication strategy. I

Could you summarize the implied QE2 strategy other than the naive belief that profound structural economical ills, like industrial base erosion with attendant disappearance of production culture and rampant unemployment, can be cured by providing more measurement units in hopes of prolonging the agony through more asset bubbles ?

don't know if the Fed is going to follow such a strategy, but I'm not going to a priori assume they're stupid.

All the empirical evidence, starting from Greenspan's tenure, indicates that the said stupidity is an a posteriori conclusion rather than an a priori hypothesis.

Thinking a bit more about the Feds paper power to buy unlimited quantity of various "assets" I must concede that you are right in the technical sense.

Why the speculative rather insignificant, quantitatively, $500 bn initial syringe shot of QE2 won't have much of an effect has been discussed above.

If, hypothetically, the Feds decides to buy say ten times more, dollar collapse as a reserve currency, due to loss of trust internationally, followed by domestic unrest, due to Zimbabwe-like environment, is a much more likely outcome than lowering 10-30 year interest rate by more than a dozen bps.

Let's hope they are not that stupid.

One may be surprised though.

Nick: "The main effect of stock prices on AD is not via wealth effects. It's Tobin's Q."

I agree with this.  The wealth effects are just a massive side effect of the Fed buying stocks.  But it feels like TARP 2.  "In order to save this economy of yours we have to give these people a trillion dollars.  You may find it distasteful, but it's what's best for you."  But it isn't. How is it preferable to a helicopter drop which is both more direct and more fair. Is it just the political reality that getting stuff done requires vast amounts of pork for special interests?

K: Is that bit about "pork for special interests" right? Any recovery will increase share prices. If the Fed buys shares now, and sells them later, it will mean the Fed and hence the US taxpayer gets some of those profits. If the Fed does something else, the taxpayer will get nothing, and will actually lose money if the Fed buys bonds.

vjk:"Could you summarize the implied QE2 strategy other than the naive belief that profound structural economical ills, like industrial base erosion with attendant disappearance of production culture and rampant unemployment, can be cured by providing more measurement units in hopes of prolonging the agony through more asset bubbles?"

Well, when you put it like that... But seriously.  The strategy I was proposing was Oct 26 @ 4:08 PM (see above).  I don't even think it's necessarily going to be sufficient to pull us out of a death spiral, as discussed in subsequent comments.  But I do think it would be stimulative.  I.e. apart from feeding pointless asset bubbles, it would have a positive effect on real investments by lowering the threshold for profitability. Real investments bring real jobs.

"All the empirical evidence, starting from Greenspan's tenure, indicates that the said stupidity is an a posteriori conclusion rather than an a priori hypothesis."

Waahaha!  There is something inherently stupid about committees, I'll agree.

I don't agree that unrest follows from devaluation.  Unrest could follow from unemployment.  Devaluation, if it happens, will reduce unemployment.

Nick:"If the Fed buys shares now, and sells them later, it will mean the Fed and hence the US taxpayer gets some of those profits."

I think it's a bit wishful.  These are the scenarios I see:

Optimally: The Fed buys $500Bn of stocks.  As they are doing that stocks rise, slowly responding to the effect of the Fed purchase.  The market gets some of the gains; the Fed gets some of the gains.  Fed gets inferior risk adjusted returns on its investment compared to the rest of the market.

Realistically:  Market sees the Fed coming from a mile away.  Front runs the Fed for the full economic impact of the expected amount of the purchase.  Fed still has go through with the purchase in order that the effect is not reversed.  Fed gets little or none of the gain.

"But you can't say that there is no reason."

Oh yes I can. What you have described is not an arbitrage, it is a kind of statistical hedge: you hope that your losses on the project will be covered by the profits on your futures position. Of course, if your project turns out to be profitable after all, you expect to give up your profits. There are many things that can go wrong with this hedge: your project may fail for idiosyncratic reasons; you may have wrongly estimated your hedge sensitivity. It is therefore an expensive and risky way to earn the risk-free rate of return. That is not a reasonable description of an arbitrage.

Even in markets with traded underlyings, such as wheat, soy, or corn, futures and spot prices sometimes fail to converge; sometimes by large amounts. If a farmer can't count on a futures position to hedge his crop, how can you expect to hedge your project with a future in an untraded quantity like NGDP?


"the EH as stated in the paper you cite, is is a hypothesis that forward rates are literally expectations of future short rates under the historical measure."

No, that is the Pure Expectations Hypothesis. EH just says that the term premia are time invariant, whereas PEH says that they are zero. These are defined against the historical measure as you point out, using ratex, as that is the only way to test the hypothesis against historical data, at least AFAICT.

But if you *define* the term premia to be price of risk then EMH will always be true under the new risk-neutral measure, regardless of what the historical term premia actually are. You are using the definition of the measure as a fudge term to make PEH true by definition.

But by doing this you lose all predictive power about the relationship between short and long rates, which I guess EMH views as a feature. And you also lose all economic content as to what may drive the term premia.

In that case, you have no reason to believe that observed longer term yields "should" be lower if the market could be convinced that ZIRP will last longer. For all you know, the market is assuming ZIRP in perpetuity, and the current yields are the appropriate ones.

But now suppose that your fudge factor does not just measure perceived risk of capital loss in the historical measure, but is also a function of general economic outlook. I.e. suppose that the term premia are of the form

(psychological factors) + expected return on investment

where here 'expected return on investment' is using the historical measure.

And in that case, falling term premia in the historical measure correspond to falling perceptions of return on capital. As others have pointed out, corporate borrowing costs are low, but you do not see investment because the return on capital is also perceived to be low.

And in that case, you want to actually improve the return on investment, rather than trying to drive borrowing costs lower. If you try to drive borrowing costs lower, then you are fighting against the liquid capital markets and against EMH. You are in effect trying to convince investors to supply excess returns to firms.

On the other hand, if the government puts in a purchase order for 1,000 planes, then Boeing's return on investment *will* increase, and it will be higher than its borrowing costs, and then more investment will be made up until the borrowing costs are equal to the new (higher) return on investment.

Now you are not trying to get capital market investors provide firms with excess returns, but you are supplying excess returns to firms so that they will invest more until those excess returns are driven to zero. Now the process of arbitrage is working *with* you, to encourage more investment, rather than having the government fight the arbitrage process.

The situation of fiscal policy rather than monetary policy _is_ _not_ symmetric.

And as a final note, in the decomposition
term premia = 'psychological factors' + 'expected return on capital'

there is a form of zero bound, in that if the second term is zero, then still the first term will be non-zero and yields will not fall lower than that, regardless of how many assets the CB purchases.

"If people did expect recovery, then nominal and real interest rates would rise ..."

Is this true? I am asking honestly, not rhetorically. How do you know this? Is it a model prediction or have you analyzed historical data? It is plausible but empirically, it seems not to have been true during the last two decades. I mean, this is what the claim of contradictory communication comes down to: the assumption that interest rates, both real and nominal, are leading economic indicators. Personally, I don't think the market is as confused about this as you say. I think they will interpret the fed's message as follows: as long as actual economic data - employment, tax receipts, durable goods orders - keeps falling, keep buying bonds. Once this reverses, start selling.

Nick, in terms of the importance of expectations, I would be careful. It could may well be the case that the expectations were wrong prior to this crisis, when everyone was assuming endlessly rising real estate prices.

That could have been the tinkerbell effect.

But if the tinkerbell effect has to eventually revert to fundamentals, then real revenues and balance sheet improvement is needed now, not more tinkerbell.

In other words, we may now be in a rational equilibria, whereas previously we were in the enthusiastic equilibria. The psychological games in which rising optimism drives rising incomes wont work, and now we need rising incomes to drive optimism.

And in terms of how the CB controls the short rate, it is not a "trick", but the fact that money is not used up when it is invested, but goes from the bank of the buyer to the bank of the seller. The total amount of cash in the system remains unchanged.

From the point of view each individual bank, it is always better to have an interest bearing cash-equivalent than non interest bearing cash. If the banking system as a whole has excess reserves, then one bank tries to replace its reserve excess with a cash-equivalent, but that merely forces another bank to have excess reserves, so it also places a bid for a cash equivalent.

As a result, the yield on cash-equivalents continues to fall until the CB removes the excess cash by selling a security.

In the same way, a bank that is in a reserve shortfall must either sell a cash-equivalent or pay the penalty rate, so there is a flurry to sell by the _banks_, driving cash equivalents to the policy rate, at which point the CB adds back the cash by buying a security.

The CB does not overpower anyone, but merely tunes the parameters of the system so that the MM participants outbid themselves until the cost of cash equivalents is what the CB wants.

None of that applies in the capital markets.

RSJ: there can be multiple equilibria. That's what i think is happening right now. Both optimisim and pessimism can be rational, if self-fulfilling.

Phil: I'm mostly coming at this theoretically. The empirical support for procyclical real and nominal interest rates is more mixed. But that's because some recessions have been driven by deliberate attempts by central banks to raise real and nominal interest rates. like 1982.

Phil Koop: You're right. I shouldn't have said arbitrage. But if the residual risks are idiosyncratic and there is a large amount of them in the economy then it is almost surely as good as arbitrage. But I shouldn't have said it because the argument doesn't depend on it.

"Of course, if your project turns out to be profitable after all, you expect to give up your profits."

No. Assuming the hedge allows you remove the systemic risk factor from your project then putting on the hedge will leave you with a profit equal to the profit that you would have had if that risk factor had gone to the forward price you sold it at. If the Fed is willing to buy it from you at 2% above the equilibrium value then your profit will be higher by 2% times the number of contracts in your hedge. Or equivalently - your profit will be what you would have made if the systemic risk factor was 2% higher than equilibrium.

The fact that you don't know your systemic risk exposure exactly doesn't change anything. You still have an estimate of it. I.e. there is an amount (not zero) of futures contracts you could trade that will make you indifferent to the systemic risk factor for the combined portfolio of the project and the futures position. This is your risk minimizing futures position. It is your best shot at an idiosyncratic risk only position. Whatever this amount is, multiply it by the difference between the price of the future and your subjective expected value of the future. This is the amount of additional profit over your expected profit the market will pay you to take on your project. Now your expected profit on the project is exactly the same as if you were expecting the system risk factor to be equal to the futures price (but you have also minimized the systemic risk).

"your project may fail for idiosyncratic reasons"

Equally with or without the hedge. But if the hedge trades above your expectation you have an additional profit buffer.

"you may have wrongly estimated your hedge sensitivity"

Yes. But not knowing it, is not an excuse for estimating it to be zero. You still need to make your best estimate of your exposure. The test of deciding at what quantity you become indifferent to the level is an excellent way to pick the right amount of any asset.

"It is therefore an expensive and risky way to earn the risk-free rate of return."

It doesn't earn the risk free rate. You decide what profit margin justifies putting it on.

"That is not a reasonable description of an arbitrage."

No, it isn't. It's a variance minimizing strategy for generating maximum risk adjusted excess returns.

"futures and spot prices sometimes fail to converge; sometimes by large amounts"

At settlement? Because they don't settle on spot? This contract would settle on spot NGDP.


I disagree with a lot of the points you are making in the above comments (and man, you make a lot). But to the extent that we agree on the only issue of consequence, i.e. that fiscal measures are far more direct, fair and efficient methods of producing the required stimulus, I think the rest is becoming a bun fight (for which I take full responsibility), and I don't have your stamina. So, with the utmost respect, thank you, and till next time.



Since you may not have had the time to consider my truce yet, that gives me a last chance to do two things:

1) Take back the martingale measure argument.  That was totally lame.  It was really a red herring bun.  Of course, any term structure with non-zero forwards is potentially arbitrage free.  "Expectations", here, ought to mean under a measure that is related to the martingale measure by a reasonable risk premium.  And this is not the case for the curves that were studied in the literature you point to.  The differences were too large. So I was wrong.  I'd also like to

2) throw another bun.  I know, it's totally undignified after offering a truce, but then, I've had a nights sleep and feel reinvigorated. It's an audacious one and it's got jam on it, so I hope it'll stick:  

The whole "EH is false" literature is flawed (Longstaff).  It appears to be largely based on government bond yields unadjusted for liquidity premia.  And the most significant results were found in the T-Bill end of the curve.  So Longstaff repeated the analysis on the term structure of GC and found small and reasonable term premia (though T-Bills revealed the same strong bias over that period as was found in previous papers).  The vast majority of the previous results were simply specific to T-Bills, which are notoriously special and idiosyncratic.

Government bonds don't anticipate the overnight target.  They anticipate their own cost of funding. In the case of bonds longer than T-Bills, this is equivalent to their own specific term structure of repo.  Repo for on-the-run government bonds is very different from GC.  But you can't observe the forward repo curve.  And for T-Bills, you can't observe the convenience yield at all, except with reference to another risk free rate.  For that risk free rate, Longstaff looked at term GC.  If I were repeating the analysis today, I would probably look at overnight fed funds vs the term structure of OIS.  

You can't blame all those old papers for using government bond yields; they didn't have any other kinds of data.  But you can blame them for not realizing that the instruments from which they were constructing their curves had additional convenience yield over and above their yield to maturity.  The Thornton paper that you pointed to, by the way, carries out their analysis on on an updated version of the same government bond data used in the original Campbell-Shiller paper and apparently completely ignores the Longstaff criticism.  

K, the problem is that you can explain away any anomaly you like with convenience yields. The question then becomes are those convenience yields efficient themselves. And they are clearly not.

K asks: "At settlement?"

Yes, in the settlement period. On the same days on the same Illinois river loading facilities. See, for instance, http://econpapers.repec.org/paper/agsaaea07/9951.htm.

As it happens, Jim Bianco has just put up a post on Ritholtz that is germaine to the present discussion: http://www.ritholtz.com/blog/2010/10/pomo-still-matters/. So there is one fund manager's view. The money quote:

The Federal Reserve believes in the portfolio balance theory. This means it does not matter what they buy with newly printed money (QE2). The market will arbitrage this new money into the market that it thinks it will have the most impact. So do not get hung up on what the Federal Reserve is buying, but rather think about where these new dollars matter the most.

Right now the markets think that newly printed dollars will benfit “risk on” markets like stocks. So, Treasury purchases are a conduit to the “risk on” markets. Treasuries will still respond to the ups and downs of the economy like they always do. Stocks, on the other hand, respond to the perceptions of newly printed dollars from the Federal Reserve and other larger macro themes (like fear or relief of a double dip recession) than the more traditional fundamentals.

@Phil Koop

NGDP futures will be cash settled. You won't be required to deliver hamburgers and entertainment services for settlement :)

TMDB:"the problem is that you can explain away any anomaly you like with convenience yields."

I don't agree.  Securities may have hidden convenience yields.  But in the case of government bonds, they aren't hidden at all.  The difference between on-the-run repo and GC is huge.  Often several percent.  That is a convenience yield.  Repo is the cost of owning the bond; not GC.  

Swaps can't be used for settlement or collateral.  They can't be repoed. They have no economic value other than their explicit payoff.  Therefore we wouldn't assume that swaps have hidden convenience yields.  Swap yields also don't exhibit non-economic returns.  

"are those convenience yields efficient themselves. And they are clearly not."

Why not?  Repo is below GC.  But that's not inefficient.  Because all you can do with a repo loan is buy a particular government bond that yields correspondingly less than the equilibrium risk free rate.


Repo is below GC

What do you mean ?

Repo is a contract to buy-back a security, and GC is a security(generic collateral) -- you cannot say "repo is below GC" -- it does not make sense.

What you are probably talking about is "on the run" repo rate vs. GC repo rate, i.e. repo spread.

But it's not clear.


Re: EH and Longstaff.

As I remember, Longstaff conclusion was that EH is not rejected for extremely short-terms (neither is it confirmed). EH was strongly *rejected* for tenors > couple of months.

K, I'm not denying that convenience yields exist, as they obviously do. But EH + convenience yields doesn't really work. For dramatic purposes, let me take an example from equities. In November 2008, Volkswagen was by far the best performing stock in DAX. Yet EH was true, as any deterioration in earnings outlook was offset by much more dramatic increase in convenience yield. EH becomes a useless truism if you introduce convenience yields.

BTW, what is your opinion about TIPS anomaly in late 2008, there was a heated debate about it here one month ago.

GC is indeed a category of securities.  It is also a type of repo loan in which the borrower gets to decide which government bond to deliver as collateral.  GC is also the name for the rate of a general collateral repo loan.  If you call for GC you don't know what government bond you are going to get as collateral - but you can be sure it wont be on-the-run; it'll be whatever's most available/cheapest to deliver.  If I call for a (reverse) repo on a particular bond, I'm going to get that bond, and receive the repo rate (typically less than GC) associated with that bond.  That's the cost of demanding a particular bond.  If you want, you can call any repo desk and ask them "where's GC?"  The answer is going to be a rate. But if you ask "where's the repo rate?" they'll ask you "for what bond?"


The debate over the EH began with my assertion that because the Fed controls the path of the short rate, it has enormous power over the near end of the curve.  This led to the response that EH is false.

But as far as I can tell, EH is false only with respect to the headline yield.  But who cares about the headline yield.  It has nothing to do with the total return of holding the asset.  All that should matter in any economic debate is whether or not the dynamics of the path of the short rate determines  the return of risk free instruments to within a reasonable term risk premium.  And I believe this to be true for government bonds.  And people will evaluate alternative investment opportunities, not with respect to the headline yield, but with respect to the expected total return of government bonds.

I didn't look at the TIPS anomaly.  But will try to have a look at it tonight.

vjk: "EH was strongly *rejected* for tenors > couple of months."

They only tested up to the end of term GC, i.e. 3 months.  They could not reject the expectations hypothesis, not even its pure form:

"In fact, except for the one week term repo rate, we cannot reject the hypothesis that the term premia in repo rates are zero."


It is also a type of repo loan

Did not know they call it just GC nowadays.

We used to say "GC repo" vs. "special repo"

But that was > 10 years ago.

Times change.

Re. Longstaff:
"1. The EH is strongly rejected by the data. The slope coefficient of both individual bonds and
aggregate bonds regressions is significantly lower than one. In one of the two specifications,
the deviations are so large that, in three cases out of four, the current slope of the term
structure of special repo spreads is negatively correlated with future changes in long term
special repo spreads. High relative levels of long term repo rates overestimate the extent to
which a bond will remain special and/or the extent of the specialness."


The whole foundation of EH (all four or is it more ? varieties) is based on insane assumptions.

Perhaps, the whole edifice of financial "engineering" is (Brownian motion, normality assumption, gaussian copula, Ito lemma, etc), but one would not go there -- too distracting. Ask Scholes about LTCM. He should know.

Funny people still believe in this stuff.


And to kick the dead horse one last time:

Using Longstaff’s (2000) primary method on in-sample data (but from an alternative
source) from May 1991 to July 1997 that covers most of Longstaff’s sample period, we reject the
expectations hypothesis for one- and two-week only. When using pre-sample repo rates from
February 1984 through May 1991, we reject the expectations hypothesis for each term.

No EH evidence in the short end either, not surprisingly.

As I said: "All that should matter in any economic debate is whether or not the dynamics of the path of the short rate determines  the return of risk free instruments to within a reasonable term risk premium. " The substantive arguments I've been making here depend on this alone. Someone else equated this to the EH in order to discredit it. Then I went back to the literature to see if the "EH is false" papers found evidence of excessive risk premia. They don't (they are however fraught with flaws), and neither does the paper you quote (which is not Longstaff - It's Buraschi and Menini (2002)). As far as the second quote, you haven't given me a reference for that either, so tough to comment.

As far as the foundations of "financial engineering" by which I assume you are referring to arbitrage pricing, it doesn't include any of the versions the EH I've come across. But then nor is there anything foundational about normality assumptions, or the Gaussian copula. Ito, I'll take, though.


It's a fun discussion! I have many opinions, but I'm not wedded to them. Often times I've found myself vehemently arguing the exact opposite belief I espoused previously.

I'm behind in reading the comments. The paper you cite, which I only looked at briefly, has a term structure that goes out 6 months, so you are basically talking about the money markets. Leaving aside for the moment that the original topic of discussion was longer term maturities, I don't think you would see data for EH at the very short end, even if EH were true.

The CB is following some Taylor rule that is well known. And the economic data that the CB uses is also known. That means that for very short time scales, the market has a pretty good chance of guessing whether there will be a cut or a hike at an FOMC meeting, whose dates are also known. Even if you don't believe in ratex, the CB is fairly predictable. Suppose that everyone knows that rates will be cut from 3% to 2% next week, and that rates would remain at this level for at least a month.

You would expect 1 month bills to be lower than 1 week bills -- an inverted yield curve. But if that actually occurred, then banks would be able to sell 1 month paper, which would be discounted against the lower bill, even today. And selling 1 month paper is about as good as selling 1 week paper in terms of obtaining funds to meet reserve requirements.

That means that their marginal cost of reserves is less than 3% today. So either you are going to charge a serious credit risk delta for the 1 week paper over the 1 month paper, which is not plausible, or the CB cannot control the marginal cost of reserves -- which is also not plausible -- or the 1 month bill is not lower than the 1 week bill.

And if you want to argue that the 1 month bill wont be lower because of a positive time invariant term, then either increase the rate cut or change the above argument to 2 week paper, etc.

I am not saying that the 1 month bill can't move, but it's not going to move in a way that confirms EH. I would be surprised if the data showed that it did.

The notion that a CB can have a predictable known path for interest rates is inconsistent with forward looking arbitrage-free asset pricing. The fact that CBs set short term rates means that someone in the economy is earning an arbitrage-free profit.

Money Market participants can be forced to take any rate, even during a period of high inflation. The CB is the monopoly supplier of cash. It sets the price, and they borrow as much as they find economic to borrow at that price. And if they don't want to lend at that price, well the alternative is a bank account. If they had funds that they didn't need to store in the form of cash equivalents, then they would be investing those funds in the capital markets. I don't like this system, and it's not necessary that the CB manipulate short term yields this way, but that is how it works.

But that monopolist power diminishes as you go out the term structure, at which point you are subject to capital arbitrage. I'm not required to pay "the market rate" for capital in the capital markets since I can make my own capital. Alternately I can refrain from re-investing and lend out my funds. And the effect of that freedom is that the market price of capital is equal to the expected return of deploying an additional unit of capital -- i.e. it is set by profit expectations based on the overall economic outlook. The capital markets are connected to the underlying performance of the economy via this arbitrage.

There are channels by which low short term rates can affect the long term rates, but these channels rely on supplying rents or imposing fees on the financial sector.

That shouldn't be surprising -- it's pretty obvious that finance is earning rents, particularly since the triumph of monetarism. But even there the amount of rents earned are finite, again because real economic forces will step in and either the risk premium of the firms are increased, or inflation forces the CB policy rate to be increased, or so many rents are extracted from productive investment that overall outlook declines up until no arbitrage is possible. The problem is that you don't know what the adjustment mechanism will be. Whatever the adjustment mechanism is, any influence on the long term rates will be via the changing expectations of economic performance as that adjustment occurs, not by EH.

OK, re: the larger point of whether short end dynamics control the long end, where we relax EH to just be the short end plus some function, then I think that its straightforward to argue that if the CB decided to hike the short end to 5% and keep it there indefinitely, then the long end would also rise to at least 5%. You would see a massive contraction in borrowing, a financial crisis, rapid deflation -- meaning real rates would rise even more, etc.

But the situation is not symmetric. For example, if the short rates are low, and believed to remain low, and this supplies more NII to banks, who pass low mortgage rates onto households, who invest more and this increases agg. demand and output and expected profit opportunities, then the long end would go up. You could say -- ex post -- that the long end went up because of anticipated rate hikes to stem inflation expectations that were the result of low rates.

And it would be observationally equivalent.

If, on the other hand, we have a Japan scenario, and the banks receive more NII but investment does not increase, the economy stagnates, outlook worsens, and long term yields continue to fall, then you could also say, ex post, that the market anticipated the policy would fail and it anticipated future low short term rates, and that this was why long rates fell. Also equivalent.

Both of the above outcomes have been historically observed, even if you disagree with the causality.

To resolve a debate about observationally equivalent outcomes, we have to discuss the plausibility of mechanisms.

For what reason would the long rate respond to a CB communications policy? What is the "CB communications channel"?

If you are relying on a NK model -- and I'm only superficially familiar with those -- then it seems that they just assume that the interest rate is the short rate, and so implicit in the assumption of that model is PEH, not EH, because they want this rate to go into the euler equation. The marginal cost of reserves does not go into anyone's euler equation.

As far as the non-financial sector is concerned -- do the NK models have a financial sector? -- the call money rate is invisible; the long term rates should go into the euler equation.

So critical to the centrality of cb communications policies in these models is the assumption of "strong" CB control of long term discount rates, in the sense that such control bounds the maximum possible credibility of the strategy. If, in your premium function, you have a term equal to something that the CB does not control, and cannot minimize, then you are putting an upper bound on the effectiveness of the communications strategy channel. I argued previously that economic outlook is in that adjustment term.

You can still argue that, forgetting about the short rate, the CB can overwhelm the market in the longer rates, and I think we all made our points there.


What's wrong with Ito's Lemma?


I have a hard time taking "cb communications strategy" seriously as a policy response to mass unemployment. Come on, don't you feel a little embarrassed? Is it just me?

TMDB:"EH becomes a useless truism if you introduce convenience yields."

I've been thinking about that point ever since my comment on Scott's blog a few days back, and his response.

The simple fact that currency is (normally) rate of return dominated by bonds says we are willing to pay for liquidity. So we have to incorporate "convenience yield". We can't avoid it. But it's unobservable of course, so there is a real danger of EH (or EMH) becoming a useless truism. But even if we can't observe convenience yield directly, if we can have some sort of theory about convenience yield, then the joint hypothesis of EH (or EMH) plus that theory of convenience yield now becomes testable. No?

TMDB says: "NGDP futures will be cash settled."

No kidding. But that fact does not improve prospects of convergence. Remember, the premise was that the traded asset is going to be artificially fixed to some arbitrary value, and the non-traded asset is the "price" you would like to move.

If you believe in K's "arbitrage", you can put it on today, because it is the same plan as hedging an individual stock with the index. Even better: both sides are traded assets. And if you are attributing magical powers to financial futures settlement, you can execute your arb that way too. Feel free to look down and laugh at me from the window of your private jet once you are wealthy beyond the dreams of avarice.


Re. the second quote:

Further analysis of the expectations hypothesis using very short-term rates

Craig R. Brown et al, Journal of Banking and Finance, 2008

Using Longstaff’s methods on a sample of repo rates that pre-dates Longstaff’s sample, we reject the expectations hypothesis for every


I have a hard time taking "cb communications strategy" seriously as a policy response to mass unemployment.

Don't we all ? A gesture of desperation, perhaps

Come on, don't you feel a little embarrassed? Is it just me?

I am puzzled by your being embarrassed. Are you one of the twelve ?

P.S. I still cannot figure out the source of profits in the message you've posted elsewhere.


Nice to have you back.   Re longer maturities:  I think it would be statistically difficult to reject EH out beyond 6-12 months. You don't have enough independent observations. As far as the term structure of tbills goes, I wasn't able to follow your argument. But I'll say this: I think it's almost impossible to say anything useful. There are too many idiosyncrasies, and you can't even bring a mispricing into line via arbitrage since you can't short a tbill. And they have huge hidden convenience yields. If you want a summary, read the Longstaff paper. Thats why you need another instrument. I would recommend OIS vs fed funds if someone wants to publish a good paper (maybe it's already out there).  

Money markets are indeed a shell game. The musical chairs version in this instance. They should be used for liquidity only, not credit, but that's not how it's set up. If you die if you can't roll your paper you shouldn't be there. 

On finance extracting rents: now you are onto something deep and profoundly important. It deserves a whole blog (not to mention revolution) of its own, and I'd like to do that some day (the blog that is). 

I'll get to some of your other points later today hopefully. 

What's wrong with Ito? I have no idea. That was my point too. 

Nick: that's exactly how I see it. And, as you said, it's about the EMH which really at the core of what I've been saying above. 

Phil Koop: "No kidding."  Laugh of the day. Thanks. 

vjk:  A version of the EH was wrong 25 years ago. 

"As far as the term structure of tbills goes, I wasn't able to follow your argument."

I'll try again :)

Assume the marginal cost of reserves is set by policy at 3%, but in 1 week it is known to fall to 2% and stay there for more than a month.

Let z_1 be the yield of 1 week financial paper and let z_2 be the yield of 1 month financial paper.

Selling 1 month paper is just as good, from the point of view of the bank, as selling 1 week paper. Both allow the bank to make up for a reserve shortfall, and both will be rolled over for 2%. It will sell whichever paper is cheaper. Therefore the marginal cost of meeting a reserve shortfall is the minimum of (z1,z2).

But that marginal cost is set at 3% today. 3% <= min (z1,z2)

Therefore the short yields can't be forward looking. z_2 is not going to fall -- it will be stuck at the policy rate until the rate is changed. Unless you ascribe radically changing risk premia, then the bill term structure wont change, either.

Arguing that yields anticipate the future Cb movements effectively says that the CB cannot directly control yields, and on the short end it can. You can't apply EH on the short end. You can only try to apply EH farther on out the curve.

Only in the special case where the CB rate is X, and markets expect future CB rates to remain at X, can you (vacuously) argue that short rates are set by both current policy and expectations of future policy. So Brown concludes "Our results imply that expectations hold when rates are less volatile and/or that we may be entering a period of lower volatility."

Under pure EH z1 is 3% and z2 is (1x3% + 3x2%)/4=2.25%. (4 weeks in a month day count basis). Assume z1 is 3%. If z2 is less than 2.25%, they should sell the 1 month bill (it's earning 3% for a week, but less than 2% for the following 3 weeks). In the following three weeks they will roll the 1 week bill at 2%. If z1>2.25% they should sell the 1week bill instead. Does this not make sense? What were you assuming for the value of z2?

Oops: "if z1>2.25%" should have been "if z2>2.25%"


Any value of z_2 < 3% is a violation of the assumption that the CB can control bank's marginal cost of reserves, so I just made it a 1 month bill and didn't average :)

If EH were true, then a bank $1 short of reserves can borrow that dollar for $1.025, in which case why would it put a bid in the interbank market for $1.03? Why would the interbank market trade at $1.03 if banks could tap the MM and obtain funds more cheaply, and with longer maturities? So the IB market is a floor underneath the positive term rates.

If you are interested in the mechanism, then one bank sells a z_2 for $1.025, but that places another bank into a reserve shortfall, so it sells a z_2, which places another bank into a shortfall, so it sells a z_2, etc. Non-bank investors, even if they believe that the z_2 is worth more, are facing an infinite supply of z_2 whenever the yield on z_2 < the yield on z_1. But of course they do not have infinite borrowing power to arbitrage the deviation between EH and supply/demand because they are not going to be able to borrow to buy z_2. The money markets is where you borrow.

The very short end of the curve is where you can't arbitrage and these supply/demand constraints reflect the supply of cash by the CB, not expected return.

And this is not to say that you can't construct futures markets that effectively bet on future policy rates, and that EH would not hold in those markets -- it would!

But no one is going to be able to borrow cash at those rates, because the CB is the one that supplies cash and the marginal cost of borrowing cash is only going to fall when the CB says it will fall, not when the CB futures markets anticipate it falling.

The mechanism by which futures prices can force spot prices to move consists of investors taking a stock of inventory off the market -- e.g. stockpiling -- or releasing a stock of inventory from the stockpile.

But in the cash markets, all the cash is in the banking system. You cannot "stockpile" it out of the system -- not in any practical way. Private sector arbitrageurs can't take any inventory out of the market, or add to the stock of cash available to be lent. The CB is the only one that can do that.

Again, that is true for cash, not capital. For longer maturities, the CB does not have this power.


...(last part!)

For rate hikes, assume that z_1 = 2% and that the EH value of z_2 should be 2.75%, since there will be a rate hike in one week, that will last at least a month.

Now, banks can borrow (from each other) in the interbank market at 2%, but lend (to each other) at 2.75%. That doesn't make sense! They will bid down the yield so that they can't arbitrage between z_1 and z_2, taking a term premia into account.

And note that as long z_2 > z_1 + term premia, banks never "run out" of money to borrow from each other or lend to each other. They have infinite firepower.

So PEH says that z_1 = 2% and z_2 = Expected (z_1) + 0

But the market equilibrium price will be: z_2 = current_(z_1) + term premium

(with no assumptions about what the premia are)

The difference between these two hypothesis will be small when 3/4*(rate hike) is small in comparison to the term premia.

I think this is why Brown argues that he can reject zero premia in periods of high IR volatility but cannot reject zero premia in periods of low IR volatility. If there are other studies that show term premia of 30bp or so, then it's certainly plausible that EH will not be rejected given rate hikes or rate cuts in the 50bp range.

What's wrong with Ito's Lemma?

The Wiener process assumption.


Oh, now my head hurts! I vaguely remember that the jumps don't have to be normally distributed, but that things can get ugly if the semi-martingale is discontinuous -- maybe that's the assumption? Does just everything collapse without normality? Still you at least can make various simple arguments about arbitrage, even if only to point out that arbitrage occurs :)

p.s. -- I made a comment over at Bilbo's re: profits.


I do not remember very well Ito-Tanaka extension re. jumps so I cannot comment.

Another thing that bothers me deeply is continuous time assumption associated underlying fin math and empirical evidence that BSM never worked quite well requiring a lot fiddling to "fit" actual market pricing. Taleb goes as far as to say that no trader ever used BSM in real life to price options.

Phil Koop:"But that fact does not improve prospects of convergence."

I dont know what you mean. If you sell the future at 5% and it cash settles at 2% then you will earn 3% of profit. Doesn't matter if it gradually converges to 2% or never even trades anywhere near there.

RSJ:"Now, banks can borrow (from each other) in the interbank market at 2%, but lend (to each other) at 2.75%. That doesn't make sense! "

Are you suggesting that interbank lending has to have the same yield for all terms?

"Are you suggesting that interbank lending has to have the same yield for all terms?"

No.. by IB lending I mean the overnight markets.

But what is the difference between an overnight loan that you know you can roll over as often as you want (since the rate is set by policy, and the policy is known) with a non-zero term loan?

The comments to this entry are closed.

Search this site

  • Google

Blog powered by Typepad