Saturday, November 15, 2014

Sign Wars

Does a lowering of a central bank's interest rates create inflation or deflation? Dubbed the 'Sign Wars' by Nick Rowe, this has been a recurring debate in the economics blogosphere since at least as far back as 2010.

The conventional view of interest rate policy is that if a central bank keeps its interest rate too low, the inflation rate will steadily spiral higher. Imagine a cylinder resting on a flat plane. Tilt the plane in one direction —a motif to explain a change in interest rates—and the cylinder, or the price level, will perpetually roll in the opposite direction, at least until the plane's tilt (i.e. the interest rate) has been shifted enough in a compensatory way to halt the cylinder's roll. Without a counter-balancing shift, we get hyperinflation in one direction, or hyperdeflation in the other.

The heretical view, dubbed the Neo-Fisherian view by Noah Smith (and having nothing to do with Irving Fisher), is that in response to a tilt in the plane, the cylinder rolls... but uphill. Specifically, if the interest rate is set too low, the inflation rate will jump either instantaneously or more slowly. But after that, a steady deflation will set in, even without the help of a counter-balancing shift in the interest rate. We get neither hyperinflation nor hyperdeflation. (John Cochrane provides a great introduction to this viewpoint).

Many pixels have already been displayed on this subject, about the only value I can add is to translate a jargon-heavy academic debate into a more finance-friendly way of thinking. Let's approach the problem as an exercise in security analysis.

First, we'll have to take a detour through the bond market, then we'll return to money. Consider what happens if IBM announces that its 10-year bond will forever cease to pay interest, or a coupon. The price of the bond will quickly plunge. But not forever, nor to zero. At some much lower price, value investors will bid for the bond because they expect its price to appreciate at a rate that is competitive with other assets in the economy. These expectations will be motivated by the fact that despite the lack of coupon payments, the bond still has some residual value; specifically, IBM promises a return of principal on the bond's tenth year.

Now there's nothing controversial in what I just said, but note that we've arrived at the 'heretical' result here. A sudden setting of the interest rate at zero results in a rapid dose of inflation (a fall in the bond's purchasing power) as investors bid down the bond's price, followed by deflation (a steady expected rise in its value over the next ten years until payout) as its residual value kicks in. The bond's price does not "roll" forever down the tilted plane.

Now let's imagine an IBM-issued perpetual bond. A perpetual bond has no maturity date which means that the investor never gets their principle back. Perpetuals are not make-believe financial instruments. The most famous example of perpetual debt is the British consol. A number of these bonds float around to this day after having been issued to help pay for WWI. When our IBM perpetual bond ceases to pay interest its price will quickly plunge, just like a normal bond. But it's price won't fall to zero. At some very low level, value investors will line up to buy the bond because its price is expected to rise at a competitive rate. What drives this expectation? Though the bond promises neither a return of principal nor interest payments, it still offers a fixed residual claim on a firm's assets come bankruptcy, windup, or a takeover. This gives value investors a focal point on which they can price the instrument.

So with a non-interest paying perpetual bond, we still get the heretical result. In response to a plunge in rates, we eventually get long term deflation, or a rise in the perpetual's price, but only after an initial steep fall.  As before, the bond's price does not fall forever.

Now let's bring this back to money. Think of a central bank liability as a highly-liquid perpetual bond (a point I've made before). If a central banker decides to set the interest rate on central bank liabilities at zero forever, then the purchasing power of those liabilities will rapidly decline, much like how the cylinder rolls down the plane in the standard view. However, once investors see a profit opportunity in holding those liabilities due to some remaining residual value, that downward movement will be halted... and then it will start to roll uphill. Once again we get the heretical result.

The residual claim that tempts fundamental investors to step in and anchor the price of a 0% yielding central bank liability could be some perceived fixed claim on a central bank's assets upon the bank's future dissolution, the same feature that anchored our IBM perpetual. Or it could be a promise on the part of the government to buy those liabilities back in the future with some real quantity of resources.

However, if central bank liabilities don't offer any residual value whatsoever, then we get the conventional result. The moment that the central bank ceases to pay interest, the purchasing power of a central bank liability declines...forever. Absent some residual claim, no value investor will ever step in and set a floor. In the same way, should an IBM perpetual bond cease to pay interest and it also had all its residual claims on IBM's assets stripped away, value investors would never touch it, no matter how low it fell.

So does central bank money boast a residual claim on the issuer? Or does it lack this residual claim? The option you choose results in a heretical result or a conventional result.


What does the data tell us, specifically the many cases of hyperinflation? As David Beckworth has pointed out, the conventional explanation has no difficulties explaining the Weimar hyperinflation; the Reichsbank kept the interest rate on marks fixed at very low levels between 1921 and 1923 so that the price level spiraled ever upwards. Heretics seem to have difficulties with Weimar—the deflation they predict never set in.

Here's one way to get a heretical explanation of the Wiemar inflation. Let's return to our analogy with bonds. What would it take for the price of an IBM perpetual bond to collapse over a period of several years, even as its coupon rate remained constant? For that to happen, the quality of the bond's perceived residual value would have to be consistently deteriorating. Say IBM management invested in a series of increasingly dumb ventures, or it faced a string of unbeatable new competitors entering its markets. Each hit to potential residual value would cause fundamental investors to mark down IBM's bond price, even though the bond's coupon remained fixed.

Now assuming that German marks were like IBM perpetual bonds, it could be that from 1921 to 1923, investors consistently downgraded the value of the residual fixed claim that marks had upon the Reichsbank's assets. Alternatively, perhaps the market consistently reduced its appraisal of the government's ability to buy marks back with real resources. Either assumption would have created a consistent decline in the purchasing power of marks while the interest rate paid on marks stayed constant.

Compounding each hit to residual value would have been a decline in the mark's liquidity premium. When the price of a highly-liquid item begins to fluctuate, people ditch that item for competing liquid items with more stable values. With less people dealing in that item, it becomes less liquid, which reduces the liquidity premium it previously enjoyed. This causes the item's purchasing power to fall even more, forcing people to once again turn to alternatives, thus making it less liquid and igniting another round of cuts to its liquidity premium and therefore its price, etcetera etcetera. In Weimar's case, marks would have been increasingly replaced by dollars and notgeld.

So consistent declines in the mark's perceived residual value, twinned with a shrinking in its liquidity premium, might have been capable of creating a Weimar-like inflation, all while the Reichsbank kept its interest rate constant.


That's not to say that central bank liabilities do have a residual value and that the heretical result is necessarily the right one. Both possibilities make sense, and both can explain hyperinflations. But to determine which is right, we need to go in and do some gritty security analysis to isolate whether central bank money possesses a fixed residual claim on either central bank assets or future government resources. Parsing the fine print in central bank acts and government documents to tease out this data is the task of lawyers, bankers, historians, fixed income analysts, and accountants. And they would have to do a separate analysis for each of the world's 150 or so central banks and currencies, since each central bank has its own unique constituting documents. In the end we might find that some currencies are conventional and others are heretic, so that some central banks should be running conventional monetary policies, and others heretic policies. 

In closing, a few links. I've taken a shot at a security analysis of central bank liabilities in a number of posts (here | here | here), but I don't think that's the final word. And if you're curious how the Weimar inflation ended, go here.


  1. JP: "Now let's imagine an IBM-issued perpetual bond."

    I think a better analogy would be an IBM stock, that pays no dividends. But IBM could use its profits to repurchase stock.

    Then imagine an IBM stock that can yield a rate of return that is below the market, or even very negative, but people still want to hold it because it is so very liquid. So the stock would have positive value even if IBM's profits were always negative.

    1. I think they call that $AMZN. ;-)

    2. I think you can use stock as an analogy too. I like the bond analogy since I don't have to get into the whole seniority thing. (Fed notes are senior to Fed stock, IBM perpetual bonds are senior to IBM stock). Also, I think it better illustrates changes in interest rates. If a stock's dividend is cut (all else staying the same), that should theoretically have no effect on the stock price since the profits are just retained and compounded for later dividends. But if a bond's coupon is cut, then the bond will have to fall in value since there is no way of getting the lost interest back.

    3. JP: If a stock's dividend is cut (present value of profits staying the same) that should have no effect on the stock price right now, but should cause an increase in the growth rate of the stock price. You get capital gains from stock buybacks, instead of dividends. And if we measure all other prices in terms of that stock, that means an increase in the deflation rate (fall in the inflation rate).

  2. Very interesting post. I have a few not very coherent thoughts.

    “Consider what happens if IBM announces that its 10-year bond will forever cease to pay interest, or a coupon.”

    This is not the way the Fed sets interest rates (at least pre QE). (pre QE) The fed would lower the interest rates by buying in the open market, until the yield it was targeting was reached (or announcing it was going to buy so that the market moved the price to the target). This would be like IBM buying its bonds until the price becomes so high the yield equals zero. So in this case you would have a bought of deflation first and then inflation as the bond decreases to par.

    So you have a very different outcomes if you consider yield instead of coupon rate. Does this mean that there is a fundamental difference between monetary policy done through open market operations (yield) and those done through interest on reserves (coupon)?

    This whole theory of the price level as the present value of the monetary base seems to be wrong somehow, maybe the direction of causality is off? With endogenous money, money is created and has value because of future economic activity. If anything the central bank is free riding on the value of money created by real activity financed by the financial system.

    “Compounding each hit to residual value would have been a decline in the mark's liquidity premium. When the price of a highly-liquid item begins to fluctuate, people ditch that item for competing liquid items with more stable values.”

    Is this really a liquidity premium? The difference between a pound and a mark in 1922 is not like the difference between the stocks of public and privately held companies. Privately held companies sell at a discount because there is no market for their stock, i.e. are illiquid. There was a market for marks in 1923. The risk you are describing here is price risk. The likelihood that the mark will fall in value is much greater than the pound, so it sells at a discount.

    1. I know the liquidity premium thing goes back to Keynes, but it has always bugged me. Price risk and liquidity risk are pretty easy to differentiate.

    2. "Does this mean that there is a fundamental difference between monetary policy done through open market operations (yield) and those done through interest on reserves (coupon)?."

      There's no fundamental difference. In either case, the central bank modifies the financial return on its liabilities in order to get the price level moving in the desired direction. They can either do it by modifying the supply of liabilities provided to the market via open market operations, which affects their non-pecuniary liquidity return, or by altering the amount of interest paid on liabilities, which affects their pecuniary return.

      I don't really understand your point on liquidity premia.

    3. What does pecuniary mean too.

      Simple example to show what I mean by liquidity vs price risk.

      Your have two assets, A and B. Both have the same expected return, but B's return has a higher variance. B would be cheaper than A. This is price risk, or interest rate risk in bonds.

      You have two assets C and D, both have the same expected return. There always a market for C, but D only trades at the end of each quarter. D would be cheaper than C. This is the liquidity premia.

  3. Surely a central bank liability has continuous nominal value in the form of an option to claim that value via exchange - due to its legal tender status - plus its value as a contingent tax credit. So its value is continuously available though either channel, which seems at least slightly reassuring.

    So how necessary is the residual value question, given the existence of these options?

    1. "So how necessary is the residual value question, given the existence of these options?"

      If residual value exists, it anchors the price so that expected deflation can set in. Neither legal tender nor tax acceptance provide such an anchor. A government may specify that dollars can discharge taxes, but it doesn't specify the quantity of services that a dollar provides. So there's no reason for tax acceptability to stop the downward roll of the price level once interest rates have set it going. Likewise, legal tender laws specify the use of dollars, but not the specific amount of debt that a dollar can discharge. (In the past I've argued that these legal tender laws are very weak and don't contribute much to the value of the dollar.)

  4. In the Weimar Republic, real rates were expected to continue falling. So there was incentive to lend. And with inflation expected to keep rising, there was incentive to borrow. Debts became easy to pay off for both lender and borrower.
    Nowadays, inflation has no chance to rise. Therefore real rates are limited on the downside. Paying off debts is more of a concern within these bounds. Investment is muted. Inflation then becomes muted. Top that off with a falling labor share and the Weimar Republic is a bad comparison to the current situation.

    1. Who's comparing Weimar to the current situation?

  5. This comment has been removed by the author.

  6. The various forms of money, credit from private and public actors must be considered. In wiemar germany debt to GDP in real terms is estimated to have been 900% of GDP. The debt service was significantly gold based war reparations and so inflating out of the problem wasn't an option.

    The Sign Wars reflect a poorly formed question and poor understanding of inflation and money. Inflation is an impact or result of all the forms of money "near money, private and public sector credit and of course currency". Ignoring the large "real" demand for hard currency gold in the Weimar story ignores the fairly limited effects rate policy could have on deflating the debt burden. Initial money printing likely accelerated money and value flow while equity type assets in nominal terms were percieved as "improving" as nominal debt burdens based in marks shrank. Unfortunately the rising inflation leading increased working capital costs and shrinking real economic flows which exacerbated a Real debt burden that was gold based and thus immune to nominal rate policy impacts. chapter 14 has a good story on this. The book, "when money dies" is also a great 1st person account of the weimar experience and how asset prices and monetary policy were mis-understood by public and policy makers foriegn and abroad.

  7. This is interesting stuff. I got inspired to write to my local central bank, in my case the Swedish Riksbank. The answer I got from the General Councel was that in case of dissolution, notes and coin would be a claim as good as any other but he also added that there was nothing specifically written about it in law. Sweden has gone from inflation to deflation with the central bank interest rate currently at 0% after a number of rate reductions.

    1. Interesting. With nothing specifically written in the law, one wonders how investors can even begin to value the krona. (Which supports the conventional view).

    2. Who needs law, or even a central bank. Somali shillings trade fine (ok at a discount). Money is a social protocol and the social norms that enable it can be formal or informal.

    3. Nick, Somali shillings do continue to trade, although they've basically become a pure commodity currency --- worth no more than the paper on which they're printed. Something keeps units like the krona far above their paper value. That the Riksbank treats these krona as an IOU might explain this differential.

    4. JP, totally agree on shilling having very limited value. Lots of factors support Krone et. al. integrity of circulation and anticipated circ growth, a demand function driven by legal tender status, tax demand, formalized legal protection and of course the network effects of acceptance in a well governed economy. I am currently working on a supply driven currency which is "earned" by the generation of solar energy at the rate of §1=1MWh verified generation. The currency is currently granted in 15 countries and we are working on growing the network of claim recipients. It's even been written up in Scientific American. The ideas is that the economic utility is determined by the network of holders. A large network will have a positive externality eventually acting as a renewable energy stimulant. It currently trades thinly in 3 markets at about §130:$1. The key insight is that bitcoin has say 5m holders creating a mix of economic and speculative utility supporting roughly $1,000/holder participant. The network utility (trade/spendability) will hopefully become autocatalytic where the average residential claimaint receives §5-9/yr creates more economic utility than they claim.

      We are shifting the code to a POW proof of work algorithm and working on the issuance (monetary policy) right now. Contstraints involve a hardcoded interest rate that minimizes inflation, but incents enough people to participate and protect the blockchain integrity. Interestingly enough transaction have a small cost which is used to prevent (transaction spam) while simultaneously acting as demurrage.

    5. Yes he also added he didn't think lawmakers ever thought about that possibility. One thing I don't get about the perpetual bond analogy: IBMs perpetual bond would find its footing at a very low level but the central banks currency is not at a very low level compared to held assets?

  8. Fantastic clear post!

    You've clearly set out just how I was trying to see things (but I was still muddled).

    Simply being able to transport purchasing power through time is an extremely sort after characteristic for an asset. Curency of a stable government can be the ultimate exemplar of that valuable characteristic.

    Hyperinflations often entail people fearing that the monetary authority isn't going to last. The currency of the loosing side in a war is a classic hyperinflation danger. I've sometimes wondered whether countries such as Switzerland and Singapore like having foreign owners of their currency and bonds as a way to have a set of powerful foreign people who want to ensure the protection of the Swiss or Singapore state as a way to protect their savings.

  9. You haven't directly mentioned the "option value" of currency here but I think it may be relevant. Perhaps the clearest description I have seen for this is from an interview with Warren Buffett’s biographer Alice Schroeder:-

    “”He thinks of cash as a call option with no expiration date, an option on every asset class, with no strike price.” It is a pretty fundamental insight. Because once an investor looks at cash as an option – in essence, the price of being able to scoop up a bargain when it becomes available – it is less tempting to be bothered by the fact that in the short term, it earns almost nothing. Suddenly, an investor’s asset allocation decisions are not simply between earning nothing in cash and earning something in bonds or stocks. The key question becomes: How much can the cash earn if I have it when I need it to buy other assets that are cheap, versus the upfront cost of holding it? “There’s a perception that Buffett just likes cash and lets cash build up, but that optionality is actually pretty mathematically based, even if he does the math in his head, which he almost always does,””

    I think the value stability that confers that attribute is anchored in the fact that so many real prices and contracts are denominated in the currency. That in itself gives it value (so long as the monetary system itself is expected to persist).

  10. You write "When our IBM perpetual bond ceases to pay interest its price will quickly plunge, just like a normal bond. But it's price won't fall to zero. At some very low level, value investors will line up to buy the bond because its price is expected to rise at a competitive rate. [..] What drives this expectation? Though the bond promises neither a return of principal nor interest payments, it still offers a fixed residual claim on a firm's assets come bankruptcy, windup, or a takeover. This gives value investors a focal point on which they can price the instrument."

    Why do you think its price is expected to rise? I assume that this price should be equal to some average discounted value as estimated by investors, which (assuming a certain interest rate) is a function of estimated likelihood of payout, estimated size of payout and estimated moment of payout. A price rise implies that, as time goes by, the likelihood of bankruptcy, windup, or a takeover should increase, and/or the expected payment should increase and/or that the moment that this could happen comes closer. But why would that be the case?