Saturday, January 11, 2014

Rowe on the Liquidity Trap

Nick Rowe argued that the zero nominal bound is an illusion and that the real issue is the desired size of the central bank's balance sheet.

I agree that the zero nominal bound is largely self-inflicted by central banks, but I don't agree with Rowe's argument.   He argues that there is a spectrum of assets that are more or less liquid.   The more liquid, the lower the yield.   This creates a continuum of assets with progressively higher nominal yields.   Rowe's point is that there is always some asset that is close to the zero bound.    An expansionary monetary policy generates a perfectly liquid asset by purchasing a somewhat less liquid asset.   If the central bank buys up the most liquid nonmonetary assets first, and then goes to progressively less liquid ones, then it will always be pressing against a zero nominal bound.   The only question is how far it must go to keep inflation (or the level of nominal GDP) on target.  

I think there is an element of truth in this argument, but I am pretty sure that with competitive issuers, a highly liquid asset can have a substantially higher nominal yield than currency.    Checkable deposits are extremely liquid, but their yield depends on the return that banks can earn on their asset portfolios and the cost of providing intermediation services.

But this high yield creates an incentive to deposit currency.   Doesn't that result in banks lending more, driving down the returns on their portfolios, and so the yield on deposits?    Not if there a nominal anchor, like a growth path for nominal GDP.   The quantity of currency is reduced to match the lower demand for it.   The implicit yield on the services provided by currency rises to match the yield on deposits.  

Now, suppose rather than conventional banking, the only option is to either hold currency or else competitively-issued bonds.   The issuers fund real investment projects.   There are costs to underwriting the bonds.   Further, there are transactions costs to managing "cash."     But does the nominal yield on these bonds have to approach zero?   That is, will people buy the bonds, driving down their yields, and I suppose fund more real investment projects until those yields are driven down?   No.  With a nominal anchor, the quantity of currency is reduced to meet the demand.  

As Rowe surely knows, the whole point of Friedman's optimum quantity of money is to generate a deflation rate so that these competitive nominal interest rates are driven to zero, raising the demand for currency.   The implicit nominal yield is driven to zero.   And, of course, the high real demand for currency in this situation implies a larger real balance sheet for the central bank.  

Still, I agree with Rowe that the zero nominal bound is largely self imposed.    Suppose the central bank issues currency.   However, it is prohibited from buying government bonds.   It solely buys corporate bonds.   The corporate bond market it entirely made up of 10 year notes.   The corporations fund risky investment projects, but they are mostly diversifiable.   The central bank's asset portfolio isn't very risky.   Even so, if the central bank were to become insolvent, the government promises to bail it out.    The interest rate on these corporate bonds is 8 percent.   The quantity of currency is $6,000 billion, and nominal GDP is $15,000 billion.

There is a national debt.   It is very small, only $1 billion.   It is funded entirely by T-bills.   The government runs a slight budget surplus, so the outstanding national debt is generally decreasing.   The T-bills are very liquid.    Some people manage their cash positions using the T-bills.   The yields are only 2 percent and have been dropping over time.

The central bank has noticed that when it expands the quantity of money by purchasing corporate bonds, the yield on T-bills falls.   And further, when people are short on money, they reduce expenditures on a variety of things.   Those who have fewer receipts are also short on cash, so they reduce expenditures.   However, those who hold T-bills to manage their cash balances sell them when they are short on money, raising the yield.

The central bank decides to target the T-bill rate.   It is pretty effective.   New Keynesian economists develop complicated models where consumption depends solely on the T-bill rate.   They even have models explaining the relationship between the yield on T-bills and the yield on corporate bonds.   The corporate bond rate is equal to the expected future T-bill rate plus a risk premium.

Now, suppose the government pays off the entire national debt.   There are no more T-bills.   There is no yield on T-bills.   Does monetary policy become ineffective?     If the only short and safe asset is currency, does monetary policy become impossible?  

Well, the demand for currency will rise because T-bills can no longer serve as money substitutes, but given the numbers I described, this is just a drop in the bucket.   Those who were managing their cash positions no longer earn a return.   Like everyone else, they just use zero nominal interest currency.

Clearly, all that changes is that the central bank's previous target can no longer be used   The central bank can still purchase and sell the corporate bonds as always.

While I can imagine New Keynesian economists insisting that the government run a budget deficit and issue some T-bills so that all of their models still work.   I can imagine central bankers taking the same position so that they can get back to business as usual.    But I can't imagine any economist thinking that monetary policy is ineffective.

Now, suppose the national debt isn't paid off, but private investors worry more about losses on corporate bonds.   Those most worried sell the bonds and buy T-bills.    The yield on T-bills is driven to zero.   That is, the $1 billion of T-bills that are outstanding now have a zero yield.    If people worried about the risk of corporate bonds still want more safety, there is no benefit to buying T-bills rather than just holding currency.   Perhaps the demand for currency rises substantially.

Is monetary policy impossible?   No, the central bank can continue to buy corporate bonds as usual.  Presumably, the sale of corporate bonds and purchase of T-bills and accumulation of currency would tend to raise the yield on corporate bonds.   The central bank can buy the corporate bonds, accommodate any increase in the demand for currency, and even reverse the increase in the yield on corporate bonds.

Of course, if there is a different story, where the supply of corporate bonds falls substantially, and the demand for them rises substantially as well, either directly by private investors or indirectly through a demand for currency.  Then, perhaps, corporate bonds would reach the zero nominal bond.   

But this account shows, much as Rowe was arguing, that monetary policy does not break down just because the shortest, safest, and most liquid security has a nominal yield of zero.


  1. If there are no safe assets to purchase, the CB can create safe assets easily enough (in your example, it would make safe overnight loans with the corporate bonds as collateral). A bank that only holds well-collateralized overnight loans doesn't even need capital.

    Any financial entity with capital can borrow overnight and buy corporate bonds. You don't need to be a central bank.

    So if a CB buys a corporate bond, it's doing two things, one monetary and one non-monetary. It's creating money, and it's using its capital to take a risk position. If interest rates are at a lower limit, then all the action is in the non-monetary part of the operation. It's not satisfying a "demand for money", it's using its capital.

    1. In the scenario I developed, currency is money. If the demand for currency rises, and the central bank creates more, it is satisfying a demand for currency.

      Your argument that if it buys corporate bonds, it isn't really satisfying a demand for currency, but is rather using its capital. And your argument is that it could develop an overnight lending market and only make well collateralized loans. And if the interest rate that would exist in that market that doesn't really exist would be zero, then it is really "using its capital" and not meeting an increased in demand for money.

      Well, currency is money, and there is a demand for it. And it is being met.

      As best I can tell, this is just the bills only norm. Central banks should be able to implement a monetary policy by trading short and safe government bonds.

      Market Monetarists claim that this is not always true.

    2. Is the bond not also a monetary asset aka "money"? It promises to pay a certain amount of $ at a later time. When I add up my monetary net worth I don't make much of a distinction whether I hold $1 million as a deposit or as a bond note. Sure, the bond carries liquidity, default, and inflation risk that is what I receive interest for. But the principal is the same and that will be what matters in my spending decisions.
      (The picture only changes when the central bank makes loans directly to businesses. Those loans will affect spending.)

  2. The precise way of formulating the ZLB constraint is to consider whether the central bank is expected to earn an economic profit over the expected holding period. Headline yield on a security is not a good indicator. The ZLB problem may appear much sooner - consider the early 2013 when the term premium of treasury securities the Fed was acquiring became negative, i.e. 10 year bond yielded less than the expected future path of interest on reserves. See my comment at Rowe's blog:

  3. Bill: this is a very good thought-experiment.

    (I'm not quite sure where you are disagreeing with me.)