Sunday, December 5, 2010

The Zero Lower Bound Does Not Bind

There are two respects in which standard theory tells us the zero lower bound on nominal interest rates matters. First, when short-term nominal interest rates are zero, conventional open market operations do not matter - there is a liquidity trap. If the central bank swaps zero-interest-rate reserves for zero-interest-rate Treasury bills, this should be irrelevant, in that no prices or quantities change. Further, a liquidity trap is also a feature of regimes like the one we currently have in the United States. When there is positive supply of excess reserves in the financial system, and reserves bear interest, then the interest rate on reserves (IROR) is determining all short-term interest rates. If the central bank swaps reserves for Treasury bills, that will also be irrelevant under current circumstances.

Of course (again, under current circumstances), if the central bank lowers the IROR, this will not be irrelevant. The problem for the Fed is that, with the IROR at 0.25%, there is little room to move. As Ben Bernanke stated in his Jackson Hole speech, a decrease of the IROR to zero would have little effect and "could disrupt some key financial markets and institutions." It certainly seems correct to say that lowering the IROR to zero would have little effect, but arguing that this would be disruptive seems wrong. In any case, this was part of the argument for QE2, the recently-embarked-upon purchase of $600 billion in long-term Treasury securities by the Fed, under the premise that something needed to be done, and there was no other avenue to pursue.

Second, the zero lower bound on short-term nominal interest rates plays an important role in New Keynesian analysis. In a typical New Keynesian sticky-price model (e.g. see Woodford's book) the nominal interest rate is the policy instrument, and in some versions of these models it can be manipulated to achieve efficiency. Under the appropriate monetary policy rule, the "output gap" can be closed, and the flexible-price equilibrium is recovered, which is the equilibrium of the underlying real-business-cycle model. However, the zero lower bound can get in the way. There can be circumstances where the real interest rate is too high, relative to what it would otherwise be with flexible prices, but the nominal interest rate is at the zero lower bound, so that monetary policy is powerless, in terms of closing the output gap. This is basically the idea in this paper by Eggertsson and Woodford.

Now, apparently some central banks in the world are not constrained by the zero lower bound in the way that the Fed appears to be. Indeed, as you can see here, the Swedish central bank set its "deposit rate," its version of the IROR, at -0.25%, from July 2009 to September 2010. If this had been possible in the United States, this would seem like a simple solution to the Fed's current policy problem. There seems agreement on the FOMC for expansionary policy, and a straightforward way to implement this, rather than engaging in a massive experiment in long-maturity Treasury purchases, would have been to charge financial institutions for the privilege of holding reserves. Is this even feasible? Well, probably not. If you follow the links from here, you will not find anything to indicate that negative interest on reserves is legal in the United States. Congress gave the Fed the power to pay interest on reserves held by depository institutions. This legislation came with at least two flaws. First, the power to change the IROR was given to the Board of Governors rather than the FOMC, where such power would appropriately reside, given the importance of the IROR as a policy tool. Second, the legislation did not permit the payment of interest on reserves held by GSEs (i.e. Fannie Mae and Freddie Mac), which in practice has created a gap between the IROR and the fed funds rate. Add to these flaws that the legislation does not permit a negative IROR. Presumably this is not allowed as it would be interpreted as a tax. So much for foresight.

Given that the IROR cannot be negative, does this mean we are stuck with the zero lower bound, and the relative price distortions that are the concern of New Keynesians? Well, no. As discussed in this speech by Narayana Kocherlakota and in this paper and this one, fiscal policy gives us a lot of flexibility. Basically, in New Keynesian models, the problems created by sticky prices are relative price distortions which lead to a misallocation of resources. What could be more natural than correcting such distortions with fiscal policy, given a sufficiently rich array of taxes? The papers I link to above show that efficiency can be achieved in sticky price environments solely with appropriate taxation. Further, there are inventive ways to manipulate tax rates so that the zero lower bound no longer matters. Unless we think that fiscal policy is somehow constrained (e.g. the legislative process is too awkward) relative to monetary policy, this casts some doubt on the New Keynesian approach to monetary policy, as set out in Woodford's "Interest and Prices." The sticky price problem, if it exists, appears to be a problem more appropriately addressed with fiscal policy instruments than monetary policy instruments. Further,the solution is not an Old Keynesian program of expansion in government purchases of goods and services, but could be a revenue-neutral change in taxes.


  1. Steve,

    For a New Monetarist, you seem to be taking an Old Keynesian view on the primacy of fiscal policy (or maybe that's the view of the papers you link to -- I haven't read them).

    This no doubt grossly oversimplifies both NM and OK thinking, but here's what I mean. In both frameworks the main source of inefficiency (sticky prices & involuntary unemployment, respectively) are more properly dealt with via fiscal rather than monetary policy.

    Is that a fair characterization?


  2. I'm wondering why Swedish people didn't withdraw all of their deposits, when the interest rate was negative. I thought that this is the reason why the central bank never set the interest rate negative not whether it is legal to do or not.

  3. Pete,

    I'm not saying I'm buying into NK models. I'm saying that this is where they lead you. However, it looks like they lead you to a different kind of fiscal policy intervention than the OK models do. NK says set distorting taxes to correct the price distortions caused by sticky prices. OK says increase the size of the deficit - doesn't much matter how you do it.


    The deposit rate is the interest rate on reserves deposited overnight with the Swedish central bank by financial institutions. I'm assuming that, as in the US, these financial institutions use reserves during the day for clearing and settlement. Quite possibly during this period financial institutions were holding zero reserves overnight, which is presumably exactly what the Swedish central bank wanted.

  4. I don't think there is a zero lower bound. All the Federal Reserve need do is conduct open market repos with primary dealers in such a way that t-bills are bought at prices above par. This will drive the fed funds rate below 0. The economics profession's obsession with the zero bound is bizarre. Old traditions may prevent them from appearing, but negative nominal interest rates are a fact of economic reality.

  5. JP,

    You have to consider the arbitrage opportunities. What is happening to the interest rate on reserves when the fed funds rate goes below zero? If IROR is zero, the fed funds rate cannot be negative, as then banks could make infinite profits by borrowing on the fed funds market and holding the reserves overnight.

  6. About a year or two back, my co-blogger Stephen Gordon was advocating almost exactly this sort of policy for Canada. He wanted the government to announce an increase in the GST (VAT) that would be postponed for about a year. The effect is to reduce the current effective real interest rate on e.g. consumer durables.

  7. The Swedish experiment with a negative deposit rate is a bit more complex than described here. While it’s true that the official deposit rate was kept at -0.25 percent from the middle of July 2009 to the beginning of September this year, the Riksbank has an additional deposit facility that banks use when the interbank market is open. Since this rate is just 10bp lower than the official policy rate, the effective interest rate on deposits was 0.15 percent during the period when the more symbolic official rate was -0.25 percent. More generally, the Riksbank has two corridors around the policy rate: The broad corridor was plus/minus 50bp from April 2009 to June this year. Since July 2009 it has been widened and is now back to plus/minus 75bp. But the more important narrow corridor has been plus/minus 10bp during the whole period.

  8. Torgeir,

    You say the official rate is just "symbolic." Does that mean there were no deposits made at that rate, or there were deposits only in unusual circumstances? What would have determined whether a deposit was made at -0.25 or 0.15?

  9. Re: Arbitrage. The fed funds rate has been 0.10-.20 percent the last year. But banks aren't arbitraging the spread by buying fed funds and collecting IOR at 0.25%.

  10. Stephen

    The deposit facility with the low rate is rarely used. During November, for example, the average amount deposited on this facility was 14 million SEK. While the amount fluctuates a little from month to month, it has been under 100 million in all but 9 months over last decade. The episode with -0.25 percent interest rate had no discernable effect on the amount deposited in this facility.

    When is this deposit facility used? The answer given when I contacted Swedish banks last year was that it’s used during the hour after the interbank market and the other deposit facility with a higher interest rate have closed down for the business day.

    The deposit facility with the higher interest rate, on the other hand, had an average amount of 14 billion SEK in November. During the height of the financial crises in December 2008, when the Riksbank intervened heavily in credit markets, 207 billion SEK was deposited in this facility at an interest rate of 0.15 percent.

    I didn’t bring up this issue to be pedantic. I think that central banks should have broken the taboo against negative policy rates. While there is a lower bound on nominal interest rates (as long as central banks issue interest free currency), storage and transaction costs in all likelihood make the effective lower bound negative. Riksbank director Lasse Svensson has publicly argued the case for negative policy rates, so I hoped we could learn something about the effective lower bound from Sweden. But unfortunately they didn’t give us the real thing.

    Using the interest rate on “reserves” as the policy rate was pioneered, by the way, by Norges Bank back in 1997. From mid 2011, however, the Norwegian central bank plans to move to the more widely used corridor system, i.e. with a deposit rate slightly below the policy rate and a lending rate slightly above the policy rate. The arguments are 1) that the current system exposes Norges Bank to too much credit risk since banks are borrowing from Norges Bank to accumulate reserves, and 2) that the current system doesn’t give incentives to trade overnight deposits efficiently among the Norwegian banks.

  11. JP,

    That's due to the fact that the GSEs do not receive interest on their reserve accounts with the Fed. You would still expect the interest differential to disappear, but there is something there limiting arbitrage that no one has successfully explained to me. If the Fed were to tighten, you would expect that differential to persist.


    Excellent. Thanks, that's very helpful.

  12. "You would still expect the interest differential to disappear, but there is something there limiting arbitrage that no one has successfully explained to me."

    Yep, it doesn't make sense. Well you've kindled my curiosity. I'll do some digging and see if there's something plausible behind it all. Will report back. Over and out.

  13. Here's a paper on this that I have not read yet:

    The GSEs do not have access to the discount window, are not supervised by the Fed, and have no reserve requirements, but they have reserve accounts with the Fed and can use Fedwire. They also are not paid interest on reserves.

  14. Stephen,

    the reason the Fed didn't lower its IROR below 25bp is because such a move would really do a lot of damage to the repo market and security lending markets, and to money market mutual funds ability to avoid breaking the buck.


  15. What damage? And what is wrong with breaking the buck?

  16. Money market mutual funds wouldn't be able to earn a living if money market yields (espcially repo and commercial paper yields) drop too low, say below 10bp.

    Repo traders wouldn't trade at super low yields.


  17. So we should guarantee money market mutual funds a living, and the same for the repo traders?

  18. Think about the damage done to the money markets. Firms basically wouln't be able to obtain short-term financing, as there would be no buyers of commercial paper and other money market instruments.

    Firms have a mismatch between revenue and costs (wage and interest payments etc.), and need the money markets to work to survive.

    Given the amounts involved, the banking system wouldn't be able to step in (assuming that they have lower transaction costs and actually would wnat to step in) and clear the market.


  19. Stephen, I'm still looking at this rather interesting problem of arbitrage, or lack thereof.

    I've run into a bit of a road-block; I no longer think I understand the fed funds market very well.

    I always thought the fed funds were basically excess reserves held at the Fed. A bank in a surplus position could trade fed funds (the surplus) to another bank in a reserve deficit position to ensure that the latter held the right amount of reserves.

    The Fed's balance sheet shows that in 2005 there was on average about $10 billion in reserve balances outstanding.

    In 2005 the GSEs - Fannie, Freddie, FHLB - had on average about $100 billion in Fed Funds sold outstanding. I pulled this off their balance sheets. According to "The Topology of the Fed Funds Market", pg 12, the "daily value" of fed fund loans in 2005 was about $300 billion.

    Here's what I don't understand. If fed funds are supposed to represent reserve balances at the Fed, how can there be so much more fed funds outstanding than actual reserves. Put differently, if fed funds are simply transfers of excess reserves, why does the end of day size of the fed funds market seem to dwarf the actual amount of reserves?

    Anyways, I don't get it, hoping you can help.

  20. JP:

    Here is what I think is going on. Suppose there are four banks in the system: A, B, C, and D. At the end of the day, bank A has $100 million in reserves, and the others have zero, so total reserves are $100 million. Suppose no reserve requirements. Now, suppose that, at the end of the day, bank B owes bank C $100. Bank B borrows $100 million from bank A on the overnight fed funds market, then takes the $100 million in reserves and pays it to bank C to extinguish its debt. Bank C then lends the $100 million in reserves overnight on the fed funds market to bank D. The total quantity of reserves held overnight is $100 million, but the total quantity of lending on the overnight fed funds market is $200 million. The reserves are the liabilities of the Fed while loans on the fed funds market are the assets (and liabilities) of individual financial institutions. The amount of lending on the fed funds market could be zero and when the quantity of reserves is large; fed funds lending could be very large when reserves are a tiny amount.

  21. Thanks Stephen, sounds like a reasonable explanation to me. It seems like the explanation has something to do with the fact that banks demand reserves for two reasons; clearing and regulatory req's. In the absence of the latter, demand is purely the former.

  22. Another little mystery I've stumbled upon. The Fed Funds rate is around 0.16-18%. This is unsecured lending. Repo rates (secured by government bond collateral) are around 0.25%. But repo rates should be below the Fed funds rate, not above, because repos are collateralized, and therefore a lender in the repo market is willing to accept a lower interest rate to compensate for the increased security of a repo.

    Anyways, going back to your comment Dec 15. Is the relationship of a Fed Funds loan to reserves the same as a chequing deposit's relationship to paper dollars? Chequing deposits are liabilities of private banks, paper dollars are the liability of the Fed. Deposits can be converted into cash, but they needn't be, so there can be a significant multiple of deposits to currency outstanding (though this need not be the case). This seems the same as your illustration of Fed funds markets. A fed fund deposit represent a private institution's promise to provide an equal amount of reserves. You can have far more fed funds liabilities outstanding than reserves, because like deposits, they are fractionally backed by the underlying asset (either reserves or paper dollars). Does that make sense?