Thursday, September 30, 2010

Kocherlakota in London

Narayana gave a speech in London, posted yesterday here. This is a new iteration on the spiel he has been giving for several months. The new and interesting stuff, from my point of view, is in his discussion of quantitative easing (QE). He makes clear that the intervention he is considering is the purchase of long-maturity Treasuries (and not private assets) by the Fed. Kocherlakota breaks down QE into four effects.

1. Commitment: Here, he says:
The first effect of QE is that it represents another form of forward guidance about the path of the fed funds rate. It is a way for the FOMC to signal—in a perhaps more striking way—that it plans to keep the fed funds rate low for an even longer time to come.
The idea here is that purchases of long Treasuries reflects a commitment by the Fed to keep short-term nominal interest rates low. Why? By acquiring a portfolio weighted toward long-maturity Treasuries, the Fed is taking on interest rate risk. If short-term interest rates go up, the Fed will have to pay more interest on reserves, which is essentially revolving short-term debt for the Fed, and will suffer a capital loss on its long-maturity assets. Of course, the Fed can (apparently for a long time) keep the interest rate on reserves (and thus the fed funds rate and all other short rates) at a low level and therefore avoid the capital loss and higher cost of revolving short-term debt. Actually acquiring the long-maturity asset portfolio is then a commitment device. Of course, the problem is that the Fed may be committing to something that it would regret sometime in the future. Should the inflation rate start to take off, the Fed might want to tighten monetary policy by increasing the interest rate on reserves. But it will be deterred from doing so because of its long-maturity asset holdings. As a result, the higher inflation rate could become self-fulfilling.

2. QE creates more outside money: This is just the standard open-market-purchase story. We would get the same effect whether the assets the Fed acquires are T-bills or long-maturity Treasuries. Kocherlakota says:
Second, QE creates more reserves in banks’ accounts with the Fed. The standard intuition is that this kind of reserve creation is inflationary. Banks can only offer checkable deposits in proportion to their reserves. Economists view checkable deposits as a form of money because, like cash, checkable deposits make many transactions easier. In this sense, bank reserves held with the Fed are licenses for banks to create a certain amount of money. By giving out more licenses, the FOMC is allowing banks to create more money. More money chasing the same amount of goods—voila, inflation.

This basic logic isn’t valid in current circumstances, because reserves are paying interest equal to comparable market interest rates.
Kocherlakota's "basic logic" is the standard money multiplier story that can be found in many money and banking and intermediate macro textbooks. If we could pass a law to ban discussion of the money multiplier, I think that would be great. First, even in the context of a banking system with binding reserve requirements, the money multiplier story is of little use for thinking about the causes of inflation. For example, suppose that we had a reserve requirement on apples. If I have a stock of apples for sale, I am required to hold a reserves, say 10% of the value of stock of apples, in a reserve account with the central bank. Suppose this reserve requirement is binding. Now, the central bank doubles the stock of reserves, thus increasing the "licenses" to sell apples. What will happen ultimately is that the nominal stock of apples for sale will double, as will all prices. In the long run, we certainly would not expect the stock of apples for sale to change. There is an apple multiplier at work here: for each unit of reserves there are 10 apples, in nominal terms, and for each extra unit increase in reserves by the central bank, there has to ultimately be an increase of 10 units (in nominal terms) in apples. However, it's not the increase in the nominal quantity of apples that is causing all prices to double ultimately, it's the increase in the quantity of outside money. In practice, even if we have binding reserve requirements, what is causing inflation is growth in the stock of outside money (currency plus reserves). Growth in the stock of the liabilities of private financial intermediaries (the apples) is neither necessary nor sufficient for inflation.

Further, even in the United States, where we still impose reserve requirements on banks, those reserve requirements have become essentially irrelevant. Why we don't do away with them entirely, and improve the efficiency of our financial system, is beyond me. In any event, if we look at pre-financial crisis times, for example 2007, the stock of reserves was hovering around $10 billion, the stock of currency outstanding was about $800 billion, and the stock of M1 was about $1.4 trillion. Thus, reserves are essentially insignificant. Why? Banks have invented ways of getting around the reserve requirements. Business have for a long time had "sweep accounts" with banks. These are transactions accounts that can be used during the day, and then are swept away overnight into interest-earning accounts not subject to reserve requirements. Similar arrangements can be made for consumers, and those types of arrangements are spreading.

I prefer to think of the US banking system as having essentially nonbinding reserve requirements, even in "normal" times. For me, the element of the outside-money-injection aspect of QE is that the reserves can potentially turn into currency. Clearly a swap by the Fed of T-bills for reserves under current circumstances (where the quantity of reserves is in the neighborhood of $1 trillion) has essentially no immediate effect, as there is little difference between T-bills and interest-bearing reserves, at the margin. A key difference between a T-bill and reserves, though, is that the reserves can be withdrawn by banks as currency, just as you can withdraw currency from your transactions account at the ATM. The potential for inflation coming from the overhang of reserves is that, if banks become less willing to hold reserves, prices and interest rates somehow have to adjust to induce consumers and firms to hold the stock of outside money, as currency and reserves. If interest rates are not adjusting (because the Fed does not change the interest rate on reserves), then ultimately the price level has to go up. Kocherlakota and I are coming to the same conclusion here, but the mechanism I have in mind is different.

3. Interest rate risk: As Kocherlakota explains, the idea here is that, by taking long-maturity Treasuries off the hands of the private sector, in return for short-maturity reserves, the Fed is shifting interest rate risk from the private sector to the Fed. This should make the private sector more willing to bear the risk associated with long-maturity assets more generally, and all long-term yields should fall. The Fed can then push down the long end of the yield curve. The problem with this argument is laid out in point 4.

4. Risk shifting: Kocherlakota says:
The Fed cannot literally eliminate the exposure of the economy to the risk of fluctuations in the real interest rate. It can only shift that risk among people in the economy. So, where did that risk go when the Fed bought the long-term bond? The answer is to taxpayers.
What he goes through here is essentially an irrelevance theorem. If we take into account the implications of the Fed's actions for fiscal policy, through the consolidated budget constraint of the Fed and the Treasury, the risk that was taken off the hands of the private sector somehow has to end up there. It could be that we are relieving some people of risk and giving others more. That may even be an optimal thing to do, but as far as I know no one has studied this. Good topic for a PhD dissertation.

One thing left out of Kocherlakota's argument is the basic natural (or fiat) advantage of a central bank. To be doing anything, or indeed to be doing any good, the Fed's intermediation of long-maturity assets must somehow be more efficient than what the private sector does. We have private financial intermediaries that convert long-maturity assets into short-maturity ones. These intermediaries are banks and mutual funds. What advantage does the Fed have in intermediating long-maturity Treasuries? Maybe this is just the ability to issue outside money. The Fed has monopolies on the issue of currency, and of the stuff (reserves) that is used in daytime clearing and settlement among financial institutions. This of course is the principle at work behind the effects of all open market exchanges of assets by the central bank.


  1. Hey Steve, I think for commitment Kocherlakota is thinking about kinda story in Lucus Stokey JME 1984, essentially the structure of nominal bond holding can support some future path of interest rate/ inflation rate time-consistently. This is observational equivalent to committing a particular interest rate policy.

  2. The Lucas-Stokey story is that there is a sequence of governments. Commitment by the government is limited, each government must make good on the debt obligations of previous governments. Each government then structures the debt to effectively commit future governments. The story has some different elements here, but I guess the idea is essentially the same. The Fed is effectively structuring the debt of the consolidated government to bind its future self. Interesting.

  3. If commitment is desirable (and I'm interested in your view if it is) then why isn't a higher inflation or price level target the way to go? (As Woodford would say).

    Kocherlakota doesn't mention it but it seems like it might be helpful.

  4. Kocherlakota does say this:

    "The Fed’s price stability mandate is generally interpreted as maintaining an inflation rate of 2 percent"

    That does more than what was in the last FOMC statement, which is vague about what the implicit inflation target is. From my point of view, an explicit inflation target is the way to go, with Fed given discretion as to how to achieve it. As I outline here, if you think of QE as giving you commitment, it could commit you in the wrong way.

  5. May I just say what a relief it is to have a blogging economist who keeps us informed with up-to-date macro. A blessed relief from the turgid regurgitations of ancient parables like the quantity theory, money multipliers and paradoxes of thrift found at the likes of deLong and Sumner. Don't these guys get it that macro has moved on since the 1950s ? Reading them is like trying to do organic chem before the discovery of the electron.

    anon2 from Tx

  6. If I understand you correctly, you write that the ability to convert reserves to currency is what makes them potentially inflationary. Presumably this is because currency in circulation would increase. The question is, by what mechanism? They could discourage deposits (by charging more fees for them?), in which case the system in aggregate would lose deposits and gain currency. In this case, why wouldn't it result in more currency hoarding by actors? Also, if the system loses deposits, wouldn't it also have to pull back on lending (which is an offset in the case of actors unwillingness to hold more currency)?

    I can see, though, how the currency-conversion ability makes existing Excess Reserves different from the system's unlimited ability to call on reserves in the future. Thanks for making that clear.

  7. anon1:

    Banks have the ability to make withdrawals from their reserve accounts in currency. I assume the way this gets done is that every bank has regular contact with the Fed and is constantly returning worn-out currency to the local Federal Reserve bank or branch, and is also receiving truckloads of cash. The Fed nets out these flows, and if the net flow is going to the bank, that is counted as a debit on the bank's reserve account with the Fed. Thus, withdrawals from the bank's reserve account are driven by withdrawals in currency from the bank's deposit accounts. Now, suppose that other assets suddenly look more attractive to the bank than the reserves it is holding. How does this translate into less reserves and more currency, particularly under the current circumstances where the interest rate on reserves is fixed by the Fed, and that rate is governing all short-term interest rates? Suppose the bank lends $20,000 to a consumer in the form of a car loan by depositing $20,000 in the consumer's account, which the consumer then transfers to the auto dealer, who has an account in another bank. When the transaction clears, the consumer's bank has $20,000 less in reserves, the auto dealer's bank has $20,000 more in reserves, and the auto dealer has $20,000 in her account with her bank. Now, the auto dealer takes the $20,000 and allocates this among a portfolio of assets. Indeed, currency is one of those assets, but the auto dealer may leave some of the funds in her bank. In any event, this process does not end until the banks are content with the reserves and all the other assets they are holding, and consumers and firms are content with the currency and other assets they are holding, given market prices and interest rates. I'm saying the process ends when the price level rises sufficiently, and what you are going to see in equilibrium is less nominal reserves and more nominal currency.

  8. Stephen,

    The mechanism you describe is clear. My observation, however, is that it requires credit demand (in the form of a car loan) to convert reserves into currency. What I think we have been discussing is whether the Fed can create inflation by providing reserves to the banking system IN ADVANCE of credit demand. So in your example, without a borrower seeking an auto loan, the reserves would not become (partially) currency.

    I wonder whether this is not an important point. Many economists seem to argue the supply of reserves at any given (constant) FF rate can create inflation; but since supply is always perfectly elastic, I don't see how that is true. I've heard some argue that the Fed's credible promise to create inflation makes reserve supply inflationary. I see the logic, but I don't think its necessarily so--it presupposes "credibility" in the absence of an exogenous mechanism for producing inflation.

    The tried-and-true means of a central bank producing inflation -- of making sure reserves are spent -- are monetizing deficits and buying foreign assets (currency devaluation). The Fed, IMO, risks talking up the inflationary consequence of QE without a means of delivering. Of course, this may all be moot since, regardless of the Fed's intentions, any QE would be indirectly monetizing a large fiscal deficit.

  9. I may have misunderstood your question. If we think about injecting reserves, if the Fed does it under current circumstances by purchasing Treasury bills, we would expect no consequences. The Fed is just swapping interest-bearing reserves for interest-bearing T-bills. What's the difference? However, the Fed seems to think that swapping reserves for long-maturity Treasuries matters. That's what Narayana is getting at here. You can see how, if you shorten the maturity of the outstanding debt of the consolidated government that this might make a difference. As Narayana argues, though, it's not so obvious.