Tuesday, October 6, 2009

Money and Interest--Comment on Sumner

As a small aside in a recent post, Scott Sumner suggested:
BTW, I wish the Fed would stop calling it "base" money; bank reserves are now essentially T-bills. Only currency is still interest-free. And monetary theories of inflation are based on explaining the supply and demand for non-interest-bearing money.

I think this is an error. Bank reserves are not essentially T-bills. And none of my monetary theories of inflation are based upon the supply and demand for non-interest-bearing money. Rather, they apply to the entire spectrum of situations--no money bears interest, some money bears interest and other money does not, and finally, where all money bears interest. When growth rates of the money supply are at issue, the special case where all money pays interest best exhibits the proportionality between money growth and inflation.

First, how are bank reserves different from T-bills? T-bills have a market price. T-bills do not serve as medium of account. While their prices are generally quoted as yields, the dollar price of a $1000 T-bill can and does change depending on supply and demand. While we might quote a price of 1% for a one year T-bill, that means that the dollar price is $990. An increase in demand might increase the price to $995. An increase in supply might reduce that price to $985. (When the price of a $1000 T-bill gets to be $1001 or so, then monetary disequilibrium can be generated because of shortages in the T-bill market.)

A one dollar reserve balance at the Fed is always worth one dollar--by definition. The price of a $1 reserve balance cannot have a price of 99 cents or $1.02. That is because base money--including both currency and reserve balances at the Fed--serves as medium of account.

Second, T-bills do not serve as the medium of exchange. The Treasury and the Fed auction them off to obtain funds to spend. Reserve balances, on the other hand, never have to be sold to obtain funds to make a purchase. They can be spent into existence. The Fed promises to make a payment, and it just credits the funds to the reserve deposit account of the seller's bank. Neither the seller, nor the seller's bank, necessarily wants to hold these funds. Both accept them because it is money--the generally accepted medium of exchange. They may well intend to spend the funds on something else.

Similarly, it is not necessary to "buy" reserve balances. Banks receive them constantly as checks and electronic payments are received by their depositors. By simply refraining from spending them--that is, buying earning assets or making loans--reserve balances are accumulated.

Balances in reserve accounts at the Fed serve as both medium of account and medium of exchange, and nothing changes when the Fed pays interest, other than the demand to hold them. Higher interest results in higher demand. Lower interest results in lower demand. It is exactly like transactions accounts that firms and households keep at banks. Higher interest raises the demand, and lower interest lowers the demand. Still, they are money.

As for monetary theories of inflation, the interest rate paid on money, among other things, determines the real demand for money. If the nominal quantity of money doubles, then the result is an excess supply of money. As that money is spent, it raises the demands for various goods and resources. As the prices of goods and resources rise, the real quantity of money falls. Equilibrium returns when the real quantity of money has returned to the real demand for money.

In response to comments similar to those above, Sumner continued:

I have two points. First, bank reserves, especially excess reserves, aren’t really a medium of exchange. I will admit that there is a derived demand for RR as a result of the demand for checking balances. But the ER number is meaningless. The are just held as assets, like T-bills. the ERs aren’t circulating, and they aren’t even backing DDs.
I think this analysis is much too dependent on the existence of reserve requirements. What would happen if there were no reserve requirements? Since sweep accounts have allowed banks to report however many "DDs" they choose, I am not sure that reserve requirements mean anything today.

Regardless of regulations requiring banks to hold reserves, as explained above, reserves are media of exchange. They can be accumulated by spending less and can be spent into existence by the issuer.

Second, if interest is paid on money, the QTM breaks down. You can no
longer assume if M doubles, P doubles in the long run. Suppose the Fed doubled
M, but paid a higher interest rate on M than alternative assets, that would
actually be contractionary.
The interest rate the Fed pays on reserves (like the interest that banks pay on transactions deposits) simply impacts the real demand to hold money. Yes, if the Fed chooses to pay higher interest on reserves, this raises the real demand for reserves. If banks choose to pay higher interest on transactions accounts, then the real demand for transactions accounts will be higher.

The price level still adjusts so that real quantity of money adjusts to the real demand for money. It is still more or less true that doubling the quantity of money will double the price level. If the nominal quantity of base money doubles and the Fed decides at the same time to pay more interest on reserves (or any number of things happen that simultaneously impact the demand for money or the money supply process) then the price level will not double.

Yes, in a sense you are right, the ERs can be costly converted into a medium of exchange (currency or DDs). But my point wasn’t technical, it was that it’s as if they were like T-bills. If you pay interest on them at a rate higher than on 3 month T-bills, then they don’t have the normal inflationary impact predicted in models. And that’s even true in the long run. Suppose in five years the T-bill yield is 4.5%, but reserves earn 5%; banks will still be hoarding ERs. There are other countries that run monetary policy this way, and they don’t have high inflation.

Private firms, other than banks, and households cannot hold reserve balances at the Fed. So, reserve balances directly serve as medium of exchange only for banks. Even so, payments with transactions accounts involve promises by the firm or household to have their agents, their banks, make payments with reserve balances at the Fed. There is a intimate relationship between the vast majority of payments and the payments banks make with their reserve balances.

If the Fed changes the interest rate it pays on reserves, this will impact the real demand for reserves. Like anything else that impacts the real demand for money, this will impact the equilibrium price level consistent with any nominal quantity of money. I am very uncomfortable with talk about "hoarding ERs." Banks choose to hold reserves. The Fed's responsibility should be to meet that demand for reserves so that nominal expenditure remains on target.

Like Sumner, I think that the Fed's interest rate policy regarding reserves has been a horrible mistake. However, I have no problem with the Fed "paying" interest on reserves. It is just that the appropriate interest rate in October of 2008 was slightly less than zero. I think the most sensible policy is to peg the interest rate on reserves at about 25 basis points below the 4 week T-bill yield.

Also like Sumner, I think that the Fed should keep nominal expenditure on a stable growth path. Paying interest on reserves impacts the nominal quantity of money that the Fed must create to meet that target. Oddly enough, under current circumstances, when a reasonable interest rate on reserves would be slightly less than zero, the quantity of base money and reserves that the Fed needs to create is a bit less than it would be if the interest rate on reserves were fixed at zero. Unfortunately, because the Fed insists on maintaining an interest rate on reserves above the T-bill rate, the quantity of base money it needs to create is larger now than it would be if the interest rate of reserves remained fixed at zero.


  1. I agree with much of what you say, but here's why I have a different take. The decision to pay interest on reserves is not a one-time thing, it is a new policy regime. Countries that follow this policy tend to adjust that interest rate as the target rate changes. Under this scenario, banks hold reserves not for transactional purposes, but as an investment, much like T-bills. If the Fed doubles the amount of reserves, the price level doesn't double, even in the long run. That is because the Fed keeps adjusting the interest rate as market interest rates change. In this respect reserves are much like T-bills. As market rates rise, the yield on T-bills rise, as does the yield on reserves. Thus the Fed can double the MB, even in the long run, with no impact on the price level. They simply pay banks enough so that they are encouraged to hold the extra reserves as an investment, not because they are required, or because they are needed for near term liquidity reasons. That's what makes the policy so different.

    It not that I disagree about any of your technical points, but the QTM way of thinking is so deeply ingrained that I worry that if we call this stuff "money" people will (wrongly) think high inflation is right around the corner.

  2. Scott:

    Thanks for the comment.

    Paying interest on reserves that varies with other interest rates makes base money more neutral in the long run, not less.

    A one time doubling of the quantity of base money should not impact interest rates in the long run. Whether the Fed changes the interest rate on reserves with others is irrelevant in that case.

    A doubling of the growth rate of base money should raise expected inflation and nominal interest rates. If the interest rate on reserves is fixed at zero, then this raises the opportunity cost of holding reserves and so should reduce its real demand. This raises the growth path of the price level, though the inflation rate should still change in proportion.

    If, on the other hand, the interest rate on reserves rises with all the other nominal interest rates, this nonneutrality is avoided. The real demand for reserves does not change, there is no change in the growth path of the price level, and the inflation rate changes in proportion to the change in the growth rate of base money.

    I see no problem with banks holding reserves for investment purposes, and consider trying to distinguish between money (or reserves) held for investment rather than liquidity purposes as a fools errand. Generally, they are held for both reasons. It is an investment with great liquidity characteristics.

    I agree that if the central bank wants, it can change interest rates on reserves as a policy tool. How that works is by changing the interest rate on reserves relative to other interest rates rather than passively following them. By using this tool, they can offset the effects of changes in the nominal quantity of base money by manipulating the real demand for reserves. So?

    I agree that it is sad that some economists, particularly those from an orthodox monetarist background, use comparison to historical changes in the monetary base and some hand waving about long and variable lags to predict immandent inflation. It just means that there is more work to do. It is supply _and_ demand.

    Please think about the following. "Orthodox" monetarists defended a money supply rule for decades. How much energy went into arguing that changes in interest rates do not significantly impact real money demand or velocity? Paying interest on reserves may play havoc with habits of though of some ecnomists, but so what? The demand for base money can change for many reasons, and the job of the monetary authority is to adjust the quantity of base money enough to keep expected nominal expenditure on its targeted growth path.

  3. That's a good point about superneutrality. It never occurred to me that interest on reserves is a way of preserving superneutrality.

    To some extent we have a argument about what is useful, not what is right. I suppose you are right that for analytical clarity reserves should be treated as base money, whether interest is paid or not. But would you agree that if the Fed moved the interest rate on reserves from below the rate on T-bills, to a rate above the yield on T-bills, at the same time they doubled the money supply, then prices would not even rise in the long run? (And that is what they did.) I realize that combines two separate policy changes, and looking at it from your side I see why it would be frustrating to mix the two issues.

    On further reflection I'll try to drop the T-bill analogy. Unfortunately this is 24 hours too late to affect a piece I have coming out in the AEI's "The American" tomorrow morning, which makes this same argument.

  4. Bill: I agree with what you say here.

    I think the fact that money pays the lowest rate of interest (or no interest) is a "contingent fact" about money, rather than an "essential fact". If any other asset paid lower interest than money, nobody would want to hold it. People hold money, even though it is rate of return dominated, because it is the medium of exchange.

    But then there's gold, which seems to contradict what i just said. Maybe people hold gold because they believe it has a negative CAPM Beta?

  5. I don't think people holding gold contradicts what you really meant by your claim. Your point that people wouldn't hold an asset paying lower interest than money is only literally true for money-denominated instruments, where interest encompasses the maximum real return. For a real asset like gold, people might just think it has a higher expected return than money. We don't resort to portfolio theory to wonder why people in Zimambwe hold gold, and I don't think the theory changes sans hyperinflation.

  6. Nick:

    If gold had a fixed nominal value, then the fact that it has no explicit yield would create a puzzle as to why people hold it. But not really, because that would be a gold standard, and there are "default" risks with paper money in a gold standard. Without credible gold clauses, those risks transfer to all debt instruments too. Gold hedges those risks.

    Because gold doesn't have a fixed nominal yield today, it can generate capital gains. Of course, we all know about gold being a traditional inflation hedge, but even if that wasn't an issue, people would hold stocks of gold for the same reason they hold inventories of any storable good--differences between current relative prices and expcted future relative prices.

  7. It is just "supply" and "demand". The critical idea is that the monetary authority must balance the supply and demand for money. Paying interests on reserves creates a demand for Excess Reserves where there once was none.

    But, interest bearing ERs are not simply t-bill like instruments. Sure they have very similar liquidity properties, but more generally they are an asset that competes for bank funds.

    The concern which you fail to name is that the Fed's portfolio will be shorter-dated than the banking system generally.

    Will the Fed generate the income necessary to sustain the sterilization of 800B in ER? If not, the system becomes unstable. The Fed must expand ER to sterilize the ER, and so forth.

    Which will it be? Well who knows, and that's precisely what's fueling the concern. No one has attempted this policy before with such a large quantity of ER.

  8. That’s interesting viewing of all this, I believe we need to keep things simple too, as without doing that, I see it extremely tough to make anything work well. Working with keeping support and demand levels, it is fairly decent way of working and with OctaFX broker; it only gets better thanks to their cTrader platform where we can work well due to their highly upgraded system with wide range of charting as well, so that is just perfect way to trade on these things.