Hummel has claimed that when the Fed pays interest on reserve balances it mixes up fiscal and monetary policy. I find this claim puzzling. The relationship between the Fed and banks holding reserve balances is exactly analogous to the relationship between banks and their depositors holding transactions accounts. Banks pay interest on transactions accounts. Why shouldn't the Fed pay interest on reserve balances?
If the clearing operations of the Federal Reserve were privatized, along the lines of the clearinghouse associations of the past, why wouldn't banks organize and join clearinghouse associations that paid interest on their clearing balances? Of course, that would require that the clearinghouses hold some kind of earning assets, and since financial intermediation isn't costless, the interest paid would need to be somewhat less than the interest earned on those portfolios, even if banks were charged specific processing fees for clearing electronic payments, checks, and any other monetary liabilities (yes, banknotes.)
Sadly, for the banks, if the earnings on the asset portfolio were low enough, the interest payments on clearing balances could turn negative. If we imagine a situation where a private clearinghouse can simply store gold or treasury currency, then, of course, sensible banks would choose a clearinghouse that would store these if it were cheaper than managing a portfolio of very low yielding earning assets. And then, the "negative' interest rate on balances would be storage costs for the gold or treasury currency.
To the degree the Fed proxies the behavior of a private clearinghouse, it should usually pay interest on reserve balances less that it earns on its earning assets, making proper adjustments for risk. Further, there should a lower limit approximately equal to the cost of storing currency. Assuming that the Fed will be bailed out by the Feds if necessary, then the risk of reserve balances is the same as government debt. And reserve balances are more liquid than the shortest T-bills. And so, a simple rule would be to set interest rates on reserve balances perhaps 25 or 50 basis points below the interest rate on four-week T-bills.
Under current conditions, the Fed is offering a financial instrument that is perfectly liquid and riskless. Why shouldn't banks pay for the privilege of holding such an appealing asset? The purpose of saving, or even holding wealth, is to fund real investment. Production processes involving capital goods are not perfectly liquid or riskless. Letting someone fund these projects without risk and with additional liquidity means that someone is taking extra risk. The margin between the interest rate paid by any financial intermediary and what it earns on its asset portfolio should reflect the difference in risk and liquidity. With the yield on the lower risk, highly liquid assets in the Federal Reserves portfolio approaching zero, negative interest on reserve balances is entirely appropriate.
Hummel is concerned that charging interest on excess reserves would tax the clearing balances banks hold for M2 and M3 liabilities. I think this is confused. What would really happen is that the sweep programs would be closed down and the banks would begin to honestly report the quantity of transactions accounts. This would increase their required reserves and so reduce their excess reserves. Sadly, it appears that no one knows what proportion of transactions accounts are falsely reported as some variant of savings account, and so the size of this effect is difficult to determine.
As Hummel notes, there would be a strong motivation to make what are truly savings accounts (beyond the transactions accounts that are falsely reported as savings accounts using sweep software) into transactions accounts. Since this would transfer what is a very near money into the medium of exchange, the economic effect is probably minimal. One can imagine the letter. Great news. You now have unlimited checking privileges on the funds in your savings account. A debit card will soon arrive in the mail. On the other hand, the motivation of banks to fund their asset portfolio with transactions accounts as opposed to certificates of deposit or other debt instruments would involve an expansion in the quantity of money.
Sumner, however, has focused more on banks expanding their asset portfolios. To the degree that banks are capital constrained, and because government bonds have a zero risk weight under current capital regulations, banks will naturally buy government bonds. The usual money multiplier process would directly expand the value of transactions accounts in the banking system, increasing the quantity of the various measures of the money supply, while increasing required reserves and so reducing excess reserves and so the interest banks must pay on excess reserves.
Conventional money multiplier analysis suggests that an expansion in transactions accounts will result in currency withdrawals from the banking system. This follows from the assumption that there is "currency/deposit ratio." However, under the unusual circumstances of an increase in the demand to hold money, it isn't at all obvious that the usual relationships would apply. That is, the currency deposit ratio could fall. However, a currency drain from the banking system would reduce total reserves, and so reduce excess reserves.
Hummel worries that taxing excess reserves would be a subsidy for money market mutual funds. I disagree. The banks are not required to hold excess reserves, and the money market mutual funds are not eligible to hold reserves balances. If banks are accumulating excess reserves because they find it a good investment, the Fed is subsidizing the banks by reducing risk and providing liquidity. Charging for that service would put banks on a level playing field relative to other financial intermediaries.
Hummel mentions that banks would never borrow from the Fed's discount window if they had to pay interest to the Fed and also pay interest on the reserves they hold. However, banks could still borrow to meet reserve requirements. Sumner advocates charging interest on excess reserves only. Of course, the Fed is lending huge amounts of money to banks at this time. However, it is a mistake to assume that the Fed is lending money to the same banks that are holding excess reserves. Currently, the banks pay a quarter percent more than they earn.
Suppose a large money center bank had been funding a huge portfolio of mortgage backed securities using off balance sheet commercial paper, overnight borrowing from other banks, and large, brokered, certificiates of deposit. There is a loss in confidence in this bank, and all of those sources of funds dry up. The Fed lends money to the bank. The bank obtains reserves immediately, but pays them out as it pays off the commercial paper coming due, pays off overnight loans from other banks, and pays off the large brokered certificates of deposit. It is other banks, ending up with those additional reserves that would have excess reserves and who would have to pay for holding them. The Fed would be bearing the risk of funding the portfolio of toxic assets, while the banks who have those toxic assets transformed into something risk free and perfectly liquid would pay the Fed for this service. If the banks would instead prefer to hold some other kind of asset, perhaps less risky and more liquid than toxic assets, but more risky and less liquid than reserve balances at the Fed, they should be able to do so, and perhaps earn interest rather than pay.
Sumner has suggested that the Fed also assess banks on their holdings of vault cash. My view is that the Fed should instead keep the charge for holding reserves less than the cost of holding vault cash. Setting up a more careful scheme for reporting vault cash and then imposing charges seems excessive. If temporary, setting up this system is more trouble that it is worth. If permanent, trying to have some kind of Fed police stamp out currency warehouses would be futile.While I don't think Hummel's notion that T-bills will replace base money makes much sense, he may have gotten an inkling of an equilibrium where all of the highly liquid, low risk assets have slightly negative yields. FDIC insured transactions accounts are perfectly liquid and riskless. Savings accounts are little different. FDIC insured certificates of deposit are highly liquid and riskless. T-bills are highly liquid and riskless. Reserve balances at the Fed are perfectly liquid and riskless. If there is an excess demand for these assets at a zero yield, why shouldn't the yields on all of them be negative? That the Fed provides riskless, perfectly liquid currency with a zero yield may make it impossible for nominal interest rates to be negative enough for the markets for these other financial assets to clear, but there is no reason to try to keep their rates positive, zero, or anything less negative that the storage cost of currency.
Hummel correctly points out that the Fed can always clear these markets by just increasing the quantity of money enough. So, suppose the Fed is purchasing zero interest T-bills to expand the quantity of money. The Fed is supposed to provide this intermediation for free? And if the Fed is instead purchasing longer term government bonds, or even private securities, then Fed is bearing the risk. Why should it provide this subsidy?
The greater the subsidy the Fed provides to those holding highly liquid and low risk financial assets, the greater the amount of quantitative easing that the Fed must undertake, and the greater the risk that the Fed must bear. In the final analysis, depending on a larger increase in the Fed's balance sheet while keeping interest rates above zero means that the taxpayer is responsible for more risk, and is providing a subsidy to those who want to hold low risk, highly liquid assets. If those holding those sorts of assets pay for the privilege, then less of a subsidy is provided, less quantitive easing is required and less risk is imposed on the taxpayer.