George Selgin testified before a Congressional Committee about the Fed's policy of paying interest on reserves.
I agree with much of what he said.
Implementing interest on reserves in 2008 explicitly aimed at restricting aggregate demand and inflation was foolish. It helped lead to disaster.
Further, paying banks to hold onto money rather than lend it is not wise during a financial crisis.
I also agree that a policy of paying interest on required reserves with no interest on excess reserves is better right now than the status quo.
On the other hand, I think reserve requirements are a bad policy and so that makes any distinction between required reserves and excess reserves irrelevant except as a compromise.
Further, I would like to see the demand for reserves by banks be independent of interest rates. The best way to do that is to both pay interest on reserves and make that interest rate vary with market interest rates.
One policy would be for the Fed to pay less on reserves than market determined T-bill yields. When T-bill yields change, the Fed would need to promptly adjust the interest rate it pays on reserves. It makes reserves into a type of money market account.
Given the liquidity and safety of T-bills, they are a close substitute to reserves for banks. If the T-bill rate rises, that would tend to increase the opportunity cost of holding reserves if the rate paid on reserves was fixed, even at zero. This would lower the demand to hold reserves. If the interest rate paid on reserves rises in proportion, there would be no impact on the demand for reserves.
More relevant to today, if the T-bill rate should fall, this would reduce the opportunity cost of holding reserves and raise the demand for them if the interest rate on reserves were fixed, including at zero. If the interest rate on reserves instead fell as well, then there would be no tendency for lower interest rates to raise the demand for reserves.
In my view, given the current very low interest rates on T-bills, the appropriate interest rate on all reserves is negative. If this creates too much of a burden on required reserves, then reserve requirements should be reduced, preferably to zero.
An alternative policy is to treat reserves like a mutual fund claim on the Fed's asset portfolio, while charging banks a management fee. The yield banks would earn on reserves would change with the yield on the Fed's asset portfolio--presumably with the yields on assets that banks can hold directly. This would make the banks' demand for reserves independent of interest rates as well.
Such a policy would expose the banks to credit risk on the Fed's balance sheet. While I don't think it is desirable for the Fed to expose banks to risk by purchasing risky assets, I consider the alternative of the Fed providing interest bearing zero risk assets to the banks while itself holding risky assets to be even worse.
With such a policy, I think a "bills only" policy for the Fed's asset portfolio would be feasible. (At least ignoring currency for a moment.) This would make the mutual fund type reserves pretty much equivalent to the money market account reserves. The yield banks earn on reserves would be slightly less than the interest rate that can be earned on the T-bills the Fed would hold in its asset portfolio.
I favor keeping the interest rate the Fed pays on reserves below T-bill rates (or having the Fed charge banks an ample "management fee.") Part of the reason is to harness the stabilization process that Selgin himself discovered. The difference between the interest rates banks can obtain on earning assets and the interest rate they receive from the Fed is the opportunity cost of holding reserves. Banks must weigh that against the transactions costs of adjusting their reserve position by trading securities. This determines the reserve balance that it is most profitable for banks to hold. However, it also depends on the variance of gross clearings by banks which in turn depends on aggregate nominal expenditure. If nominal expenditure grows at a slow steady rate, this should keep the demand for bank reserves also growing at a slow steady rate. Normally, then, the Fed could maintain monetary equilibrium by keeping the quantity of bank reserves growing at a slow, steady rate.
As I understood Selgin's first version of the argument, he claimed that a fixed quantity of bank reserves would result in fixed equilibrium level of nominal expenditure on output. The version I describe above has a slightly different emphasis but is based on the same reasoning.
However, if the Fed is providing an asset with a fixed interest rate and no risk, then shifts in market interest rates or even the risk of bank assets will cause fluctuations in the demand for reserves interfering with the process described above. Of course, it would simply require that the Fed manipulate the quantity of reserves. More risk or lower market interest rates, the Fed would need to create more reserves for the banking system. Less risk or higher market interest rates, the Fed would need to create fewer reserves. In my view, it is better to avoid the need to make these sorts of adjustments in the quantity of reserves, and so that is the rationale for the mutual fund type reserves.
I have always had less confidence than Selgin in the process he discovered. I think there is an important element of truth in the argument, but I don't favor either a fixed quantity of reserves or a constant growth path for reserves. And so, I also favor index futures convertibility--more or less like Scott Sumner's proposals. (Oddly enough, I also favor more central bank discretion than either of them!)
Finally, I reject any notion that the Fed or any other central bank should create "costless" reserves, paying interest on the reserves equal to "the" interest rate. I have no idea why anyone would think that having the central bank manage all credit allocation would be desirable or wise.