I am pleased that Greg Mankiw has recently given the interest penalty on reserves idea more visibility. (It is also nice to be called a “prominent economist” by Mankiw.) Unfortunately, like the rings of Saturn, this issue is much more complicated than it appears at first glance. Many economists try to look at the entire liquidity trap issue by boiling it down to a few fundamentals; the perfect substitutibility of cash and T-bills at zero rates, or the problem of credibly promising a permanent expansion in the money supply. Unfortunately none of those easy thought experiments are enough. The problem must be examined from multiple perspectives at once. Hence I plan a long post that will have readers crying “No mas, just try Sumner’s plan so I don’t have to read any more of this.”
Any examination of this issue must start with a few descriptive statistics. These are off the top of my head and may be a bit off, but are close enough.
Cash = $800 billion, Required reserves = $80 billion, Excess reserves = $8 billion
Cash = $850 billion, Required Reserves = $90 billion, Excess reserves = $800 billion
T-bills are somewhere around $2 trillion.
So from a monetarist excess cash balance perspective, the problem is the hoarding of excess reserves by banks. This is partly due to the payment of interest on reserves, which although only 1/4 percent, is slightly above the rate on T-bills. And reserves are superior to T-bills in maturity and equal in default risk. So it’s no surprise that lots of reserves are hoarded. The experience of Japan in the early 2000s and the U.S. in the 1930s, however, indicates that substantial reserve hoarding can occur even without explicit interest payments.
We also know that banks hold very low levels of ERs any time the opportunity cost (in terms of the T-bill alternative) is even modestly positive. Thus in the summer of 2008 when the target rate was only 2%, ERs were still very low. So a penalty rate on excess reserves of say 4% should bring ERs down to extremely low levels. In my view the Fed would then have to dramatically reduce the monetary base in order to prevent high inflation. But Mankiw is not convinced, so let’s look at this issue more closely.
Let’s start by assuming that even with a negative 4% rate on ERs, the yield on T-bills stays near zero, or perhaps just negative enough reflect the cost of storing cash in safety deposit boxes. (BTW, if you are wondering what would happen if T-bill yields also fell to minus 4%, one answer is that it would end the liquidity trap. So we could stop right there, QED.) But I think most would agree that that T-bill yields will go, at best, slightly negative. Mankiw’s argument is that the interest penalty would have to be passed on to depositors in the form of negative rates on bank deposits, and that this would cause people to simply hoard cash in safes.
Even if Mankiw is correct, I will later make a number of arguments that this may be much less of a problem than he thinks. But before doing that, let me question this assumption. Suppose the government pays 4% positive interest on required reserves and negative 4% interest on excess reserves. The penalty rate on excess reserves affects behavior at the margin, and should reduce ER holdings to the very low levels that existed before reserve and T-bill rates were nearly equalized. In that case let’s say we go back to $90 billion in RRs and $9 billion in ERs. Now assume that people begin to move away from T-bills because of slightly negative yields. Do they have to go into cash? I don’t see why. When banks expand their balance sheets to have more deposits, they will expand everything else proportionately. Because RRs are ten times bigger than ERs, the net flow of government cash to banks will increase, not decrease. So if RRs are 1/10th of deposits; and if ERs are 1/100th of deposits, then the flow of interest will be 3.6% of reserves, and 0.36% of deposits. Just enough to be slightly better than cash; and it’s FDIC-insured with no need to buy a safe!
There is the possibility that if T-bill rates go slightly negative, that cost would also have to be passed on to consumers. But of course this means the Fed would also pay less in interest on its debt, so it would not be a problem to slightly raise the interest rate on RRs to make sure people put hoards into deposits, not safety deposit boxes (where there is no multiplier effect.) Ten trillion in new bank deposits should be more than enough to get the monetarist excess cash balances effect working at many different margins, not just T-bills (which total around $2 trillion) but all sorts of other assets as well. Others can check my numbers, but I think it would be much less expensive to the Treasury than fiscal stimulus, even if pursued for a year. And I would expect it to push the U.S. out of a liquidity trap much sooner than that.
Obviously this idea might not work, but even if all the ERs that banks no longer wanted to hold went out into cash held by the public, I believe the effect would be much better than Mankiw anticipates. Remember that the whole point of monetary expansion is to stuff more Federal Reserve notes down the publics’ throat, then what they want to hold. (Call it a foie gras policy.) As the public tries to get rid of excess cash balances, AD starts rising. I think even Krugman would accept the fact that if enough money was injected into the economy, inflation expectations would rise. The last time I made that claim he denied holding this view, so let me be more specific—I think even Krugman would accept that if the Fed printed enough cash to buy up all the assets on Earth, then inflation expectations would rise. Of course he might reasonably object that this would be very risky. It would either lead to hyperinflation in the long run, or would have to be later withdrawn, which could easily lead to massive capital losses by the Fed.
If the real problem with QE is that to be effective it would have to be so large as to be risky, then let’s make it as small as possible. The Fed has already announced that they intend to do a sort of QE-type policy, but they are starting from a position of $800 billion in extra base money that is currently doing absolutely nothing for the economy. If economists think QE is an idea worth pursuing, why not start by making things as easy for the Fed as possible, why not stop the one institution from hoarding that they are able to stop from hoarding, the commercial banking system. The penalty rate on excess reserves would reduce bank demand for base money by roughly $800 billion, meaning that we could inject an extra $800 billion into circulation with no increase at all in the monetary base, and thus no increase in the Fed’s exposure to possible capital losses. That’s what I’d call a good start. Maybe it will all be put in safes (although I doubt it.) But even if so, wouldn’t it be nice to find out now, instead of waiting until the Fed injected another $800 billion in new base money, above and beyond the current bloated monetary base?
But the preceding $800 billion head start on current policy is just the beginning of the advantages offered by an interest penalty on reserves—there are many more. For instance, the demand for cash is probably far less elastic at zero interest rates than most academic economists suspect. Most economists work with abstract models that totally ignore the real reasons why people hold cash. For most of the public, cash and T-bills are not close substitutes. People basically hold cash for two reasons: transactions, and much more importantly, for tax evasion purposes. Total cash in circulation is close to $3000/person. Admittedly some is held overseas. But even in smaller countries whose currency is not hoarded overseas, the cash ratio is fairly high. What explains this? It is the underground economy—cash is very anonymous and thus a good way of hiding wealth from the authorities (especially the IRS.) For this reason, the hoarding demand for cash is not highly elastic in the short run at near zero rates. Don’t believe me? Then ask yourself why as interest rates have fallen almost to zero, the cash to GDP ratio has only increased slightly.
People who hoard cash will do so even at 5% nominal rates, as they can avoid paying 30% of unreported gains to the IRS. Indeed the ratio of tax rates to interest rates correlates closely with tax hoarding in the long run, as it represents how long you can hoard cash before you would have been better off paying your taxes in the first place. It’s true that in the very long run zero nominal rates would lead to considerably more hoarding by the public, but the other problem faced by underground economy hoarders of cash is that it is very costly to adjust their cash holdings in the short run. They can’t go in and out of T-bills; rather they must adjust their flow of (cash) consumption purchases. Look at the picture here, and ask yourself how easily the drug dealer could double his cash holdings, or cut them in half, and whether he even cares if T-bill yields go slightly negative. And the short run is all important for the liquidity trap, as it is widely believed to be a short run problem (indeed this is the essence of Krugman’s “expectations trap” argument.) The Fed needs a monetary expansion that is expected to last until the economy exits the liquidity trap. If that exit is expected to be soon, then the problem is much easier to solve.
The other part of real world cash demand is transactions balances. Most people walk around with very small cash holdings in their wallets. And the amounts don’t seem particularly sensitive to interest rates in the short run. I believe this may have to do with fear of losing their wallets and/or having them stolen. If so, then these transactions balances still might not balloon very much at zero rates, or even slightly negative rates. In that case Mankiw is speculating that an entirely new type of cash demand will occur at slightly negative rates, legal economy individuals holding cash as an investment. This type of cash demand is probably very small right now, but could obviously become large if cash was seen as a desirable substitute for T-bills.
The question is how much cash is the public actually likely to hoard in the short term? How much food can we force down the goose’s throat before it can’t take any more? My hunch is that inflation expectations would rise significantly before anywhere near $800 billion was moved out of banks and into circulation. I cannot prove that, but I believe I can prove that it would at least make QE more effective. My proposal would offer the following three advantages over the current QE program:
1. Cash hoarding would only occur at one margin (the public), not two.
2. Penalty rates on excess reserves would eliminate cash hoarding on the margin where it currently is far more of a problem (excess reserves.)
3. The QE program would start from a monetary base position roughly $800 billion lower than the current base, thus greatly reducing future potential capital losses by the Fed.
To those who don’t believe me, I would ask the following question: If eliminating positive interest on excess reserves and going to interest penalties doesn’t help, then what would be the harm from raising the current positive interest rate even higher? Why not raise it high enough so that the extra $1 trillion the Fed is contemplating injecting through QE is all hoarded by banks as excess reserves? Most people would say that’s a terrible idea, it would insure that QE would definitely fail. But if that is a terrible idea, insuring the failure of QE, then my interest penalty idea must be a good one, making it more likely that QE will succeed. Notice I didn’t say guaranteeing it would succeed, just making success more likely.
Keep in mind that even people like Paul Krugman have said QE is worth a shot, and his model predicts it can work if the Fed’s inflation target is credible. If even economists known as being particularly skeptical about monetary stimulus in a zero rate environment think it is worth a try, why not do it in the way most likely to succeed? And again, one can’t argue that my plan wouldn’t make success more likely, unless one is willing to argue that the exact reverse policy (much higher interest rates on reserves to insure the extra QE reserve injections are hoarded) would not make the failure of QE more certain. And I can’t imagine anyone making that argument.
Now we have exhausted all the monetarist transmission arguments that I can think of, and yet we have barely scratched the surface of the much more complex problem of liquidity traps. So now let’s return to Krugman’s expectations trap, the argument that the real problem during a liquidity trap is a lack of policy credibility, the fact that the public might not believe the Fed’s promises to raise the price level at a particular rate over time. Indeed, this may be the problem we face right now, as the Fed hints that it wants roughly 2% inflation, but the five year TIPS spreads show only about 0.8% expected inflation.
The good news is that no determined central bank has ever failed to inflate. The bad news is that very few have tried at near zero rates. Thus many people wrongly believe that near zero rates represent a barrier to inflation, when they actually represent the market correctly understanding that the central bank is engaged in deflationary policies (the U.S. in the 1930s and Japan more recently.) In neither case did the central banks do the things (like currency depreciation) that you would expect from a central bank determined to reflate. The only exception was the 1933 dollar depreciation program, which was successful in reflating prices.
Apart from all the advantages that I have already discussed, an interest penalty would be a strong signal from the Fed that it was serious about reflating. So apart for the monetarist excess cash balance mechanism, you would also have the mechanism that Krugman called for—higher inflation expectations. But the Fed should not stop there. They must back up this action with a firm promise to follow a nominal aggregate trajectory (level targeting) of either a rising price level (100, 102, 104, 106, etc.) or rising NGDP (100, 105, 110, 115, 120, etc.) They must also promise to try to “catch up” should they fall short. This will further increase inflation expectations.
The preceding is almost certainly enough to get the job done, but I’m just getting started. They have to be willing to engage in the “nuclear option,” if necessary pegging the price of a real asset such as gold or a foreign currency. And then adjusting that peg until prices started rising. In my view those two nominal anchors are not desirable (especially exchange rates, which could cause all sorts of diplomatic problems.) Instead, the Fed should announce that if all else fails, if one month after their new policy is announced inflation expectations remain below 1.8%, they will start pegging the price of a 12-month CPI futures contract at 103, and then in following years 105, 107, 109, etc. Why three percent in the first year? To insure that you scare markets enough so that the nuclear option is not necessary. Remember you are still doing all the other things discussed earlier (interest penalties, QE, explicit nominal targets, etc.)
As readers of this blog know, I would much prefer a NGDP growth target of 5% per year, but I am trying to reach a broader audience after the Mankiw link, and thus am using the much more generally accepted idea of an inflation target. This futures targeting idea has been around for several decades, and no one has yet found a flaw. As an aside, a slightly different proposal for futures targeting was criticized by Bernanke and Woodford in 1997 for having a circularity problem, but even they said the concept could work if you elicited a market forecast of the instrument setting required to equate CPI expectations with the policy goal. This is exactly what my 1989 and 2006 proposals do. The Fed offers to buy and sell unlimited CPI futures at the target price. Each transaction triggers a parallel OMO, so that purchases and sales continue until the monetary base is at a level where the publics’ inflation expectations equal the policy target. Like other “foolproof” escapes from liquidity traps, it contemplates targeting the nominal price of a real asset. But unlike other proposals it is not subject to the problem of policy lags.
I don’t think the Fed would enjoy doing something as experimental as futures targeting right now. But because it is a foolproof plan, just the threat of doing so, when combined with the many other steps proposed above, would mean that they would not actually have to use it. And consider all the wild experiments they have already done. Doesn’t it make more sense to commit to a highly experimental policy that there is less than 1 in 100 chances that you’d actually have to implement, and which will definitely work, than to continue on with other highly experimental and extremely costly policies that we already know (from TIPS spreads) are not even expected to work?
1. Unlike other posts this was rushed out in response to the Mankiw link. Thus I reserve the right to modify any ideas discussed here that have not been “vetted” yet, and are later shown to be foolish.
2. Some people ask about vault cash. That is not a problem. It could be treated as reserves (and thus subject to the ER penalty), or exempted but capped at no more than X% of deposits.
3. I’m not a banking expert. If the current crisis means banks need more reserves than usual, then raise the reserve requirement to a level appropriate. Or keep it as is, but only charge the penalty on reserves X% above the required rate. Nothing of substance would change in my argument.
Update, 4/25/09: JKH pointed out that the average reserve ratio is closer to 1%. Bill Woolsey said that that reflected a 10% rate on transactions accounts, and 0% on savings accounts. In that case I would prefer a 1% reserve ratio on all accounts, and thus an equal subsidy to all accounts. As a practical matter that would mean actual hoarding would occur in saving accounts, where lower transactions cost would mean a higher interest rate. If the money was hoarded by the public in savings accounts, $800 billion would support $80 trillion in M2 growth. Where would all the bank assets come from? Hyperinflation.
Bill has argued strongly that the interest subsidy is not needed. I still think it refutes Mankiw’s point, but I also think Bill’s probably right. The interest penalty on excess reserves is enough, when combined with QE and an explicit nominal target. So I’ve come around to Bill’s view that there is no point in having an interest subsidy on RRs.