Not long ago, nearly everyone thought that nominal interest rates could not go below zero. Now, we have negative policy rates in the euro area and Japan, while in Sweden and Switzerland, the lowest controlled rate is below -1%. And government securities worth trillions of dollars bear negative rates, too.
When we first wrote about negative rates a year ago, we argued that the effective lower bound (ELB, rather than ZLB) for nominal rates was determined by the transactions costs of storing and transferring cash. While Fed staff estimate pure storage costs -- exclusive of security and insurance -- at about 0.35%, we reasoned from the behavior of money market mutual funds with broader transactions costs that the ELB might be in the range of -0.50% (minus one-half percent). Below that, we thought, there would be a move into cash, facilitated by banks and others who would efficiently manage the notes for clients.
The pure storage costs seem modest. After all, SFr 500 million of 1,000 Swiss Franc notes takes up only about a cubic meter of space. (For €500 euro notes and $100 bills, a cubic meter is on the order of €300 million and $90 million, respectively.) Why wouldn't banks offer accounts that were claims on a part of this relatively small pile of cash in their vault? During the day, the funds would be available just like a checking account, but would be swept into cash currency rather than being held as reserves at the central bank. Overnight, the piles of currency belong to customers, while during the day they are the asset balancing the cash account liability on the bank's own account.
But, at the negative rates that we have seen so far, cash in circulation has not spiked. In Switzerland, for example, after more than one year of subzero rates, cash holdings have risen by less than 9%, only slightly above the average growth rate of 6½% over the previous five years. (There have been news stories about increases in the holdings of SFr1000 notes, but the reported increase of SFr5 billion in late 2015 is still small, accounting for just over 1% of the total monetary base.)
This leads us to ask two questions. First, how much further can nominal interest rates fall (that is, how low is the ELB)? And second, what role should negative rates play in the future?
The answer to the first question is that we still don't know (for example, see here). However, given that rates have fallen this far without a currency surge, we are inclined to think that negative interest rates (at least this low and possibly lower) should become a regular part of the central bank toolkit; and for reasons we will explain, this tool is probably best used before resorting to quantitative easing (QE), not after.
To understand these conclusions, we start with a discussion of the economics of negative rates. When it comes to the transmission of monetary policy to the real economy -- how it is that reducing interest rates translates into higher growth and inflation -- there is nothing special about zero. A simple way to see this is to recall that the economic impact of monetary policy is based on real interest rates. Whenever expected inflation is higher than the nominal interest rate, the real interest rate is negative. Indeed, short-term real interest rates have been less than zero in much of the world since 2009. So, when nominal interest rates also go below zero, there really isn't anything unusual going on -- at least, in theory. All the channels of transmission -- reductions in saving and increases in borrowing, upward pressure on asset prices, and downward pressure on exchange rates--should still be at work, so long as people don't prefer cash to other assets. (For a cross-country comparison of negative interest rate policies, see here.)
To get a sense of what happens in practice, consider the following chart of Swedish interest rates. The plot shows the policy rate (labeled "repo rate," in black), the two-year Swedish government bond rate (in red), and the two-year rate mortgage bond rate (in blue). As the policy rate goes down, the sovereign and mortgage rate go down with it. It certainly seems that when policymakers change their interest rate target, it moves the entire constellation of interest rates. And, zero does not seem to be any impediment. At least, not yet.
Interest rates in Sweden (monthly), 2000-2016
Source: Sveriges Riksbank.
To understand how this works, and why other market rates follow policy rates below zero, we need to think about the decisions of a bank and an investor. Given an additional deposit, a bank has several options. It can: (1) deposit the funds in its reserve account at the central bank and receive the rate paid on excess reserves; (2) purchase a government security; or (3) purchase a risky security (or make a risky loan). Focusing on the first two, which are both free of default risk, the bank's arbitrage activity will drive short-term government bond rates down in line with reductions in the interest rate on excess reserves (see the left-hand side of the diagram below). As the chart for Sweden shows, risky rates that earn a spread over the benchmark sovereign usually will follow.
Because investors also hold government bonds, their decisions are important as well. Not having access to the central bank, an investor who receives an extra dollar can: (1) purchase a government security; (2) make a deposit in a bank; or (3) hold physical currency (see the right-hand side of the figure above). The implication is that, as the sovereign interest rate rises or falls, the bank deposit rate should follow. When interest rates are positive, this is what happens (eventually).
But here we have the first complication created by negative policy rates: banks are reluctant to reduce their deposit rates below zero. If you ask bankers why they resist passing negative interest rates on to their depositors, they usually express concern about losing customers to competitors. Yet, as far as we can tell, no one has ever shown even the slightest resistance to paying negative real interest rates. We suspect there is a first-mover problem with passing on negative nominal rates. Once everyone does it, the taboo will be gone.
You might also hear the claim that bank IT systems can't handle negative deposit rates, but this is doubtful. Since they can clearly handle negative rates on assets, it is not a big leap to managing negative rates on liabilities. And, if there's a fixed cost of changing software, banks would make the change if negative rates become a normal part of the monetary policy toolkit.
So, it looks to us as if cutting the policy rate to current negative levels can work just like cutting the rate when it is positive. There is, however, another potential obstacle. This one is associated with the introduction of quantitative easing before cutting rates below zero. To put it bluntly, after a central bank forces banks to hold massive excess reserves, how can it then tax those same excess reserves by setting a negative deposit rate?
Japan is the most extreme case. Over the past three years, as a consequence of the central bank's quantitative and qualitative easing program (QQE), excess reserves have increased six-fold and currently stand at ¥230 trillion, or nearly one-quarter of banks' ¥950 trillion in assets. With a return on assets of 20 basis points, the Japanese banking system has profits of roughly ¥2 trillion. If the Bank of Japan were to lower its deposit rate to -1%, earnings would be wiped out, unless this cost can be quickly passed on to liability holders.
For this reason, most central banks implementing negative rate policies have done it using tiered systems. As the figure above shows, the BoJ is paying a positive rate on a base level of reserves, zero on an amount that will increase with the total level outstanding, and a negative rate only on the marginal reserve holdings show in pink. The Danes, Swedes and Swiss are all using comparable mechanisms that shield "infra-marginal reserves" from the tax. The only outlier (at this writing) is the ECB, which charges the negative rate on the entirety of banks' excess reserves.
For readers familiar with the economics of public finance, you can think of this tiered approach as an excise tax in which all of the revenues -- except for the last bit received on the marginal reserve holdings -- is repaid as a lump sum. The marginal tax rate determines behavior while the re-distribution of the revenue offsets the income effect that the tax would otherwise have. The key is that the re-distribution mechanism cannot be linked to the recipient's current or future behavior: the central banks' solution is to pay the subsidy based on the level of past reserve holdings.
Why provide any subsidy? Policymakers appear to be reluctant to push commercial banks into charging negative rates to their clients so long as assessing the negative reserve rate only on marginal holdings is sufficient to achieve their goal of lowering market interest rates. There is a catch, however. The subsidy to the commercial banks shows up on the central bank's balance sheet as a negative interest rate spread between the zero interest rate paid on infra-marginal reserves and the negative rate earned on short-term government debt. (One might rightly ask whether this isn't a fiscal policy, but that is for another day.)
This leads us to ask whether, if central banks wish to use negative interest rates in their standard tool kit, wouldn't it make sense to do so when the quantity of reserves before QE causes excess reserves to balloon? Put differently, once the target interest rate falls to zero, so that the level of reserves is determined by the central bank supplier, reducing the target rate down to the ELB may be a more effective policy than introducing QE.
Finally, we return to the question of just how low rates can go. Where is the ELB? As government rates go down to -1 or -2 or even -3 percent, won't someone start to create cash facilities that provide a better return for clients? To be sure, we don't see private individuals or most nonbank financial firms getting into this business. The reason is the cost of enforcing very strict anti-money laundering (AML) regulations, especially the "know your customer" rules. Not only are the operational costs substantial, but the penalties for unwittingly facilitating criminal activity are very high. What this means is that if a private person or a nonbank firm were to attempt withdraw a few billion dollars in cash, they would almost surely incur heavy monitoring by the authorities -- both criminal and tax. (Credit card network operators like Discover and Amex, as well as money transfer firms like Western Union, already deal with AML issues, so they may be willing to create stored value cards that provide access to a pile of cash sitting in a vault somewhere.)
That leaves us (mostly) with banks. Here, assuming the willingness of the central bank to supply large-denomination notes in unlimited quantities, space is not the constraint. Instead, we suspect that insurance is. It may not currently be feasible to insure massive amounts of cash against loss. In Denmark, for example, the maximum insurance one can get is for DKr25,000 (less than US$4,000). As a result, banks self-insure their cash positions. Perhaps large insurers will enter this business if the price is right, but it is unlikely to be cheap. And, it may be inconvenient. Not wanting the concentration risk, the insurer may require that the pile of cash be broken up and placed in multiple locations. But that would raise transport (and related security) costs.
Another consideration is that central banks could depress the ELB by discouraging banks from demanding a massive volume of large-denomination notes. Such "moral suasion" probably can work when interest rates are only modestly or temporarily negative, but should rates fall sufficiently far for an extended period, a central bank would be forced to impose quotas if it wished to contain cash demand.
As we continue trying to figure out what the transactions costs are for storing, transporting, and insuring large amounts of cash, we have two further thoughts. First, on a per unit basis, these costs are probably lower the larger the jurisdiction. That means that the floor on interest rates is likely higher in the euro area, Japan and the United States than in Sweden, Switzerland or Denmark (that is, even if the Fed ultimately concludes that it has the legal authority to set the interest rate on excess reserves below zero). Second, the floor is probably a soft one because the eventual prospect of higher interest rates limits the potential profit from a large investment in cash management. That is, because of the fixed costs involved in setting up cash accounts, and the threat that rates will rise, banks will hesitate to pay the start-up costs until rates are expected to stay low enough long enough to warrant the risk.
The bottom line: international experience suggests that negative interest rates, at least as low as we are seeing today and (in some places) significantly lower, will become a permanent part of the monetary policy toolkit. If that's right, we need not worry quite so much whether a 2% inflation target is too low.
An earlier version of this post appeared on our blog www.moneyandbanking.com.