The academic J.W. Mason did an interesting piece of analysis in "Interest Rates and Bank Spreads." He was responding to an internet debate, which I am not directly addressing. In Mason's article, he crunches the published average bank interest rate charges (both lending and borrowing), and shows that they are consistent with a relatively steady spread regardless of the level of interest rates. Luckily, I do not have to download the data and analyse it myself; he did the work for me. Instead, I want to explain why we should expect this spread behaviour to occur, and an interested reader can then consult his analysis to see that the theory matches observed behaviour.
Why Take Duration Risk?It is certainly true that banks "lend long and borrow short," that is, they own assets with longer average maturities than their liabilities. This is then converted into believing that banks have a perpetual duration* mismatch on their balance sheets, and so they are exposed to interest rate risk. This is augmented by the experience of the early 1990s, which I discuss at the end of this article. Within the economics community, that early 1990s experience has entrenched itself in the received wisdom, and no attempt has been made to keep up with developments within the financial system.
In practice, vanilla interest rate and currency risks are the only risks that can be accurately measured by risk management tools. As soon as we start to incorporate credit, equity or optionality risks, risk models are pretty much an elaborate exercise in validating the dictum "garbage in, garbage out." Regulators and bank executives can monitor the duration risk of a bank quite accurately, and they keep the risk "small" relative to the size of bank.
Of course bank treasury units are not major players in interest rate trading. The risk limits they have are absolutely large when compared to smaller players in fixed income; but even so, their risk is small relative to their parent bank's balance sheet.
Although it is clear that regulators would want banks to be duration neutral, one might cynically think that bankers would want to get around the stodgy regulators and secretly run duration risk. However, this does not take into account real world behaviour. I started in finance in 1998, and during that period, the overwhelming consensus was that bond yields were always "too low." As a result, investors probably underperformed in aggregate versus their benchmarks on the basis of their interest rate exposure (but they tend to bailed out by their credit carry positions). Therefore, bank executives either agreed with their economists and did not want to be long duration, or else they realised that interest rate forecasters are at best unreliable. As a result, the best course of action is neutralise interest rate risk and concentrate on whatever the bank's "core competencies" are supposed to be.
Interest Rate Risk Versus Liquidity RiskThe "borrowing short/lending long" practices of banks do not expose them to interest rate risk, rather liquidity risk. Unfortunately, academics cannot model liquidity risk with a random process, and so they really have no way of grasping the concept. People who believe in perfectly liquid markets in all assets (including forwards out to infinity) where the prices are set by a single perfectly rational household are going to have a very hard time visualising liquidity risk.
- If you buy a 10-year bond at 3%, and are guaranteed to fund it by rolling 3-month loans every 3 months, your trade will be profitable if the average funding cost is less than 3% over the next 10 years, and unprofitable if the average is above that (assuming no default). These profits or losses are an interest rate risk, and it was standard economic/financial models deal with.
- If you buy a 10-year floating rate bond which pays a of 3-month LIBOR +1%, and you are guaranteed rolling funding, your position will be profitable if the average funding cost is less than 3-month LIBOR + 1%. Your risk in this case is the possibility of your floating rate funding costs rising, presumably due to a weakening of your balance sheet. This could be termed funding cost risk. A bank that is worried about funding cost risk is a bank that is about to go out of business; you cannot survive as a financial intermediary if your cost of funding is higher than what you are lending at. As a result, this is not normally a topic of discussion.
- If you buy a 10-year floating rate bond that pays 3-month LIBOR+1% by borrowing for 3 months, you are exposed to the possibility that your lender will not renew the loan. Since the bond will not have matured, you will need to find a means of paying back the original lender. You might be forced to sell the bond in the market, quite possibly at a distressed price. This risk is known as refinancing or liquidity risk.
How Banks OperateAs always, the real world is complicated. But as a simplification, you can view a bank as being a combination of a number of different lending businesses, with a treasury acting to coordinate activity and manage aggregate interest rate risk.
Each unit has a cost of funding (which depends upon the riskiness of its assets), which it is charged, and holds assets. If those assets have a long duration, the unit creates an internal swap with the bank treasury to convert the asset to a floating rate bond plus a spread. (Other than for units that routinely operate in the capital markets, this presumably would not be an explicit swap. Rather it would be an accounting convention that acts as the economic equivalent of a swap. I refer to it as a swap to give the underlying financial market equivalent of the accounting convention.) This way both sides of the lending units' balance sheets are both floating rate, and immunised against policy rate changes.
The treasury ends up facing all of these "internal swaps," and it then has to manage the aggregate risk position of the bank. If the lending units own a lot of assets with fixed interest rates, the Treasury will end up with an unbalanced duration position. It will need then to trade with other fixed income investors in order to balance it's (and the aggregate bank's) books.
It can do this by:
- issuing term deposits;
- using swaps and futures;
- issuing fixed coupon bonds or preferred shares;
- taking a small amount of proprietary trading risk.**
This means that only the Treasury is managing aggregate interest rate for the bank. It would make no sense to have one lending unit going massively long duration, and another going short with the net effect of the two bets balancing out. If interest rates moved, one group or the other would be successful (and paid bonuses), while the bank in aggregate would make no money on interest rates. By centralising interest rate risk at the treasury unit, it is clear who is responsible for interest rate risk management.
The data shown in J.W. Mason's post are consistent with this type of operating procedure, in which interest rate risk is largely hedged out.
Post-Script: What About The Early 1990s?
The early 1990s cycle saw the Greenspan Fed keeping interest rates as "unsustainably" low interest rates as a means of allowing banks to rebuild their balance sheets after previous misadventures. Since the level of rates was presumably affecting bank profitability, it is clear that banks then did not have the duration of their liabilities and assets matched.
As this affected the level of the policy rate, this greatly impressed monetary economists, and so this episode is burned into their memories. As a result, this mechanism ("low interest rates are good for bank net lending spreads") is kept in mind.
However, they did not notice the Great Bond Bear Market of 1994. A few firms blew up when interest rates where "renormalised," and it scarred a generation of financial market participants. (It possibly explains why senior people in finance are dedicated bond bears.) At the same time, advances in digital computing allowed for widespread risk measurement. (Before 1994, it in unclear how many non-specialists even understood what duration was, never mind having the ability to measure the DV01 of a swaps book.) As a result, regulators and the banks themselves clamped down on interest rate risk, leading to the environment that I described above.
* Duration is a measure of the sensitivity of an asset's price to changes in interest rates. A bond with a higher ("longer") duration will have a greater percentage loss in value than one with a lower duration, if the yields on both bonds increase the same amount. Increasing the maturity of a bond will increase its duration, all else equal. (A bond could have a lower duration than a bond with a closer maturity date, if its coupon was higher.)
** In the United States, the residential mortgage market is somewhat unusual in that it features long-term fixed interest rates, where the borrower can prepay with little penalty. This creates a hard-to-model optionality in residential mortgages. To a certain extent, the banks can offload that optionality onto the bond market by securitising the loans into mortgage-backed securities (MBS). However, my guess is that many smaller banks would not attempt to deal with the optionality embedded within mortgages that are left on their balance sheet; it is unclear which business unit within the bank would face the optionality.
(c) Brian Romanchuk 2015