So my question is about maturity mismatch and what is the appropriate method to calculate it. I have seen some simplified examples in academic books but, when I tried to approach it in a holdings bank, I could not really find a solution. Do I do a duration gap analysis? Cause for example in the balance sheet I am looking, the liabilities are less 1 year, between 1 and 5 years, and more than 5 years. Do I just look for the short term A-L ? Can you suggest me to an approach?
There’s two things of interest: interest rate risk, and liquidity risk.
For liquidity risk, you could compare short-term liabilities to short-term assets. This would be just a ratio. The long-term buckets (1-5) and (5+) aren’t going to add much information. Banks are in the business of rolling over 1-5 year term deposits, so that bucket will always be big. And since bank loans are often amortising, while the deposits have a bullet maturity, the mismatch between a 5-year deposit (bank liability) and a 10-year loan (bank asset) is much smaller than the maturity date suggests. This is why I would just look at ratios with the mid-long buckets merged.
For interest rate risk, the duration gap is what matters (plus interest rate shock simulations; see below). (Although I write duration, that is just habit. Technically, you need something like a DV01: how much money is gained/lost on a 1 basis point rise in yields? This can be inferred from an estimated duration.)
A part of the question is unclear - you are looking at a financial report of a real world bank? There might be a memo item buried in the notes on the financial statement that gives the exact mismatch. The bank and the regulators know what the mismatch is, it is unclear whether that is public information. Since it is sensitive commercial information, it is highly possible that the net duration information is not public. (I obviously do not know.)
You could try approximating the duration for each bucket of assets and liabilities, and then add them up. The problem is that this answer could easily be wildly incorrect, and the exercise is probably a waste of time.
- It ignores the optionality of mortgages (for a US bank). So straight duration is somewhat misleading, you also need to look at large shifts in the yield curve (which is another standard stress test applied by regulators and the bank’s risk management.)
- A bank can own or issue long-term floating debt. So even if the maturity is 10 years, the duration is at money market length.
- It ignores the fact that banks use interest rate derivatives (swaps) to hedge their duration risk.
If you dig into the regulator’s documentation (obviously depends on the bank’s jurisdiction), they should publish the risk limits for interest rates that they allow. Obviously, the bank’s internal models will give a number within the limits, as even the most incompetent regulator would notice a deviation. It may be that the bank has some exotic interest rate risk that does not show up on those stress tests, but you would never know by looking at public information anyway.