How to Read a Bank's Capital DisclosureHard

A bank publishes its own solvency and liquidity arithmetic. Four ratios carry it, and each answers a different failure — value, leverage, cash, and funding.

5 min read · 846 words

Read it in this order

• A bank can fail two ways — it runs out of value, or it runs out of cash — and this document answers both, in four numbers that are usually read as one.
• The order that works: the capital ratio → the leverage ratio beside it → the two liquidity ratios → the asset quality tables → the buffers and what happens as they are used.
• Everything here is a description of how the arithmetic is built. Nothing on this page is a view about any bank.

1. The capital ratio, and where the work is done

• The numerator is the loss-absorbing money: ordinary shares and retained earnings, less a list of deductions. It is the part that can go to zero before anybody else loses.
• The denominator is not the balance sheet. Every exposure is multiplied by a weight meant to reflect its risk, so two banks with identical assets can report very different ratios if their weights differ.
• Which is why the weights are the interesting half. Banks using their own models and banks using the standard weights are not directly comparable, and the disclosure says which is used where.
• Read the movement, not only the level. A ratio that improved because risk-weighted assets fell is a different event from one that improved because capital was raised or earnings retained.

2. The leverage ratio, which ignores the weights on purpose

• Capital over total exposure, unweighted. It exists because a risk weight is a model and a model can be wrong in the same direction across a whole industry.
• It binds where the weights are lowest. A book of assets everybody agrees are safe produces a high capital ratio and can still be a great deal of borrowed money, and this is the number that says so.
• Read it beside the capital ratio, never instead of it. One assumes the risk model is right and the other assumes nothing; the pair is the information.

3. Liquidity: the other way to fail

• The liquidity coverage ratio compares assets that can be sold or pledged quickly against an assumed outflow over a short stress. The assumptions about which deposits run are the whole content of it.
• Net stable funding asks the same question over a year: is long lending funded by something that will still be there.
• Both are stress scenarios with stated rules, not forecasts. A bank comfortably above both can still fail if the run is faster or more concentrated than the rules assume, which is what 2023 demonstrated.
• Deposit concentration is the missing variable. A book of many small balances and a book of few large ones can produce the same ratio and behave nothing alike.

4. Asset quality

• Exposures by stage separate loans performing normally from those where credit has deteriorated and those already impaired. Movement between the stages is the leading indicator, not the totals.
• Coverage is provisions against impaired exposures — how much of the expected loss has already been taken through the accounts.
• Expected credit loss is three estimates multiplied together: probability of default, loss given default, exposure at default. The disclosure gives the inputs, and a small change in any of them moves the provision a long way.
• Concentration tables — by sector, by geography, by counterparty — answer the question the averages cannot: whether the bad outcomes arrive together.

5. Buffers, and what happens as they are used

• The minimum and the buffers are different things. Buffers sit above the minimum and are designed to be usable in a downturn; that is what they are for.
• Using them has consequences that are written down: as capital falls through the buffer, restrictions on distributions apply automatically. That is why the distance to the restriction, rather than the distance to the minimum, is the number that constrains behaviour.
• The loss-absorbing debt stack — the instruments designed to be written down or converted in a resolution — is disclosed separately, and its order is the order in which people lose money.

6. Reading two banks against each other

• Check the basis before the number. Consolidated or solo, transitional or fully loaded, and which modelling approach — three choices that move a ratio without anything changing underneath.
• The stress test is a scenario, so a result says what a defined shock would cost rather than what is likely. See supervision for who sets it and why.
• The disclosure is a floor on what is known publicly. It is detailed, it is periodic, and it says nothing about what happened after the reporting date.

What is missing on purpose

• The composition of the deposit base in any useful detail. How sticky the funding is is exactly the question the ratios approximate and do not answer.
• Positions taken since the reporting date. Every figure is a snapshot with a date on it.
• Whether the models are right. The document reports outputs; the argument about the inputs happens between the bank and its supervisor.