Bank CapitalMedium

Capital is not money set aside. It is the share of the balance sheet that nobody has to be repaid — the part that can be lost without anybody defaulting.

8 min read · 1 446 words

The misunderstanding worth clearing first

A capital ratio sounds like a reserve: a pile of money in a vault, untouched, ready for a bad day. It is not, and the vault image gets almost every conclusion wrong.

Capital is a claim on the balance sheet, not an asset on it. A bank funds its lending with deposits, with borrowing, and with money that nobody can demand back — shareholders' money and certain instruments written to behave like it. That last slice is capital. Every euro of it is already invested in loans and securities alongside everything else. What makes it different is not where it sits but who is owed it: nobody. So it can absorb a loss without anyone being unpaid, which is exactly what a buffer is for.

This is why "the bank should hold more capital" never means "lend less". It means fund the same lending with a larger share of money that carries no promise to repay.

Why the denominator is risk-weighted

The headline ratio is capital divided by risk-weighted assets, and the weighting is the interesting half. A hundred of cash at the central bank and a hundred of unsecured lending to a start-up are both a hundred on the balance sheet and are not the same exposure, so the framework scales each by an estimate of how risky it is before adding them up.

• Standardised weights come from a table: exposure class, and often an external rating. Simple, comparable across banks, and blunt — every borrower in a class is treated alike.
• Internal-model weights come from the bank's own estimates of default probability and loss given default, subject to supervisory approval. More sensitive to real risk, and dependent on models the bank calibrates itself. Model risk is the obvious objection, and the framework's answer is an output floor: the modelled figure may not fall below a set share of what the standardised table would give.

The trade-off is permanent and there is no version of it that is free. A risk-insensitive denominator rewards a bank for loading up on whatever is under-weighted; a risk-sensitive one hands the bank a say in its own capital requirement.

The leverage ratio: the check on the denominator

Because the denominator can be argued with, there is a second ratio that refuses to be: capital divided by total exposure, unweighted. Every asset counts as itself, plus an allowance for derivatives and off-balance-sheet commitments.

It is deliberately crude, and that is its function. A portfolio of assets the models regard as nearly riskless can be enormous relative to the capital behind it while the risk-weighted ratio looks comfortable. The leverage ratio is a floor that does not care what any model thinks. Whichever of the two binds first is the one actually constraining the bank — and which one binds tells you a great deal about what that bank does. See leverage for the same arithmetic outside a bank.

How the stack is built

• Common equity tier 1 — ordinary shares and retained earnings, minus deductions for things whose value would evaporate in a failure (goodwill, certain deferred tax assets). This is the quality tier, and it is the one every meaningful requirement is expressed in.
• Additional tier 1 — instruments that are perpetual, whose payments can be cancelled without triggering a default, and which convert to equity or are written down when equity falls through a trigger. Debt in form, loss-absorbing by contract. Contingent convertibles are these, and what happened to them in March 2023 is the clearest illustration of what "loss-absorbing" means in practice.
• Tier 2 — dated subordinated debt. It absorbs losses only in a resolution or a liquidation, which makes it a slower-acting layer.
• The buffers on top — a conservation buffer that every bank carries, a countercyclical buffer that supervisors raise in good times so it can be released in bad ones, a systemic buffer for banks whose failure would spread, and a supervisor-specific requirement from the annual review.

The part that changes behaviour: the buffer is meant to be used

A buffer sitting above a minimum is not a second minimum. It is designed to be run down in a downturn — that is the entire point of building it in the good years. But falling into the buffer range triggers automatic restrictions on distributions: a formula caps the share of earnings that may leave as dividends, buybacks and certain bonus payments.

That mechanism produces a well-documented tension. A bank approaching the buffer has a strong incentive to avoid entering it, and the cheapest way to avoid entering it is to shrink the denominator — lend less — which is the opposite of what a released buffer was supposed to achieve. Supervisors say plainly that buffers are usable; the distribution restriction says something a treasurer hears more loudly. Both statements are true at once, and the gap between them is one of the live questions in prudential policy rather than a settled matter this page could resolve.

Capital is not liquidity, and confusing the two explains most bank failures

• A capital failure is losses exceeding the money nobody has to be repaid. The bank is worth less than it owes. This is solvency, and it is slow: it accumulates over quarters and is visible in the accounts if anybody reads them.
• A liquidity failure is being unable to pay what falls due today, whatever the balance sheet says. It is fast, it can happen to a solvent bank, and it is what actually closes the doors. Northern Rock in 2007 and Silicon Valley Bank in 2023 are both this failure, with different triggers.

The framework has separate ratios for the second problem: a liquidity coverage ratio, requiring enough unencumbered high-quality assets to survive a defined thirty-day stress, and a net stable funding ratio, requiring long-dated assets to be funded by something that will not disappear in a week. Liquidity covers the mechanism; the point here is that a bank can pass every capital test and fail on funding, and no capital ratio would have warned anybody.

What a capital ratio does not tell you

• Whether the weights are right. The ratio is only as good as the estimate in its denominator, and sovereign exposures in particular have historically been weighted in a way that is a policy choice rather than a risk measurement.
• What the assets are actually worth. An amortised-cost portfolio carried above its market value flatters equity until it is sold — how a position hits the books is the mechanism, and it is why a capital ratio and a fair-value note can tell two different stories about one bank.
• Whether the buffer will be used. See above.
• How quickly deposits can leave. Nothing in a capital ratio is a measurement of depositor behaviour, and depositor behaviour is what decides how much time a troubled bank has.

Who reads this, and why

Every seat in risk, compliance and audit reads it because the ratios are the constraint their firm operates under. Treasury and ALM manages the balance sheet against it. A supervisor sets it. Debt capital markets sells the instruments that fill the tiers, and research and anybody writing about a bank has to know that "well capitalised" is a statement about one ratio and not about survival. The capital disclosure playbook walks through the document these numbers are published in.

What to take away

• Capital is funding that nobody has to be repaid, not a reserve of cash.
• The ratio is capital over risk-weighted assets, and the weighting is where the arguments are.
• The leverage ratio exists because the denominator can be argued with; whichever binds first is the real constraint.
• Buffers are meant to be used, and the distribution restriction that comes with using them pulls the other way.
• Capital answers solvency. Liquidity answers whether the bank opens tomorrow. They are different failures with different ratios.