Bank capital buffers

Why must a profitable bank hold capital that it is not allowed to put to work?

Bankers describe capital requirements as money sitting idle — a pile the regulator forbids them to lend, dragging on returns. The complaint has an obvious appeal: capital that earned nothing would be waste, and waste in a competitive industry ought to be competed away.

But the premise deserves a hard look before the conclusion does. Is capital a pile of unlent money at all? And if it is not, what exactly is the requirement asking a bank to give up?

Start with the balance sheet, because the whole confusion lives there. Assets — loans, securities, reserves — sit on the left: what the bank owns. Liabilities and equity sit on the right: where the funding came from. Deposits and bonds are borrowed funding. Capital is the portion funded by shareholders rather than creditors.

Now notice what that placement implies. Capital is not a category of asset, so it is not something a bank can lend or not lend. Raising the capital requirement does not shrink the left side; it changes the mix on the right. A bank funded 6% by equity and one funded 12% by equity can hold exactly the same loans. So the idle-pile picture is not a small inaccuracy — it names the wrong side of the sheet.

Then what does the buffer actually do? Ask what happens when loans go bad. Losses reduce assets. Someone's claim has to shrink to match, and the ordering is fixed: equity absorbs first, then subordinated debt, then senior creditors and depositors. Capital is the depth of the layer that can be destroyed before a depositor's claim is touched. That is its entire function — and note that it functions precisely by not being committed to anything.

This is where the idea earns its name. A buffer is redundancy: capacity held so that a shock is absorbed locally instead of propagating. Aircraft carry a second hydraulic system, bridges are built past their expected load, data is stored in more copies than are needed to read it. In each case the redundant capacity does no work in normal operation. That is not a defect in the design; it is the design. Capacity that is fully employed has nothing left to give when the unusual arrives.

Now ask the sharper question: why must this redundancy be mandated? Why would a bank not choose it? Because of who bears the shock. If a bank fails, it does not fail alone — deposits are insured, payment systems seize, and other banks holding its paper take losses that force them to sell into a falling market. Much of the cost of a failure lands outside the bank. A private calculation therefore rationally lands on less redundancy than the system needs, and the gap between the two is what the requirement fills.

So how is it actually set? Under the Basel III framework, common equity tier 1 must be at least 4.5% of risk-weighted assets, with a capital conservation buffer of a further 2.5% on top, plus a countercyclical buffer that supervisors can raise from zero to 2.5% when credit growth looks dangerous, plus a surcharge for globally systemic banks. The weighting matters: assets are scaled by assumed riskiness, so a government bond and an unsecured business loan do not demand the same capital.

The conservation buffer is the piece that answers the question's word "must." It is not a hard floor. A bank may dip into it — that is the point of holding it — but as it does, it faces automatic limits on dividends, share buybacks and discretionary bonuses. The design deliberately makes the buffer usable and makes using it visible.

Here is the honest complication. That design did not fully work. During the March 2020 shock, supervisors encouraged banks to draw on their buffers to keep lending, and banks largely declined — fearing that dipping in would be read by markets as distress, and would trigger the distribution restrictions. The Basel Committee and the Bank of England have both examined this "buffer usability" problem since. Redundancy that its holder is afraid to spend is not functioning redundancy, and the fix is still being argued over.

Three refinements.

The first corrects the framing directly: higher capital does not mean less lending. It means the same lending funded with more shareholder money and less borrowed money. Empirical work is genuinely mixed on transition effects — banks adjusting quickly to a new ratio may shrink assets rather than raise equity — but the steady-state claim that capital crowds out credit does not follow from the accounting.

The second is that shareholders are not simply worse off per unit of risk. A less leveraged bank is a safer bank, so its equity is less risky and its required return should fall. That is the Modigliani-Miller logic; it holds imperfectly in practice, mainly because debt interest is tax-deductible and deposit insurance subsidises borrowing — both of which make leverage cheap for reasons that have nothing to do with efficiency.

The third is a limit worth naming. Capital absorbs losses. It does not supply cash. A solvent bank can still fail if depositors leave faster than assets can be sold, which is why liquidity rules exist alongside capital rules and why the 2023 failure of Silicon Valley Bank was not principally a capital-ratio story.

• Why risk weights are contested, and what a simple leverage ratio catches that they miss.
• How liquidity coverage rules address the failure mode capital cannot.
• Whether countercyclical buffers can be raised early enough to matter.

Every term the collection defines is gathered in the glossary.