SupervisionHard

3 min read · 529 words

What the seat actually does

Somebody decides how much capital a bank must hold against each thing it owns, what an insurer may count as an asset, and what happens when a firm fails anyway. That is supervision, and it is the constraint every seat on the sell-side and buy-side of this map works inside.

Two halves that are often confused. Prudential supervision asks whether the firm can survive; conduct supervision asks whether the customer was treated properly. They use different tools, sit in different bodies in many countries, and a firm can pass one while failing the other completely.

• Capital — how much loss the firm can absorb, and what counts as capital in the first place.
• Liquidity — the buffers and the ratios behind them. See treasury and ALM, where they bite.
• Stress testing — running the firm's own book through a scenario the supervisor wrote.
• Resolution — planning in advance how a failing firm is taken apart without stopping the payment system.

A day, and where it goes

• Returns. The regulatory reporting firms file, and what looks different from last quarter.
• Meetings with the firm, which is where most supervision actually happens.
• Model approvals — whether a bank may use its own model to compute a requirement, and on what evidence.
• Policy, which is the slower half and the one that changes behaviour across a whole market.

What it is measured on

• Failures, and how they were handled — an orderly one is a success, not an embarrassment.
• Whether the requirement bound. A ratio every firm clears comfortably is not constraining anything.
• Consistency across firms, since a requirement applied unevenly is an incentive to be the exception.
• Whether the stress was severe enough, which is only answered by the real one.

What it touches on this site

• Where the rules land — compliance, risk management and treasury.
• What a rating boundary does — insurance investment, where a capital charge forces a sale.
• The instruments written for the rules — contingent convertibles and significant risk transfer.
• When it goes wrong — 2023, where an instrument did exactly what it was written to do and surprised the market anyway, and 2023.

How it goes wrong

• Risk weights that are gameable. A rule that scores an exposure lightly creates a market in exactly that exposure.
• Regulating the last crisis. Rules are written from the failure that happened, and the next one arrives somewhere unmeasured.
• Activity moving rather than stopping. Tighten a bank and the business appears somewhere the rules do not reach.
• A stress scenario that is not stressful, which produces a passing grade and no information.

Concepts to master

• Capital absorbs loss; liquidity meets payments. A firm can have plenty of the first and fail on the second.
• A risk weight is a judgement, written as a number and therefore easy to mistake for a measurement.
• Resolution changes who bears a loss, which changes what every instrument is worth — see who gets paid.
• The rules are a constraint on the market, so they are part of why prices are where they are.