The Deposit Limit Is a Policy

Deposit insurance is the most successful piece of financial regulation ever enacted, and it works by being uninteresting. Nobody thinks about it, which is the entire mechanism.

The provision that sets it up contains a number: the amount per depositor that is guaranteed. That number is the whole policy.

Below it, a saver has no reason to care which institution holds their money, and a bank has no reason to fear that ordinary savers will leave in a panic.

Above it, both of those facts reverse, and they reverse suddenly rather than gradually, because the guarantee has an edge rather than a slope.

The edge is where the interesting behaviour happens. A depositor near the limit behaves like an insured depositor; one comfortably above it behaves like a creditor.

Business deposits are the case the limit handles worst, and they are a growing share of what banks hold.

Creditors monitor. They watch the institution, they move at the first sign of trouble, and they move faster than any supervisor can respond.

A firm meeting a payroll cannot split its operating balance across a dozen institutions, so it sits above the limit for entirely mechanical reasons rather than because it is wealthy.

The insurer's published data shows how the balance between insured and uninsured deposits varies enormously between institutions, which is the variable that decides how a run behaves.

When such a firm moves, it moves the whole balance at once, and the deposits most likely to leave in a panic are therefore the ones least able to stay.

A bank funded largely by insured deposits is close to run-proof. One funded by balances above the limit is exposed to a coordination problem that has nothing to do with its solvency.

A separate treatment for transaction balances has been proposed repeatedly and never adopted, largely because it is hard to define without opening a route around the limit.

That is the property the last several crises have demonstrated, and it is a property of the limit rather than of any bank's decisions.

Raising the limit would reduce the exposure and would also extend a public guarantee to a wealthier set of depositors, which is a distributional choice.

Lowering it would restore market discipline and would make more institutions vulnerable to runs that have nothing to do with whether they are sound.

The implementing regulations handle the mechanics of how the limit is applied per depositor, per institution, per ownership category, and the mechanics are where a great deal of the real coverage sits.

A sophisticated depositor can multiply coverage by splitting balances across categories and institutions, which is entirely legal and entirely unavailable to anyone who does not know it is possible.

So the effective limit is not one number. It is one number for people who have advisers and a much lower number for everybody else.

There are firms whose business is arranging exactly this, which is a market that exists solely because of where a line was drawn.

The central bank sits behind all of it as the lender of last resort, and the relationship between the two instruments is rarely made explicit.

Insurance stops the depositors running. Lending of last resort handles the institution that is illiquid anyway. They address different failures and are discussed as though they were one.

The funding side is the part with the least public attention. The insurance fund is built from levies on the institutions, so the industry pays for its own guarantee.

That is elegant in principle and procyclical in practice: levies rise after losses, which is when the institutions can least afford them.

The number itself moves by legislation, rarely, and usually in an emergency, which is the worst moment to choose it and the only moment anyone attends to it.

Setting it in a crisis means setting it against the failure in front of you rather than against the structure, and the figure then stays for a decade or more.

What would improve this is a scheduled review with a published rationale — what the limit covers, who sits above it, and what the exposure looks like — so the number is chosen rather than inherited.