The Sweep

TAM-ARB.29 · Arbitrage · The Approximate Mind

A checking account paying nothing, in a year when short rates pay meaningfully more than nothing, is a spread.

It has never been treated as one, because it never looked like one. It looks like a service, a habit, a place the paycheck lands. But the arithmetic is the series’ oldest shape: the household holds a balance, the institution pays a fraction of what the balance earns, and the difference is revenue, collected continuously, at a scale that makes it one of the largest quiet transfers in consumer finance. The industry’s own term for the asset is candid in the way accounting sometimes is. Sticky deposits: balances that do not move when the price says they should, valued precisely for the not-moving, funded by households whose attention was always going to be somewhere else.

Stickiness was never loyalty. It was the cost of watching rates, comparing offers, opening accounts, and moving money, multiplied by the smallness of each month’s difference, and the cost was real when a person had to pay it in hours. The optimizing instrument pays it in nothing. A household system with ledger authority sweeps idle cash to the best available yield continuously, as a background process, the way the taken-advice essay established everything executable gets executed. The spread does not narrow under that pressure. It closes, at the pace of enrollment, and the closing is pure gain to the household, the kind of dissolved spread that was somebody’s revenue on one side and nobody’s virtue on either.

The machine, in one sentence: continuous automated cash placement converts the deposit base from an inert liability repriced at the bank’s discretion into a rate-sensitive flow repriced at the market’s, and the funding model built on inertia loses its funding at exactly the speed the instrument gains adoption.

Who Was Standing on the Spread

#

The uncomfortable finding is not that banks lose a margin. The series has priced the loss of margins many times over and owes nobody a candle. The finding is who loses it, and where they were standing.

The deposit franchise is not evenly held. The largest institutions fund from many sources, hold market share in dense places, and can pay for deposits when they must. Community and regional banks are the ones funded disproportionately by sticky retail deposits, and community and regional banks are, in thin counties, frequently the last institution of any kind still holding local knowledge, local credit judgment, and a balance sheet with the county’s name on it. The stickiness that the optimization dissolves was not distributed like a tax on everyone. It was concentrated, as a subsidy nobody designed, on precisely the institutions that the corpus’s serving-unit work identifies as among the last bodies in thin geography able to hold anything at all.

So the collision: household deposit optimization at scale removes the funding base of the institutions that are the last serving units standing in thin counties. The optimization that helps the household removes the institutional holder from the county. Not partly true, not true from an angle. Both completely true, at the same time, about the same account. The household is not wrong to sweep; the household was paying the subsidy, invisibly, out of its own idle cash, and no theory of fairness obliges a family to fund a bank’s deposit margin as a civic act. And the county is not wrong to notice that the branch that closes behind the sweep was underwriting the diner, the equipment loan, and the one banker who knew which farms flood.

A spread can be extraction and infrastructure at the same time, and closing it delivers the gain to one ledger and the loss to a geography.

The Speed Term

#

One property distinguishes this closing from every other in the arc, and it is the property that turns a pricing into a warning.

Sticky deposits were the banking system’s shock absorber, and the stickiness was measured in months and years: a rate mismatch could persist because the deposits stayed while the balance sheet adjusted. Swept deposits move at the speed of the instrument, which is to say at the speed of software, and they move together, for the reason the disagreement essay declared: instruments converge, and converged instruments respond to the same signal in the same week. The recent record already holds the preview, a large institution unwound in days by depositors coordinating at phone speed, and that episode ran on humans forwarding messages. An optimized deposit base is that episode as a standing condition: fully rational, fully synchronized, and continuously armed. The funding that was stable because moving it was tedious becomes funding that is stable only while no signal says otherwise.

The claim, dated. Horizon: end of 2031. By then, deposit betas at retail-funded institutions will have moved materially toward market rates in optimized-household markets, and at least one supervisory framework will have formally repriced retail deposit stickiness in its liquidity assumptions, treating swept retail money as market-sensitive funding. Miss condition: if in 2032 retail deposit betas at community institutions remain near their historical bands despite broad household-instrument adoption, then either stickiness has sources the inertia account missed, habit, trust, the schools account nobody re-plumbs, or enrollment stalled, and the mechanism overstated the flow.

The Frame

#

Decision rule: fund no institution whose deposit franchise is valued on stickiness without pricing the sweep’s arrival date into the franchise; and in thin geographies, treat the community bank’s funding base as what it actually is, a serving unit’s life support wired through an arbitrage, so that any position on the bank is also, disclosed or not, a position on whether the county keeps its last institutional holder.

The spread closes, the household keeps the difference, and somewhere a county discovers that the margin it never knew it was paying was holding the door of the last building in town with a lender inside.

The Approximate Mind is a series exploring what AI actually does to human life. Arbitrage prices what the audit does to every spread it reaches.