Comparison
Retained Interest Across Credited Rates: How Much of What Your Cash Earns Never Reaches You
When uninvested cash earns one rate where it is held and is credited another, what share of the interest does the holder never receive — and what is that worth as an ordinary annual charge?
The short answer
The share retained is one minus the ratio of the two rates — and it is a charge that appears in no fee schedule. Where the cash earns 4.00% and 1.00% is credited, the holder never receives 75.00% of the interest their money earned. Restated in the units a fee schedule uses, that is 3.00% of the cash balance a year.
Those two sentences are the same fact, and that is the point of publishing both. A retention sounds like a proportion of something that was never promised; an annual charge on a balance sounds like a fee. Every other cost on an investment account is levied on a balance or on a trade and is therefore disclosed somewhere. This one is levied on a cash flow the holder never sees credited, so there is nothing for a schedule to disclose — the interest simply arrives smaller.
- Who it applies to
- Any account where uninvested cash earns a rate somewhere and the holder is credited a different one — a brokerage cash balance is the ordinary case. It compares two rates, not brokers, banks or products.
- What this does not tell you
- It establishes what a stated pair of rates implies about the share of the interest a holder never receives. It does not establish what any broker credits, what any bank pays, what the short-term rate is, or where cash should be held instead.
An exact identity rather than a projection: one ratio of two rates the reader states, so no market figure, horizon or assumed return enters the answer, and nothing here goes out of date.
What is being compared
Two rates, held against each other across a ladder. Nothing varies between rows except the rate the cash earns where it is held, nothing varies between columns except the rate credited to the holder, and no amount of money appears anywhere on this page.
| Earned rates compared | 2.00% · 3.00% · 4.00% · 5.00% |
|---|---|
| Credited rates compared | 0.00% · 0.50% · 1.00% · 1.50% · 2.00% |
| Quantity compared | The share of the interest retained, and what it is worth as a charge |
| Amounts | None. The balance cancels out of both quantities |
| Period | None. Both rates are quoted over the same period, so it cancels too |
| Where the cash is held | Taken as given. No sweep arrangement is described or assumed |
| Tax | Not modelled. It falls on what is credited, not on what is retained |
| Other charges | Not modelled. A platform, custody or trading charge is a separate cost |
Every rung on both ladders is a stated rate entered into the identity, chosen so the shape can be read. Neither ladder is a survey of what is credited or earned, no broker or bank is named, and nothing here reports a market rate on any date.
Where the retention comes from
Uninvested cash does not sit still. It is swept to a bank, where it earns something close to the short-term rate. The holder is credited a rate the broker sets, and the difference is kept:
retained share of the interest = 1 − (rate credited ÷ rate earned)
The balance appears on both sides of that division and cancels, sonothing about how much cash is held enters the answer. The period cancels for the same reason, provided both rates are quoted over it — which is the ordinary case. What is left is a ratio of two rates, and it is exact rather than an approximation that holds for large balances.
The identity is refused above the earned rate rather than computed there. A credited rate higher than the rate the cash earns is a subsidy paid out of the broker's own funds, not a retention, and printing a negative retention would answer a question about a different arrangement in a table titled for this one.
The share of the interest the holder never receives
Read across a row and the retention falls as more is credited, as it must. Read down a column and it rises as the cash earns more — which is the direction most readers do not expect, and the one that matters when rates move.
| Rate the cash earns | 0.00% credited | 0.50% credited | 1.00% credited | 1.50% credited | 2.00% credited |
|---|---|---|---|---|---|
| 2.00% | 100.00% | 75.00% | 50.00% | 25.00% | 0.00% |
| 3.00% | 100.00% | 83.33% | 66.67% | 50.00% | 33.33% |
| 4.00% | 100.00% | 87.50% | 75.00% | 62.50% | 50.00% |
| 5.00% | 100.00% | 90.00% | 80.00% | 70.00% | 60.00% |
The first column is the one least often stated anywhere. Where nothing at all is credited, the retention is the whole of the interest — 100.00% of it, at every rate on the ladder — and a holder comparing accounts on their published charges would find no trace of it.
The diagonal is the other end. Where the credited rate equals the rate the cash earns, nothing is retained, and the figure is a genuine zero rather than a rounding: the holder receives the interest their money earned.
The same retention, as an ordinary annual charge
A share of an unstated quantity is hard to hold against a cost that is stated. Below is the identical retention expressed the way a fee schedule expresses a fee: as a percentage of the cash balance, charged once a year.
| Rate the cash earns | 0.00% credited | 0.50% credited | 1.00% credited | 1.50% credited | 2.00% credited |
|---|---|---|---|---|---|
| 2.00% | 2.00% | 1.50% | 1.00% | 0.50% | 0.00% |
| 3.00% | 3.00% | 2.50% | 2.00% | 1.50% | 1.00% |
| 4.00% | 4.00% | 3.50% | 3.00% | 2.50% | 2.00% |
| 5.00% | 5.00% | 4.50% | 4.00% | 3.50% | 3.00% |
Nothing in this table is new arithmetic. It is the difference between the two rates — 3.00 percentage points in the worked case — and the only thing that has changed is the unit. That change is the point: a charge nobody publishes becomes a number that can sit beside the charges everybody publishes.
A banded rate is this identity applied twice
Some arrangements credit one rate up to a threshold and another above it. That is a second application of the same identity to a second band rather than a second mechanism, so the tables above still contain the answer: read the row for the rate the cash earns once at each credited rate, and the overall retention is the balance-weighted average of the two figures.
The weighting is not published here, and the omission is deliberate. It needs the balance and the threshold, and both belong to the reader rather than to a reference set. What the tables do establish is the bound: a banded arrangement's overall retention necessarily lies between the two band figures, so neither band read alone is the answer and both read together are.
A retention is not a scandal, and it is not disclosed
Both halves of that sentence are meant.Holding cash somewhere costs something to arrange, and a broker keeping part of the interest is one of the ordinary ways an account is paid for. Some arrangements say so plainly, in a rates page a holder can find. The finding here is not that the retention is hidden by design; it is that it is structurally absent from the one document a holder compares accounts on, because a fee schedule lists charges levied on balances and trades and this is neither.
The figures are also conditional on cash actually being held. A holder whose account runs near zero uninvested pays none of this, whatever the two rates are, and the tables above say nothing about how much anyone holds or for how long.
Two holders facing the same two rates can still be in different positions, depending on:
- how much cash is actually held uninvested, and for how long in the year
- whether the credited rate is banded, and where the thresholds sit
- whether the credited rate moves when the earned rate moves, and by how much
- whether interest is credited at all below a stated balance
- what deposit protection applies to the cash, and to how much of it
- the platform, custody and trading charges that sit alongside this one
No account, broker or arrangement is presented here as good or bad value, and nowhere to move the money is named or implied. What the tables price is a cost; whether it is worth paying depends on everything else the account does, which this page does not know and cannot see.
What this comparison does not determine
The figures on this page are exact arithmetic on two stated rates. They are not quoted rates, market conditions, published schedules or a forecast of anything. The comparison cannot determine:
- what any broker credits on cash, or whether it credits anything at all
- what any sweep bank pays, or what the short-term rate is on any date
- whether cash is swept at all, or held in the broker’s own balance sheet
- whether the arrangement is disclosed, and where a holder would find it if it is
- whether the cash is protected by any deposit-guarantee scheme, and up to what limit
- what tax is due on the interest that is credited
- what any alternative would earn, cost, or risk
- whether holding cash at a broker is sensible, careless or suitable for anyone
It also holds both rates still. A credited rate that moves later than the rate the cash earns retains more in the interval and less afterwards, which is a path, and this identity is a statement about one pair of rates rather than about a path.
None of this is financial, legal or tax advice, and no product, provider, broker or bank is recommended, ranked or named anywhere on this page.
Where these figures come from
Every number above is produced at build time from retained share of the interest = 1 − (rate credited ÷ rate earned), applied to the two ladders in the comparison set. Nothing on this page is typed by hand, and nothing is calculated in the browser.
There is no calculation-model version behind them, and that absence is the point. A ratio of two rates is not a projection, so it inherits no horizon, no assumed return and no compounding convention. Its regression suite holds every published figure against a second, independently written implementation of the same identity — two implementations that agree are evidence, one checked against itself is not.
The build refuses a credited rate above the rate the cash earns, and refuses an earned rate of nothing, rather than reporting a figure for either. Neither is a case the identity has content at, and printing a boundary that is not a quantity would be worse than printing none.