Tools
Currency Drag on Overseas Dividends and Trades
Your platform takes a currency spread on every overseas dividend as well as on every trade, and the comparison tables publish one number for both. What is that costing you a year as a share of your overseas holding, split into the leg that recurs whatever you do and the legs that depend on how you trade — and at what dividend yield does it start to matter more than the annual fee you are ranking it against?
Every comparison table gives a platform one currency number. A currency charge is triggered by a conversion, and three quite different things cause one: a dividend arriving, the two trades at the ends of the position, and whatever you trade in between. The first recurs every year whether or not you ever trade. The second is paid twice and gets cheaper the longer you hold. Platforms genuinely differ between them, so the sum cannot rank two of them — and the published figure does not say which leg it is about. This page keeps the three apart, prices them against an annual fee you name, and says at what dividend yield the ranking between them flips.
What this tool does not decide
- Which platform to use. This prices one mechanism against one annual cost you entered. A platform is a great deal more than that — what it charges to deal, what it costs to leave, what it does when something goes wrong — and a lower modelled cost does not on its own establish that the lower-cost arrangement is better for a particular person.
- What your platform charges. This tool publishes no spread, names no platform, broker, venue or currency pair, fetches nothing and has no field to name one with. Both rates are yours to enter, and that is why nothing on this page can go out of date.
- Where the exchange rate is going. A spread is a charge on a conversion, taken whichever way the rate has moved. There is no rate in this arithmetic and no field for a view about one.
- What the holding returns. Everything here is a cost, expressed as a share of the holding a year. No price movement and no total return appears in it anywhere.
- Anything about tax withheld on the dividend, in any jurisdiction.
Your figures
Your holding and your platform, in seven numbers
Both spreads are yours to enter and this page publishes neither. The comparison tables carry one number per platform and do not say whether it applies to dividend receipts, which is the leg that recurs — so the rate you need is on your own platform’s charges page and nowhere else. Take both from it. Because nothing is published here, there is nothing on this page that can go out of date, and nothing you type is sent anywhere.
These entries need attention before the calculation can run
The holding, and what it pays
The part of your portfolio held in a foreign currency, not the whole of it. There is no currency here and no field for one: the two amounts on this page come back in the unit you enter this in, and every share is a share of it.
Before anything is withheld or converted — 3.2 for three point two per cent. Enter 0 if it pays nothing: the answer is then the trading legs standing alone, which is a real reading and is the number the table on this page opens with.
The two spreads, which are two different numbers
This is the leg the comparison tables do not tell you about, and it is the one that recurs every year whether or not you ever trade. Enter 0 if your platform states that its currency charge does not apply to dividends — that is a real arrangement, and it is the case in which no dividend yield changes this answer at all.
The figure a comparison table usually means. Enter 0 if your platform charges nothing on trades. It is often not the same number as the one above, and which of the two is larger differs between platforms — which is why one figure per platform cannot rank them.
How you hold, and what you are ranking against
The entry and exit conversions are paid once each and divided by this, so a one-year holding makes them an expensive annual charge and a twenty-year holding makes them almost nothing. Fractions are admitted — 0.5 is six months. Nothing else on this page depends on it, and no return is projected over it.
A custody or platform fee is the usual case, and it has to be a share of the same holding a year for the comparison to mean anything — a fee in money is not this figure until you have divided it by the holding yourself. Enter 0 if the arrangement you are comparing with has no annual charge; every unit of currency drag is then an excess over it.
Result
Jump to resultEvery maintained comparison table publishes one currency figure per platform, and there are three. A spread taken when a dividend is converted recurs every year whether or not you ever trade. A spread taken when you buy and when you sell is paid twice in the life of the position, so it gets cheaper the longer you hold. A spread on the trading you do in between is neither of those. Platforms genuinely differ between the legs — one charges on dividends and little on trades, another the reverse — so two platforms cannot be ranked from the sum, and the published single figure does not say which leg it is about. This page keeps the three apart, and it never publishes the total without them.
Enter your seven figures above and select Calculate. Nothing is sent anywhere: the calculation runs in this browser, and no value is stored, shared or placed in the address bar.
- The dividend leg — recurs every year whether or not you trade
- Not yet calculated
- The entry and exit conversions, spread over your holding period
- Not yet calculated
- A year of your trading
- Not yet calculated
- All three together, a year
- Not yet calculated
- What that is in money, a year
- Not yet calculated
- The comparator you entered
- Not yet calculated
- What the comparator costs on the same holding
- Not yet calculated
- The difference, signed
- Not yet calculated
- The drag as a multiple of the comparator
- Not yet calculated
- The dividend yield at which the ranking flips
- Not yet calculated
- How much more yield that would take
- Not yet calculated
Every share here is a share of the overseas holding, a year. That is what makes the currency drag and the comparator comparable at all — and it is why the comparator has to be an annual cost on the same base. A flat account fee in money is not this figure until you divide it by the holding yourself, and a fee charged on a whole portfolio is a share of the portfolio rather than of the overseas part of it. Nothing here is a return: a spread is a charge on a conversion, and what the exchange rate then does is not in this arithmetic anywhere.
This is arithmetic on seven numbers you supplied. It holds the yield, both spreads and the turnover constant across the holding period, which none of them will be. It assumes both conversions actually happen — a platform that lets you hold the foreign currency, or that nets purchases against dividends, converts less than this. And it prices one mechanism: two platforms differ in dealing charges, custody, spreads on the instruments themselves, service and access, and the comparator field carries exactly one annual cost rather than all of that.
A drag above a comparator is not advice to switch, and a crossover is not a ranking. This page prices one mechanism on one holding against one annual cost you named, and a platform is a great deal more than that: what it charges to deal, what it holds, what it costs to leave, what it does when something goes wrong. A lower modelled cost does not on its own establish that the lower-cost arrangement is better for a particular person. What is published here is what the currency legs cost. What to do about it is not a question seven numbers can answer.
How this is worked out
Why one figure per platform is not enough
A comparison table gives a platform one currency number. A currency charge is triggered by a conversion, and there are three quite different things that cause one:a dividend arriving, the two trades at the ends of the position, and whatever you trade in between. They behave differently over time, and platforms genuinely differ between them — one charges on dividend receipts and comparatively little on trades, another the reverse. So the sum tells you what this year cost and does not tell you which platform is cheaper for the way you actually hold.
| Leg | Triggered by | How often | Effect of holding longer | What moves it |
|---|---|---|---|---|
| The dividend leg | Every dividend that is paid to you in a foreign currency | Every year, whether or not you ever trade | Unchanged by how long you hold | The yield on the holding, and whether your platform charges on dividends at all |
| The entry and exit conversions | Buying the position, and selling it | Twice in the life of the position | Falls as an annual cost the longer you hold | How long you hold |
| The turnover leg | Every conversion caused by trading in between | As often as you trade | Unchanged by how long you hold | How much you trade |
The second row is why a long-horizon income investor is ranked wrongly by a single figure. Their two end conversions are divided over decades and come to almost nothing a year, while their dividend leg recurs in full every year — so the platform that looks dear on a table quoting one trading spread can be the cheaper one for them, and the platform that looks cheap can be dearer.
The quantities that follow
- The three legs
- What each of the three costs a year, as a share of the overseas holding. The decomposition is the finding, and the total is never published here without it
- The total
- Their sum. It is the figure a comparison table publishes as one number per platform, and it is the one a reader cannot rank two platforms from
- The excess over the comparator
- The total less the annual cost you named, signed. Positive is this arrangement costing more on this mechanism, and it is a statement about one mechanism rather than about a platform
- The crossover
- The gross dividend yield at which the total equals the comparator. There may be none, for three reasons that are three different answers, and every result names which
The arithmetic
Write V for the overseas holding, y for the gross dividend yield,sd for the spread on a dividend conversion, st for the spread on a trade conversion, t for conversions through trading in a year as a share of the holding,h for the holding period in years and f for the comparator:
dividend leg = y × sd
entry and exit = 2 × st ÷ h
turnover leg = st × t
total = y × sd + (2 × st ÷ h + st × t)
crossover y* = (f − 2 × st ÷ h − st × t) ÷ sdThe whole thing is a straight line in the yield, with slope sd and an intercept made of the two trading legs. That is not a simplification of something more elaborate — it is what a spread on income plus a spread on turnover actually is, and it is why the crossover has an exact answer rather than one found by trying values.The intercept is the part no table gives you: it is what the arrangement costs at a yield of nothing, and it is where the table on this page starts.
The three reasons there may be no crossover
They are three different answers and this page never merges them.Your platform takes nothing on dividends — then no yield moves the answer at all, and the whole of your drag is trading. The trading legs alone already cost at least the comparator — then the arrangement is not the cheaper one even at a yield of nothing, and the yield is not the lever. No yield up to a hundred per cent is enough — then the comparator is large relative to the spreads, and it is worth checking that it really is an annual cost on the same holding.
What you enter, and what this page will never enter for you
A holding, a yield, two spreads, a turnover, a holding period and a comparator.This page publishes no spread and names no platform. The research behind this tool found that the comparison tables carry one figure per platform and do not say whether it applies to dividend receipts — so a rate published here could only be somebody's, and would be wrong for everybody else. Because it holds no such fact, there is nothing on this page that can go out of date.
The limits this calculation imposes on itself
- The yield, both spreads and the turnover are held constant across the whole holding period. None of them will be, and a rising yield on an unchanged holding walks a reader towards the crossover without them doing anything.
- Both end conversions are assumed to happen. A platform that lets you hold the foreign currency, or that nets dividend receipts against purchases, converts less than this and the entry and exit row overstates it.
- The crossover and the comparison against the comparator are reported on exact equality rather than within a tolerance. Reporting two arrangements as level while one is measurably dearer is the one reading arithmetic about a ranking must not produce.
- Turnover counts conversions once each, so a round trip is two. Entering a round trip as one halves the turnover leg.
- Nothing here is conditional on any platform, broker, venue or currency pair, and no figure below describes anybody’s actual charges.
What is not modelled
- where any exchange rate is going — a spread is a charge on a conversion, and the rate itself is not in this arithmetic
- dealing commission, custody, platform or account fees beyond the single annual comparator you enter
- a platform that lets you hold the foreign currency rather than converting every receipt
- a scheme that nets dividend receipts against purchases so that fewer conversions happen
- withholding tax on the dividend, or any relief on it — a separate deduction on the same payment
- the spread on the instrument itself, as distinct from the currency conversion
- a holding spread across more than one foreign currency at different rates
- the cost of transferring the holding to another platform, or of closing it
- the identity of any platform, broker, venue or currency pair
The mechanism behind the leg that recurs — what a small deduction does once it repeats every year — is taught in the explainer:how investment fees compound.
Calculation model and corrections
- Calculation model
- Platform FX Drag v1.0
- Last reviewed
- Why there are three legs and not one
- A spread taken when a dividend is converted recurs every year whether or not you ever trade. A spread taken when you buy and when you sell is paid twice in the life of the position, so this page divides it by how long you hold and its annual cost falls as the holding lengthens. A spread on the trading you do in between is neither. Platforms genuinely differ between the legs — one charges on dividends and little on trades, another the reverse — so two platforms cannot be ranked from the sum, and this model never makes the sum available without the parts
- Whose spreads they are
- Yours, both of them, and this page publishes neither. Every maintained comparison table this candidate’s research read publishes one currency figure per platform and does not say whether it applies to dividend receipts. The rate you need is on your own platform’s charges page, and it is often on a different page from the trading one
- What a drag is a share of
- The overseas holding, a year. That is the same base an annual custody fee is quoted on, which is what makes the comparator comparable at all. A fee stated in money is not this figure until you have divided it by the holding yourself, and a fee charged on a whole portfolio is a share of the portfolio rather than of its overseas part
- How turnover is counted
- As conversions, once each, as a share of the holding. A round trip in a year is two, so buying and selling the whole position once in a year is two hundred per cent. The ceiling of ten times the holding admits an active trader while refusing an arithmetic mistake, and nothing about the model degrades above one
- What the crossover is, and the three reasons there may not be one
- The gross dividend yield at which this platform’s total currency drag equals the comparator. There may be none, for three different reasons that are three different answers: the platform takes nothing on dividends, so no yield moves it; the trading legs alone already cost at least the comparator; or no yield up to a hundred per cent is enough. Every result names which of the three applies rather than reporting an absence
- Both ends of every rate are admitted
- Nought is a real arrangement on every control here — a holding that pays no dividend, a platform whose currency charge does not apply to dividend receipts, a platform that charges nothing on trades, a reader who does not trade, and a comparator with no annual fee. Refusing any of them would hide the shape of the mechanism at its limit behind a validation message
- A spread is not an exchange rate
- It is a charge on a conversion, taken whichever way the rate has moved. Where the rate then goes is not in this arithmetic, there is no field for a view about it, and nothing on this page is a return
- Excluded
- Every platform, broker, venue, currency pair and published spread; dealing commission and any platform charge beyond the single annual comparator you enter; a platform that lets you hold the foreign currency rather than converting; netting dividends against purchases; withholding tax on the dividend; the spread on the instrument itself; more than one foreign currency at different rates; the cost of transferring or closing; and whether either arrangement is better
- Rounding
- Display only; intermediate values remain unrounded
Correction history
- The obvious implementation of this model publishes one currency drag figure and offers the decomposition as detail. It was built the other way round on purpose: the sum is not obtainable from the model without its three parts, because the frozen record’s entire finding is that a single figure per platform is what the maintained comparison tables already publish and what a reader cannot rank two platforms from. A model that made the total available alone would have reproduced the defect it was commissioned to correct, with better arithmetic behind it.
- The crossover is nullable three ways rather than one, and that was a deliberate widening during implementation. A platform that takes no spread on dividends, a platform whose trading legs alone already exceed the comparator, and a platform no admissible yield could push past it are three different situations with three different answers — change nothing, change platform, and stop worrying. A single nullable number would have given all three the same answer, which is the failure mode this model’s type file was written to make unrepresentable rather than merely to avoid.