Tools

Dividend Withholding Drag at Two Rates

Tax is withheld on a fund’s dividends inside the fund, before any figure you ever see. If two arrangements suffer different blended rates across the source countries a global index holds, what is the difference worth a year, and what has it come to over your horizon?

A dividend is taxed by the country it came from, inside the fund, before it reaches the fund’s own accounts and long before it reaches a figure on your statement. Two arrangements holding the same index can suffer different blended rates across those source countries, and the difference is real money over a long horizon. What nobody publishes is the blend itself — so this page publishes none, asks you for two, and reports what each takes, the difference between them, and what that difference has come to by the end. It also reports the two side by side on purpose: the level is uncertain and the difference is exact, and they look identical on a page.

What this tool does not decide

Which arrangement is cheaper. This computes one mechanism, and a fund’s domicile differs in more than one. The ongoing charge, the tracking difference and the spread are all outside this arithmetic, and any of them can be larger than what this page returns.
What either rate is. No authority publishes an effective blended rate across source countries. A tax authority publishes rates between pairs of countries; an index provider’s assumption is a non-treaty institutional investor’s; and for a synthetic fund the rates in its index can be a commercial parameter agreed with swap counterparties. A realised figure comes from a fund’s own annual report.
Whether a fund gets the rate its domicile is entitled to. That is decided per source country, turns on treaty residence and beneficial ownership, and is published by neither side in a form a holder can look up. It bears on every figure here and this page cannot reach it.
What you would be worth. Every figure counts the dividend stream and nothing else. Price return is excluded by name rather than assumed to be zero.
Anything about your own tax, in any jurisdiction.

Your figures

The interactive calculation runs in your browser and needs local JavaScript, which is currently switched off. Activating the Calculate button does nothing at all: no request is sent and nothing is submitted.

Everything else on this page works without it. The model, the formulas, the timing convention, the worked example, the assumptions and the limitations are all in the methodology below, and the companion explainer covers the mechanism in full.

Your figures, in five numbers

No authority publishes an effective blended withholding rate, and the four things that look like one each fail differently. Tax authorities publish rates between pairs of countries, which is all a tax authority has any reason to publish. An index provider publishes a withholding assumption, but it is a non-treaty institutional investor’s — an upper bound rather than a measurement of what any fund suffers. For a synthetically replicated fund the withholding rates in its index can be a commercial parameter agreed with swap counterparties and reviewed each year against competing products’ performance rather than against tax law, and a provider publishes that process in plain terms. What is left is the one document that reports withholding actually suffered against dividend income received: a fund’s own annual report. That is where a realised blended rate comes from, and this page publishes none of its own.

The fund’s gross dividend yield

Gross: before any withholding and before any charge. A factsheet publishes this figure. Enter 0 if the fund distributes nothing — that is the boundary at which this mechanism has nothing to act on, and it is an answer rather than an error.

The blended withholding rate at the first arrangement

Blended across every source country the fund receives dividends from, not the treaty rate with any one of them. There is no suggested figure here because nobody publishes one. A realised blended rate for a real fund can be read from that fund’s own annual report, which reports withholding suffered against dividend income received.

The blended withholding rate at the second arrangement

The same quantity, at the arrangement you are comparing against. Neither field is the high one and there is no ordering rule: the sign of the difference is the answer, and requiring an order would ask you to know it before asking the question.

How many years you are comparing over

The period the cumulative gap runs over. One year is where the two conventions below agree exactly, which is worth seeing once.

What happens to the net dividends

There is no pre-selected answer on purpose. Without reinvestment the gap is exactly the annual difference multiplied by the horizon; with it the gap compounds, and the linear figure understates it. Which of those describes your holding is a fact about it rather than a preference, and the difference between the two is itself informative.

Result

A treaty rate is a rate between two countries. A fund tracking a global index receives dividends from twenty-three of them, and a single treaty rate describes only the source country that treaty covers. The trap is that the share looks obvious and is not: on the index factsheets this page’s research read, the United States is 72.03 per cent of the index by market capitalisation and approximately 53.4 per cent of it by dividends, because the markets outside it yield 2.55 per cent against the index’s 1.53. Withholding is charged on dividends. So the figure a reader reaches for describes about half the base, and applying it to a whole global fund is the mistake this page exists to make visible.

Enter your five 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.

What the first rate withholds each year, as a share of capital: Not yet calculated
What the second rate withholds each year, as a share of capital: Not yet calculated
The difference between those two, a year: Not yet calculated
The dividend yield that actually reaches the fund, at the first rate: Not yet calculated
The same, at the second rate: Not yet calculated
What one unit of capital becomes over the horizon, at the first rate: Not yet calculated
The same, at the second rate: Not yet calculated
The gap between those two, per unit of capital: Not yet calculated
The horizon those last three figures run over: Not yet calculated
The reinvestment convention they were computed on: Not yet calculated

Every figure here is per unit of capital, so none of them depends on the size of your holding and there is no amount and no currency anywhere in the arithmetic. Every figure also counts the **dividend stream and nothing else**: price return is excluded by name rather than assumed to be zero, because this page has no authority to ask you for one and no basis on which to assume one. The gap over the horizon is therefore the gap on the dividends. It is not a difference in what you would be worth, and a price return common to both arrangements would scale both sides of it.

This computes one mechanism, and a fund’s domicile differs in more than one. The ongoing charge, the tracking difference, the spread, the securities-lending arrangements and the size of the fund are all outside this arithmetic, and any of them can be larger than the number this page returns. So is the holder’s own tax: whether an amount withheld inside a fund is reclaimable, creditable or simply lost is a question about the holder’s jurisdiction rather than about the fund’s, and it is not modelled here. There is also a question this page cannot reach at all — whether a particular fund qualifies for the treaty rate its domicile is entitled to, which is decided per source country and is published by neither side in a form a holder can look up.

A difference computed here is not a verdict about either arrangement and this page names neither. It says what two rates you supplied would do to a dividend stream you supplied, which is arithmetic and is exact given those rates — and says nothing about whether those rates are the right ones, whether either fund would obtain them, or which is the better thing to own. The whole reason the difference is publishable while the level is uncertain is that the difference is arithmetic on two numbers you chose. Reading it as a ranking would put all of that uncertainty back in, invisibly.

How this is worked out

Why the obvious figure describes about half the problem

Tax is withheld on a fund’s dividends by the country the dividend came from, inside the fund, before any figure the holder ever sees. A reader who wants to size that reaches for a treaty rate — and a treaty rate is a rate between two countries, while a fund tracking a global index receives dividends from twenty-three of them.The share that matters is not the share a factsheet leads with. A market can be the largest part of an index by size and a much smaller part of it by dividends, and it is dividends that withholding is charged on.

The same index on the same date, weighted two ways. Both figures are the index provider’s own; the second is derived from two published yields and the published weight.
Weighted how	The largest market’s share	Where it comes from	Is it the base?
By market capitalisation	72.03 per cent	The figure on the factsheet, and the one a reader reaches for	Not the base. Withholding is not charged on market capitalisation
By dividends	approximately 53.4 per cent	Derived from the two published yields — 1.53 per cent for the index and 2.55 for the index excluding that market — and the published weight	This is the base. Withholding is charged on dividends

Where a blended rate could come from, and why this page ships none

A blended rate is the single effective rate across every source country, weighted by where the dividends actually came from.No authority publishes one, and the four things that look like one fail differently — three of them are somebody else’s number computed for somebody else’s purpose.

The last row is the one that works, and it is the reason this page asks you for a rate instead of publishing one.
Where you would look	What it publishes	Why it is not your blended rate
A tax authority’s treaty table	A rate between two countries	It is one country pair. There is no row for “a fund domiciled in X holding a global index”, because that is not a quantity a tax authority has any reason to publish
An index provider’s net-return assumption	A withholding rate the index applies before reinvesting dividends	It assumes an investor who does not benefit from double-taxation treaties. A treaty-eligible fund suffers materially less, so the gross-minus-net spread is an upper bound rather than a measurement
A synthetic fund’s custom index rates	A published table of per-country rates	They can be a commercial parameter. One provider publishes that its custom rates have been agreed with the swap counterparties and are reviewed each year against competing products’ performance rather than against tax law
A fund’s own annual report	Withholding actually suffered against dividend income received	Nothing, and this is the one that works. It is retrospective and specific to that fund in that year, which is why nobody publishes it as a general figure — and why this page asks you for a rate rather than shipping one

The arithmetic

Write y for the gross dividend yield, w for a blended withholding rate andn for the horizon in years. Five lines, and the whole calculation:

drag(w)        = y × w
net(w)         = y × (1 − w)

factor(w)      = (1 + net(w))ⁿ        dividends reinvested
factor(w)      = 1 + n × net(w)       dividends taken as income

difference     = drag(wB) − drag(wA)  =  y × (wB − wA)
gap over n     = factor(wA) − factor(wB)

That it is five lines is the point rather than an admission. The thing that would have made it wider — a weighting across source countries — is a figure no authority publishes and therefore one this site may not invent.The difference is computed from the two rates rather than from the two drags, which is the same quantity by a better-conditioned route: subtracting two rates you stated is exact, where subtracting two products means subtracting two numbers that have each already been rounded.

The level and the difference are not the same quantity

A level — what one rate withholds a year — inherits the whole of your uncertainty about what a blended rate across every source country actually is. A difference between two rates you stated is exact given those rates, whatever they are worth. Both are published on every result and neither substitutes for the other, because the failure they prevent is invisible: a level and a difference look identical on a page, and one of them carries an error term the other does not. The clearest case is one rate applied to both arrangements, which gives a level of 0.229 percentage points a year and a difference of exactly nothing.

What the horizon figure counts, and what it does not

The net dividend stream, and nothing else. There is no price return in this arithmetic, and it is excluded by name rather than assumed to be zero — this page has no authority to ask you for one and no basis on which to assume one.So the gap over the horizon is the gap on the dividends and it is not a difference in what you would be worth. A price return common to both arrangements would scale both accumulation factors, so it would not change which is larger and it would change by how much.

The limits this calculation imposes on itself

Both rates are yours and this page publishes neither. It therefore holds no figure that can go out of date, and no claim about any fund, domicile or index.
This computes one mechanism. A domicile differs in more than one — the ongoing charge, the tracking difference, the spread and the size of the fund are all outside this arithmetic, and any of them can be larger than what this page returns.
Whether a particular fund actually qualifies for the treaty rate its domicile is entitled to is a separate question, decided per source country, and published by neither side in a form a holder can look up. It bears on every figure here and this page cannot reach it.
The holder’s own tax is not modelled. Whether an amount withheld inside a fund is reclaimable, creditable or simply lost is a question about the holder’s jurisdiction rather than about the fund’s.
Price return is excluded by name rather than assumed to be zero. The gap over the horizon is the gap on the dividend stream, and a price return common to both arrangements would scale both sides of it.

What is not modelled

any composition weighting, country list, market-capitalisation share or dividend share — no authority publishes one on a dividend basis and a UBWHY figure for it would be an estimate of a tax fact
any published withholding rate, treaty rate or index withholding assumption
the fund’s ongoing charge, its tracking difference, its spread and every other cost
whether either fund actually qualifies for the treaty rate its domicile is entitled to, which is decided per source country
the holder’s own tax — whether the withheld amount is reclaimable, creditable or lost
price return, and therefore any statement about total wealth
currency, exchange-rate movement and the currency the dividends are declared in
the identity of any fund, manager, index, provider, domicile or regulator
whether either arrangement is cheaper overall, or better

The mechanism underneath all of this — that a small deduction taken every year is not a small number over a horizon — is taught in the explainer:how investment fees compound.

Go deeper

Why a small annual number is not a small number
How Investment Fees Compound Into Lost Wealth
Why a small annual investment fee can create a much larger long-term difference, why the second component of that difference is signed, and where the arithmetic stops.
The difference this page computes is a few hundredths of a percentage point a year, and the reason it is worth computing at all is what a deduction that size does over a horizon. That is the explainer’s whole subject, in the more familiar form of a charge.
Read the explainer: How Investment Fees Compound Into Lost Wealth

Calculation model and corrections

Calculation model: Withholding Differential v1.0
Last reviewed: 27 August 2026
What “blended” means, and why it is not a treaty rate: The single effective rate withheld across every source country a fund receives dividends from, weighted by where those dividends actually came from. A treaty rate is a rate between two countries and describes only the one that treaty covers. On the index factsheets this page’s research read, the United States is 72.03 per cent of the index by market capitalisation and about 53.4 per cent of it by dividends, because the markets outside it yield more — and withholding is charged on dividends
Whose rates they are: Yours, both of them, and this page publishes neither. No authority publishes an effective blended rate: tax authorities publish pairwise treaty rates, an index provider’s assumption is a non-treaty institutional investor’s and therefore an upper bound, and a synthetic fund’s index rates can be a commercial parameter agreed with swap counterparties. A realised blended rate comes from a fund’s own annual report, which reports withholding suffered against dividend income received
What a drag is a share of: Capital, not the dividend. A yield of 1.53 per cent with 15 per cent withheld is a drag of 0.2295 percentage points of capital a year. Every drag, differential and gap on this page is per unit of capital, so none of them depends on the size of a holding and there is no amount or currency anywhere in the arithmetic
The level and the difference are different quantities: A level inherits the whole of the uncertainty about what a blended rate actually is. A difference between two rates you stated is exact given those rates, whatever they are worth. Both are published on every result and neither substitutes for the other — the failure this prevents is invisible, because a level and a differential look identical on a page
What the horizon figure counts: The net dividend stream and nothing else. Price return is excluded by name rather than assumed to be zero: a price return common to both arrangements would scale both accumulation factors, so it would not change which is larger and it would change by how much. The gap is therefore the gap on the dividends, and it is not a difference in total wealth
The two reinvestment conventions: Reinvested, the accumulation compounds and the gap is larger in magnitude than the annual difference multiplied by the horizon. Taken as income, the gap is exactly the annual difference multiplied by the horizon. Neither is offered as a default, because the difference between them is itself informative
Both ends of a withholding rate are admitted: Nought is a source country that withholds nothing, and a reader comparing an arrangement against a hypothetical no-withholding baseline is asking a real question. A hundred is the boundary at which the net yield is nothing. Neither inverts the mechanism, and refusing either would hide the shape of it at its limit behind a validation message
Excluded: Every composition weighting and country list, every published withholding or treaty rate, the fund’s ongoing charge and tracking difference, whether a fund actually qualifies for the treaty rate its domicile is entitled to, the holder’s own tax, price return, currency, and whether either arrangement is better
Rounding: Display only; intermediate values remain unrounded

Correction history

The model’s frozen record anticipated a calculation that would have to carry a composition weighting across source countries and publish the assumption that produced it. The research found that UBWHY may not carry one at all — no authority publishes an effective blended rate on a dividend basis, and a UBWHY weighting would be an estimate of a tax fact rather than a stated assumption about one. So the model was made narrower than it was commissioned to be: no composition, no country list, no market-capitalisation share and no dividend share, and no field that could hold any of them. The arithmetic that survives is over rates the reader states, and the thing the model publishes that no observed source does is the warning that a single treaty rate describes one source country.
The check that carries the most weight in this model’s verification is not numerical. It asserts that a non-zero level with a differential of exactly zero is reachable — one rate applied to both arrangements gives 0.229 percentage points a year and a difference of nothing — and it exists so that the two fields cannot be quietly collapsed into one by a later change. The two quantities have different error terms, and a surface that computed one and labelled it the other would be wrong in a way no reader could see. That is the only kind of wrong this site has no defence against, so it is checked rather than reviewed.