Key takeaways
- In calendar 2025 the Invesco S&P 500 UCITS ETF returned 17.75% against its index's 17.43%, a tracking difference of 0.32 points, on an all-in cost of 0.12%.
- The cheapest share class in the same group charges 0.03% and beat that identical index by 0.16 points, roughly half the surplus the costliest one delivered.
- The iShares MSCI UK Small Cap fund trailed its index by 0.77 points in the year to July 2025, and still posted a realised tracking error of only 0.08%.
- The iShares MSCI Japan fund trailed by 0.06 points over the same year, with a realised tracking error of 0.41% against an anticipated ceiling of 0.15%.
- Securities lending earned the iShares Core S&P 500 fund $3.01m in that year, about 0.25 basis points on a share class holding $121.8bn.
The cheapest S&P 500 tracker in 2025 was not the one that tracked best
You're looking at two S&P 500 ETFs. One charges 0.03% a year, the other 0.07%. Which one tracked the index better?
On last year's published figures, neither. The share class that beat the S&P 500 by the widest margin in calendar 2025 was the one with the highest all-in cost, at 0.12%. That's four times the cheapest fund in the group, and it won by roughly double.
The number that captures this is tracking difference: how far a fund's return landed from its index's return over a stated period. It isn't the ongoing charge, and it isn't tracking error either. Those two get used interchangeably in fund comparisons, and they measure genuinely different things.
Tracking difference and tracking error answer two different questions
Both terms have regulatory definitions, and they sit in the same glossary. ESMA's Guidelines on ETFs and other UCITS issues define annual tracking difference as "the difference between the annual return of the Index-tracking UCITS and the annual return of the tracked index".
One number, one period, all causes included. It's the whole shortfall or surplus, net of everything the fund did and paid.
Tracking error is defined in the same document as "the volatility of the difference between the return of the Index-tracking UCITS and the return of the index or indices tracked". It measures how much the gap wobbled, not how big the gap was.
A fund can trail its index by an identical sliver every single day. Its tracking difference is large and its tracking error is close to zero. Another fund can finish the year level with its index after swinging either side of it for months. Tracking difference near zero, tracking error large. Both figures are public: ESMA's paragraph 11 requires the annual report to state the tracking error and to "disclose and explain the annual tracking difference".
Two iShares funds show the two numbers pulling apart
The iShares VII plc annual report for the financial year to July 2025 puts both figures in one table, fund by fund, with the causes ticked off beside them. Two rows make the point better than any definition.
The iShares MSCI UK Small Cap UCITS ETF returned 5.67% against a benchmark return of 6.44%. It trailed by 0.77 points, more than its 0.58% total expense ratio. Its realised tracking error was 0.08%, among the lowest in the whole table. The fund missed by a wide margin, and it missed steadily.
The iShares MSCI Japan UCITS ETF did the reverse. It returned 6.07% against 6.13%, trailing by 0.06 points. Its realised tracking error was 0.41%, against an anticipated level of "up to 0.15%" set in the prospectus. BlackRock's own footnote attributes the miss to "a difference in valuation between the Fund and the benchmark index, caused by a differing holiday treatment".
Rank those two funds on tracking error alone and you pick the one that cost 0.77 points. That is the case for tracking difference in a single comparison, and it comes from the manager's own audited report.
Five S&P 500 share classes, one index, one calendar year
Every share class below tracks the same series: the S&P 500 Net Total Return index, Bloomberg ticker SPTR500N. All report in US dollars, net asset value to net asset value, with income reinvested. The index returned 17.43% in calendar 2025.
| Share class | Charge | 2025 return | Tracking difference | Replication |
|---|---|---|---|---|
| SPDR S&P 500 UCITS ETF (Acc) | 0.03% | 17.60% | +0.17 | Physical |
| SPDR S&P 500 UCITS ETF (Dist) | 0.03% | 17.59% | +0.16 | Physical |
| iShares Core S&P 500 UCITS ETF (Acc) | 0.07% | 17.58% | +0.15 | Physical |
| Vanguard S&P 500 UCITS ETF (USD Acc) | 0.07% | 17.58% | +0.15 | Physical |
| Invesco S&P 500 UCITS ETF (Acc) | 0.12% all-in | 17.75% | +0.32 | Synthetic |
The chart above plots that last column. Four of the five sit within 0.02 points of each other while their charges differ by more than a factor of two. The one that stands apart is the one that costs the most, and the 0.15 points it opens over the cheapest share class exceed any single charge in the table.
The index assumes 30% withholding tax, and no European fund pays that
Start with the benchmark, because that's where the surplus is born. Vanguard's factsheet spells out the convention in a footnote: "The S&P 500 Net Total Return Index represents price-plus-net cash dividend return. Net cash dividend equals reinvested dividends less 30% withholding tax."
So the index is calculated as though a US withholding tax of 30% were deducted from every dividend. An Irish-domiciled fund doesn't pay that rate. The IRS treaty table lists the withholding rate on dividends paid by US corporations to residents of Ireland at 15%.
The arithmetic follows. The fund keeps 15 percentage points more of its dividend income than the index assumes it keeps. Vanguard reports a dividend yield of 1.1% on the portfolio, so that differential is worth roughly 0.17 points a year. The physical funds in the table delivered 0.15 to 0.17.
The synthetic fund takes one further step. Invesco's ETF holds a basket of securities and swaps their return for the index return with a counterparty. Under US regulation, a contract referencing a "qualified index" is treated as "a single security that is not an underlying security", which puts it outside the dividend-equivalent withholding rules. The S&P 500 clears the diversification tests: the rule requires an index that "References 25 or more component securities" and "References no component underlying security that represents more than 15 percent of the weighting". iShares reports NVIDIA, the largest constituent, at 7.51% of the portfolio, with the top 10 holdings at 36.39%.
None of this is a new finding. Blitz, Huij and Swinkels, writing in European Financial Management, studied European index funds and exchange-traded funds and concluded that the explanatory power of dividend withholding taxes as a determinant of their underperformance was "at least on par with fund expenses". Their sample and their benchmarks differ from the table above, so the sign of the gap differs too. What carries across is the ranking of causes: tax sat alongside the fee, not behind it.
A swap-based S&P 500 fund can therefore be paid something close to gross dividends while the index it's measured against assumes 30% went to the IRS. That is the most plausible source of the 0.32 in the table. The fund isn't cheaper than its rivals; it's taxed differently, and the ongoing charge has no way of telling you so.
The ongoing charge is not one number, and providers do not compute it alike
Comparing headline charges assumes they cover the same things. They don't. Vanguard's factsheet defines its ongoing charges figure as covering "administration, audit, depository, legal, registration and regulatory expenses". Invesco publishes a 0.05% ongoing charge and, separately, a 0.07% swap fee, with a footnote stating that "the total cost is the sum of the ongoing charge figure and swap fee".
A screener that reads the first field and stops shows Invesco at 0.05%, cheaper than iShares and Vanguard at 0.07%. The honest figure is 0.12%, the dearest of the five. Either way the fee ranking failed to predict the tracking ranking, and the second reading fails it harder.
None of these figures include what it costs to buy the thing. Dealing spreads, platform custody and currency conversion sit outside every number here, and for a retail holder they can be larger than all of them. We took that apart in the real toll of FX, custody and withholding charges.
BlackRock's own report names the drivers and ranks them fund by fund
The iShares VII table doesn't stop at the outcome. Each row carries ticks against the causes. For the iShares Core S&P 500 UCITS ETF, the ticks fall on "net income difference and tax" and "securities lending", and the report defines the first as "withholding tax rate differential, tax reclaims and income timing differences between the Fund and the benchmark index".
That fund returned 16.03% against a benchmark 15.87% in the year to July 2025, on a 0.07% total expense ratio. Gross of that fee the tracking difference was 0.23%; net of it the fund kept 0.16 points. Its realised tracking error was 0.02%, against an anticipated "up to 0.10%". Both numbers are good, and they are good for different reasons.
Securities lending is real money, and on an S&P 500 fund it is very small
Lending portfolio stock to borrowers for a fee is the second driver on that row, and it's the one most often oversold in marketing material.
In the year to July 2025 the iShares Core S&P 500 UCITS ETF earned $3.01m of securities lending income. Its US dollar accumulating share class held $121.8bn at the year end. That works out at roughly 0.25 basis points of return, and about 0.21% of the dividend income the fund collected over the same period.
The reason is borrow demand, not policy. Only 2.55% of the fund's lendable assets were out on loan at the year end. Nobody pays much to borrow the largest companies in the United States. In the same report, the iShares Nikkei 225 UCITS ETF had 32.17% of its lendable assets on loan.
Where lending does earn, the split matters. BlackRock's securities lending agent "will receive a fee of 37.5% of such securities lending revenue" and pays third-party operational costs out of that fee. On an S&P 500 tracker that division is an argument about a rounding error. On a Japanese or small-cap fund it isn't.
Sampling and cash are the other two levers, and they cut both ways
The third tick on the iShares table is "investment technique", which the report defines as "cash management, trading costs, currency hedging, futures held and sampling techniques".
Sampling means holding a subset of the index rather than every line in it, which is normal where the index has hundreds of small or illiquid members. The MSCI UK Small Cap fund is listed in the same report as "Index tracking - non-replicating". Its tracking difference gross of the fee was negative 0.19%. The sampling, dealing costs and cash didn't merely fail to offset the 0.58% charge; they added to the shortfall.
It runs the other way too. The iShares FTSE MIB UCITS ETF returned 27.19% against a 26.08% benchmark, a tracking difference of 1.44% gross of its 0.33% fee. Cash drag is the mildest of the levers: money sitting uninvested earns nothing while the index stays fully invested, which shaves returns in a rising market and cushions them in a falling one.
The gap shows up over a decade, not just over one year
A single year proves very little. The distributing SPDR share class has a 10-year record: the index returned 14.51% a year and the fund 14.71%, a surplus of 0.20 points annually against a 0.03% charge. Its annualised tracking error over 3 years was 0.02%.
Invesco's synthetic fund compounded 317.43% over the same 10-year span against 301.92% for the index, a difference of 15.51 percentage points of cumulative return. Two funds, two structures, and in both cases the direction held for a decade.
The counter-case: one index, one convention, one tax rule
Three objections deserve a proper hearing, and the last is serious.
First, the sign of every tracking difference here is an artefact of the benchmark convention. These funds beat their index because the index is built net of a 30% deduction. Measured against a gross total return version, all five would trail. Providers label the series inconsistently, too: iShares calls it "S&P 500 Index", State Street shows "Index Type: Net Total Return", Vanguard writes "S&P 500 Net Total Return". All three published 17.43% for 2025, so it's one series with three names, and the label alone can't tell you which you're looking at.
Second, tracking error isn't useless. It tells you whether the gap is stable, and a fund that promised "up to 0.10%" and delivered 0.02% kept a specific commitment. If you might sell on a particular day rather than at a year end, day-to-day dispersion is the risk you carry, not the annual average. Anyone sizing a position against a benchmark faces the same issue, which is a tracking-error budget question rather than a cost question.
Third, and this is the one that matters, the synthetic advantage rests on a tax regulation and a treaty rather than on anything the manager does well. The qualified-index exception is a rule that can be rewritten, and so can the 15% treaty rate. A swap also introduces a counterparty. Invesco's own factsheet warns that the fund's ability to track "is reliant on the counterparties to continuously deliver the performance of the benchmark in line with the swap agreements". An edge of 0.15 to 0.17 points is a thin reward for accepting that exposure, and a reasonable reader can decline it.
What this evidence cannot tell you
The sample is narrow. One index, one currency, one calendar year for the five-fund comparison, plus one fiscal year of BlackRock's cross-fund table. US large-cap equity is the easiest tracking problem in the business, with 500 liquid names and a published rebalance calendar. None of it transfers to emerging markets, small caps or corporate bonds, and the MSCI UK Small Cap row already shows the limitation: a 0.77-point shortfall on a 0.58% fee.
Past tracking difference is history, not a forecast. A fund that gained 0.32 points in 2025 is not promising the same next year, because the drivers are dividend yields, tax rules and borrow demand, and all three move. The annual report figures also cover a representative share class in the report's own words, so another class of the same fund can differ.
And the whole comparison is measured at net asset value. What you actually receive depends on the spread you trade at and the charges your platform adds on top, which is the ground covered in the costs that sit beyond the expense ratio.
What would change the conclusion
If the US removed the qualified-index exception from the dividend-equivalent rules, the synthetic advantage disappears and the ranking in the table inverts. The same follows if the US-Ireland treaty rate on portfolio dividends moves off 15%. Neither is inside a fund manager's control, and neither is announced in a factsheet.
If a provider switched its published benchmark from the net version to a gross version, every tracking difference above would flip sign without one thing changing inside the fund. That is worth sitting with. The headline number depends on a choice the fund company made about what to measure itself against, which is why the index ticker matters as much as the number beside it.
LedgerTouch reports what your holdings actually returned rather than what their factsheets charge, which is the same distinction one level up. The place to check any of this is the annual report, where ESMA requires tracking difference and tracking error to be published together every year, with an explanation whenever the realised error misses the anticipated one. The fee lives on the marketing page. The outcome lives in the accounts.