Cheapest Index Fund vs Tracking Difference: 10-Year Test

Key takeaways

• Across four S&P 500 UCITS ETFs with published charges from 0.03% to 0.12% all-in, of the three printing a ten-year return the cheapest fund finished last and the dearest finished first.
• Invesco's swap-based tracker beat the S&P 500 net total return index by roughly 0.43 percentage points a year over the decade to July 2026; State Street's 0.03% fund beat it by 0.20.
• All four funds outran their own benchmark, because the net index assumes 30% dividend withholding while the US-Ireland treaty, effective from 1998, caps source-country tax on portfolio dividends at 15%.
• $10,000 tracking the net index over those ten years grew to $38,745; the same money in the dearest of the four funds grew to $40,238, a gap of $1,493.
• Morningstar's 2016 study found the cheapest fifth of US equity funds succeeded 62% of the time against 20% for the priciest, across a 1.55 point fee spread, not the 0.09 points separating these four.

The fee table points one way, the ten-year record points the other

Does the cheapest index fund give you the best return? It feels like it must. If four funds hold the same 500 American companies and differ only in what they charge, the one charging least should keep the most.

Over the ten years to 31 July 2026, that's not what happened. The cheapest of the four large S&P 500 UCITS trackers, State Street's SPDR fund with a 0.03% total expense ratio, returned 14.71% a year net of fees. Vanguard's fund charges more than double that, 0.07%, and returned 14.72%. The dearest, Invesco's swap-based fund with an all-in cost of 0.12%, returned about 14.94% a year, the widest margin over the index in the set. The fourth fund, iShares at 0.07%, prints no ten-year row at all, so the ranking below is of the three that publish one.

Every figure comes from the issuer's own factsheet, all four carrying performance data as at 31 July 2026, and every fund measures itself against the same benchmark, the S&P 500 net total return index. The chart above plots the three ten-year figures against the index. Invesco's lead isn't one decade's luck, either: it repeats in every calendar year from 2021 through 2025, in the table further down.

So on this evidence the fee row on a factsheet told you very little about the return row. The cheapest index fund was a perfectly good buy that finished behind two dearer ones. What decided the ranking was tracking difference, the gap between what a fund delivered and what its index did, and that gap has a cause worth understanding before anyone trusts it to continue.

Four S&P 500 ETFs, and the fee schedules they published for July 2026

The four funds here are large Irish-domiciled S&P 500 trackers with UK factsheets, reporting net fund assets from $42.4bn for the State Street fund up to $154.1bn for the iShares one. State Street, Invesco and Vanguard all print the same index ticker, SPTR500N, the net total return version; the iShares sheet names its benchmark only as the S&P 500 Index, though the benchmark returns it prints match the net series to the basis point. Their published charges, from performance data as at 31 July 2026:

• State Street SPDR S&P 500 UCITS ETF: total expense ratio 0.03%, physical replication.
• Invesco S&P 500 UCITS ETF: ongoing charge 0.05% plus a swap fee of 0.07%. Invesco's footnote is explicit that "the total cost is the sum of the ongoing charge figure and swap fee", which makes 0.12% all-in. Synthetic replication.
• iShares Core S&P 500 UCITS ETF: total expense ratio 0.07%, physical replication.
• Vanguard S&P 500 UCITS ETF: ongoing charges figure 0.07%, physical replication.

Tracking difference is the number those fees are supposed to predict: the fund's return minus the index's return over the same period. It isn't the same thing as tracking error, which measures how much that gap wobbles from month to month; the distinction between tracking difference and tracking error is worth a guide of its own.

A fee is one input to tracking difference, and it's the only input printed in bold. Everything else that moves the gap, withholding tax treatment, securities lending income, sampling, cash drag, swap pricing, sits in the fine print or nowhere at all. The total cost of owning a fund runs through those hidden layers one by one; this piece is about what they added up to in practice.

Every fund beat its own index, because the index assumes a tax the funds don't fully pay

The strangest thing in the ten-year record isn't the ordering. All four funds, fees and all, returned more than their benchmark. An index fund is supposed to lag its index by roughly its fee. These didn't lag at all.

The explanation sits in a footnote on Vanguard's factsheet: "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."

The net index assumes that 30% of every dividend is lost to US withholding tax. An Irish-domiciled fund doesn't lose that much. The US-Ireland tax convention, signed at Dublin on 28 July 1997 and effective from 1 January 1998, caps source-country tax on portfolio dividends at "15 percent of the gross amount of the dividends", and its residence article names an Irish "Collective Investment Undertaking" among the residents of Ireland for the convention's purposes. Half the tax the benchmark assumes never leaves a qualifying fund. Those retained dividends form a cushion, and the fee is paid out of the cushion rather than out of your return relative to the printed index.

For the two physical funds that print calendar-year tables, the cushion covered the fee with room to spare: SPDR and iShares each beat the net index by between 0.15 and 0.25 percentage points in every year from 2021 to 2025. Vanguard reports rolling twelve-month years instead, and finished above its benchmark in all ten of them. The synthetic fund did better still. Invesco's fund doesn't hold the 500 shares and collect their dividends; it holds a substitute basket of equities and exchanges that basket's return for the index's through swap agreements with banks, a structural choice with consequences beyond fees, which is why the physical vs synthetic ETF question gets a guide of its own. Invesco's sheet states the fund "aims to track the net total return performance of the S&P 500 Index, less fees", and then reports beating that index in every period shown.

The cheapest index fund posted the smallest ten-year tracking difference

Here's the decade, fund by fund, against the index's 14.50% a year. State Street prints the same index at 14.51%; hold that basis point of disagreement in mind, because it matters later.

• Invesco, 0.12% all-in: 14.94% a year, about 0.43 percentage points above the index. Its ten-year cumulative return of 302.38%, against the index's 287.45%, is the widest ten-year gap printed on any of the four sheets.
• Vanguard, 0.07%: 14.72% a year, 0.22 points above.
• SPDR, 0.03%: 14.71% a year. State Street's own table puts the net-of-fees gap at 0.20 points a year, or 6.92 points cumulative.
• iShares, 0.07%: its current sheet prints no ten-year row, so it cannot be ranked here. Its calendar-year record from 2021 to 2025 beats SPDR's in three of the five years, on a fee more than twice as high.

In money, in the dollars these funds report in: $10,000 left tracking the net index for those ten years becomes $38,745. In the SPDR fund it became $39,437, which is $692 more than the index the fund was supposedly losing 0.03% a year to. In the Invesco fund it became $40,238, $1,493 more. The spread between the cheapest and dearest fee schedule, 0.09 percentage points, was overwhelmed by the spread in what the structures delivered.

The year-by-year record, in percentage points of fund return above the net index. Invesco and iShares figures are differences between the fund and index returns printed on their factsheets; State Street prints the differences directly. Vanguard reports rolling August-to-July years rather than calendar years, so it sits this table out.

Consistency is the striking part. None of the fifteen fund-years in that table is negative, and Invesco's column is the largest in all five years. State Street's sheet also reports a three-year annualised tracking error of 0.02, meaning the gap barely wobbles around its average. Whatever generates the outperformance recurs every year, in up markets and in 2022's fall alike, which points to structure rather than luck.

Morningstar's fee evidence is the strongest counter-case, and it measures a different spread

The case for choosing on fee has serious evidence behind it, and its best-known statement is Morningstar's. In a May 2016 study, Russel Kinnel called the expense ratio "the most proven predictor of future fund returns". From 2010 to 2015, the cheapest fifth of US equity funds went on to survive and beat their category 62% of the time; the priciest fifth managed 20%. International equity showed 51% against 21%. The study counted dead funds too, so the gap isn't survivorship playing tricks.

Both findings are true at once, because they measure different spreads. In Morningstar's US equity sample, the cheapest quintile averaged a 0.65% expense ratio and the priciest 2.20%: a 1.55 percentage point gap, wide enough to swamp everything else a fund does, and wide enough that fund fees compounded over decades decide the outcome on their own. The four trackers here are separated by 0.09 points. At that distance the fee stops being the biggest force on the page, and withholding treatment, replication method and lending income take over. Fees predict outcomes across price tiers. Within this tier, they decided nothing.

There's a second counter-argument, and it's sharper. The fee is a contract; the tracking difference is a history. The 0.03% is what State Street's factsheet says the fund charges now. The +0.43 is what Invesco's structure happened to deliver over one particular decade. A reader who weights the promised number over the realised one is weighting certainty, and that's a defensible weight. What the record shows is narrower: for ten years the realised number has been larger than the promised one, more stable than a fee cut, and ordered the opposite way.

What the tracking difference record cannot tell you

These are NAV returns. Invesco's sheet says its figures "do not reflect the actual share price, the impact of the bid/offer spread or broker commissions". Bought at a premium, sold at a discount, wrapped in a platform's percentage fee, the basis points this piece has been counting can disappear inside a single badly timed trade.

The sample is one index, one domicile, one decade. US large-cap equity held through Irish funds sits at the friendly end of withholding treatment; a tracker of another market faces different treaty rates and can sit the other side of its net benchmark. Ten years to July 2026 is one path through history, and the whole game is played on a dividend yield that State Street prints at just 1.13%. A higher-yielding index would make the withholding cushion bigger; a growthier one would shrink it.

Precision has limits here too. Vanguard prints the ten-year index return at 14.50% and State Street prints it at 14.51%. When two administrators disagree by a basis point about the same index, differences of a few hundredths between funds are inside the measurement noise. The separation between the Invesco fund and the rest, at roughly 0.2 points a year, is not.

And the sources are the sellers. Every performance table above is the issuer's own marketing document. The figures cross-check, four issuers printing consistent index rows is reassuring, but nobody publishing these sheets is neutral. Securities lending adds return and adds a risk the table doesn't show. A swap adds counterparty exposure that a decade of smooth tracking difference says nothing about, and that exposure is precisely what the extra 0.43 a year is partly paying you for.

What would change the conclusion

The cushion is a tax rule, so tax can take it away. The 30%-versus-15% gap rests on a treaty and on how US law treats the dividend leg of index derivatives. A renegotiated convention, or a US move to tax dividend-equivalent payments on swaps more heavily, would shrink every number in the table above, and would shrink the synthetic fund's edge first and hardest.

A wider fee gap restores the fee's power. Move from comparing 0.03% against 0.12% to comparing either against the 0.65% and 2.20% tiers in Morningstar's data, and cost is once again the whole story. This finding lives inside the tracker price war, not outside it: the reason fees couldn't separate these funds is that competition has already crushed them to within 0.09 points of each other.

The record itself can also turn, because a tracking difference is re-earned every year. The number worth watching isn't the fee row but the most recent two or three calendar years of fund-minus-index on each factsheet, the same rows this piece is built on. A portfolio tracker such as LedgerTouch shows the return you actually received beside the index you meant to own, which is where a fading edge would first become visible. If the cheapest index fund starts printing the widest tracking difference, the easy answer and the right one finally agree.