Key takeaways
- UCITS rules cap a fund's risk exposure to a single OTC derivative counterparty at 10% of assets when that counterparty is a bank, and 5% in every other case.
- Xtrackers manages the net exposure of its unfunded swap funds to 0%, calling collateral from the counterparty by the end of the next trading day whenever it drifts above that.
- ESMA lets one collateral issuer account for 20% of net asset value, aggregated across counterparties, so a fully collateralised fund still carries concentration you can look up.
- A swap on a qualified index sits outside US dividend withholding, which Invesco says is worth an extra 15% of dividend value against an Irish physical fund.
- Invesco's swap-based S&P 500 ETF beat its net-total-return index in all ten calendar years from 2016 to 2025, by margins of 0.31 to 0.67 points.
Three ways to track an index, and only one of them buys the shares
You've found two S&P 500 trackers. One holds the American shares. The other holds a basket of unrelated equities and swaps that basket's return for the index return with a bank. You want to know what the second one exposes you to, and whether the exposure is priced.
Here's the short answer. Under UCITS rules a fund's risk exposure to one counterparty in a derivative that isn't centrally cleared can't exceed 10% of its assets when that counterparty is a bank, or 5% in other cases. That's the legal ceiling, and the large European swap providers don't sit anywhere near it. Xtrackers' prospectus says it manages the net exposure of its unfunded swap funds to 0%, and calls collateral from the counterparty by the end of the next trading day if the number moves above that. So what you're carrying isn't the headline 10%. It's one trading day of price movement, the adequacy of a haircut, and a legal claim on a collateral pool you'd have to enforce.
What you get in exchange is specific, and in US equity it's measurable. The rest of this piece takes each replication method apart, then shows where the counterparty number actually comes from and what the swap is buying.
Full replication and optimised sampling differ in what a fund is allowed to leave out
Physical replication has two settings, and the prospectuses name them. Xtrackers splits its physical funds into "Full Replication Funds", which hold "all, or a substantial number of, the constituents of the relevant Reference Index broadly in proportion to the respective weightings", and "Optimised Replication Funds", which "may not hold every constituent or the exact weighting of a constituent in the Reference Index".
Read the full-replication definition again. "A substantial number of" is not "all". The same document says full replication funds "may from time to time not contain all of the constituents of the Reference Index", and lists the reasons: illiquid constituents, limited availability of shares, legal or regulatory restrictions, the size of the fund. Full replication is a target with named exceptions, not a promise.
Optimisation abandons the target deliberately. An optimised fund seeks a similar return by investing "in a sub-set of the constituents of the Reference Index", by "utilising optimisation techniques", or by holding "securities that are not part of that Reference Index". It's a statistical bet that a smaller portfolio behaves like the index, and it's how funds handle benchmarks with thousands of small, thinly traded members. The bet fails when the excluded names move differently from the included ones. What that failure looks like in published figures is the subject of our piece on tracking difference versus tracking error.
A swap-based fund still owns securities, it just swaps their return away
Synthetic replication isn't the absence of a portfolio. In the dominant European structure, the unfunded swap, the fund buys real securities and then trades away their performance.
Invesco's supplement for its S&P 500 ETF sets the mechanism out plainly. The fund invests in "a basket of global equity securities and equity related securities" and, "in exchange for the performance/return of the Basket with an Approved Counterparty, will receive the return of the unhedged Reference Index" through unfunded swaps. The fund owns the basket. The swap settles the difference between what the basket did and what the index did. The S&P 500 itself, on the same factsheet, "comprises around 500 companies".
Xtrackers describes the same two options and puts numbers on them. For unfunded swaps the maximum proportion of net asset value subject to the swap "is 110 percent", with an expected proportion of "100 percent". The funded version, where the fund hands over cash rather than holding a basket, carries the same ceiling — and there the collateral "will represent at least 100 percent" of the fund's gross total counterparty risk exposure, marked to market daily. Amundi says the same thing in retail language: its fund "will invest into a total return swap (financial derivative instrument) delivering the performance of the Index against the performance of the assets held".
That difference between funded and unfunded matters more than it sounds. In an unfunded swap the fund is the beneficial owner of the securities it holds. In a funded swap the collateral sits pledged in a ring-fenced account, which the BIS noted in April 2011 "can potentially lead to delays in realising the value of collateral assets if the swap counterparty fails".
The 10% counterparty cap is a ceiling nobody in this market runs at
The limit itself is short. A UCITS fund's risk exposure to one counterparty in a non-centrally-cleared derivative "shall not exceed either: (a) 10 % of its assets when the counterparty is a credit institution ...; or (b) 5 % of its assets, in other cases." Swap counterparties are banks, so the operative number is 10%. Amundi states it in its own words: "the exposure to the counterparty cannot exceed 10% of the total assets of the fund."
Two further limits sit on top. A fund can't combine securities, deposits and derivative exposures with one body past 20% of assets. And ESMA's guidelines require the exposures from OTC derivatives and from securities lending to be added together before the counterparty limit is tested, so a physical fund that lends stock is using the same allowance. Total return swaps are measured against Article 52 of the UCITS Directive, and the guidelines say the swap's underlying exposures "shall be taken into account" when the limits are calculated.
The chart below sets those figures beside what the providers actually run. The gap is the story. A cap of 10% is what the law permits; 0% is what Xtrackers says it targets, resetting through collateral by the end of the following trading day.
What's left after collateral is a day, a haircut and a legal claim
"Fully collateralised" is not the same as "no exposure", and ESMA's collateral rules tell you exactly where the residue sits.
Collateral must be highly liquid, valued at least daily, of high quality, and "issued by an entity that is independent from the counterparty". It must be "capable of being fully enforced by the UCITS at any time without reference to or approval from the counterparty" — the clause that decides whether you get paid quickly or litigate. Non-cash collateral "should not be sold, re-invested or pledged". And a fund receiving collateral for at least 30% of its assets must run a stress-testing policy on the collateral's liquidity.
Now the concentration. A basket can carry "a maximum exposure to a given issuer of 20% of the UCITS' net asset value", and where a fund faces several counterparties, the baskets "should be aggregated to calculate the 20% limit". So a fund can be collateralised to the last cent and still have a fifth of its net asset value riding on one issuer's paper. That's a smaller risk than an uncollateralised swap and a larger one than the word "collateralised" suggests.
The prospectus language is careful about the rest. Xtrackers writes that its collateral management is "subject to market risk and price movements prior to the end of trading day T+1 and settlement risk". Invesco's supplement is blunter about concentration: there is "no requirement for the Fund to execute transactions with more than one Approved Counterparty", and no agreement obliging a surviving counterparty "to make good any losses which the Fund may incur as a result of a counterparty default". Invesco says elsewhere that it uses "up to six swap counterparties per ETF". The practice and the permission are different documents.
The reason swap funds persist in US equity is a line in the US tax rules
Start with what a foreign fund pays. The statutory US withholding rate on US source income paid to a foreign recipient is 30%. An Irish-domiciled fund does better, because the US-Ireland treaty caps the tax on dividends at "15 percent of the gross amount of the dividends in all other cases".
A swap changes the question. US rules would normally treat a payment referencing a dividend on an underlying security as a dividend equivalent, taxed as if it were a US dividend. But the section 871(m) regulations carve out indices: "a qualified index is treated as a single security that is not an underlying security". No underlying security, no dividend equivalent, no withholding on that leg.
Qualified status is a test, not a label. An index must reference "25 or more component securities", no single component above "15 percent" of the weighting, no five components above "40 percent" together, rebalancing only by "publicly stated, predefined criteria", a dividend yield no greater than "1.5 times" that of the S&P 500, and futures or options traded on a recognised exchange. A broad US large-cap index clears all of them comfortably. A concentrated sector index may not.
That's the whole edge. Invesco puts a figure on it: a synthetic ETF domiciled in Ireland "can benefit from an additional 15% of dividend values, compared to a physically replicated ETF also domiciled in Ireland". Its own fund supplement says the same thing in fund-document language — the index return used to calculate the swap "may reflect a lower rate of withholding tax than ordinarily applied within the Reference Index". Withholding is one of three drags a cross-border holder faces, alongside dealing and platform costs; we took the full toll apart in our piece on FX, custody and withholding charges.
Ten calendar years, ten small positive gaps
Structural arguments are cheap. The published record is not.
Invesco's S&P 500 ETF is measured against the S&P 500 Net Total Return index, which is calculated after an assumed withholding deduction. On the factsheet dated 30 June 2026, the fund's calendar-year returns run 17.75%, 24.91%, 26.18%, -18.17%, 28.64%, 18.30%, 31.37%, -4.47%, 21.41% and 11.54% for 2025 back to 2016. The index returned 17.43%, 24.50%, 25.67%, -18.51%, 28.16%, 17.75%, 30.70%, -4.94%, 21.10% and 11.23% over the same years.
The fund finished ahead in all ten. The margin ranged from 0.31 points to 0.67 points, and it did that while charging a 0.05% ongoing charge plus a 0.07% swap fee — a total cost of 0.12%, which Invesco's own footnote insists on adding together. Amundi's swap fund on the same index reports 0.15% in management and administrative costs. Positive tracking difference after costs isn't skill. It's a tax treatment showing up in a return series, year after year, on a share class holding $56.3bn.
Note what the comparison is not. This is a fund against its own benchmark, not a fund against a physical rival, and the benchmark's withholding assumption does much of the work. The mechanics of how any ETF's market price stays tethered to that net asset value are a separate question, covered in our piece on ETF creation and redemption.
Stamp duty and transaction taxes do the same job outside the US
The US case is the loudest, but it isn't the only one. Invesco lists the markets where synthetic structures carry a structural edge: US, UK, European, global and China A-shares indices. A physically replicated UK fund has to "pay Stamp Duty" on its purchases; funds tracking Italian and French shares meet a financial transaction tax. Swap counterparties are typically exempt on their hedging trades, so the cost never reaches the fund.
This is why a global index can compound several exemptions at once. On Invesco's account a synthetic MSCI World fund "should not be liable for dividend withholding tax on the US components, nor should it pay Stamp Duty or FTT on its UK and Italian and French constituents". The pattern holds: synthetic replication survives where a tax or a transaction cost falls on the holder of the shares, and the swap moves the holding somewhere else.
The strongest case against synthetic ETFs was made in April 2011
The serious objection isn't hypothetical, and it came from the regulators. In April 2011 the Financial Stability Board flagged synthetic ETFs, then "45% of that market" in Europe, as a source of contagion risk. The argument was structural. The swap counterparty was typically the same bank that ran the fund, so "problems at those banks that are most active in swap-based ETFs may constitute a powerful source of contagion and systemic risk". Worse, because "there is no requirement for the collateral composition to match the assets of the tracked index", the FSB argued the structure could become a way to fund an illiquid inventory the repo market wouldn't take. The BIS made the same point that month, describing the same limit from the other side: collateral covering "at least 90%" of net asset value, "limiting the swap counterparty risk to a maximum of 10% of the ETF's market value".
Much of that critique was answered by the rules that followed. ESMA's guidelines imposed issuer diversification on collateral, required it to be independent of the counterparty, banned re-use of non-cash collateral, and forced the prospectus to name the counterparties and describe "the risk of counterparty default and the effect on investor returns". The unfunded structures the big providers run now leave the fund owning the basket outright.
The critique isn't dead, though. Invesco's supplement still permits a single counterparty. The collateral machinery has never been tested by a large swap counterparty actually failing. And the conflict the FSB identified — bank as manager, bank as counterparty — is mitigated by disclosure rather than removed by structure. A reader who treats the 2011 objection as fully retired is going further than the evidence does.
What these documents cannot tell you
Almost everything above is a rule or an issuer's description of its own policy. Rules state a ceiling, not an outcome. Policies use words like "generally require" and "with a view to ensuring", and both Xtrackers and Invesco explicitly reserve the intra-day gap, the settlement gap and the market-risk gap. None of that is evasive. It's just not a guarantee, and it shouldn't be read as one.
The performance evidence is thinner than it looks. It's one fund, one index, one currency, ten calendar years, taken from a marketing factsheet the manager produced. A different decade, a different index or a different provider's swap pricing could produce a different sign. The swap fee is set by negotiation with counterparties, and nothing obliges it to stay at 0.07%.
The counterparty question has no live test data at all. No large UCITS swap counterparty has defaulted since the collateral guidelines took effect, so the collateral rules have been stress-tested on paper and not in a real failure. Haircuts, valuation timing and enforceability all sound adequate in a prospectus. What they're worth is a question about a week nobody has lived through yet.
The market-share figure is a period piece. The FSB's 45% describes Europe as it was, not as it is, and the composition of the synthetic market has shifted heavily towards the handful of indices where the tax argument bites.
What would change the conclusion
If the qualified index exception were narrowed. The whole US case rests on one regulatory carve-out, and qualified status is retested at the start of each calendar year against the weighting and dividend-yield conditions. A change to the section 871(m) rules, or an index drifting past the 15% single-name weighting test, would remove the edge without changing anything about the fund.
If a swap counterparty actually failed. A default that recovered collateral cleanly, at the marked value and inside the reset window, would settle the argument in the industry's favour. One that realised meaningfully below the reset value would show that 0% net exposure was an accounting position rather than an economic one.
If withholding relief moved the other way. The 15-point gap between the statutory 30% and Ireland's treaty rate is what makes the physical fund's tax drag worth avoiding. A treaty change, or a broader relief for foreign funds, would compress the swap's advantage towards its costs.
The figure worth tracking isn't the replication label on a factsheet. It's the counterparty section of the annual report — which counterparties, how many, and what the collateral was on the reporting date. Xtrackers and Invesco both publish the basket. Once a fund is more than a small slice of your equity book, that page is worth more than the ongoing charge, and it takes about as long to read.