What exactly happens if the swap counterparty defaults?

Unfunded: the fund keeps its owned basket, claims the unsettled MtM (small by design) against margin/estate, and re-swaps with another bank or converts to physical tracking. Funded: the fund enforces on the segregated collateral. Disruption and days of imperfect tracking: yes; the loss of the fund: no, by construction.

Why do substitute baskets hold stocks unrelated to the index?

Because the basket's only jobs are being owned, liquid and lendable; the swap supplies the index. Resemblance would add cost without adding safety.

Is the 10% cap per fund or per counterparty?

Per counterparty: a panel of four could theoretically run 40% gross, which is exactly why reset policies and actual reported exposures (typically ~0–2%) are the numbers to read, not the legal ceiling.

Do swap fees eat the withholding edge?

They take a bite (competitive panels shrink it); the net result is empirical and public: Chapter 35's TD comparison of synthetic vs physical peers on the same index. On S&P 500 UCITS funds the net edge has persistently favored synthetics; on indices without a withholding story, physical usually wins on simplicity.

Are US-listed ETFs ever synthetic like this?

Mainstream US index ETFs, no (Chapter 11). Swap-heavy US products exist in the leveraged/inverse aisle: a different machine with a different manual (Chapter 72).

Synthetic ETFs Deep Dive: Swaps and Collateral

Architecture one: the unfunded swap (the modern default)

The fund takes investors' cash and buys a substitute basket it fully owns: typically large, liquid, boring equities (often European blue chips, chosen for liquidity and lending value, not for resembling the target index). Simultaneously it enters a total return swap: fund pays the basket's return, receives the target index's return (the gross-dividend version, where the Chapter 11 tax edge lives).

The exposure arithmetic: at any moment, counterparty exposure = the swap's mark-to-market in the fund's favor, i.e., how far the index return has outrun the basket return since the last settlement. If the counterparty vanished, the fund keeps its basket (it owns it outright: this is the structure's bedrock) and loses only that unsettled increment. Managed practice keeps the increment small:

Resets: the swap is settled (marked to zero) on triggers: daily at several providers, or whenever exposure crosses a threshold (commonly well under the 10% legal cap; many target ≤ a few percent, some effectively zero via daily resets).
Collateral on top: many programs additionally exchange margin on the swap under EMIR-era practices, pushing net exposure toward or below zero.

The unfunded machine. The vault is the fund's property; the cable's unsettled reading is the exposure.

Architecture two: the funded swap

Here the fund transfers its cash to the counterparty, which delivers the index return and posts collateral into a segregated custody account, pledged to (or title-transferred to) the fund, typically over-collateralized (105–120% depending on asset quality). The fund's protection is the collateral pool rather than an owned basket. Functionally similar; legally different in ways that matter in a default: pledge structures require enforcing on collateral, and collateral composition becomes the whole ballgame. The market's evolution has favored unfunded structures for mainstream products precisely because "we own the basket" is the cleaner sentence; funded structures persist in commodity ETCs and some legacy ranges.

Collateral quality: the paragraph that separates funds

Whatever the architecture, read the collateral/basket disclosure with three questions:

What assets? G10 government bonds and large-cap developed equities are the right answers. UCITS rules impose eligibility, diversification and correlation constraints; provider policies publish the actual mix, often daily.
What haircuts? Collateral is valued at discounts (haircuts) by asset type: equities at 90–95%, bonds nearer par: producing the over-collateralization percentages. Published in the policy.
Correlation sanity: the historical sin (pre-2011 reform era) was equity exposure collateralized by illiquid or correlated junk: exactly what fails when needed. Post-ESMA-guidelines practice is dramatically cleaner, and the daily-published baskets let you verify rather than trust.

The multi-counterparty evolution deserves its mention: flagship synthetic funds increasingly swap with panels of several banks (three to six), diversifying the residual exposure and putting counterparties in fee competition. One more line in the checklist.

The recurring honest comparison

Every risk here has its physical-fund mirror (Chapter 36): a physical fund lending 40% of assets against collateral runs the same species of exposure as a synthetic fund's swap, managed with the same tools (over-collateralization, daily marking, quality rules). The structures differ in where the counterparty sits, not in whether one exists. The clean-hands option is a physical fund that doesn't lend; it also, on US indices, pays the withholding toll the synthetic avoids. Adults choose between priced trade-offs; this chapter's job is making sure the price tags are legible.

The five-item checklist, answerable in fifteen minutes

From the prospectus (swap section), the provider's collateral pages, and the annual report:

Architecture: unfunded (fund owns basket) or funded (pledged collateral)? Prefer understanding it to preferring one.
Reset policy: daily / threshold X% / at creations-redemptions? Lower and more frequent = smaller residual exposure.
Counterparties: named? Panel or single? (Single-counterparty with the provider's own affiliated bank is the configuration deserving extra attention: permitted, disclosed, and worth noticing.)
Collateral/basket quality: composition published, haircuts stated, correlation rules per ESMA?
The scoreboard: multi-year TD versus the gross-return-relevant benchmark (Chapter 35): is the structural edge actually being delivered to you, net of swap fees?

The five-item audit. Every answer is public; almost nobody looks them up.

A worked example

Elias (Chapter 21's synthetic S&P 500 buyer) does the fifteen minutes:

Unfunded; substitute basket: ~180 European large caps, published daily.
Swap reset daily and at any creation/redemption: residual exposure historically ~0.0–0.2% of NAV per the annual report.
Panel of four counterparties, named, none affiliated with the issuer.
Basket haircuts per policy; ESMA-compliant diversification; he spot-checks Tuesday's basket file: Nestlé, SAP, TotalEnergies... boring, as designed.
5-yr TD −0.04% vs the gross-relevant comparison, i.e., the ~15bp withholding edge over the physical peer, delivered, after swap costs.

Fifteen minutes, five public answers, and the word "synthetic" has been replaced in his head by a specific machine with published gauges: which is this book's definition of informed consent.