Physical vs synthetic ETF replication: what your fund actually holds

A physically replicating ETF does what most people assume all ETFs do: it buys the securities in its index, in index weights. For large liquid indices this is done in full, position by position. For indices with thousands of constituents — MSCI ACWI, broad bond benchmarks — full replication becomes expensive, so funds use optimised sampling: they hold a subset of securities chosen to match the index's risk characteristics, accepting tiny tracking deviations in exchange for lower trading costs.

Physical funds carry one commonly cited risk that deserves context: securities lending. Most large physical ETFs lend a slice of their holdings to short sellers for a fee, which offsets costs and often improves tracking. The loans are collateralised — typically over-collateralised — and UCITS rules cap counterparty exposure, so the realistic worst case is friction, not loss of the fund. Issuers disclose lending revenue and collateral policy; the better ones return most lending income to the fund.

A synthetic ETF holds a substitute basket of liquid securities — often large European equities — and enters a total return swap with a bank. The bank pays the fund the exact index return (including dividends) in exchange for the return of the substitute basket. Your exposure to the S&P 500 is therefore a contractual promise, not direct ownership of US stocks.

The obvious question is counterparty risk: what if the swap bank fails? UCITS rules cap uncollateralised swap exposure at 10% of fund assets, and in practice issuers reset swaps far more frequently, keeping exposure near zero. In a bank failure, the fund still owns its substitute basket — the realistic loss is a few days of index divergence, not the portfolio. It is a real risk, but a bounded and managed one, materially different from the popular image of a fund that could simply evaporate.

Here is the part factsheets undersell. When a physical Irish-domiciled fund receives a US dividend, it pays 15% withholding tax. A synthetic fund receiving the index return via swap can be paid the gross index return: under US rules (Section 871(m) carve-outs for deep, liquid indices), swaps on indices like the S&P 500 reference the gross dividend. That is worth roughly 0.2% per year at current yields — often more than the entire TER of the fund.

This is why the best-tracking S&P 500 UCITS ETFs are frequently synthetic, and why sophisticated investors who compare tracking differences rather than replication labels regularly end up holding them. The advantage is specific to US equity indices; for European or global indices where withholding leakage is smaller, physical funds usually track just as well.

The practical checklist: for US large-cap exposure, put synthetic funds on the shortlist and compare five-year tracking difference against the physical alternatives. For everything else, physical replication at scale is the default — simpler to reason about, no swap dependency, and competitive on cost. Either way, judge funds by delivered tracking, fund size and fee, not by the replication label alone.

Three lines settle the question. Replication method: "physical (full)", "physical (sampled)" or "synthetic (swap-based)" — stated plainly on every UCITS factsheet and KID. Securities lending: whether the fund lends, what share of revenue it keeps, and the collateral policy. Swap counterparties: for synthetic funds, which banks and how many — multi-counterparty structures diversify the (already small) exposure.

None of these should override the fundamentals of fund selection — index choice, cost, size, domicile. But when two funds tie on those, replication mechanics are the tiebreaker, and for US equities specifically, the synthetic structure's withholding advantage is one of the few free lunches left in index investing.