Physical vs synthetic replication: what really changes, and when each one wins

Every ETF factsheet has a line most people skip: replication method. It says "physical" or "synthetic", and the usual reaction is to assume synthetic is worse. It is not that simple — on some indices synthetics systematically do better, for one very specific reason.

Physical replication: buying the shares

The fund actually buys the index constituents. Two variants:

• Full. It buys all 500 S&P 500 companies in their exact proportions. Workable when the index is manageable and liquid.
• Optimised or sampled. It buys a representative selection instead of everything. Standard for indices with thousands of constituents or illiquid names — an MSCI World has ~1,400 positions, and a bond index can have tens of thousands.

In this catalog, 220 of the 226 products replicate physically. It is the default.

Synthetic replication: a swap

The fund does not buy the index. It buys a basket of arbitrary securities (the "substitute basket") and signs a swap with a bank: the fund hands over its basket's return and the bank hands back the index's. What you get is the index return, generated by contract rather than by ownership.

Why they exist: US dividends

Here is the part almost no article explains. An Irish fund that owns US shares bears a 15% withholding tax on the dividends it collects. That is the best deal available, but it is still a leak.

A swap on a US index can be structured so that this withholding does not apply. The bank delivers the index return in its gross version. On an S&P 500 dividend yield of roughly 1.3%, the 15% you stop losing is worth about 0.20 percentage points a year — more than many funds' entire TER.

That is why on the S&P 500 and Nasdaq-100 synthetics tend to beat physical ones despite similar fees. It is not magic: it is tax.

The real synthetic risk, and how capped it is

If the counterparty bank fails without having paid what it owes, the fund has a hole. That is true. But UCITS rules cap it tightly:

• Net exposure to a single counterparty cannot exceed 10% of the fund's assets.
• In practice providers go much further: they collateralise the swap daily, often over-collateralising it, and spread across several counterparties.
• The collateral sits in segregated custody, like the rest of the portfolio.

The risk is real, but it is a capped and managed risk, not an open bet.

The physical small print: securities lending

And here is the nuance that evens things out. Many physical ETFs lend their shares to third parties (usually for short selling) in exchange for a fee. That fee improves the fund's return — sometimes enough to offset part of the TER.

But it means a "physical" ETF can have part of its portfolio out on loan at any moment, with its own collateral and its own counterparty risk. In other words: the line between "physical, safe" and "synthetic, risky" is far blurrier than the label suggests.

Every provider publishes its policy: the maximum share it lends, how it is collateralised, and what cut of the income the provider keeps. It is in the prospectus, and it is worth a look.

When each one wins

Situation	Usually wins	Why
S&P 500, Nasdaq-100	Synthetic	Avoids US dividend withholding
Very broad or illiquid indices	Synthetic	Physical replication is costly and imprecise
Frontier markets and commodities	Synthetic	Physical access is hard or impossible
Large European or global indices	Physical	No offsetting tax advantage, and more transparent
If counterparty risk bothers you	Physical	You own the securities, with their own small print

In short

• Physical buys the securities; synthetic delivers the index return by contract.
• Synthetics win on US indices because they avoid the 15% dividend withholding: roughly 0.20 points a year.
• Counterparty risk is capped at 10% by UCITS, and in practice far below that.
• Physical funds lend securities too, so the label is not a seal of absolute safety.

Educational content, not investment advice. Check each fund's prospectus for its specific replication, collateral and securities-lending policy.