Physical vs synthetic ETFs: replication compared

Why this comparison matters

The physical-versus-synthetic question is usually framed as safe versus risky. That framing misreads both structures. The replication method shapes tracking precision, tax efficiency, and the location of counterparty risk, but it does not determine whether counterparty risk exists. Understanding the distinction clarifies why synthetic replication collapsed in Europe after 2008 yet remains the better-tracking option for specific exposures. The broader logic is traced in the Eco3min mapping of passive management and ETF market structure.

What physical ETFs are

A physical ETF buys and holds the securities of the index it tracks, either in full (full replication) or through a representative sample (optimized replication). Authorized participants create and redeem shares in kind, exchanging baskets of underlying securities for ETF shares, which keeps the market price close to net asset value. According to LSEG (Lipper, April 2026 review), physical replication held 87.02% of European ETF assets at end-December 2025, with full replication dominating equity ETFs. Many physical ETFs also lend out their holdings to generate revenue, a practice that introduces its own counterparty exposure.

→ Complete breakdown: How do ETFs differ mechanically from mutual funds?

What synthetic ETFs are

A synthetic ETF does not hold the index securities. Instead, it enters a total return swap with one or more counterparties, typically large banks, that agree to deliver the index’s return in exchange for the return on a collateral basket the fund holds. Under the UCITS directive, exposure to a single swap counterparty cannot exceed 10% of net asset value; the swap is marked to market daily and reset when exposure approaches that cap, and the collateral basket must meet liquidity and diversification rules. Synthetic replication is now concentrated in commodities, where physical storage is impractical, and in US equities held within European wrappers.

→ Related: Why do ETF prices diverge from NAV during stress?

The key differences

Mechanism. Physical ETFs own the constituents and rely on in-kind creation and redemption by authorized participants; synthetic ETFs own a collateral basket and rely on a swap contract to convert that basket’s return into the index’s return. The creation-redemption channel is the mechanical core of the physical model (what an authorized participant does).

Where counterparty risk sits. This is the distinction that matters, and it is widely misstated. Synthetic counterparty risk is explicit and capped: UCITS limits it to 10% of NAV, and US Federal Reserve research (2017) found that synthetic ETFs disclosing their baskets were over-collateralized by about 2% on average. Physical ETFs are not free of counterparty risk: an ECB analysis (2018, citing Morningstar) noted that roughly a quarter of ETFs had more than 50% of their portfolio out on loan at some point in 2011, exposing investors to the borrower. The difference is visibility and collateralization, not the presence or absence of risk.

Tracking and tax. The “riskier” structure often tracked indices at least as tightly. Vanguard documents that synthetic ETFs tend to show lower tracking error than physical peers, because the swap delivers the index return contractually. For US equities held in European UCITS funds, swaps are exempt from the 15-30% dividend withholding tax that physical Irish or Luxembourg funds incur, which is why one synthetic S&P 500 ETF reported outperforming the largest physical equivalent by roughly 0.43% over one year and 1.25% over five (ETF Stream, 2023).

How they behave across regimes

The structural choice is invisible in calm markets and decisive in counterparty-stress regimes. Before the 2008 crisis, European ETF assets were split roughly evenly between physical and synthetic replication; after Lehman, counterparty fear and a 2011 Financial Stability Board warning drove investors out of swap-based products, and synthetic’s share of European equity ETF assets fell from about 46% in 2010 to 17% by end-2020 (Morningstar). In acute liquidity stress the two structures fail differently: a physical ETF holding illiquid bonds can see its market price diverge from a stale NAV as the ETF prices in real time while the underlying does not trade, whereas a synthetic ETF’s pressure point is the solvency of its swap counterparty and the value of its collateral when that collateral is hardest to sell. The switch parameter is counterparty-solvency perception and funding stress, not inflation or growth. In the United States the question is largely settled: the SEC effectively halted synthetic ETF issuance in 2010, leaving physical replication at roughly 98% of the market. Read alongside: our mapping of ETF selection criteria.

Neither structure removes counterparty risk; each only decides where it sits, inside a swap or inside a securities-lending book.

The common confusion

The common assumption is that physical ETFs are safe because they hold real assets and synthetic ETFs are dangerous because they rely on derivatives. In 2011, at the height of the “replication war,” BlackRock’s chief executive captured the prevailing fear, describing a buyer of one synthetic product as effectively “an unsecured creditor” of the swap bank. The collateralization and daily-reset rules that govern UCITS synthetic ETFs, and the securities-lending exposure embedded in many physical ETFs, complicate that clean split. The structures carry different risks located in different places; the data does not support treating one as categorically safe.

Practical observation

Go deeper

Related guides

• Active vs passive investing: what the evidence documents
• Bitcoin ETF vs self-custody: structure compared

Frequently asked questions

How do physical and synthetic ETFs differ mechanically?

A physical ETF holds the index’s securities, in full or as a representative sample, and uses in-kind creation and redemption to stay near net asset value. A synthetic ETF holds a collateral basket and signs a total return swap with one or more counterparties that deliver the index return in exchange for the basket’s return. The first owns what it tracks; the second owns collateral and a contract. Tracking, tax treatment, and the location of counterparty risk all follow from that single structural choice.

Are physical ETFs free of counterparty risk?

No. Many physical ETFs lend out portions of their holdings to generate revenue, which exposes shareholders to the borrower if it fails and the collateral falls short. An ECB analysis (2018, citing Morningstar) found roughly a quarter of ETFs had more than half their portfolio on loan at some point in 2011. Synthetic counterparty risk, by contrast, is explicitly capped by UCITS at 10% of net asset value, marked to market daily, and reset when it approaches that limit. The honest comparison is not risk versus no risk, but disclosed-and-capped swap exposure versus often-less-visible securities-lending exposure. The companion comparisons continue this approach across the site.

Why has synthetic replication lost market share in Europe?

Before 2008, European ETF assets were split roughly evenly between the two structures. After Lehman, counterparty fear and a 2011 Financial Stability Board warning prompted large outflows from swap-based products, and providers converted many funds to physical replication. By end-2025, physical replication held about 87% of European ETF assets (LSEG/Lipper). Synthetic replication persists where it is structurally advantaged: commodities, where storage is impractical, and US equities in UCITS wrappers, where swaps avoid dividend withholding tax.

Last updated — 12 July 2026