Reconstructed equity ETFs

When an equity ETF's real quotes don't cover the whole analysis period, its history is extended backwards. The result is an estimated track for past scenarios, not an official market price, and it is labelled as such everywhere it appears.

Five methods, depending on the product

Factor replication

For factor ETFs (value, momentum, quality, small cap, UCITS country indices), the product's historical profile is estimated against market and styles, and the missing history is generated by combining factors and profile.

The factors used are those of the five-factor model plus market:

Factors, denominated in USD, are converted to euro. They serve only to extend history by a few years before real quotes begin. The ETF's TER is applied to the reconstructed past, so product cost is simulated there too.

Industry portfolios (US sectors)

For US sector ETFs and indices (Health Care, Industrials, Nasdaq 100) the approach is different: it uses Ken French's 49 daily industry portfolios, with no factor regression. The sector portfolio (value-weighted or equal-weighted) is selected and spliced to the real UCITS ETF when quotes begin.

Same principle Testfolio applies to SPDRs, on a different source.

Research series on the Italian market

For products tied to the Italian market, missing history connects to the reconstructed series from Professor Coletti's research.

Geographic composition

For ETFs and indices covering geographic areas (Europe, Pacific, emerging, world) history is reconstructed by combining country indices.

Reconstructed country indices

For local or niche markets, or where factor regression proves unreliable, the extension uses country equity indices from a database of reconstructed indices.

Why this isn't a one-off trick

Factor-based reconstructions are the same approach used by widely adopted backtesting platforms and by the academic literature (Fama-French, AQR, Vanguard Research). On broad, liquid indices, monthly deviations from the official benchmark are often on the order of a few basis points.

For equities, Fama-French factors plus momentum typically explain over 98% of the monthly variance of a diversified ETF: on the S&P 500, MSCI World and similar, the reconstructed track sits very close to the official benchmark.

Limits worth knowing

Reconstruction quality isn't uniform, and it's useful to know where it degrades:

• The more concentrated the product, the less reliable the replication. An ETF on a single niche sector or a small market carries idiosyncratic variance no factor model captures.
• The track reproduces the strategy's behaviour, not its real frictions: bid-ask spreads, limited capacity, trading suspensions.
• The TER applied is today's, while product costs tended to be higher in the past. That's an optimistic assumption over the reconstructed period.

None of these invalidates the exercise: they would make the backtest worse, not better, and knowing the direction of that bias is useful when reading results.

The full catalogue of reconstructed products, with start dates, is in Reconstructed product catalogue.