How we extend ETF history before inception

The problem

TLT launched in 2002. GLD launched in 2004. QQQ launched in 1999. If you measure their drawdowns only from their ETF inception, you miss the 1973-74 crash (SPY's worst bear market), the 1980-82 bond disaster (TLT's worst-ever period), and the entire history of gold before its secular bull run. A 2008-only view of risk is misleading because 2008 isn't the worst any of these asset classes ever did.

To give honest drawdown numbers, we need the underlying asset-class return series, not the ETF ticker's price history. The asset classes existed long before the ETFs that track them today.

The solution: proxy chain

For every ETF, we define an ordered fallback chain. We take the real ETF's adjusted-close series where it exists, then splice in the pre-inception history from the closest-tracked predecessor — usually an institutional index mutual fund that held the same assets under the same mandate. Returns, not levels, are what we stitch; the level series starts at the ETF's actual inception price and backs into the predecessor's returns for earlier dates.

Source: backend/app/backtest/fallback_table.py. Every backtest and drawdown run records its per-ticker chain in the build log, so you can always see which source tickers contributed to any period.

Per-ticker chains

Why you can trust this

• Real institutional data. VFINX, VUSTX, VFITX, PIGIX, FEMKX etc. are index mutual funds with decades of audited NAV history. They are not synthetic.
• Same mandate. VFINX tracked the S&P 500 via full replication from day one in 1980 — the same exposure SPY does today. VUSTX tracked long Treasuries. The asset-class return series is the return series, wrapper aside.
• Transparent. Every drawdown result carries its proxy chain in the build log. If a result seems off, you can verify which source data generated it.
• We disclose scaling. A couple of chains (VWO pre-2001, TIP pre-2000) apply a small volatility scaler to account for the emerging-markets vs developed-markets cap distribution, or nominal vs inflation-protected Treasuries. The scaler is in the fallback table and in the build log.

Known caveats

• Pre-ETF returns carry the predecessor fund's expense ratio, not the modern ETF's. For VFINX → SPY that's ~16bps vs ~9bps — small but nonzero.
• Intraday volatility is not comparable before ETFs existed. Mutual funds priced once per day; our series is daily only.
• Gold before 1975 was not freely tradable for US investors. We still show the series because the asset existed and its price history is well-documented, but treat pre-1975 drawdowns as reference context, not backtestable trades.
• The chain is ETF-level, not strategy-level. If a strategy uses a leveraged ETF (e.g. TQQQ, SSO), we synthesize leverage from the underlying index return minus a financing cost — see the backtest engine docs for that path.