Survivorship Bias in Equity Backtests

If you backtest a strategy on today's S&P 500, you are testing on the survivors — the companies that made it. The losers were quietly removed. Your backtest will look much better than it would have at the time, and the gap is called survivorship bias.

The Mechanism

Indices like the S&P 500 are not static. Companies enter when they grow, exit when they shrink, get acquired, or go bankrupt. Today's 500 is the result of decades of this filtering. If you take that list and backtest on it from 1990 forward, every company in your sample survived 35 years. By construction.

The ones that died — Lehman, Enron, Blockbuster, hundreds of dotcoms — are simply absent from your dataset. Their drawdowns, their bankruptcies, their failed turnarounds: none of it appears in your backtest.

The Magnitude

Empirical studies place the survivorship-bias overstatement at roughly 1–2% per year for naive equity backtests, with much larger effects in small-cap, emerging-market, and mutual-fund universes. Multiplied over a 10-year backtest, this is the difference between "great strategy" and "honest result."

Where It Hides in Practice

Index membership lists. "Backtest on S&P 500 from 2000" almost always means today's 500. The 1999 S&P 500 had different members.
Sector universes. "All US tech stocks" today excludes companies that failed and were delisted.
Mutual fund and hedge fund return databases. Funds that closed are often dropped retroactively, leaving only successful funds in the historical record.
"Top N by market cap" universes. Today's top 100 are not yesterday's top 100.

How to Reduce the Bias

Use point-in-time index data when available. Some data providers offer historical index membership lists; with these, you can scan the actual S&P 500 of June 2010 instead of today's 500.
Include delisted tickers. Use a data source that retains delisted equities. Their data is usually noisier but the bias correction is large.
Discount your backtest results. If you cannot fix the data, mentally subtract 1–2% annualized from your reported returns to get closer to the truth.
Test on a clearly-fixed universe. If you backtest on "AAPL, MSFT, GOOG, AMZN" specifically, you have hand-selected survivors and you know it. The result is a study of that basket, not a strategy.

What QuanterLab Does Today

QuanterLab's scanners typically use today's universe (e.g., today's S&P 500). This is honest about its limitation but does not correct it. When testing on broad universes, treat the historical performance as upper-bounded by the survivorship gap. Strategies that look marginal in this setting are very likely losing strategies on a real point-in-time universe; only strategies with substantial edge in survivor-only backtests are worth carrying forward to validation.

The Bottom Line

Survivorship bias is silent, structural, and free of moral judgement — it is a feature of how indices are maintained, not a flaw in any researcher. Knowing it exists is half the battle; mentally discounting your headline numbers is the other half.

Brown, S. J., Goetzmann, W. N., Ibbotson, R. G. & Ross, S. A. (1992). Survivorship Bias in Performance Studies. Review of Financial Studies, 5(4), 553–580.
Elton, E. J., Gruber, M. J. & Blake, C. R. (1996). Survivorship Bias and Mutual Fund Performance. Review of Financial Studies, 9(4), 1097–1120.

Textbook references

Campbell, J. Y., Lo, A. W. & MacKinlay, A. C. (1997). The Econometrics of Financial Markets. Princeton University Press.
López de Prado, M. (2018). Advances in Financial Machine Learning. Wiley.

Try it in QuanterLab

When you scan today's S&P 500, you are scanning the survivors. To estimate how much survivorship bias inflates your backtest, compare results on a static index list vs. a point-in-time list (when available).