Survivorship bias in investing, with examples

Quantooo Lab ยท 2026-10-22

During the Second World War, analysts studied bullet holes on returning bombers and proposed armouring the most damaged areas. The statistician Abraham Wald pointed out the flaw: the planes hit elsewhere never came back. The holes showed where a plane could be hit and survive. Investing data has exactly the same blind spot.

What is survivorship bias in investing?

Survivorship bias is the distortion you get when you study only the things that are still around. In investing, it usually means testing a strategy on today's list of companies or funds, which excludes the ones that went bankrupt, were taken over, or shrank out of the index. The test then looks better than it could have in real time.

How does survivorship bias inflate a backtest?

Take today's S&P 100 and test a strategy on these stocks back to 2015. Every company in the list is, by definition, one that grew large enough to be there today. Some were small or struggling in 2015. Your backtest is quietly holding tomorrow's winners. Even a simple "buy everything, equal weight" portfolio looks better than any investor could have achieved, because the losers are missing from the start.

Can you give a real example from this site?

Two, in fact.

  • Walgreens (WBA) was in our US universe and was later delisted. We kept its filing history so that past scores still include it, rather than letting it vanish from the history.
  • India's Nifty 50. Our simulated India portfolios use today's 50 members going back to 2020. Several companies joined the index because they grew. That flatters every simulated India result, including the equal-weight benchmark, and we say so on the portfolios page.

How do you avoid survivorship bias?

Use point-in-time membership: for each past date, test only the companies that were actually in the index on that date, including ones that later left. This needs historical constituent lists and data for delisted companies, which are often paid datasets. When they are not available, the next best thing is to say clearly that the bias exists and in which direction it pushes the results.

Does survivorship bias affect fund performance data too?

Yes, heavily. Funds that perform badly are often closed or merged into others, and their records disappear from many databases. Average fund returns computed only from surviving funds overstate what investors really earned. Studies of mutual fund databases have found this effect to be meaningful, especially over long periods.

What should you look for in any backtest?

Ask whether the stock list was fixed at the start of each period or taken from today. If it was taken from today, treat the results as an upper bound. Our live portfolios avoid the problem entirely because they only trade on scores published in advance: there is no way to pick tomorrow's survivors when tomorrow has not happened yet.

Open questions for students and researchers

Questions without a single right answer. Pick one and share your reasoning in the discussion below.

  1. Estimate how much survivorship bias inflates an equal-weight backtest of today's Nifty 50 members since 2020. What data would you need for a precise answer?
  2. Is survivorship bias always upward? Construct a strategy whose backtest would be biased downward by using today's index members.
  3. How should a researcher handle a company that was acquired at a premium: as a success, a failure, or neither?

Debate: three perspectives

An illustrative debate written by the author to lay out the main viewpoints. The three voices are roles, not real people. Your view is welcome in the discussion below.

The Sceptic: Any backtest built on today's index list is fiction. Without point-in-time membership, do not publish it.

The Practitioner: Point-in-time membership is often paywalled. Flagging the bias clearly and estimating its direction is better than not testing at all.

The Academic: Survivorship is one form of selection bias, which runs through all of empirical finance, from fund databases to published anomalies. The general question is always: what process decided which observations I can see?

Which view is closest to yours, and what evidence would change your mind?