Survivorship Bias in Investing Performance Claims

Alpha Sentinel is a software platform, not a registered investment adviser. This article is educational, not investment advice. Full disclaimer at the end of this post.

A brochure lands on your desk showing a group of funds that have beaten their benchmark every year for a decade, and the question a careful reader asks is a quiet one: what happened to the funds that are not on this page? Survivorship bias is the error of studying only the winners that remain while the losers have disappeared from view, and it makes performance claims look better than the full record would support. The questions below walk through how it works, where it hides and what to ask whenever you see a track record.

Key Takeaways

• Survivorship bias occurs when a performance record includes only the investments that survived, leaving out those that closed, merged or were removed.
• Because poor performers are more likely to disappear, the surviving group's average return looks better than the original group's.
• The bias shows up in fund comparisons, index histories, backtests and advisors' track records.
• A performance claim is only as meaningful as the starting group it was drawn from, so ask what the full group looked like.
• No claim of past performance describes what will happen next.

What is survivorship bias?

Survivorship bias is a form of selection error. When you look at a group of things that made it through some process, you see only the successes, and you draw conclusions about the process as though the failures never existed. In investing, the process is the passage of time, and the failures are funds that were shut, companies that were delisted and strategies that were abandoned.

The idea is often explained with a story from the Second World War, in which analysts studying damaged aircraft that returned from missions recommended reinforcing the areas with the most bullet holes. The correction was that those planes had survived being hit in those places; the planes hit elsewhere had not come back to be counted. The missing data held the real answer.

In finance, the missing data is quiet. A fund that performs poorly is often closed or merged into another fund, and when it disappears, its record can disappear with it. What remains is a group that looks, on average, as though it did well, because the ones that did not were removed.

How does it inflate fund performance claims?

Imagine a fund company that starts the year with twenty funds. Over several years, the weakest few are closed or folded into stronger ones. When the company later advertises the ten-year record of its funds, the figures describe only those that still exist. The average return of the survivors is higher than the average return of the twenty that began.

The same thing happens in data services that compare funds. If a database includes only the funds currently operating, any average it reports leaves out every fund that closed along the way. Researchers who study this have found that the effect can add a meaningful amount to reported average returns, and the size depends on the period and the type of fund. The general direction is not in doubt: the bias flatters.

The longer the period, the larger the effect, because more funds have had time to drop out. A claim that a category has beaten a benchmark over twenty years may sound impressive, but if the category's casualties are not in the sample, it says less than it appears to. The same caution applies to rankings of the best funds of the past decade, which by construction include only funds that lasted.

Where else does survivorship bias hide?

It appears wherever a record is built from what remains. Index histories are one example. A stock index is periodically updated, with struggling companies removed and successful ones added, so a calculation of how today's index members performed in the past, using only the companies that are in the index today, overstates what an investor who held the index at the time actually experienced.

Backtests are another. Someone testing a trading idea on a list of today's well-known companies has, without meaning to, selected companies that did well enough to remain well known. Companies that failed, were acquired at a low price or dropped out of the market are absent, so the idea looks better than it would have in real time.

Track records of individual managers and advisors raise the same issue. Those who performed poorly leave the industry, change jobs or stop reporting, and the people who remain are, in part, the ones who were lucky or skilled enough to stay. Even personal anecdotes follow the pattern: you hear from the friend whose investment worked, and rarely from the one whose investment disappeared.

The bias can also affect your own memory. People recall the holdings they still own more readily than the ones they sold at a loss and forgot, which makes their own history look better than it was.

What should you ask when you see a performance claim?

A few questions address most of the problem:

• What was the starting group, and how many members are no longer in the data?
• Does the record include funds or strategies that closed, merged or were removed?
• Is the period chosen to start at a favorable point, and what do other periods show?
• Were the results measured after fees and costs, and how were they calculated?
• Would the claim still look as strong if the missing members were included?

If the answer to the first two is that no one knows, treat the claim cautiously. Credible reports describe their methodology and acknowledge what is excluded. Be wary of results that give no way to learn what was left out.

Remember also that a record with no survivorship bias still describes the past. Strong figures do not carry forward automatically, and funds that top one period often lag in the next. Questions about how a specific product fits your finances belong with a qualified professional, and working with advisors and self-directing are both legitimate approaches.

What does this look like in practice? An illustrative example

The following is a hypothetical example, not a result or a recommendation. An attorney receives a summary describing a family of ten funds whose average ten-year return exceeds a broad benchmark.

She asks one question: how many funds did the family offer ten years ago? The answer, after some digging, is fifteen. Five were closed or merged along the way. She writes down the results she can find for those five, which are lower than the survivors', and recomputes the average across all fifteen. The advantage over the benchmark shrinks considerably and in this example nearly disappears.

She does not conclude that the family is poor or that the claim was dishonest. She concludes that the original number answered a narrower question than it seemed to. She files the comparison in her notes and adds to her checklist the first question she will ask of every track record.

Frequently Asked Questions

What is survivorship bias in simple terms?

Survivorship bias is drawing conclusions from the winners that remain while ignoring the losers that disappeared. In investing, it makes groups of funds or strategies look better than they were.

How much does survivorship bias affect fund returns?

The effect varies by period and fund category, but it consistently raises the reported average return of a group, because poorer performers are more likely to close. The longer the period, the larger the potential distortion.

Can survivorship bias affect backtests?

Yes, if a backtest uses only companies that exist today, it leaves out those that failed or were removed, which flatters the strategy. Credible tests use data that includes companies as they were at each point in time.

How can I avoid survivorship bias when evaluating performance?

Ask what the starting group was and what happened to the members that are no longer there. A record that includes the closed and merged funds is a more complete picture of what an investor in the whole group experienced.

How The Shield works

On weekdays from 4:00 AM to 8:00 PM ET, The Shield checks the stocks on your watchlist every few minutes, using prices delayed about 15 minutes. When one of them moves into downward momentum below its moving average, it emails you what changed and the price. Alerts are informational, not recommendations, and may be delayed or missed.

Looking at what is missing is part of owning your decisions, and the decisions are yours. The 7-day pass for The Shield from Alpha Sentinel is the single door in if you want a steadier routine around your holdings. We watch. We alert. You decide.

The ShieldYour watchlist, watched during market hoursStart your 7-day pass$29 today. Your subscription renews automatically at $249/month after the 7-day trial unless canceled before renewal. Cancel in account settings.

FULL DISCLAIMER: Alpha Sentinel is a software platform that provides informational tools, including market scans and position-monitoring outputs. Alpha Sentinel is not a registered investment adviser, broker-dealer, certified public accountant, financial planner, or law firm. Nothing on this page or produced by the platform constitutes investment, tax, legal, or accounting advice; an offer or solicitation to buy or sell any security; or a personalized recommendation of any kind. The platform is not a substitute for the services of qualified professionals, and nothing in this article should be read as a recommendation that any reader manage their portfolio without professional guidance. All investments involve risk, including the risk of total loss of principal. Tax treatment depends on the user's individual circumstances and applicable law; consult a qualified tax professional regarding any tax-related question. The output of any software filtering or monitoring rule is not a prediction of future market behavior, and no software tool can be relied upon to warn users in advance of losses or to prevent losses. Past performance, whether of any security, strategy, or software signal, does not guarantee future results. Alpha Sentinel does not publish or warrant any hit rate, win rate, or similar performance metric, and users should not infer any such metric from the platform's outputs.

Frameworks, examples, schedules, and references to user types in this article are illustrative only and are not representations of typical results. Notification delivery times and service availability depend on third-party systems and may be subject to delays, errors, or failures.