Survivorship Bias: Why the Average Fund Looks Better Than It Actually Performed
Pull up a list of mutual funds with 15-year track records and the average return usually looks impressive. What the list does not show is every fund that closed, merged, or quietly vanished along the way, almost always because it performed poorly, and that missing data systematically flatters everything you can actually see.
The core principle
Survivorship bias occurs when a dataset includes only the entities that made it to the end of the measurement period, silently excluding the ones that failed, closed, or were merged away along the route. In fund investing, this happens constantly: poorly performing funds are routinely closed or absorbed into a better-performing sibling fund within the same fund family, a decision the fund company controls, and this practice erases the weak track record from most public databases entirely rather than merely flagging it as closed.
The mechanism is worth walking through concretely. A fund family launches ten small-cap funds in the same year, each run by a different manager with a somewhat different style. Over the following decade, three perform well, four are mediocre, and three lose enough money and shareholder assets that the company quietly closes them or merges their remaining assets into one of the stronger surviving funds. Ten years later, a fund screener searching for small-cap funds with a full ten-year history returns only seven funds, because the three closed funds have no ten-year track record left to display. The average return across those seven surviving funds looks solid, but it excludes the three failures completely, not just from casual notice, from the underlying data itself. The true average performance across all ten original funds the family actually launched was meaningfully worse than what any screener today will show.
How the math works
Two worked examples show how large the distortion can be.
Example 1: a fund family's advertised average. A fund family runs 20 actively managed equity funds launched together 15 years ago. By today, 6 have been closed or merged, each having underperformed its benchmark by an average of 4 percentage points per year before being shut down. The 14 surviving funds have averaged an annualized return of 8.5% over the 15 years. If the 6 closed funds are included at their actual (worse) performance of roughly 4.5% annualized before closure, the honest asset-weighted average across all 20 original funds works out closer to ((14 x 8.5) + (6 x 4.5)) / 20 = (119 + 27) / 20 = 7.3%. The visible, survivors-only average of 8.5% overstates the family's true long-run performance by roughly 1.2 percentage points per year, compounding into a materially different picture over 15 years: $10,000 at 8.5% becomes about $33,700, while $10,000 at 7.3% becomes about $28,700, a gap of $5,000 that exists purely because of which funds got counted.
Example 2: industry-wide fund closures. Industry research on US mutual funds has found that somewhere around 5% to 10% of funds close or merge in any given year, a rate that compounds sharply over long horizons. If a category starts with 100 funds and 7% close annually (a simplified constant rate for illustration), the number of original funds still reporting a full track record after 15 years is roughly 100 x (0.93)^15 ≈ 34 funds, meaning a database built from currently reporting funds with 15-year histories is missing the data from roughly two out of every three funds that originally existed in that category, virtually all of them closed for underperforming.
How it shows up in real portfolios
Retail investors most commonly encounter survivorship bias when choosing among actively managed funds using a screener or a "top funds" list, unaware that the very act of screening for funds with a long, unbroken history already excludes every fund that failed to survive that long. A fund family's flagship fund with an exceptional 20-year record often looks like proof of manager skill, when in reality it may simply be the one surviving fund out of a dozen the family launched around the same time, the others closed and forgotten.
The same bias distorts backtested trading strategies and stock-picking research, a subject covered in more depth under backtesting: a researcher who tries hundreds of strategy variations on historical data and publishes only the ones that "worked" is exhibiting a close cousin of survivorship bias, since the strategies that failed in testing are simply discarded and never discussed, leaving an observer to believe the published strategy reflects skill rather than one lucky draw among many attempts.
Stock market indexes themselves carry a milder version of the same effect. Broad indexes periodically remove companies that go bankrupt, get acquired, or fall below inclusion thresholds, and add new, often successful, replacements, which means the long-run historical return of "the market" as commonly reported is not quite the return of buying and holding a fixed set of companies from decades ago; it reflects a continuously curated, self-cleaning list, a nuance that matters for anyone extrapolating very long-run historical index returns into a precise forecast.
A related but distinct pattern shows up among individual investors evaluating their own stock-picking history. It is easy to remember the handful of winning picks that are still held and doing well, since they remain visible in the current portfolio, while forgetting the losing picks that were sold years ago and quietly stopped being tracked mentally. This personal version of survivorship bias tends to inflate an investor's own sense of their stock-picking skill over time, which is one reason keeping a written, dated log of every trade, wins and losses alike, is a more honest way to evaluate personal investing performance than relying on memory of a portfolio's current holdings.
Survivorship bias also distorts how "successful investor" stories get told and consumed. A magazine profile of a self-taught trader who turned a modest account into a fortune is, almost by definition, a survivor: the profile exists because that particular trader succeeded, while the far larger group who tried a similar approach and lost money generates no comparable article, since a failed trading story rarely gets written up as inspiration. Reading widely circulated success stories as representative of what a given strategy typically produces, rather than as one visible outcome plucked from a much larger and mostly unseen population of attempts, is one of the more subtle ways this bias shapes ordinary investing decisions rather than just fund selection.
Actionable breakdown
- Ask whether a track record includes closed and merged funds
- Reputable data providers disclose survivorship-bias-adjusted returns
- A "top funds of the decade" list rarely does this
- Treat long, clean track records with mild skepticism
- A fund family's best surviving fund is not a random sample
- Compare it to a broad, unmanaged benchmark instead
- Apply the same lens to backtested strategies
- Strategies that failed testing are rarely published or discussed
- What you see is a curated set of survivors, not a fair sample
- Prefer index funds where the bias mostly disappears
- Index funds do not depend on picking the surviving manager
- The benchmark's own construction rules handle turnover transparently
Common pitfalls
- Chasing a fund family's flagship fund because its long-term record looks exceptional, without asking how many sibling funds from that same family were closed along the way and are no longer visible in any screener.
- Applying survivorship logic only to funds and forgetting it also distorts stock index history, since indexes periodically drop failed companies and add new winners, quietly improving the apparent long-run record of the market as a whole.
- Assuming a strategy that worked in a backtest reflects a fair, unbiased test, when the strategy itself may have been shaped by a researcher who kept adjusting parameters until finding one that happened to survive the available historical data.
- Overweighting recent "best fund" lists, which by construction only include funds that have not yet been closed, a status that says more about past performance than future results.
Related concepts
For the related trap in strategy research, see backtesting and reversion to the mean. For why picking a winning active fund is so difficult even without this bias, see active management and benchmark. Our behavioral finance guide covers this alongside other data traps that distort investor decision-making.
The bottom line
Any performance record that only shows the winners is telling you an incomplete story, so always ask what got excluded, and why, before trusting the average.