If you bought today’s top ten funds by ten-year return, you would be buying a statistical illusion. Not because the numbers are falsified, but because those ten funds are the survivors of a much larger group whose worst members are no longer around to be counted. Survivorship bias is one of the quietest errors in personal finance: it doesn’t distort a single decision, it distorts the information behind every decision you make.
What survivorship bias actually is
Survivorship bias happens when we study a group and draw conclusions about it by looking only at the members that still exist, ignoring the ones that dropped out along the way. The classic example isn’t financial at all: during World War II, the US military examined planes returning from combat to decide where to add armor, and found that bullet holes clustered on the wings and tail. The intuitive conclusion — armor those spots — was wrong. Statistician Abraham Wald pointed out that the armor belonged where the damage wasn’t, because planes hit in those other areas were precisely the ones that never made it back.
In investing, the mechanism is identical. When a fund database shows “10-year category average return,” that average only includes funds that have existed for the full ten years. The ones that closed, merged into another fund at the same management company, or simply vanished due to poor performance are not part of the calculation. Their negative returns haven’t been erased from the world — they’ve been erased from the sample you happen to be looking at.
Why bad funds disappear from the statistics
Fund managers have no incentive to keep a fund alive once it has spent years losing money and losing investors. When a fund racks up mediocre returns, almost always one of two things happens: it gets merged into a better-performing fund from the same company — its track record absorbed and effectively erased from the system — or it gets liquidated outright once assets under management fall below the threshold that makes it profitable to run.
Studies of the fund industry in the US and Europe are consistent about the scale of this: somewhere between 6% and 8% of mutual funds disappear every year, whether through merger or closure. Projected over fifteen years, that means half or more of the funds that existed at the start of the period are no longer present at the end. And the ones that disappear are not a random sample: the overwhelming majority underperformed their category average. In other words, the “cleanup” process systematically removes the worst performers, leaving an artificially high average among those left standing.
S&P Global’s SPIVA report, which compares active management against benchmark indices, explicitly corrects for this because ignoring it would inflate the results of active management. When funds that disappeared are folded back into the calculation — the so-called survivorship-bias adjustment — the share of active funds that beat their index over long periods drops noticeably compared with the headline, unadjusted figures. The gap between looking only at survivors and looking at the full universe can amount to one or two percentage points of annual return, which compounded over twenty years is the difference between a comfortable retirement and a tight one.
The same mistake hiding inside stock indices
Survivorship bias isn’t limited to funds; it also shapes how we interpret the indices themselves. An index like the S&P 500 or the Ibex 35 isn’t a fixed snapshot of the same companies since its creation. Companies that go bankrupt, get acquired, or fall below capitalization or liquidity requirements are removed from the index and replaced by stronger ones. The index you look at today is, by construction, made up of companies that survived.
That has a real practical consequence: when someone argues that “stocks always deliver positive returns over the long run” using an index’s historical chart as proof, they’re using data that has already passed through a survivorship filter. The companies that went bankrupt along the way — and that at some point were part of the index — don’t drag down its historical return, because they were removed before reaching zero. The index, as a whole, continuously renews itself by discarding losers, which is not something a basket of individual stocks bought and held indefinitely without any criteria would do automatically.
This doesn’t invalidate index investing as a strategy — in fact it’s one of its structural advantages, since the index itself does the work of rotating toward viable companies — but it does invalidate the naive reading of “stocks always go up” as if it were a law of physics rather than the outcome of a continuous selection process.
How this bias reaches you without a label
Survivorship bias rarely announces itself by name. It arrives disguised as seemingly useful content:
“Best funds of the last 10 years” rankings. By definition, only funds that have existed for ten years can appear on the list. A fund that launched, performed poorly, and was liquidated after six years will never compete for that ranking, even though its return would have dragged down the category’s real average over that period.
Bank and platform comparison tables. When an institution shows you the historical returns of “the funds we offer,” it rarely includes the ones it stopped offering because they underperformed. The current lineup is already pre-filtered by the institution’s own survivorship process.
Success accounts on social media. The trader who has been posting profit screenshots for three years is visible precisely because things went well; the hundreds of people who tried the same strategy and lost money simply stopped posting, or never built an audience in the first place. You never see the denominator, only the numerator.
Startups and venture capital. When the average return of venture capital funds is quoted using only the ones that ended up reporting results, it systematically excludes the ones that failed without leaving a public trace — which are the majority.
In every case the mechanism is the same: visibility correlates with success, so any sample built from “what I can see today” over-represents the winners.
What to ask before trusting a historical return
You can’t eliminate survivorship bias, but you can neutralize much of its effect with a handful of questions before basing a decision on historical data.
Does this figure include funds that no longer exist? Serious data providers like Morningstar do offer survivorship-adjusted figures if you explicitly ask for them; the number that shows up by default in your broker’s app almost never does.
How long has the entire category existed, not just this particular fund? If a fund category has a twenty-year track record but the specific fund you’re being offered has only existed for five, ask what happened to its predecessors at the same firm.
Would this fund’s or stock’s past return be reproducible if it had to compete against the ones that are no longer around? It’s a counterfactual question, but it forces you to remember that today’s winner competed against a much larger field of contenders than the ones that now show up in any comparison.
Does the strategy depend on picking winners in advance, or on holding a broad, diversified basket? The more your outcome depends on correctly picking “future survivors,” the more exposed you are to survivorship bias in past data inflating your sense of how good your odds actually are.
The underlying lesson isn’t that historical data is useless — it’s that almost no source of historical data is neutral by default. Nearly all of it has passed, to some degree, through a selection process that favors whoever is still standing. Knowing that filter exists, and explicitly asking whether it has been corrected for, is the difference between comparing real returns and comparing the most flattering possible version of history.
A numerical exercise helps put the size of the effect in perspective. Imagine a hundred funds launching the same year in a given category. Fifteen years later, based on the industry’s observed attrition rate, it’s reasonable that only forty to fifty are still around. If you calculate the average return using only those survivors, you get a figure that entirely excludes half of the original group — precisely the half that performed worst. That’s not a technical footnote: it’s the difference between measuring the outcome of an investment decision made fifteen years ago and measuring the outcome of having guessed correctly, with the retrospective benefit of already knowing who would survive. The first question is the one that actually matters for your future money; the second is the one you’re almost always being answered without asking.
