Survivorship bias means judging a group by its survivors only. Failures stay missing, invisible, or dropped from the data. The result looks better than the whole group, and the cause of success looks clearer than it is.
This article explains the WWII plane story behind the bias. It shows where you meet it in finance, study advice, and research. It gives three checks to see the missing failures.
- Survivorship bias is a selection problem: end cases are not a random subset of start cases.
- The plane fix: protect undamaged areas on survivors, because missing planes carry fatal-hit data.
- In finance, closed funds vanish from tables, which overstates returns.
- In research, positive results publish more, dropouts answer less, and advice comes from stayers.
- Three checks help: name the filter, study failures, and vet the source.
What is survivorship bias
Survivorship bias happens when a filter removes some cases, and you judge from the cases that remain. You see winners, survivors, repeat responders, and published results. You miss losers, missing planes, closed funds, dropouts, and unpublished studies.
The bias is a missing-data error in your own words. A filter lets some cases through. You study only those cases. You treat them as the whole group. Britannica defines it as a logical error from focus on survivors of a filter.
Place it next to its neighbours, because students mix them. The table shows the difference in one line each.
| Bias | What is missing or distorted | Simple test |
|---|---|---|
| Survivorship bias | Non-survivors after a filter; data are gone | Ask who did not survive the filter |
| Selection bias | Sample was never random at the start | Ask how the sample was drawn |
| Availability heuristic | Memory favors vivid cases you recall fast | Ask if vivid cases match the base rate |
Survivorship bias is a kind of selection and sampling bias. Availability describes recall. Survivorship describes missing data. For recall effects, see the availability heuristic. For ten biases told apart, see ten biases told apart.
The plane story Wald actually worked on
During World War II, analysts studied damage on aircraft that returned. Many returning planes showed hits on wings, tail, and center body. The first plan was to armor the most-hit areas.
Abraham Wald worked on this problem in the Statistical Research Group at Columbia University. He wrote eight memoranda in 1943 on estimating vulnerability from survivors. The Center for Naval Analyses reprinted them in 1980 as still useful.
Wald started from a simple gap. Teams knew planes sent, planes returned, and hits on each returned plane. They did not know hits on missing planes. An exposition states the gap as the basic difficulty. The commander lacks the distribution for planes that did not return.
His answer used careful limits. He assumed hits spread in a uniform way, and an unhit plane returns. From survivors, he estimated loss chance after each hit. He then inferred what missing planes likely showed.
The practical point was direct. Hits on weak parts show less on survivors, because planes hit there return less. So protect areas with little damage on survivors. A mathematics summary gives the same point in plain form.
Treat internet dialogue with care. Many retellings invent short quotes about motors. The reliable record is the memoranda plus summaries by Wallis, Mangel, and Samaniego. The moral needs no invented lines: study the missing planes too.
Where you meet it
You meet the same structure far from airfields. A filter removes cases. Reports use survivors only.
In finance, closed funds vanish from most tables. A fund that fails often merges into another fund in the same family. Its poor record drops from standard sources. Studies of survivors only then overstate returns.
A classic fund study states the cause in one line. Funds that vanish tend to do so from poor performance. The authors tracked all funds from the start and kept merger terms. See the abstract of the fund study.
In study and business advice, you hear survivors only. Profiles list dropouts who built large firms. They skip the larger group who left and did not build one. Feeds show top scorers with one method.
They skip scorers with the same method and low marks. Testimonials come from stayers who liked a habit. Quitters rarely post. For the related error of blaming character and missing luck, see the fundamental attribution error.
In research, three filters matter.
First, publication favors positive results. Journals print striking positives more than nulls. A modeling essay warned that many claimed findings may be false. Small studies, small effects, many teams, and flex in design raise the risk. Read a success as selection for positivity, not proof alone.
Second, surveys lose the people they need most. In the United States COPE study, one-survey responders had more symptoms than four-survey responders. Adjusted ratios were 1.30 for anxiety and 1.43 for depression. New symptoms raised odds of missing later waves, 1.68 for anxiety and 1.56 for depression.
The authors warn that repeat-only trends look too optimistic. Screens do not equal diagnosis. If low mood lasts, talk to a doctor or a mental health professional.
Third, career advice comes from stayers. A Nature column calls this flaw by name. Stayers remain visible. Leavers fade. The path then looks more repeatable than it is.
Use the same lens for study tips. Top-scorer tips form a survivor sample. They miss users of the same plan with low scores. Save survivorship bias plus neighbours to one list in Psychology World on iPhone and iPad. Drill that list with Flashcard, Question and Typing until each term stays. A list with failures next to successes keeps the base rate in view when you revise.
Why the mind misses the failures
Successes stay visible. Failures fade. Media repeats winners. Survivors post and teach. What you recall fast feels common. That recall effect is the availability heuristic.
Confirmation adds force. With a plan in mind, you notice fitting stories. You skip breaking stories. For this pattern, see confirmation bias.
Cause and correlation blur adds a third push. Survivors often share a trait. Early rising, dropping out, or one routine look causal. You never see non-survivors with the same trait. Treat shared traits as clues until you see both groups.
Risk tilt helps the story stick. Costs draw more mind share than equal gains. That tilt alone is not the bias. It makes vivid winners outweigh quiet base rates.
Three checks to see the missing failures
Use these checks for a success story, a fund table, a survey trend, or a study tip.
- 1.Ask what did not survive.
Name the filter. Who dropped out, merged, stopped answering, or stayed unpublished. If the source hides this, mark the claim as incomplete.
- 1.Study failures alongside successes.
Find starters with the same trait who failed. Compare both groups. If failures shared the trait, the trait did not decide it. If you find no failures, lower confidence.
- 1.Vet the data source before you act.
Check start sample, span, and end sample. Check loss reasons. For funds, check full start tracking with merger terms. For surveys, check follow-up rates and repeat-only use.
For advice, check leavers plus stayers. For test quality, see reliability and validity. For setting effects, see demand characteristics.
If a check fails, hedge. State that data cover survivors only. Note the true rate may sit lower. Seek a source that tracks the full start group.
Frequently asked questions
What is survivorship bias in one sentence
It is judging a group by cases that passed a filter, while missing cases bias you toward over-optimism and false causes.
How is survivorship bias different from selection bias
Survivorship bias loses cases after a filter, while selection bias starts with a sample that was never random.
What is the WWII plane example
Analysts saw hits on returning planes and planned to armor hit areas, but Wald showed missing planes matter more and pointed to undamaged areas as likely fatal zones.
Does survivorship bias affect psychology research
Yes. Positive results publish more, symptomatic dropouts answer less, and stayers give most advice, so survivor-only views can flatter trends and tips.
How can I avoid it when studying
Name the filter, find same-start failures, compare both groups, and check start sample, dropouts, and follow-up before you copy a method.
Sources
- 1.Encyclopedia Britannica. Survivorship bias.
- 2.Center for Naval Analyses (1980). A reprint of A method of estimating plane vulnerability based on damage of survivors.
- 3.Mangel, M. and Samaniego, F. (1984). Abraham Wald's work on aircraft survivability.
- 4.American Mathematical Society. The legend of Abraham Wald.
- 5.Elton, E. Gruber, M. and Blake, C. (1996). Survivor bias and mutual fund performance.
- 6.Oxford Academic (1996). Survivor bias and mutual fund performance, abstract.
- 7.Czeisler, M. et al. (2021). Uncovering survivorship bias in longitudinal mental health surveys during the COVID-19 pandemic.
- 8.Hemprich-Bennett, D. et al. (2021). Beware survivorship bias in advice on science careers.
- 9.Ioannidis, J. (2005). Why most published research findings are false.
- 10.Wikipedia. Survivorship bias.
- 11.The Decision Lab. Survivorship bias.