What self-selection bias means
Wikipedia's definition (checked 20 October 2026) says self-selection bias arises in any situation in which individuals select themselves into a group, causing a biased sample with nonprobability sampling. The problem is not that people choose. It is that the traits that lead them to choose can also shape the result being measured.
The Wikipedia entry (checked 20 October 2026) adds that a poll suffering from this bias is sometimes called a self-selected listener opinion poll, or SLOP, and that the problem is a major one in sociology, psychology, economics and other social sciences. It separates self-selection from other selection problems by cause: self-selection may involve purposeful intent by respondents, while other forms of selection bias may arise more inadvertently, possibly as the result of mistakes by those designing a study.
The Catalogue of Bias, a collaborative project that documents biases in health research, covers the same ground under the name volunteer bias. Its definition is short: participants volunteering to take part in a study intrinsically have different characteristics from the general population of interest.
How it fits with related biases
Several overlapping terms describe the same family of problems. The table below lists the definitions as each source gives them.
| Term | Short definition | Source |
|---|---|---|
| Selection bias | Study participants differ systematically from the population of interest, producing a systematic error | Catalogue of Bias |
| Self-selection bias | Individuals select themselves into a group, producing a biased, nonprobability sample | Wikipedia |
| Volunteer bias | Volunteers for a study have intrinsically different characteristics from the general population | Catalogue of Bias |
| Nonresponse bias | The people who respond have different answers from those who do not | Wikipedia (as a close relative) |
The Catalogue of Bias defines selection bias more broadly and explains why it matters for cause and effect: when participants differ in ways other than the treatment or exposure being studied, the results are biased by confounding. Our guide to confounding variables covers that mechanism. Self-selection bias is also closely related to nonresponse bias, which is about who answers rather than who signs up.
Five examples from research
Each case below comes from the cited source. The first is a teaching example; the rest are from published health research summarized by the Catalogue of Bias.
| Case | Who selected themselves | What went wrong | Source |
|---|---|---|---|
| Test-prep course | Students who chose to take the course | Higher scores may reflect motivation, socioeconomic status or prior test experience, not the course | Wikipedia |
| Hormone replacement therapy and heart disease | Women who chose HRT | Observational studies suggested HRT reduced heart disease; later randomized trials suggested it might raise the risk | Catalogue of Bias |
| Probiotic trial for childhood atopy | Mothers who stayed in the trial | Representation of the most deprived fell as the trial went on | Catalogue of Bias |
| Sexuality research | People who volunteered | Volunteers reported more positive attitudes, less guilt and more experience than non-volunteers | Catalogue of Bias |
| Lifestyle cohort studies | People who signed up | Participants showed lower mortality than the general population, a healthy participant effect | Catalogue of Bias |
The test-prep case is a standard teaching illustration: students who choose to take a course may differ from those who do not in motivation, socioeconomic status or prior test-taking experience, so a gap in scores could appear even if the course did nothing. Those who chose the course might have scored higher anyway.
The hormone replacement therapy (HRT) case shows what is at stake. According to the Catalogue of Bias, several observational studies found HRT reduced coronary heart disease, but later randomized controlled trials found it might increase the risk. The women in the observational studies who were on HRT were more health conscious, more physically active and of higher socioeconomic status than those who were not. The Catalogue calls this self-selection a cause of confounding and a healthy-user bias.
Who tends to volunteer
The Catalogue of Bias summarizes what is known about volunteers: they tend to be more educated, come from a higher social class and be more approval motivated. It cites one report finding that women are more likely to volunteer than men and are healthier and adhere to treatment more often.
It also notes that differences are not limited to demographics. Volunteers can hold different attitudes toward the trial and the institutions running it, and people may be less likely to put themselves forward for studies of conditions or behaviors regarded as less socially acceptable, such as drug-related conditions. Volunteer bias can occur at every stage, from recruitment through retention to follow-up.
The probiotic trial in the table illustrates that last point. The Catalogue reports that as the trial went on, the most deprived participants were more likely to be lost to follow-up at six months and two years, while mothers interested in probiotics were more likely to attend research clinics.
Self-selection in polls and online surveys
Polls are where most readers meet this problem. Pew Research Center describes online opt-in sampling as an approach where people are not selected randomly but are recruited from online sources such as banner ads or social media. Pew says opt-in polls can perform similarly to traditional methods for some purposes, such as election polling, but that several studies have documented large errors in them.
Pew's own example is specific. A December 2023 opt-in poll reported that 20 percent of US adults under 30 agreed that the Holocaust is a myth. When Pew repeated the question in January 2024 on its American Trends Panel, which is recruited by random sampling of residential addresses, it found 3 percent, the same as every other age group. Pew says the gaps are what it would expect if many bogus respondents, people who answer insincerely to earn rewards, claimed to be under 30. That is a different mechanism from classic self-selection, yet it is still a question of how people enter the sample, and Pew contrasts opt-in recruitment with its own random address sampling. In another Pew study comparing three opt-in samples with three probability-based panels, young or Hispanic opt-in respondents were far more likely to say yes to 10 or more of 16 yes/no questions, while very few respondents on the probability panels gave such implausible answers. In an earlier Pew experiment, 12 percent of opt-in respondents under 30 claimed to be licensed to operate a nuclear submarine, a share that in reality rounds to 0 percent.
The same logic applies to informal polls on social media or news sites. Anyone who chooses to click is self-selected, so the result describes the people who answered, not the public.
How researchers reduce it
The Catalogue of Bias says it is difficult to estimate the impact of volunteer bias or even the direction of its effect. Separately, it lists preventive steps. It notes that the likelihood of volunteer bias increases as the refusal rate increases, so any technique that raises volunteer numbers is likely to reduce it, and that ensuring anonymity and confidentiality is essential to increase participation.
In the HRT example, later randomized controlled trials reached a different conclusion from the observational studies. For surveys, Pew's approach is probability-based recruitment, where nearly all adults have a chance of selection, followed by weighting to match the population.
In the test-prep example, a relative measure of improvement might make the study somewhat more reliable, but only partially.
Questions to ask when you read a result
Did the participants choose to take part, or were they selected at random? If they chose, what might make the people who joined different from those who did not?
Was the treatment, course or product assigned, or did people pick it? A comparison between choosers and non-choosers is open to the HRT problem.
For polls, how were respondents recruited? In Pew's replication, an opt-in survey and its own random address sample gave very different numbers on the same questions.