Nonresponse bias

Nonresponse bias is systematic error that arises when the people who respond to a survey differ meaningfully, on the trait actually being measured, from those who were sampled but did not respond, so the completed sample no longer represents the target population even when selection into the sample was random.

Unlike ordinary sampling error, which shrinks as sample size grows, nonresponse bias does not shrink just by collecting more responses if the same kind of person keeps opting out; it depends on who is missing, not how many are missing. It is often confused with a low response rate, but the two are distinct: a survey can have a low response rate with little bias if nonrespondents resemble respondents, or a high response rate with meaningful bias if one specific subgroup systematically declines to answer.

This is one of the central threats to validity in political and public-opinion research, for example if supporters of one side of an issue are more or less willing to talk to pollsters than the other. Researchers try to detect and correct it through survey weighting on known correlates of nonresponse, but weighting can only adjust for traits it actually measures, not for unmeasured ones, which is why a well-built sampling frame matters as much as effort spent chasing a higher response rate, and why the resulting distortion is often grouped under the broader term sampling bias.