Behavioral Science Dictionary

Non-response bias

Also known as: Nonresponse bias

Methods & Evidence

When the people who don't answer differ from those who do, the sample lies.

What it means

Non-response bias is a distortion that arises when individuals selected for a survey or study who do not respond differ systematically on the variables of interest from those who do. It is a specific form of selection bias operating at the level of participation rather than recruitment: a high response rate is not in itself protective, and a low rate is not necessarily fatal, because what matters is whether the propensity to respond is correlated with the outcome being measured. The mechanism is that responding is itself a behavior — busy, disengaged, dissatisfied, or hard-to-reach people opt out at different rates — so the realized sample is no longer representative even of a well-drawn frame. Researchers probe it by comparing respondents to known population benchmarks, contrasting early versus late responders, and weighting or modeling the missing-data process. It matters because it can silently invalidate polls, customer-satisfaction scores, and any inference generalized from a self-selected subset.

Examples

A restaurant emails a satisfaction survey, but mainly delighted and furious customers bother to reply, so the average rating misrepresents the typical diner's experience.

An engagement survey goes out during a hiring freeze; the most disaffected staff cannot be bothered to fill it in, so morale scores look healthier than the office feels.

A phone poll reaches mostly people happy to talk to strangers at dinnertime; if that willingness tracks how they vote, calling more numbers will not fix the tilt.

Where this comes up

← All 1001 terms