Non-Response Bias Estimator

Estimate how much non-response may be skewing your survey results.

Response rate 80%

3

Estimated full-sample mean: 57 (respondent-only mean: 60)

Estimated bias: 5.3% — a wave-extrapolation estimate, not a direct measurement.

How the non-response bias estimate works

Non-response bias happens when the people who respond to a survey differ systematically from those who don't, skewing the results. Because non-respondents' true values are unknown, this calculator uses the standard wave-extrapolation method: treating late respondents (those who needed more reminders or effort to include) as a proxy for non-respondents.

Enter the mean of your outcome variable among all respondents, the mean among just the late-responding subgroup, how many people responded, and how many were invited. The calculator estimates what the full-sample mean might be, and how far your respondent-only mean may be from it.

This is a well-established estimation method (Armstrong & Overton, 1977), not a direct measurement of true non-response bias — treat the result as a diagnostic signal, and interpret a large estimated bias as a reason to investigate further, not a precise correction factor.

The formula

Estimated full mean = (Respondents × Respondent mean + Non-respondents × Late-respondent mean) ÷ Invited

Non-respondents are estimated as invited minus respondents. The late-respondent mean stands in for the unknown non-respondent mean. The bias estimate is the respondent-only mean minus this bias-adjusted estimate.

Worked example

800 of 1,000 invited responded (mean 60); late respondents averaged 45

Estimated full mean = (800×60 + 200×45) ÷ 1,000 = 57. Bias estimate = 60 − 57 = 3 points higher among respondents than the bias-adjusted estimate.

Frequently asked questions

Why use late respondents as a proxy for non-respondents?

Late respondents needed more effort or reminders to participate, similar to non-respondents. This "continuum of resistance" assumption is a well-documented, widely used approximation — not a guarantee that the two groups are identical.

What counts as a "late respondent"?

Typically the last wave of responses — for example, those who responded only after a second or third reminder, or in the final days of fieldwork.

What should I do if the estimated bias is large?

Treat it as a signal to investigate — consider weighting adjustments, additional outreach to under-represented groups, or reporting the limitation alongside your results rather than treating the respondent-only figure as unbiased.

Next steps in your analysis

Built by PanelRoster

PanelRoster is an enterprise operations management platform for organizations coordinating distributed teams, assignments, workflows, quality assurance and operational execution.

Explore PanelRoster