Specialty Digest

This Week’s Edition — Aug 31, 2026
Specialty Digest Weekly  ·  Global English Edition

SPECIALTY DIGEST

DISCOVER IDEAS THAT SHAPE OUR WORLD

Nonresponse bias

nounStatisticsalso non-response bias, nonresponse error, non-response error

In one line

Nonresponse bias is the error that appears when the people who answer a survey differ from those who do not in ways that matter.

In simple terms

Every survey misses people. Nonresponse bias is what happens when the people it misses differ from the people it reaches.

If the difference has nothing to do with the question being asked, the estimate can still be sound. If it does, the number is wrong, and a larger sample will not rescue it.

How it works

Agencies track the response rate, which the US Census Bureau defines as the share of units that should have been interviewed and actually were. It measures nonresponse because a high level raises the chance that estimates are biased, when the units that do not respond differ from those that do.

The rate on its own, though, predicts surprisingly little. AAPOR, the professional body for survey researchers, reports no straightforward positive relationship between response rates and bias, and notes that some of the least biased results have come from surveys with less than optimal rates. It says there is still no consensus on why.

Pew Research Center measured the gap directly. Its telephone response rates fell from 37 percent in 1996 to about 9 percent by 2017. Even so, estimates of party affiliation and ideology sat within roughly 1.5 percentage points of high-response benchmarks, and across 14 demographic items the average difference was 3 points.

Other measures were far off. Volunteering was overstated by about 16 percentage points, and neighborhood involvement by 38. People willing to answer a survey turn out to be more civically engaged than people who are not.

Why it matters

Bias attaches to the question, not to the survey. The same poll can be close on voting intention and badly wrong on volunteering.

Weighting is the usual repair. Responses are adjusted so the sample matches known population figures for age, education, race and similar traits. It corrects only for what has been measured. Where nonrespondents differ in some unrecorded way, the tilt survives the weighting.

That is why statistical agencies also test directly, comparing respondents against administrative records to see whether nonresponse actually moved an estimate.

Where you’ll see it

  • Methodology notes explaining how a poll was weighted.
  • Official statistics releases that publish response rates beside the estimates.
  • Post-election reviews asking why polls missed.
  • Workplace and customer surveys where only the delighted and the furious reply.

Example

A staff survey draws replies from a third of employees. If the busiest people were least likely to answer, the results on workload will look better than the reality.

Often confused with

Margin of error. Margin of error describes random variation from sampling and shrinks as the sample grows. Nonresponse bias is a systematic tilt, and a bigger sample does not reduce it.

Key facts

  • The US Census Bureau defines the survey response rate as the ratio of units interviewed to all units that should have been interviewed, and measures nonresponse because high rates increase the likelihood of bias.3
  • AAPOR reports no straightforward positive relationship between response rates and bias, and says results showing the least bias have in some cases come from surveys with less than optimal rates.2
  • AAPOR also notes there is currently no consensus about the factors behind the disjuncture between response rates and survey quality.2
  • Pew Research Center telephone response rates fell from 37 percent in 1996 to about 9 percent by 2017.1
  • At that response rate, Pew estimates of party affiliation and ideology differed from high-response benchmarks by about 1.4 to 1.6 percentage points, and across 14 personal and demographic items by an average of 3 points.1
  • The same surveys overstated volunteering by about 16 percentage points and neighborhood involvement by 38 percentage points.1
  • Nearly all high quality surveys use statistical weighting, which cannot correct for differences on characteristics that were never measured.1
Go deeperWhy Do Election Polls Get It Wrong?
Quick checkWhy does a low response rate not automatically mean a survey is biased?Show answer

Bias depends on whether non-responders differ on the thing being measured. If they do not, a low response rate can still give an accurate estimate.

Sources

  1. Pew Research Center. What Low Response Rates Mean for Telephone Surveys. 15 May 2017 (accessed 15 September 2026)
  2. American Association for Public Opinion Research. Response Rates: An Overview. Undated (accessed 15 September 2026)
  3. US Census Bureau. Response Rates Definitions (American Community Survey). Undated (accessed 15 September 2026)

Editorially reviewed by Specialty Digest Editorial TeamLast reviewed September 16, 2026Researched and drafted with AI assistanceReport an issue