Skip to content
AP Statistics · Exploring One-Variable Data and Collecting Data

Nonresponse bias

Nonresponse bias appears in 19 reviewed AP Statistics questions and sits in Exploring One-Variable Data and Collecting Data. Below are worked examples showing exactly how the exam tests it, with the answer marked and every distractor explained.

How Nonresponse bias is tested

Example 1Medium
A survey company mails questionnaires to 1,000 randomly selected households to ask about home renovation plans. Of the 1,000 surveys sent, only 150 are returned. The company reports results based on the 150 responses. What is the primary concern?
  1. AUndercoverage, because renters may not receive the survey.
  2. BVoluntary response bias, because respondents chose to participate.
  3. CSampling frame bias, because not all households were on the mailing list.
  4. DNonresponse bias, because the 150 households that responded may differ systematically from the 850 that did not.Correct

Why D is right

When a large fraction of the selected sample does not respond (here, 85%), there is substantial risk of nonresponse bias. The households that returned surveys may be those with more renovation interest or activity, differing systematically from non-respondents. The final 150 responses likely do not represent the originally selected 1,000 households.
Why the other 3 choices miss
A
Undercoverage is possible but the question describes the mailing going to 1,000 randomly selected households. The primary concern given the 15% response rate is nonresponse bias.
B
Voluntary response bias occurs when people volunteer to be in a study. Here, 1,000 were randomly selected and sent surveys; the issue is with those who chose not to return them (nonresponse), not self-selection into the initial sample.
C
Sampling frame bias would be relevant if the mailing list failed to cover certain households. The question states 1,000 were randomly selected, implying a reasonable sampling frame.
Example 2Harder
A technology company wants to understand user satisfaction across its customer base of 2 million users. The company sends surveys through its app, but 90% of responses come from users aged 18-34, while this age group represents only 40% of the total customer base. This discrepancy illustrates
  1. Aresponse bias because younger users give different answers than older users
  2. Bselection bias from using a stratified sample
  3. Cundercoverage and nonresponse bias among older usersCorrect
  4. Dmeasurement error caused by poorly worded satisfaction questions in the app survey

Why C is right

The disproportionate representation of younger users reflects two overlapping problems. First, older users may have lower smartphone/app usage (undercoverage — they are less likely to be in the digital sampling frame). Second, even among those who see the survey, older users may be less likely to respond (nonresponse bias). Together, these biases produce a sample that dramatically overrepresents younger users.
Why the other 3 choices miss
A
Incorrect. While younger and older users may give different answers, the primary concern is that younger users are dramatically overrepresented in the sample. Response bias refers to inaccuracy in answers, not to who responds.
B
Incorrect. There is no indication of stratified sampling. The company sent surveys through the app to all users, and the disproportionate response is due to differential engagement, not a sampling design choice.
D
Incorrect. The issue is who responds to the survey, not how the questions are worded or how they measure satisfaction. The satisfaction questions themselves are not the source of the discrepancy.

Studied alongside

Other concepts from Exploring One-Variable Data and Collecting Data.