Skip to main content

Social desirability bias in AP Statistics

Social desirability bias is a type of response bias where people answer a survey question the way they think makes them look good instead of truthfully. Because everyone leans toward the same "respectable" answer, the error pushes the sample statistic consistently above or below the true parameter.

This definition is part of Fiveable's AP Statistics review, which also has guided practice questions.

Verified for the 2027 AP Statistics exam•Last updated November 2026•Fiveable Content Team

Why the error leans in one direction

Ask people how much they exercise, and the answers come back too high on the whole, not randomly scattered around the truth. That pattern belongs to Unit 1 (Exploring One-Variable Data and Collecting Data), Topic 1.12 Potential Problems with Sampling, which asks where bias in a data collection plan could come from. Social desirability bias is one answer, filed under response bias.

Response bias means answers or measurements drift away from the "true" value in a single direction. Social desirability bias explains why they drift. People overreport things that are admired (exercise, voting, studying, recycling) and underreport things that carry a stigma (drinking, cheating, screen time, skipping class).

That one-direction push is what makes it bias in the statistical sense. It's a systematic error built into the procedure, so it doesn't average out the way random variation does.

Spotting it in a survey scenario

The course pairs response bias with "answers individuals give about themselves," and that's where social desirability bias shows up. Look for three ingredients in the stem:

  • The data are self-reported, not measured directly (asking people how much they exercise versus tracking their steps with a device).
  • The topic has a socially "right" answer, like honesty, health, grades, or following rules.
  • Respondents might feel judged, especially in face-to-face or phone interviews, or when answers aren't anonymous.

Here's an example. A school asks seniors in person, "Have you ever cheated on a test?" Even with a perfect simple random sample, many cheaters will say no. The sample proportion who admit cheating will come out lower than the true proportion of seniors who have cheated.

The fix lives in how the question is asked, since the problem has nothing to do with who gets picked. Anonymous, self-administered surveys tend to reduce it. A bigger sample doesn't help, because more people giving the same flattering answer just makes the wrong estimate more precise.

Keeping it separate from nonresponse and wording bias

Seniors who say "no" when asked in person about cheating did respond, so labeling that problem nonresponse bias is wrong. Nonresponse means selected people never answer at all. With social desirability bias, people do answer. They just don't answer truthfully. If the scenario says people "refused" or "didn't return the survey," think nonresponse. If it says they "exaggerated" or "underreported," think response bias.

It's also easy to confuse with question wording bias. Wording bias comes from a leading or confusing question that steers everyone. Social desirability bias can happen with perfectly neutral wording, because the pressure comes from the topic and the setting. A leading question about a sensitive topic can cause both, and naming both is fine as long as you explain each.

It's also different from undercoverage or voluntary response bias, which are about who ends up in the sample. Social desirability bias is about what the people in the sample say.

Naming the direction in Question 1 or Question 4 answers

A survey asking students to report their weekly study hours is the kind of data collection scenario that shows up in free-response Question 1 (Multi-Focus, Practices 1 and 2) and can also come up in Question 4 (Multi-Focus, Practices 2, 3, and 4). When a part asks you to identify a potential source of bias, a strong answer does three things in context:

  1. Names the bias (response bias due to social desirability).
  2. Explains the mechanism: why these respondents would shade their answers on this topic.
  3. States the direction: whether the estimate would be too high or too low, and which statistic.

The third step is where answers often fall short. "The results could be inaccurate" isn't enough. Something like "Because students want to appear hardworking, many will overstate their weekly study hours, so the sample mean study time will likely overestimate the true mean for all students at the school" hits all three.

Multiple-choice questions test the same reasoning with a summary statistic. If students tend to exaggerate study time, the sample median will tend to overestimate the true population median, and the sample mean will tend to overestimate the true population mean. A random sample doesn't change that conclusion, since randomization protects against selection bias and does nothing about dishonest answers.

Where this sits in AP® Statistics

Keep studying AP® Statistics Unit 1