Social Desirability Bias

Category: Social

You shade your answers toward what makes you look good, so self-reports drift toward the flattering version of you instead of the real one.

How it works

Social desirability bias is the tendency to answer questions in a way that makes you look good to others (and to yourself) rather than in a way that is true. Researchers split it into two motives: impression management, where you knowingly polish your answer for an audience, and self-deception, where you actually believe the flattering version. That two-part model was developed by Delroy Paulhus in 1984, building on earlier work by Sackeim and Gur, and it refined the older idea (from Crowne and Marlowe) that social desirability was a single "need for approval."

The bias bites hardest on questions with an obvious "right" answer that society rewards or punishes, so you over-report the virtuous stuff (voting, exercise, charitable giving, condom use) and under-report the shameful stuff (racism, drug use, cheating, how much you actually drink). The mechanism is not lying so much as a fast, mostly automatic edit: your brain runs the answer through a "how does this land?" filter before you ever say it. The more visible the interviewer and the more sensitive the topic, the stronger the pull.

Where you'll see it

  • Handwashing: in observational studies run for the American Society for Microbiology, around 90 to 96 percent of people SAY they wash their hands after using a public restroom, but observers watching discreetly counted only about 66 to 85 percent actually doing it. The gap is pure social desirability, because nobody wants to admit they skipped the sink.
  • Voter turnout: surveys consistently show more people claiming they voted than voter rolls confirm. When Stanford's Jon Krosnick and Allyson Holbrook used the item count technique (which lets people report anonymously without ever saying 'I voted' out loud), reported turnout dropped in a telephone survey, exposing the direct question as inflated by the desire to look like a good citizen.
  • The Bradley effect (1982 California governor's race): polls had Black candidate Tom Bradley about seven points ahead on election eve, yet he lost to George Deukmejian by roughly a point. One leading (and contested) explanation is that some white voters told pollsters they would vote for Bradley to avoid appearing prejudiced, then did otherwise in the booth. The same dynamic resurfaced in debates over 2016 and 2020 US polling misses.
  • Sensitive behavior in surveys: list experiments on condom use have reduced misreporting by roughly 17 percentage points versus direct questions, and self-reported income, church attendance, and charitable donations all run high while drug use and problem drinking run low, precisely because respondents edit toward the answer that flatters them.

Where it comes from

The concept was pinned down by University of Washington psychologist Allen L. Edwards, first in a 1953 paper and then in his 1957 monograph "The Social Desirability Variable in Personality Assessment and Research." Edwards showed something damaging: how socially desirable a personality-test item was rated to be predicted, at a correlation in the .85 to .90 range, how many people would endorse it about themselves. In other words, personality inventories like the MMPI were partly measuring the urge to look good rather than the trait itself. In 1960 Douglas P. Crowne and David Marlowe answered a flaw in Edwards's approach (his scale was tangled up with psychopathology) by publishing "A New Scale of Social Desirability Independent of Psychopathology" in the Journal of Consulting Psychology. Their 33-item Marlowe-Crowne scale, built from behaviors that are culturally approved but statistically rare, became the standard yardstick and has since been cited in more than 5,000 papers.

How to counter it

Kill the audience before you ask. The bias runs on being watched, so switch sensitive questions to self-administered, anonymous formats: online forms, sealed entry, computer-based surveys. Face-to-face and phone interviews reliably inflate the virtuous answers and bury the shameful ones, so stop collecting reputation and start collecting data.

When the topic is loaded, ask sideways. Use indirect methods that let people admit a behavior without pointing at themselves: a list experiment buries the sensitive item inside an aggregate count, and randomized response hides the answer behind a coin flip. Both shrink the gap between what gets reported and what is actually true.

Reach for the ugly specific. When you are the one answering, notice the too-clean answer arriving too fast and force yourself to the concrete number instead of the rounded-up self-portrait: the exact count of drinks last week, the actual date you last worked out. If your answer flatters you and comes with no messy detail, you are reciting your reputation, not your behavior.

Write questions that give people an out. Drop the loaded framing, add a face-saving preamble like "many people find they don't manage to...", and never make someone perform virtue in front of you. A question that makes the honest answer feel shameful is a question that guarantees a lie.

The tell

When your answer arrives too clean and too fast, and it happens to be exactly the answer that makes you look good, you are probably reporting your reputation, not your behavior.

Related biases

References

  1. Crowne, D. P., & Marlowe, D. (1960). A new scale of social desirability independent of psychopathology. Journal of Consulting Psychology, 24(4), 349-354
  2. Edwards, A. L. (1957). The Social Desirability Variable in Personality Assessment and Research. New York: The Dryden Press
  3. Holbrook, A. L., & Krosnick, J. A. (2010). Social desirability bias in voter turnout reports: Tests using the item count technique. Public Opinion Quarterly, 74(1), 37-67
  4. Paulhus, D. L. (1984). Two-component models of socially desirable responding. Journal of Personality and Social Psychology, 46(3), 598-609
  5. Krumpal, I. (2013). Determinants of social desirability bias in sensitive surveys: a literature review. Quality & Quantity, 47(4), 2025-2047