Chain of custody · waved through on who handled it
Social Desirability Bias
The tendency for survey and self-report respondents to give answers that present them favorably to others, over-reporting approved behaviors and under-reporting disapproved ones, rather than reporting what is accurate.
Entered into evidence“There is an almost perfect relationship between the probability of endorsement of personality statements and the social desirability scale values of those statements.”
→ follow the claim down the corridor, hand to hand
Filed · origin
Allen L. Edwards, 1953
The relationship between the judged desirability of a trait and the probability that the trait will be endorsed
Handler 01 · study
Edwards (1953) trait-endorsement correlation
Probability of self-endorsement tracked rated social desirability almost perfectly, showing self-reports are driven substantially by desirability rather than accuracy.
Allen L. Edwards, 1953 · Reported correlation of approximately r = .87 between judged desirability and endorsement probability
Handler 02 · study
Crowne & Marlowe (1960) scale construction
Produced the Marlowe-Crowne Social Desirability Scale, the field's dominant measure; respondents high on the scale systematically distort self-presentation.
Douglas P. Crowne, David Marlowe, 1960 · Internal consistency reported around KR-20 = .88 in the original paper; test-retest ~.89
Handler 03 · study
Jones & Sigall (1971) bogus pipeline
When respondents believe a 'lie detector' can reveal their real attitudes, reports on sensitive/prejudicial topics shift toward less socially desirable, presumably more honest answers.
Edward E. Jones, Harold Sigall, 1971 · not reported (empirical paradigm paper; no single summary effect size for bias reduction is reported; quantitative effects compiled in later reviews)
Handler 04 · study
Kreuter, Presser & Tourangeau (2008) mode experiment with validation
Self-administration reduced socially desirable misreporting: Web respondents disclosed undesirable items more and had lower false-negative rates than interviewer-administered CATI.
Frauke Kreuter, Stanley Presser, Roger Tourangeau, 2008 · Undesirable affirmative reports: CATI 25.8%, IVR 27.7%, Web 30.5% (CATI vs Web p<.02); pooled odds ratio ~0.8 for Web misreporting vs CATI
Handler 05 · study
DeBell, Hillygus, Shaw & Valentino (2024) 'genuine pipeline' on turnout
Warning respondents that their answer could be checked cut self-reported turnout sharply, closing much of the gap between reported and actual turnout, evidence of social desirability inflation in turnout reports.
Matthew DeBell, D. Sunshine Hillygus, Daron R. Shaw, Nicholas A. Valentino, 2024 · Reported turnout fell from 74.2% (control) to 68.5% (explicit pipeline), a 5.7-percentage-point reduction, p<.001
Final disposition
Admitted w/ caveat
The core phenomenon, that self-reports shift toward socially approved answers when an audience is present and away from it when privacy increases, is robust and repeatedly confirmed against external validation data (turnout records, university grades, drug toxicology). What is genuinely mixed/contested is moderators and remedies: meta-analyses disagree on whether web/computer administration reliably reduces the bias. Gnambs & Kaspar (2015), analyzing 460 effect sizes (N=125,672), found computerized self-administration increased disclosure of sensitive behaviors relative to paper, but the SAME authors' Gnambs & Kaspar (2017) meta-analysis (30 studies, 3,746 participants) found a near-zero overall difference (Δ≈0.01) between computerized and paper modes on social-desirability scales. The validity of using social-desirability scales as statistical 'controls' is also disputed.
Replication — mixed. The core phenomenon, that self-reports shift toward socially approved answers when an audience is present and away from it when privacy increases, is robust and repeatedly confirmed against external validation data (turnout records, university grades, drug toxicology). What is genuinely mixed/contested is moderators and remedies: meta-analyses disagree on whether web/computer administration reliably reduces the bias. Gnambs & Kaspar (2015), analyzing 460 effect sizes (N=125,672), found computerized self-administration increased disclosure of sensitive behaviors relative to paper, but the SAME authors' Gnambs & Kaspar (2017) meta-analysis (30 studies, 3,746 participants) found a near-zero overall difference (Δ≈0.01) between computerized and paper modes on social-desirability scales. The validity of using social-desirability scales as statistical 'controls' is also disputed.
Handling note · how it gets waved through
The dominant account treats the bias as a response-editing process: when a true answer is embarrassing or norm-violating and an audience (real or imagined) could judge it, respondents shade their answer toward the approved direction. Tourangeau & Yan (2007) frame this within the survey-response model as the final 'editing' stage and show misreporting is situational, larger when the respondent actually has something embarrassing to report and when privacy is low. Paulhus (1984) splits the construct into two components: impression management (deliberate, audience-driven faking) and self-deceptive enhancement (non-conscious, genuinely held but inflated positive self-views).
Competing account — A long-running debate concerns whether 'social desirability' is a contaminating bias to be removed, a substantive personality trait (need for approval), or partly a valid signal. Some researchers (e.g., the Paulhus tradition) argue self-deceptive enhancement is not deliberate distortion at all, while others question whether statistically controlling for social-desirability scales actually improves validity rather than removing real variance. There is also debate over whether anonymity/computerization reliably reduces the bias.
Seen in the field
Politics / public opinion · 2024
Voter turnout over-reporting in the ANES (2020)
In the American National Election Studies 2020 Time Series Study, a randomized experiment showed self-reported turnout was inflated by social desirability: simply warning respondents that turnout could be validated cut reported turnout from 74.2% to 68.5%, a 5.7-point drop, narrowing the gap to validated turnout.
Education / survey methodology · 2008
Misreporting of academic records by survey mode
University alumni under-reported low GPAs and failing grades when interviewed by phone versus a self-administered Web survey, verified against actual institutional records; false-negative rates for 'low GPA' were 83.3% by phone (CATI) versus 61.5% by Web.
Medicine / public health · 2007
Under-reporting of illicit drug and substance use in surveys
Reviews of survey methodology document that respondents systematically under-report illicit drug use, drinking, abortion and other sensitive health behaviors relative to validation data (e.g., toxicology), with the degree of under-reporting depending on topic sensitivity and survey privacy.
To break the chain
Reduce the perceived audience and the incentive to posture: use anonymous self-administered (computer/web) collection, neutral non-judgmental wording, indirect techniques (list experiment/item-count, randomized response), and where ethical the bogus/genuine pipeline (signaling answers can be validated). Validate against external records where possible rather than treating self-reports as ground truth.
Kreuter, Presser & Tourangeau (2008) found Web self-administration cut socially desirable misreporting versus interviewer phone (odds ratio ~0.8); DeBell et al. (2024) cut inflated turnout 5.7 points with a validation warning. Note remedies are imperfect: Gnambs & Kaspar (2017) found near-zero mode differences on SD scales, so anonymity alone is not a guaranteed fix.
Catch it in the act
Watch for it whenever a question touches identity, virtue, or norms and an answer could embarrass: turnout, charitable giving, exercise, drinking, drug use, prejudice, hand-washing, condom use. The tell is a gap between what people say and what records show, and the size of that gap shrinking as you remove the audience, anonymous, self-administered, or 'we can verify this' framing pulls answers back toward the truth.
Cross-referenced files
- Impression ManagementchildPaulhus's deliberate, audience-driven component of socially desirable responding, conscious self-presentation to look good to others.
- Self-Deception Positivity (Self-Deceptive Enhancement)childPaulhus's non-conscious component: honestly believed but inflated positive self-views, not deliberate lying.
- Acquiescence BiassiblingAnother systematic survey response style (tendency to agree regardless of content) often discussed alongside social desirability as a source of self-report contamination.
- Demand Characteristicseasily-confusedRespondents inferring and conforming to the experimenter's expectations; overlaps with but is broader than the look-good motive of social desirability.
- Self-Report BiasparentBroad umbrella of systematic errors in self-reported data, of which social desirability is a leading specific cause.
- Hawthorne Effectmechanistically-linkedBehavior change due to being observed; shares the 'audience presence alters response' mechanism.