This study examines how social desirability (SD) bias undermines the validity of Likert-type surveys in sociology and social policy research, and provides a methodological review of its psychological foundations, empirical manifestations, and mitigation strategies for sensitive survey contexts.
A narrative methodological synthesis was conducted using PsycINFO, Scopus, Web of Science, and Google Scholar (1960–2026). Findings are organized into four mitigation families, including questionnaire design, timing-based interventions, indirect questioning, and analytical controls, and are evaluated against Stocké (2004) three-condition framework, which identifies non-anonymity, approval-seeking disposition, and item desirability differentials as necessary co-conditions for SD bias activation.
SD bias arises from both deliberate impression management and automatic self-deception, distorting correlations, group means, and prevalence estimates in policy-relevant surveys. Experimental evidence reveals a curvilinear relationship between response time and SD bias, with both fast- and slow-responding amplifying bias, undermining time-pressure interventions as a general remedy. Layered combinations of anonymity assurances, indirect questioning, careful item wording, and statistical controls provide more robust, though still partial, protection.
Unlike prior reviews that treat psychological mechanisms and practical mitigation in isolation, this study integrates both within a single theoretically anchored framework spanning classical instruments and emerging approaches, including evaluative neutralization, ICT, and digital behavioral detection. It presents a structured matrix comparing mitigation techniques by their strengths, trade-offs, and implementation requirements, and provides a theory-derived rationale for why layered strategies that target different triggering conditions simultaneously offer greater practical robustness than any single technique.
