A new 2-stage approach to control social desirability bias

• Applicant: Dr. Julia Meisters
• Subject Area: Personality Psychology, Clinical and Medical Psychology, Methodology
• Term: since 2024
• Project identifier: Deutsche Forschungsgemeinschaft (DFG) - Project number 529232907

Dishonest responses jeopardize the validity of direct surveys on sensitive topics. Indirect survey techniques guarantee confidentiality through randomization and thus allow more honest answers. While this usually leads to higher estimates for the prevalence of sensitive attributes, it decreases comprehensibility and statistical and temporal efficiency, and sometimes even the proportion of correctly classified individuals. We propose a new 2-stage approach based on a multinomial model applicable to all indirect questioning techniques to meet these objectives better. This new approach solves several problems of indirect questioning techniques identified in previous research and thus promises to contribute to improved control of social desirability in self-reports. It reduces comprehension problems and increases statistical and temporal efficiency by using randomization only when needed while preserving the benefits of randomization for those for whom it is required. The project's main goal is to validate this improved 2-stage approach experimentally. First, it is intended to test whether the new approach is capable of increasing validity, the proportion of correct classifications, and the statistical and temporal efficiency of current indirect survey techniques (Exp. 1). In addition, possible moderators of its performance (Exp. 2), as well as the perceived comprehensibility and confidentiality of the new approach will be investigated (Exp. 3). In Experiment 4, the hitherto often neglected question of retest stability of individual responses and prevalence estimates will be examined. Subsequently, experiment 5 will clarify for the crosswise model, which is currently most frequently the focus of indirect surveys, whether the 2-stage approach is capable of reducing the dreaded influence of random response behavior due to its improved comprehensibility and whether it is superior to an alternatively proposed correction procedure for the crosswise model in this respect. Finally, in Experiment 6, the 2-stage approach will be used to obtain population-representative prevalence estimates for two particularly sensitive and socially relevant attributes, namely, prejudice against transgender persons and against women in positions of power. With the planned experiments, we want to contribute to further developing current indirect survey techniques. The results of the planned studies will be relevant for many fields of research in which estimates of the prevalence of socially (un)desired attributes are needed.

• DFG Programme: Research Grants
• Co-Investigators: Dr. Adrian Hoffmann; Professor Dr. Jochen Musch