The Crosswise Model for Surveys on Sensitive Topics: A General Framework for Item Selection and Statistical Analysis

Area of Science:

• Survey Methodology
• Psychometrics
• Statistical Modeling

Background:

• Direct questions on sensitive topics can lead to response bias.
• Indirect questioning techniques aim to improve data accuracy by concealing responses.
• The Crosswise Model uses paired sensitive and non-sensitive items to elicit truthful answers.

Purpose of the Study:

• To develop an integrated methodology for selecting baseline items in the Crosswise Model.
• To propose novel Bayesian estimation methods for these models.
• To demonstrate improved efficiency and relaxed assumptions compared to existing Crosswise Model applications.

Main Methods:

• Developed a methodology for baseline item selection based on conceptual and statistical criteria.
• Introduced four distinct statistical models within this framework.
• Proposed new Bayesian estimation techniques for implementing the models.

Main Results:

• The new methodology offers improved efficiency over common Crosswise Model applications.
• The proposed models may relax previously required statistical assumptions.
• Empirical application on LGBT attitudes demonstrates the Crosswise Model's potential.

Conclusions:

• The integrated methodology and Bayesian estimation enhance the Crosswise Model's utility.
• Improved efficiency and relaxed assumptions facilitate broader application.
• Available tools (app, code) support wider adoption of the methodology.