Fully Randomized Conjoint Reduces Social Desirability Bias by Two-Thirds

🔎 What This Asks

A test of whether conjoint analysis can elicit more honest survey responses by reducing social desirability bias (SDB) on sensitive attributes.

🔬 How the test was set up

A novel experimental comparison contrasts three designs:

• a standard, fully randomized conjoint design;
• a partially randomized design where only the sensitive attribute varies between the two profiles in each task;
• a control condition that accounts for any confounding from greater attention to the varying attribute under the partially randomized design.

📊 Where this was implemented and what was measured

• Two empirical studies: one on attitudes about environmental conservation and one on preferences for congressional candidates.
• The primary outcome is the average marginal component effect (AMCE) of the sensitive attribute.

• Key empirical result: in both studies, estimates indicate the fully randomized conjoint design could reduce SDB for the AMCE of the sensitive attribute by about two‑thirds of the AMCE itself.

⚠️ Caveats and next steps

• Findings are encouraging but exploratory. Results show sensitivity to alternative model specifications.
• Additional confirmatory evidence using the same experimental comparison is recommended before strong conclusions are drawn about generalizability.

Why it matters: the proposed design provides a direct way to evaluate whether conjoint designs mitigate SDB and offers a concrete path for future confirmatory tests in survey research.