Description
Uses a private randomising device to let respondents answer sensitive questions truthfully, recovering population prevalence while protecting individuals.
When asked directly about sensitive behaviour, people under-report. In randomised response, each respondent privately uses a randomiser (e.g. a die) that decides whether they answer the sensitive question truthfully or give a forced answer. The researcher never knows which applied, but because the probabilities are known, the prevalence can be estimated. Variants include Warner's original, forced response and unrelated question designs. Meta-analyses show RRT yields more valid prevalence estimates than direct questions in many contexts, though comprehension matters. For UX, it measures sensitive behaviours like password sharing, account sharing, policy violations, cheating or risky use.
Attributes
- Discipline
- Survey & Psychometrics
- Loss type
- Unmigrated
- Goal
- Measure behaviour
- Stage
- Discovery
- Data
- Quantitative
- Effort
- Days
Steps
- Choose the sensitive question and the RRT variant (forced response is easiest to explain).
- Set randomisation probabilities (e.g. 75% answer truthfully, 25% forced 'yes' or 'no').
- Implement an online randomiser that respondents trust (e.g. a birthday-based rule or visible die) and explain protection clearly.
- Include comprehension checks; consider a direct-question control group for comparison.
- Estimate prevalence and its standard error using the design formulas, accounting for the larger variance.
- Size the sample accordingly (RRT needs several times the sample of direct questioning).
Sources
- Warner, Randomized response: A survey technique for eliminating evasive answer bias (1965), JASA
- Lensvelt-Mulders et al., Meta-analysis of randomized response research (2005), Sociological Methods & Research
- Blair, Imai & Zhou, Design and analysis of the randomized response technique (2015), JASA
Also known as
- RRT
- Warner's randomised response
- Forced response design
Categories
- Analytics & Behaviour
- Survey & Psychometrics