ACSPRI Conferences, RC33 Eighth International Conference on Social Science Methodology

Font Size:  Small  Medium  Large

Techniques to measure sensitive information

Sebastian Sattler, Peter Graeff

Building: Law Building
Room: Breakout 6 - Law Building, Room 022
Date: 2012-07-12 03:30 PM – 05:00 PM
Last modified: 2012-06-22


Precise question: How can sensitive behavior (such as plagiarism, shoplifting or intoxicated driving) be measured correctly in order to get valid and unbiased estimates of its prevalence? „A question is sensitive when it asks for a socially undesirable answer, when it asks in effect, that the respondent admits he or she has violated a social norm” (Tourangeau & Yan 2007, p. 860). E.g. plagiarism can be defined as the presentation of ideas or quotations of others as own work. This behavior violates social norms (e.g. concerning fairness) and is of sensitive content in survey research with an increased likelihood of item-non-, unit-non-response or response bias. Knowledge of precise prevalence rates is important for policy making and for judging the validity of prior analysis.

Methods/Data: The prominent Randomized Response Technique (RRT, e.g. Warner 1965; Greenberg, Horvitz & Abernathy 1974) preserves objective anonymity for respondents by applying a random device (such as a spinner or coin) when respondents have to choose between optional answers in a questionnaire or an interview. This classical technique has several drawbacks. New techniques such as Triangular Technique and Crosswise Technique (Yu, Tian & Tang 2008) warrant anonymity not by applying a random device but by providing item options. These techniques are easier to apply since a random device is no longer necessary. In an ongoing experimental study, five techniques (direct response, crosswise, triangular and two randomized response versions) are compared using data from a paper and pencil class room survey study among students. Expected sample size within the next 4 months: 4,500 participants (actually N=2,500).

Results: Preliminarily results imply that respondents tend to answer more truthfully if anonymity techniques are applied compared to direct response.

Implications: Theoretical explanations imply that RRT bears potential for improvement that must be considered in forthcoming studies, but are harder to apply compared to non-randomized variants. So, the application of the newer techniques is recommended.

Greenberg, BG, Horvitz, DG & Abernathy, JR 1974, ‘Comparison of randomized response designs’, in F Prochan & RJ Serfling (eds.), Reliability and biometry, statistical analysis of life lengt, Philadelphia, SIAM, pp. 787-815.
Tourangeau, R & Yan, T 2007, ‘Sensitive questions in surveys’, Psychological Bulletin, vol. 133, no. 5, pp. 859-83.
Warner, SL 1965, ‘Randomized response: a survey technique for eliminating evasive answer bias’, Journal of the American Statistical Association, vol. 60, no. 309, pp. 63-9.
Yu, JW, Tian, GL & Tang, ML 2008, ‘Two new models for survey sampling with sensitive characteristics: design and analysis’, Metrika, vol. 67, no. 3, pp. 251-63.