Surveying Multiple Sensitive Attributes using an Extension of the Randomized-Response Technique

The randomized-response technique (RRT) protects the privacy of respondents by adding random noise to their responses, such that there is no direct link between an individual's response and her true status. Although the RRT has repeatedly been shown to outperform direct questioning, it has rarely been used in survey research. First, it is difficult to survey multiple issues simultaneously. Second, traditional RRT models do not take the problem of nonadherence to the instructions into account. We describe a modification of the RRT that is capable of surveying multiple attributes using just a single randomization process and controls for nonadherence. An empirical application demonstrates the superiority of this approach over both, direct questioning and the forced-response variant of the RRT.