Randomized response: a survey technique for eliminating evasive answer bias.

A survey technique for improving the reliability of responses to sensitive interview questions is described. The technique permits the respondentto answer "yes" or "no" to a question without the interviewer knowing what informationis being conveyed by the respondent. The privacy of the interviewee is protected by randomizing his response. For example if all members of a population belong either to group A or to group B and the investigator wants to determine the proportion of group A individuals in the population, this information can be elicited by using the following procedures. Before each interview, the respondent is provided with a spinner marked with a point A and a point B. The spinner is marked off in such a way that the spinner's marker will stop at point A with a probability of p and at point B with a probability of 1-p. When the interviewer asks the sensitive question concerning group membership, the respondent spins the spinner out of the sight of the interviewer. The marker will either stop at point A or point B. The respondent then indicates whether or not he belongs to the group to which the marker is pointing. The respondent does not tell the interviewer where the marker is pointing. Assuming that these responses are truthful, it is then possible to determine maximum likelihood estimates of the true proportion of As in the population. The formula for calculating these estimates is provided. Both the degree of truthfulness which can be expected from the respondents and the sample size required for given levels of precision depends on the parameter p.