1982: the Covariate Unrelated Question Randomized Response Model

The randomized response technique (Warner, 1965) is a data collection procedure that allows researchers to collect sensitive information while guaranteeing privacy to respondents. For example, if one question from a pair of questions that have the same response options is selected randomly and in private, the question answered by a given individual will not be known to the researcher. However, there is sufficient information from a sample of respondents ( e.g., the total number of "yes" responses and the probability of selecting the sensitive question) to estimate the proportion of the population with the sensitive characteristic. Additional variance for the estimator of the population with the sensitive characteristic results from sampling items with the randomized response technique. The technique described in this paper allows one to improve accuracy of estimation by using a covariate that correlates with the sensitive characteristic but which is itself not sensitive. Theory has been developed for a covariate randomized response model which is an extension of the Warner procedure (Dayton and Scheers, 1980); the present study involveS a similar extension for the unrelated question randomized response technique (see Greenberg, et al., 1969). This study comprises three stages: (i) theoretical development of the covariate model for both IT known and ~y unknown cases, (2) conY struction of efficiency tables that compare the covariate model to the unrelated question model and (3) a demonstration study of the covariate model.