A unified framework for analysis and comparison of randomized response surveys of binary characteristics

The randomized response (RR) procedures for estimating the proportion (π)$(\pi )$ of a population belonging to a sensitive or stigmatized group ask each respondent to report a response by randomly transforming his/her true attribute into one of several response categories. In this paper, we present a common framework for discussing various RR surveys of dichotomous populations with polychotomous responses. The unified approach is focused on the substantive issues relating to respondents’ privacy and statistical efficiency and is helpful for fair comparison of various procedures. We describe a general technique for constructing unbiased estimators of π$\pi$ based on arbitrary RR procedures, from unbiased estimators based on an open survey with the same sampling design. The technique works well for any sampling design p(s)$p(s)$ and also for variance estimation. We develop an approach for comparing RR procedures, taking both respondents’ protection and statistical efficiency into account. For any given RR procedure with three or more response categories, we present a method for designing an RR procedure with a binary response variable which provides the same respondents’ protection and at least as much statistical information. This result suggests that RR surveys of dichotomous populations should use only binary response variables.