Relations among statistical methods for assuring confidentiality of social research data

Abstract

Two statistical methods have been proposed for ameliorating the problem of respondent embarassment in surveys which involve sensitive questions. These methods are directly related to a third device developed to assure that responses are confidential in a particular sense, i.e., they permit only a stochastic link between respondent's identification and response. The purposes of this paper are to examine conditions under which the methods are equivalent, to relate all methods to familiar concepts of misclassification error, and to argue that one of the more important benefits of these methods is protecting the researcher's records from subpoena. © 1972.