As a tool for eliminating evasive answer bias and the bias due to respondents’ refusal to cooperate , randomized response technique (RRT) is extensively applied in many areas of sociological, medical, biological and psychological research involving a study about the sensitive items. Kim and Warde (2004) proposed a mixed RRT assuming equal probability selection of population units and this was subsequently extended to complex survey situations by Amitava (2005). Here we propose an improvement of Kim and Warde’s (2004) RRT in complex survey situations and illustrate the superiority of proposed model over the Kim and Warde’s (2004) procedure.