Randomized response technique (RRT) models are commonly used to estimate proportion of sensitive trait in survey sampling. In this study, we propose an improved class of estimators for sensitive trait when population level information is available on a non-sensitive auxiliary variable. This class of estimators is based on a generalization of the famous Kuk’s (1990) estimator and a difference estimator proposed by Diana and Perri (2009). We validate the results using survey data collected at the university of Wah, Pakistan. We also use a data set used earlier by Zaizai (2006).