This paper proposes a new stratified technique to address the problem involving estimation of the population proportion of people with sensitive attribute(s). Studying the proposed technique under proportional and Neyman allocations shows that the proposed technique is more efficient than (outperforms) the Singh & Gorey [1] and Tarray & Singh [2] stratified randomized response models. Applying the proposed technique to a survey on drug use disorder also shows the applicability of the model.