This study investigates the incorporation of the jackknife procedure into the three-stage cluster sampling design in the estimation of finite population totals and confidence lengths. A particular attempt is also made in deriving the asymptotic properties of the three-stage cluster sampling design. Also, a comparison of the performance of the proposed estimator with some already existing estimators such as Nadaraya-Watson, Horvitz-Thompson estimator, jackknifed Nadaraya  -Watson estimator and the ratio estimators is investigated. Five (5) simulated datasets were used in the empirical study. Based on the empirical study, the jackknife procedure in both the jackknifed Nadaraya-Watson estimator and the three-stage cluster sampling design is recommended. The good performance in the bias, mean squared errors, and the tight confidence lengths sets them apart from the other already existing estimators. Note that the good performance in the two estimates (JNW and three-stage cluster sampling design) is attributed to their ability in capturing the auxiliary statistic.

cluster, three-stage, jackknifing, efficiency, robustness, bias.