<p>A new improved calibration estimator for the population mean in a stratified sampling in the presence of an outlier in the auxiliary variable was proposed in this paper. The median of the auxiliary variable was used to define the calibration constraints. The choice of the median is because it possesses the ability to be insensitive to the presence of outliers compared to the mean used in the literature. A simulation study as well as an empirical study was performed to establish the performance of the proposed estimator over some existing estimators. The results of both the simulation and empirical studies show that the proposed calibration estimator performed better and more efficiently when compared to all the existing calibration estimators considered in this work based on the absolute bias (<i>ABS</i>) and relative root mean square error (<i>RRMSE</i>) criteria.</p>