This paper identifies and estimates the relative average treatment effect in the presence of misclassification. We propose consistent estimators based on nonparametric methods. The simulation results reported illustrate the performance of the proposed estimators.