This paper is concerned with the auto-covariance function (ACVF) of a regime switching AR (1) process. In this model, two independent Markov chains govern on auto-regressive coefficient and standard deviation of white noise process. Our approach to solve this problem is to obtain the ACVF of a AR(1) model with time varying parameters and then to extend this result to regime switching case. An application of our formulae in model selection is proposed. Finally, a conclusion section is also given.