The problem of joint ego-motion estimation and multiple object tracking (MOT) in automotive multiple-input and multiple-output (MIMO) radar has been studied. The 3D ego-motion estimation is performed based on phase changes of the raw signal caused by relative movement between objects and the radar, and the ego-motion-induced velocities are compared with the detected ones to label static vs moving objects. The static objects are used for ego-motion estimation again to improve the accuracy, while the moving objects are used for MOT. The performance of the algorithm has been studied on simulated data and evaluated using different tracking algorithms, proving the feasibility of this approach.