Turbo pumps for rocket engines often equipped balance piston (BP) systems at the back-shroud of the impellers for cancelling their axial thrust. The BP system is self-balancing and stable under quasi-static conditions, but it is known that the BP systems can be unstable under certain dynamic conditions. The performance characteristics of turbo pumps equipped with unshrouded impellers might be affected by the axial position of the rotor. Thus it is necessary to consider this effect when calculating the balance of axial thrust. Few experiments have determined the characteristics of unshrouded impellers equipped with BP systems yet. In this research, an experimental study of a model turbo pump for rocket engines was carried out. This pump had an unshrouded impeller, a BP system, a vaned diffuser, and a volute. Axial forced oscillations were applied on the rotor of the pump by an active magnetic bearing (AMB) test facility. This setup can oscillate with freely-selected amplitude and frequency applying thrust to the rotor. During the oscillations, the fluctuation of axial thrust under the operating conditions was monitored using strain gauges. The axial thrust compensation ability and the response of the BP system were evaluated by analyzing the magnitude, amplitude and phase delay of the axial position of the rotor. Moreover, 3D simulations and 1D simulations were carried out for the model pump. In the 3D simulations, computational fluid dynamics (CFD) was used to calculate the internal flow of the model pumps. The BP system was equipped with an impeller on which were applied forced oscillations. The impeller movement was modeled using a mesh morphing method. The 1D simulation predicted the axial thrust by calculating the mass flow balance using the geometry of the model pump. The phase lag between the axial position and the thrust was dominated by the pressure fluctuation at the BP chamber caused by the mass flow balance. The 3D simulations well predicted the fluctuation, but the characteristics of the BP system estimated by the 3D simulations were more stable than those determined by the experiments. On the other hand, the characteristics estimated by the 1D simulation was less stable than those by the experiments. However, these simulations grasped the tendency of the BP system to become unstable as the oscillation frequency increases, and are effective in predicting the characteristics.