Achieving high efficiency in hospital bed operations becomes an important aim of hospital management, as managerial environment of hospitals has been increasingly competitive in Korea. Many authors have made studies on higher occupancies, but very few about systems characteristics of the time serial data as well as decision behavior on bed operations. For those topics overlooked, this paper applies Wu's Dynamic Data Systems Analysis, for which an ARMA model is developed and its characteristic roots are analyzed. The results include such information as periodic nature of data, strength and effects of disturbances, based on which interpretations are provided tuned to the target system's operational aspects. For this work, data were collected and processed for two different periods of time, October 1990 to April 1991 and October 1993 to April 1994. While the average occupancies between two periods are not statistically different, the effect of error disturbances was managed better in 1990 than in 1993. And the restoring forces from fluctuations are almost the same between the periods of comparison. With this analysis, conclusions are summarized: in both periods, short-term flexibility in patient admissions and discharges are good, but long-term management of hospital census trend needs to be improved by developing the system of controlling patient admissions and discharges. For the improvement, some possible approaches are suggested for future studies.