Abstract
In the study of graded manpower. systems using Markov models, a common method for evaluating manpower policies is to examine the staffing structures, e.g. in terms of the proportion of staff in each grade, which result from these policies. In cases where the staffing structure does not exhibit any obviously undesirable feature, this limited analysis may not be sufficient. It is suggested that in such cases a useful form of further analysis is the examination of the length of service distribution of staff in each grade in the system. A method for determining these length of service distributions is outlined and some results for a particular manpower system are presented.