A Time Series Analysis of Australian Regular Army Enlisted and Officer Separations

Abstract

Accurately determining end strength is important to be able to plan future accessions in a manpower system. Predicting separations is vital to end-strength modelling. Predicting separation rates within the Australian Amy is an identified area of required research to ascertain the best models for aiding reporting and as a decision support tool. In support of the Australian Regular Army end-strength model, this thesis examines the use of time series analysis on enlisted and officer separations over an eleven-year period. This thesis develops multiple time series models using ten of the eleven years of data to forecast Australian Regular Army separation numbers for the eleventh year. The observed separation numbers of the eleventh year are used to compare the accuracy of each of the models developed. Models developed include moving average, autoregressive, exponential smoothing, Winter’s method additive, and autoregressive moving average. This thesis finds that Autoregressive Integrated Moving Averages models are the most accurate time series models in predicting separation rates, outperforming the seasonal exponential smoothing and Holtz-Winter models.