Forecasting and Latent Heterogeneity in Multi-Departmental Manpower ‎Systems: A Hidden Markov Framework

Abstract

Recent advances in manpower planning have emphasized the need to model workforce systems ‎within departmentalized frameworks, reflecting the structural reality of most organizations. ‎While departmentalization serves as a form of disaggregation that can address observable ‎heterogeneity, it does not eliminate heterogeneity arising from latent, unobservable factors and ‎may instead introduce additional forms of hidden variability associated with intra-departmental ‎and inter-departmental transitions. This study aims to develop and apply a generalized hidden ‎Markov manpower model for forecasting and analyzing latent heterogeneity in multi-‎departmental manpower systems. A recursive forecasting structure is derived and applied to a ‎real departmentalized manpower dataset obtained from a university non-academic workforce, ‎under the assumption of stable recruitment and promotion policies. The empirical results reveal ‎a dominance of diagonal elements in the estimated transition probability matrix, indicating slow ‎promotion dynamics across departments and grades. Forecasts over multiple periods show a ‎gradual decline in lower-grade manpower stocks and relative stability in higher-grade ‎categories, with marked differences in promotion propensities across latent subclasses within ‎the same observable categories. These findings confirm the significant role of latent ‎heterogeneity in shaping long-term manpower evolution. In summary, the generalized ‎departmentalized hidden Markov manpower model developed in this study offers a more ‎suitable framework for describing and forecasting manpower systems. By explicitly ‎incorporating both departmental structure and unobservable differences in personnel behaviour, ‎the model produces more reliable forecasts and provides insights that are useful for informed ‎manpower planning and policy decisions‎.