The effects of type of forecasting model and aggegation procedure on the accuracy of managerial manpower predictions

This article examines Markov chain models in the prediction of future needs for managers in a particular organization. It compares the accuracy of predictions made using a Markov chain model to predictions obtained by regression methods using data on personnel movementa during 13 years in a large corporation. Two methods for estimating the transition probabilities between the different levels of the organization were presented and the a m m y of the respective Markovian models was compared. The effects of aggregating the employees into different hierarchical groups on the accuracy of prediction wae also examined. The Markov chain models yielded, in general, more accurate predictions than the regression models. The predictions that were obtained based on transition probabilities estimated from past movementa between levels were more accurate thaa those based on transition probabilities estimated from time series of the number of employees at each level. The accuracy of all models, however, depends on the forecasting of new employees entering the firm at each level The results also suggest that higher aggregation leads to more accurate predictions. The utility ofusing Markovian vs. regression models from both the accuracy aspect and the availability and costs of data are discussed.