Evaluation of Corporate Modelling Approaches Aiffering with Regard to the Level of Aggregation in a Hypothetical Multi-Plant Case.

Abstract

Study of many papers on corporate modelling indicated that companies are starting to use detailed, bottom-up models rather than the previously more popular top-down type of models. However, no empirical justification for this change appears to have been attempted. In this research the two approaches are applied to a hypothetical multi-plant manufacturing company. The corporate planning decisions include price, level of marketing expenditure, level of production, and the allocation of production to each plant. Many criteria of the relative efficacy of alternative courses of action are introduced, for example the satisfaction of the stakeholders. The rate of return on capital employed is used, however, to evaluate the two approaches. The required calculations were incorporated into BASIC programs on an HP2000 ACCESS computer, and rigorously tested. The thesis is divided into four parts. The first surveys the literature, identifies the problem and suggests a method of attack, and describes the features of the case study. The second discusses the mathematics and logic of the models, and describes the computer programs. The third is devoted to testing the validity of the programs of the models. The fourth is allocated to testing the hypotheses and evaluating the two models. Five hypotheses are formulated and analysed using correlated sampling techniques. Appropriate statistical tests were applied in each case. The results suggested that top-down models were perfectly adequate for the 'tactical' decisions such as level of marketing expenditure and product pricing, and much cheaper than the bottom-up models. However, 'operational' decisions such as the level of production and allocation of production to each plant were significantly improved by a bottom-up approach. The applicability of these results to other situations is argued on the grounds of the wide range of values experimented with. However, the approach followed is quite general, and could be applied to any situation where the necessary data was made available.