A goal programming approach to estimating performance weights for ranking firms

This paper proposes several goal programming (GP) models for estimating the performance measure weights of firms by means of constrained regression. Since some single-criterion performance measures are usually in conflict, we propose two opposed alternatives for determining multiple-criterion performance: the first is to calculate a consensus performance that reflects the majority trend of the single-criterion measures and the other is to calculate a performance that is biased towards the measures that show the most discrepancy with the rest. GP makes it possible to model both approaches as well as a compromise between the two extremes. Using two case studies reported in the literature and introducing another one examining non-financial companies listed in Ibex-35, we compare our proposal with other methods such as CRITIC and a modified version of TOPSIS. In order to improve the comparisons a Montecarlo simulation has been performed in all three case studies.Scope and purposeThe study falls into the area of multiple-criteria analysis of business performance. Firms are obliged to report a vast amount of financial information at regular intervals, and for this there is a wide range of performance measures. Multicriteria performance is calculated from the single-criterion measures and is then used to draw up rankings of firms. As a complement to the other multicriteria methods described in the literature, we propose the use of GP for implementing two quite different strategies: overweighting the measures in line with the general trend or overweighting the measures that conflict with the rest. Besides the use of Spearman's correlation, we introduce two other measures for comparing the solutions obtained.