A caveat on the measurement of productive e$ciency

The correspondence between used performance measures and enterprise objectives, such as pro"t maximization and cost minimization, is fundamental for manufacturing companies. This paper identi"es, and critically examines, a minimal set of relevant properties that a productivity index needs to satisfy to rightly assess performance development of a decision-making unit. Commonly applied and suggested productivity measurement techniques, such as partial e$ciencies, total factor productivity (TFP), index number approaches, integrated partial e$ciencies and operational competitiveness ratings, are analyzed in order to assess the alignment with superior objectives. There is a considerable spread in the results of this class of models and the interpretation may prove di$cult or misleading. As these apparently less complicated productivity measures increasingly are employed as a component in evaluation of manufacturing e$ciency, the question is of high managerial relevance. In particular, the paper points out inconsistencies with properties related to commensurability, monotonicity, and implications of maximizing behavior. Based on this viewpoint, issues such as the consistency with pro"t maximization will be shown extra interest. The paper also provides a critique of previous work in non-parametric e$ciency analysis where properties have been postulated or based on other arguments. The "ndings suggest that there is no globally superior measurement technique to be found in this class and that care should be taken when evaluating managerial performance not to penalize rational behavior. ( 2001 Elsevier Science B.V. All rights reserved.