Monitoring the productivity change of retailing stores

The purpose of this paper is to evaluate the productivity change of stores from an European retailing organisation. The Malmquist index, complemented with bootstrapping, is used to measure the changes in store productivity between the years 2002 and 2004. It also investigates the differences between two distinct store formats (supermarkets and hypermarkets), and the impact of scale size on productivity change. This paper describes a case study of the application of the Malmquist index and bootstrapping to retailing stores. From a methodological point of view, it describes an enhanced approach to explore the relative position of frontiers from two di erent time periods, which enables determining if the frontier of one period dominates the other, or if the frontiers are crossed. The analysis of performance changes over time should take into account two e ects: the variation of technical e ciency of each store and the change in the position of the best-practice frontier. The Malmquist index correctly captures these two e ects. The variation in technical e ciency measures changes in the ability of each store to approach the best performance levels observed in the reference units. The changes in the frontier re ect technological developments in the practices of the best shops. The results of the case study showed that hypermarkets had a more favorable performance than supermarkets between 2002 and 2004. The stores improved overall productivity levels, mainly due to improvements in the productivity levels of the frontier. There were some supermarkets which moved further away from the best-practice frontier (both technical and scale e ciency levels declined), leading to a decrease in their overall productivity levels. Concerning the relative position of the frontiers, we concluded that for some regions of the production possibility set there is statistically signi cant evidence that the frontier of 2004 is more productive than the frontier of 2002. Nevertheless, for other input-output mixes the frontiers of 2002 and 2004 are equally productive, such that in those regions the frontiers may crossover