PRODUCTIVITY ASSESSMENT OF MULTIPLE RETAIL OUTLETS

One of the main problems in assessing the relative efficiency of multi ple retail outlets is the fact that customers are an essential compone nt of the ''production process'' in retailing. Hence, the efficiency o f a particular retail store depends both on the seller's and the buyer s' efficiency. While there is often data an each store's inputs to the production of retail services, the contribution of each store's custo mers to this process is difficult to measure, and often goes unmeasure d. In the absence of data on variation in customer efficiency across s tores, it is difficult to make comparisons of the relative efficiency of the retailer's operation across stores In this paper, we propose to solve the problem of missing data on customer inputs by treating stov es as members of different latent classes, where the classes are defin ed by differences in customer inputs and other market characteristics. We consider the clusterwise estimation of multiple translog cost func tions, which identifies sets of retail outlets operating under similar conditions and simultaneously estimates multiple cost functions for t hese classes. The efficiency of each outlet would then be evaluated re lative to others in its class, which allows for a more equitable evalu ation of each retail outlet, in comparison to other units operating un der similar conditions. We apply this approach in the evaluation of mu ltiple branches from a commercial bank in Latin America, and compare t he efficiency measures obtained from it with measures obtained from ot her methods, using the bank's central managers' classification of mark ets as a benchmark. Our results indicate that the clusterwise estimati on of translog cost functions leads to a more equitable assessment of the branches more in tune with the market differences perceived by the bank managers.