摘要：A method is developed to integrate the efficiency concepts of technical, allocative, and scale inefficiencies (TI, AI, SI) into the variable returns to scale (VRS) frontier approximation in Data Envelopment Analysis (DEA). The proposed weighted DEA (WDEA) approach takes a weighted average of the profit, constant returns to scale (CRS), and VRS frontiers, so that the technical feasibility of a VRS frontier is extended toward scale- and allocatively-efficient decisions. A weight selection rule is constructed based on the empirical performance of the VRS estimator via the local confidence interval of Kneip, Simar, and Wilson (2008). The resulting WDEA frontier is consistent and more efficient than the VRS frontier under the maintained properties of a data generating process. The potential estimation efficiency gain arises from exploiting sample correlations among TI, AI, and SI. Application to Maryland dairy production data finds that technical efficiency is on average 5.2% to 7.8% lower under the WDEA results than under the VRS counterparts.
