GRIPS 政策研究センター Policy Research Center


Nov 28, 2011 Report No:11-15

Underestimation of Inefficiency in Social Efficiency Benchmarking with Non- Parametric Methods of Production Technology Identification: A Note

  • Hun Koo HaAsia Pacific School of Logistics, Inha University
  • Masashi YamamotoUniversity of Toyama
  • Yuichiro YoshidaGRIPS
  • Anming ZhangSauder School of Business, University of British Columbia
Field Economics
Language English

In the conventional social productive efficiency measurement, a DEA-based non-parametric method
is typically employed to identify the piece-wise-linear production possibility frontier. Applying the
directional distance-function approach a-la Luenberger (1992) to the production possibility frontier
obtained in this fashion can, however, lead to an underestimation of inefficiency for a DMU with
relatively large undesirable outputs. This underestimation becomes more acute if the sample size is
small or data are clustered. This paper reveals the mechanism behind this underestimation bias, and
then quantifies the degree of underestimation using nine-year panel data of rail and aviation sectors in
Japan. Through a comparative analysis between parametric and non-parametric methods, we find,
among others, that the underestimation of the aviation sector’s productive inefficiency is as large as
80%, which the non-parametric method failed to detect.

attachment 11-15.pdf