Chance-constrained cost efficiency in data envelopment analysis model with random inputs and outputs

Rashed Khanjani Shiraz, Adel Hatami-marbini, Ali Emrouznejad, Hirofumi Fukuyama

Research output: Contribution to journalArticlepeer-review


Data envelopment analysis (DEA) is a well-known non-parametric technique primarily used to estimate radial efficiency under a set of mild assumptions regarding the production possibility set and the production function. The technical efficiency measure can be complemented with a consistent radial metrics for cost, revenue and profit efficiency in DEA, but only for the setting with known input and output prices. In many real applications of performance measurement, such as the evaluation of utilities, banks and supply chain operations, the input and/or output data are often stochastic and linked to exogenous random variables. It is known from standard results in stochastic programming that rankings of stochastic functions are biased if expected values are used for key parameters. In this paper, we propose economic efficiency measures for stochastic data with known input and output prices. We transform the stochastic economic efficiency models into a deterministic equivalent non-linear form that can be simplified to a deterministic programming with quadratic constraints. An application for a cost minimizing planning problem of a state government in the US is presented to illustrate the applicability of the proposed framework.
Original languageEnglish
Pages (from-to)1863–1898
Number of pages36
JournalOperational Research
Early online date20 Feb 2018
Publication statusPublished - Sept 2020

Bibliographical note

© 2018 Springer Publishing. This is a post-peer-review, pre-copyedit version of an article published in Operational Research. The final authenticated version is available online at:


  • Data envelopment analysis Weight restrictions Random input–output Cost efficiency Quadratic programming


Dive into the research topics of 'Chance-constrained cost efficiency in data envelopment analysis model with random inputs and outputs'. Together they form a unique fingerprint.

Cite this