EURO 2024 Copenhagen
Abstract Submission

EURO-Online login

221. Stochastic multi-period and multi-stage efficiency measures

Invited abstract in session MC-48: DEA and stochastic models, stream Data Envelopment Analysis and its Application.

Monday, 12:30-14:00
Room: 60 (building: 324)

Authors (first author is the speaker)

1. Shiang-Tai Liu
Graduate School of Business & Management, Vanung University

Abstract

Many uncontrollable factors will affect company operations. In addition, the collected data is often snapshot observations, and if collected earlier or later, the values may differ. A company's inputs and outputs over time are thus stochastic variables. Several approaches have been proposed to deal with stochastic observations in efficiency measurement, including stochastic frontier analysis and chance-constrained programming. Conventionally, each DMU's input and output variables are assumed to be independent. This assumption is unrealistic because DEA requires isotonicity, which implies that the outputs correlate with the inputs. Ignoring this correlation will produce misleading results. In a network system, the merit of efficiency decomposition is that a relationship between the system and division efficiencies is obtained, based on which the division affecting the entire system's performance the most can be identified. When the observations are stochastic variables with multivariate normal distributions, they can be transformed into a variable with a standard normal distribution. In addition, only looking at a single time efficiency may lose the general picture of the overall performance for a multi-period. This paper aims to develop a model to measure the efficiency of DMUs with several divisions performing specific operations over a time period. This model is able to handle multi-period stochastic data and decompose the system's efficiency into division efficiencies.

Keywords

Status: accepted

Back to the list of papers