[Paper Review] Directional Congestion in Data Envelopment Analysis

Guo-liang Yang|arXiv (Cornell University)|Oct 25, 2015

Efficiency Analysis Using DEA31 references3 citations

TL;DR

This paper introduces directional congestion in Data Envelopment Analysis (DEA), proposing two estimation methods based on input and output directions to measure inefficiency. It establishes relationships with classic and strong/weak congestion, and demonstrates applicability through a case study of Chinese Academy of Sciences research institutes, showing improved efficiency evaluation precision.

ABSTRACT

First, this paper proposes the definition of directional congestion in certain input and output directions in the framework of data envelopment analysis (DEA). Second, two methods from different viewpoints are also proposed to estimate the directional congestion in a DEA framework. Third, we address the relations among directional congestion and classic congestion and strong (weak) congestion. Finally, we present a case study investigating the analysis of the research institutes in the Chinese Academy of Sciences (CAS) to demonstrate the applicability and usefulness of the methods developed in this paper.

Motivation & Objective

• To define directional congestion in DEA within specific input and output directions.
• To develop two distinct methods for estimating directional congestion from different analytical viewpoints.
• To clarify the theoretical relationships between directional congestion and established concepts like classic, strong, and weak congestion.
• To validate the proposed methods through a real-world application in Chinese research institutes.
• To enhance the precision and practical relevance of efficiency analysis in multi-input, multi-output production systems.

Proposed method

• Proposes a new definition of directional congestion based on input and output directions in the DEA framework.
• Develops two estimation methods: one based on input-oriented and another on output-oriented directional distance functions.
• Uses the directional distance function (DDF) to model how inputs can be reduced and outputs increased in specific directions.
• Applies the concept of congestion as a form of inefficiency where reducing certain inputs leads to disproportionate output increases.
• Integrates directional congestion into the standard DEA model by incorporating directional vectors into the efficiency measurement.
• Employs a case study using data from 11 research institutes under the Chinese Academy of Sciences to test the methodological framework.

Experimental results

Research questions

• RQ1 How can congestion be redefined in a directional manner within the DEA framework?
• RQ2 What are the two distinct methodological approaches to estimating directional congestion in DEA?
• RQ3 How does directional congestion relate to classic congestion and strong/weak congestion concepts?
• RQ4 In what ways does directional congestion improve the accuracy of efficiency evaluation in multi-input, multi-output settings?
• RQ5 What empirical evidence supports the practical utility of the proposed directional congestion model?

Key findings

• Directional congestion provides a more nuanced measure of inefficiency than traditional congestion by considering specific input and output directions.
• The two proposed methods offer complementary perspectives—input-oriented and output-oriented—for estimating directional congestion.
• Directional congestion is shown to be a generalization of classic congestion, with stronger theoretical consistency across different efficiency contexts.
• The case study of 11 Chinese Academy of Sciences institutes reveals that directional congestion identifies inefficiencies overlooked by standard DEA models.
• The results demonstrate that directional congestion enhances the discrimination power of DEA in real-world institutional efficiency analysis.
• The application confirms that directional congestion is both operationally feasible and practically valuable for policy and management decisions in research organizations.