Composite Indicators Construction by Data Envelopment Analysis: Methodological Background

Composite Indicators Construction by Data Envelopment Analysis: Methodological Background

Gordana Savić (University of Belgrade, Serbia) and Milan Martić (University of Belgrade, Serbia)
DOI: 10.4018/978-1-5225-0714-7.ch005
OnDemand PDF Download:
$30.00
List Price: $37.50

Abstract

Composite indicators (CIs) are seen as an aggregation of a set of sub-indicators for measuring multi-dimensional concepts that cannot be captured by a single indicator (OECD, 2008). The indicators of development in different areas are also constructed by aggregating several sub-indicators. Consequently, the construction of CIs includes weighting and aggregation of individual performance indicators. These steps in CI construction are challenging issues as the final results are significantly affected by the method used in aggregation. The main question is whether and how to weigh individual performance indicators. Verifiable information regarding the true weights is typically unavailable. In practice, subjective expert opinions are usually used to derive weights, which can lead to disagreements (Hatefi & Torabi, 2010). The disagreement can appear when the experts from different areas are included in a poll since they can value criteria differently in accordance with their expertise. Therefore, a proper methodology of the derivation of weights and construction of composite indicators should be employed. From the operations research standpoint, the data envelopment analysis (DEA) and the multiple criteria decision analysis (MCDA) are proper methods for the construction of composite indicators (Zhou & Ang, 2009; Zhou, Ang, & Zhou, 2010). All methods combine the sub-indicators according to their weights, except that the MCDA methods usually require a priori determination of weights, while the DEA determines the weights a posteriori, as a result of model solving. This chapter addresses the DEA as a non-parametric technique, introduced by Charnes, Cooper, and Rhodes (1978), for efficiency measurement of different non-profitable and profitable units. It is lately adopted as an appropriate method for the CI construction due to its several features (Shen, Ruan, Hermans, Brijs, Wets, & Vanhoof, 2011). Firstly, individual performance indicators are combined without a priori determination of weights, and secondly, each unit under observation is assessed taking into consideration the performance of all other units, which is known as the ‘benefit of the doubt' (BOD) approach (Cherchye, Moesen, Rogge, & van Puyenbroeck, 2007). The methodological and theoretical aspects and the flaws of the DEA application for the construction of CIs will be discussed in this chapter, starting with the issues related to the application procedure, followed by the issues of real data availability, introducing value judgments, qualitative data, and non-desirable performance indicators. The procedure of a DEA-based CI construction will be illustrated by the case of ranking of different regions of Serbia based on their socio-economic development.
Chapter Preview
Top

Background

Composite indicators (CIs) represent an aggregation of a set of sub-indicators for measuring multi-dimensional concepts that cannot be captured by a single indicator e.g. competitiveness, sustainability, single market integration, etc. (OECD, 2008). Dobrota, Savic, and Bulajic (2015) gave the example of using 12 single indicators for evaluating the European countries’ educational structure development is given in the paper. The main focus of that paper is on the selection of indicators based on the defined research goals and available data, as well as the choice of an appropriate aggregation method. Thus, the construction of CIs includes several steps: defining a theoretical framework for creating a composite indicator, followed by data and individual performance indicators selection, the selection and imputation of missing data, data normalization, weighting and aggregation (OECD, 2008). The controversial issues can appear in any step of CI construction since results are significantly affected by the method used in data normalization, weighting, or aggregation. A decision maker is faced with a challenge to choose the right metrics for normalization (Cherchye, Moesen, Rogge, & van Puyenbroeck, 2007), the right weighting scheme (fixed or equal weighting), and the right method of aggregation (Munda, 2012).

Complete Chapter List

Search this Book:
Reset