Measuring ecological efficiency with data envelopment analysis (DEA)
Introduction
`Organizations of all kinds are increasingly concerned to achieve and demonstrate sound environmental performance by controlling the impact of their activities, products or services on the environment [...].' (see introduction to ISO 14000). This quotation reflects the fact that environmental protection has become a ‘trendy’ topic. One reason for this development is that companies want to realise the idea of ‘sustainable development’. But more importantly, there are increasingly more benefits to be gained due to for example growing concern from the stakeholders, expanding environmental legislation, and cost savings.
An important instrument within the companies’ environmental management is the measurement of ecological efficiency (also: environmental performance measurement). Greatly varying definitions of ‘ecological efficiency’ are given in the literature, depending on the application context and subjective influences. The best-known definition is the one of the world business council for sustainable development (WBCSD): `Eco-efficiency is reached by the delivery of competitively-priced goods and services that satisfy human needs and bring quality of life, while progressively reducing environmental impacts and resource intensity throughout the life cycle, to a level at least in line with the earth’s estimated carrying capacity' (DeSimone and Popoff, 1997, p. 47). In order to measure ecological efficiency, it is necessary to operationalise such ‘global’ definitions in terms of environmental performance indicators (EPIs) and to draw a comparison of these indicators. ISO 14031 differentiates between operational performance indicators, management performance indicators, and environmental condition indicators. Ecological efficiency is usually measured by operational performance indicators which are based on material and energy balances. EPIs can be used for either target-performance comparisons, comparisons over time, internal or external eco-benchmarking. It should be noted that the measurement of ecological efficiency may include some typical ‘economic’ indicators.
There are several problems regarding comparisons of EPIs. The data base usually contains numerous, but non-standardised EPIs. Furthermore, even if companies have reliable and standardised environmental data, they might not publish it. Therefore, usually just a small number of companies are included in the comparison of EPIs. These deficits may be overcome in future by e.g. initiatives of the European Union regarding the standardisation and publicity of environmental data. As a technical problem, the literature does not address methods to support the comparison of EPIs. The difficulty is that weights are needed to aggregate the data into indicators and to aggregate those indicators into top-indicators, where the data and the indicators are extensive, complex, and measured on different scales. The derivation of weight coefficients (revealing prices, some measure of equivalence, trade-offs, or functional relationships) turns out to be very difficult. This is due to subjective and political influences and to lack of knowledge on the impacts of consuming natural resources and of emitting pollutants into nature.
Data envelopment analysis (DEA) is a well-established methodology for the assessment of performance of a homogeneous set of decision making units (DMUs), e.g. economies, bank branches, schools or hospitals, which are described by their input and output quantities. DEA shows a high potential regarding the aggregation problem mentioned before, as DEA does not require explicit weights. The main problem when using DEA in the ecological context is the fundamental assumption that for a DMU, to achieve efficiency, ceteris paribus larger quantities of outputs and smaller quantities of inputs are preferable. In only a small number of articles the underlying assumption, ‘all objects are goods’, is modified in order to consider undesirable outputs. Undesirable outputs are prominent in the ecological context (waste or emissions), but may also appear e.g. in health care (complications of medical operations) and business applications (tax payments). Most of the applications of and theoretical papers on DEA-models in the ecological context presented so far show some deficits, especially arbitrary approaches regarding the treatment of such ‘bads’. Up to now, the ecological extension of DEA does not have a theoretical foundation. Such a foundation would, amongst other things, reveal the ‘mirror’ case of desirable input, e.g. the destruction of waste in a waste-burning power plant or its transformation in a recycling plant.
To improve the state-of-the-art, this paper aims at a systematic approach to derive models for measuring ecological efficiency with DEA. A certain degree of generalisation allows to open DEA to objects other than ‘goods’, or to put it another way, for more complex preference structures. In order to generalise DEA, it is necessary to start from the axioms and foundations of DEA in production theory and activity analysis, as well as to consider other methods and findings, e.g. from decision theory and vector maximum theory.
The paper is organised as follows. Section 2 outlines the state-of-the-art of the ecological extension of DEA. Especially, an ecologically extended additive model is presented in detail and compared to alternative approaches. In Section 3, a systematic approach for deriving ecological DEA-models is developed. The main idea of this approach is to generalise a DEA-model by allowing for a more complex preference structure of the inputs and outputs by integrating a multi-dimensional value function f. For illustration, we present the generalisation of the additive model by Charnes et al. (1985) and two different specifications of f. The rest of the section discusses the theoretical and practical relevance of the approach presented. In Section 4 we summarise the findings of the paper and point out some ideas for future research.
Section snippets
Literature review
The difficulties of an ecological extension of DEA have already been outlined in Section 1: DEA measures the efficiency of DMUs described by their input and output quantities, where more output and less input ceteris paribus yield a higher degree of efficiency. In other words, DEA assumes that inputs and outputs are ‘goods’. This ‘preference structure’ or assumption is not valid any longer in the ecological context as illustrated above by examples for ‘bads’, e.g. ecologically undesirable
Assumptions
For a systematic generalisation of DEA, it is necessary to start from its axioms and foundations in production theory and activity analysis. DEA describes a DMU ρ with the empirically observed quantities of its inputs xρ=(x1ρ,…,xmρ)T and outputs yρ=(ym+1ρ,…,ym+nρ)T. Each DMU or input/output-vector represents a (production-) activity in the sense of Koopmans’ (1951) activity analysis.
The generalisation takes its starting point from seven assumptions underlying common DEA-models. In the
Conclusions
The original motivation for this paper was to enable an ecological extension of DEA, constituting a framework for applications of DEA in the ecological context. A generalised DEA model was developed using the classical DEA assumptions as a starting point. The generalisation is predominantly related to the underlying preference structure but also to the integration of user-specified weights into the conventional model. A multi-dimensional value function f was incorporated in an additive model
Acknowledgements
Financial support by the Deutsche Forschungsgemeinschaft (DFG) is gratefully acknowledged. The authors are indebted to two anonymous referees for helpful comments on an earlier draft of this paper. KA wants to expressly thank the DEA working group at Warwick Business School, University of Warwick, England, for two fruitful visits and the great organisation of the EURO Summer Institute “DEA – 20 years on”.
References (35)
- et al.
Estimation of returns to scale using data envelopment analysis
European Journal of Operational Research
(1992) - et al.
Measuring the efficiency of decision making units
European Journal of Operational Research
(1978) - et al.
Foundations of data envelopment analysis for Pareto–Koopmans efficient empirical production functions
Journal of Econometrics
(1985) - et al.
An activity analysis model of the environmental performance of firms – application to fossil-fuel-fired electric utilities
Ecological Economics
(1996) - et al.
An enhanced DEA Russell graph efficiency measure
European Journal of Operational Research
(1999) DEA in the ecological context – An overview
- et al.
Weights restrictions and value judgments in data envelopment analysis − Evolution, development and future directions
Annals of Operational Research
(1997) - et al.
Some models for estimating technical and scale inefficiencies in data envelopment analysis
Management Science
(1984) DEA on relaxed convexity assumptions
Management Science
(1996)- et al.
DEA – Theory, Methodology and Applications
(1994)