Callens Tyteca 1999 - Towards Indicators of Sustainable Development for Firms

Ecological Economics 28 (1999) 41 – 53

METHODS

Towards indicators of sustainable development for firms A productive efficiency perspective Isabelle Callens, Daniel Tyteca * Centre Entreprise En6ironnement, Institut d’Administration et de Gestion, Uni6ersite´ Catholique de Lou6ain, Place des Doyens, 1, B-1348 Lou6ain-la-Neu6e, Belgium Received 7 July 1997; received in revised form 25 February 1998; accepted 3 March 1998

Abstract Sustainable development can be reflected by various economic, social and environmental factors that are closely interconnected with each other, and with the additional dimension of time, which stresses the long-term perspective of several factors. Due to their central role in human activities and development, firms should play an important part in the attainment of sustainability goals. The purpose of this paper is to contribute to the methodology of indicators that allow for the assessment of business participation into sustainable development. A fundamental standpoint adopted is to view economic, social and environmental efficiency as a necessary (but not sufficient) step towards sustainability. To work out indicators, we build on both the concepts of cost – benefit analysis and the principles of productive efficiency. We assume that we have observations on economic, social and environmental factors for a set of decision making units (DMUs), e.g. firms in an industrial (sub-) sector. The efficiency of each DMU is computed from a set of observed data, using mathematical programming techniques, resulting in DMUs that are ‘efficient’ and define the efficiency frontier among the set of DMUs, and DMUs that are ‘inefficient’. To cope with the multidimensionality of sustainable development, it is important not to base decisions on one unique, aggregate sustainability indicator; instead, it is suggested to develop two or three partial indicators that stress different aspects of the problem. The proposed indicators could be used as an aid to detect so-called factors of unsustainability, and hence to provide for recommendations as to the regulations and incentives, or managerial practices, that will contribute to overall sustainability. © 1999 Elsevier Science B.V. All rights reserved. Keywords: Sustainable development; Indicators; Measurement; Productive efficiency; Business

* Corresponding author. Tel.: + 32 10 478368/478375; fax: + 32 10 478324; e-mail: [email protected]@qant.ucl.ac.be 0921-8009/99/$ - see front matter © 1999 Elsevier Science B.V. All rights reserved. PII S0921-8009(98)00035-4

42

I. Callens, D. Tyteca / Ecological Economics 28 (1999) 41–53

1. Introduction In the last few decades, environmental problems faced by our modern society have given rise to a considerable wealth of research. More recently, environmental preoccupations appeared to become part of a more global and fundamental context, that of sustainable development. The culminating point of that new awareness was reached at the Earth Summit in Rio in June 1992. Many nations have subscribed since then to the principles of sustainable development. Its most widespread, general (and vague) definition had the merit to create quasi-unanimity: it states that sustainable development is a ‘‘development that meets the needs of the present generation without compromising the ability of future generations to meet their own needs’’ (World Commission on Environment and Development, 1987). Such a broad definition is likely to give rise to various different interpretations, since people all have different goals and sensitivities and will generally not agree on what to sustain (Norrthon, 1995). Thus we must be very careful in defining the goals and options of sustainability. Moreover, the so-called ‘‘goals’’ of sustainability cannot be unique and defined once and for all; rather, sustainability should be viewed as a dynamic process in which the targets have to be continuously checked and improved, or as a philosophy that permanently tends towards improvements. According to Welford (1995), ‘‘sustainable development is made up of three closely connected issues and each one of these needs to be addressed by industry. Firstly, the environment must be valued as an integral part of the economic process and not treated as a free good’’. This implies ‘‘minimal use of non-renewable resources and minimal emission of pollutants’’, as well as protection of ecosystems in order to avoid the loss of plant and animal species. ‘‘Secondly there is a need to deal with the issue of equity’’. Equity ‘‘applies not only to relationships between the First and Third Worlds, but also within countries between people’’, which, e.g. implies the reduction of unemployment. ‘‘Thirdly, sustainable development requires that society, businesses and individuals operate on a different time scale than

currently operates in the economy. This is the issue of futurity’’. This implies longer term, intergenerational considerations; therefore, ‘‘longer planning horizons need to be adopted and business policy needs to be proactive rather than reactive’’ (Welford, 1995). Thus society has to formulate sustainability goals, incorporating all three aforementioned issues, and these goals have to be permanently checked and improved. The extent to which given goals are met can be measured by activity or performance indicators. The information obtained from indicators is especially valuable in the assessment of tools such as taxes, regulations, or voluntary agreements, as to their validity and effectiveness towards meeting sustainability objectives. For example, the information obtained from life cycle analysis or life cycle assessment of products (SETAC, 1992) can be reflected in the form of price signals to reflect their whole environmental or societal costs (Portney, 1993, 1994). As another example, at the national level, environmental laws and regulations can be improved by using the information obtained from such indices as NNP (net national product, i.e. GNP corrected to account for degradation of natural capital). Indicators can also be used to compare similar existing units (products, technologies, plants, firms, sectors and countries) in terms of their performance with respect to some specified targets (economic, environmental, etc.). In this case their utility is to identify ‘laggers’ among the units considered as well as the causes of lagging, and to adopt relevant corrective actions. More generally, indicators can contribute to discover so-called factors of unsustainability (Callens and Wolters, 1998) and to give recommendations as to means to reduce their influence. Conversely, indicators can be used to select the best among a set of possible alternatives still to be implemented. As an example of the latter, one could think of environmental impact assessment methods (Devuyst, 1993). Other uses of indicators include monitoring the performance and progress of given units (products, firms, etc., see above) over time. Until now, sustainability indicators have been developed mainly at the most global level, i.e. the state or country level (Daly and Cobb Jr., 1989;


Pearce and Atkinson, 1992, 1993; Pearce, 1994). Due to their central role in human activities and development, firms should play an important part in the attainment of sustainability goals. A recent methodology, based on so-called sustainable development records (SDR), was developed for the firm or plant level (Bergstro¨m, 1993; Block and Ho¨gstro¨m, 1995). That approach is primarily devoted to the appraisal of a given firm’s contribution to sustainability. Ragas et al. (1995) also proposed to develop sustainability indicators for production systems. They start from the concept of environmental space, which would allow the measurement of sustainability, through the definition of an adequate distribution scheme of the various components of environmental space among production systems. The latter approach can be viewed as a ‘topdown’ definition of sustainability indicators, i.e. the measurement is performed with respect to some global, predetermined level of sustainability. In this paper we have a different, somewhat complementary ‘bottom-up’ focus, that starts from the production systems themselves. We set out to compare various firms or plants in a given set, in terms of their performance as regards sustainability, in order to be able to detect which ones of them are the laggers, what are the reasons for lagging (e.g. which are the factors of unsustainability involved (Callens and Wolters, 1998)) and therefore what are the possibilities towards improvement. In this scheme, efficiency with respect to economic, social and environmental resources is viewed as a necessary (but not sufficient) step towards sustainability. This is discussed in the following section.

2. Economic, social and environmental efficiency as a necessary step towards sustainable development Before conceptualizing and implementing sustainability indicators for firms, we would have to raise the following question: what is a sustainable enterprise? However, the answer to such a question in absolute terms appears impossible, because it depends on so many factors that are hardly

43

controllable. For example, the geographic and socio-demographic environment in which a plant is located will play a determinant role. Thus, a pulp and paper plant isolated in the Canadian forest, making extensive use of forest products, and discharging wastes into a large river, may turn out to be sustainable because ecosystems are only locally affected and long-term equilibria are not affected. On the other hand, the same plant, using the same resources and with the same polluting activities, but located in a heavily populated area, is not sustainable because it would deplete natural resources (forest, water) and have negative impacts on the communities living in the neighbourhood. Therefore we must look for a definition in relative terms. As mentioned in the introduction, we might refer to specific targets stated for resource uses and external impacts, e.g. with reference to the concept of environmental space. Or we might compare similar production units placed in similar contexts, to detect whether some of them behave in a more appropriate way than others, with respect to specific economic, social or environmental characteristics. Then the initial question can be more appropriately stated as ‘in what way would a given enterprise be more or less sustainable than another’? If we admit that this is a valid question, the next step will be to define the kind of information that should appear in the sustainability indicators in order to allow for such comparisons. The discussion conducted so far implies that we would hardly define any indicator based on sufficient conditions for sustainable development. Instead, we start from investigating a set of necessary conditions that firms must fulfill in order to be sustainable. Necessary conditions are viewed as being efficient in the use of resources, in the pollutants released to the environment, in the social role played by firms as reflected by their rate of employment, the working conditions, and in the care taken with respect to future generations in the setting of long-term objectives. Since there are no benchmarks, for any of these characteristics, that would indicate from which level firms could be declared sustainable, the focus is on comparing the firms between themselves. Some are more efficient, or inefficient, in all respects;


44

most often, however, they might be efficient in specific respects but inefficient in others. Therefore we need a methodology to take appropriate tradeoffs. This is what is developed in the paper. Table 1 indicates a set of possible factors for which information would be required in the framework of efficiency. This list is not meant to be comprehensive and is provided here only as an example. It is likely to vary across the various industrial sectors. The factors considered embrace various kinds of quantities; for example, some of them are representatives of stocks, other ones stand for flows, while a third part of them would reflect transformations. One other categorization would distinguish between factors that are viewed as emissions (e.g. pollutants that are present in a given firm’s effluents), and others that can be characterized as immissions (i.e. wastes that are actually discarded to the environment, whose influence and quantification depend on the local characteristics of the receiving bodies). The concept of immission does not only apply to environmental, physical characteristics, but can be adapted and extended as well to socio-economic components. For example, locating a firm in a given region may induce employment not only in the firm itself, but also indirectly in services or administrations in charge of controlling the ecological or economic activities of the firm, or in the transportation sector, etc. Thus we are not trying to specify rigid boundaries to the systems whose sustainability is being tested. Instead, the emphasis herein is on the methodology of the approach, whatever the investigated unit may be. When dealing with a specific application,

however, setting the boundaries is fundamental: then we would have to specify exactly which system is under investigation (e.g. a plant with its inputs/ outputs, a plant or a company with its raw material uses and its environmental impacts and immissions). The list of factors to be accounted for, among those indicated in Table 1 as an example, will depend upon the specification of boundaries. It can be seen in Table 1 that some of the factors considered reflect sustainability goals for the firm considered (mainly economic goals, which underlie its ability to survive), while others (i.e. socio-economic and ecological factors) mainly reflect sustainability goals for society. It is also apparent that some of the items cover more than one category of aspects. For example, the use of natural (renewable or non-renewable) resources clearly has economic incidences, ecological impacts, but also social impacts, related to resource availability for future generations. Some items might have contradictory influences depending on the aspects considered. For example, employment would tend to be minimized from a business point of view but should rather be maximized from a social point of view. Also, the intervention of certain factors might evolve with the level of scientific knowledge, which will influence the way in which they should be accounted for in the elaboration of indicators. As an example, one may think of the CFCs, once considered a useful, inert auxiliary in various productions, but now identified as one of the major pollutants contributing to the ozone layer depletion.

Table 1 Sample list of information required in the development of sustainability indicators at the firm level

Economic aspects Social aspects

Ecological aspects

Short term

Long term

Turnover, value added, output production, resources used as inputs (including recycled products and energy) Employment, salaries, labor intensiveness or productivity, injury risk noise, odour

Profitability, competitiveness, market shares, product durability; research and development efforts

Natural resources, wastes, pollution, transportation modes and distances

Welfare, education, availability of (non-) renewable, resources (including energy), size (SME vs big), personnel rotation rate Global impacts: biodiversity, global warming, acid deposition, landscape, ultimate waste disposal, product recycling ability


Dealing with sustainability, some authors (Ashford and Meima, 1993; Welford, 1995) also insist on the importance of such factors as the state of technology, information structures, societal values and culture. These are hard factors to measure directly but they can be captured in an indirect way in some of the items listed in Table 1. For example, it might be important to satisfy cultural preferences for small, decentralized centers of production, ‘‘using local resources and good technology to increase productivity to make products and provide services that satisfy the fundamental needs of local people without destroying the environment’’ (Ashford and Meima, 1993). Such cultural or social preoccupations can be reflected by variables such as employment, labor intensiveness, and plant size (Table 1) which should therefore be given the adequate orientation (i.e. the emphasis should be more on small size, labor intensive plants in order to achieve sustainability).

3. Sustainability indicators based on efficiency standpoints Now we come to the difficult question of how to account for, and aggregate, such diversified information as listed in Table 1, provided that it is available. Data on the factors listed in Table 1 can be exploited in two possible ways. First, we could use past observations on the values of these factors; the indicator would then be a static tool that would help us to identify efficient and less efficient units among a given set. Second, we could use projected values that would reflect a set of possible alternatives; in this case the indicators would be a prospective tool. In both cases, we have to assume that we have observations on the factors for a set of (existing or possible) decision making units (DMUs), i.e. plants in a firm or firms in an industrial (sub-) sector. Indeed, to be operational, the indicator should compare units that are really comparable, in the sense that they produce similar products that are designed to fulfill analogous usages. We also assume, without loss of generality, that the available data cover one time period (1 year). Let us designate the quantities enumerated in

45

Table 2 Notation used in the definition of sustainability indicators Variables to maximize Economic aspects ECONmax,i i= 1,…, I1 Social aspects SOCmax,i i= 1,…, I3 ENVmax,i i= Environmental aspects 1,…, I5

Variables to minimize ECONmin,i i= 1,…, I2 SOCmin,i i= 1,…, I4 ENVmin,i i= 1,…, I6

Table 1 by the symbols defined in Table 2. For each of the three classes of aspects, i.e. economic, social and environmental, we consider factors whose value should be minimized (all other things being equal) and those whose value should be maximized, in order to reach efficiency, or more generally, in the perspective of sustainability. For example, natural resources used should clearly be minimized, from an economic as well as ecological standpoint, while durability is a factor we would tend to maximize in a sustainability framework. Note that in the following, we will use the terms ‘factors’, ‘variables’, ‘quantities’ and sometimes ‘aspects’ as synonyms. First, if we consider the definition of indicators in a strict cost–benefit framework, the problem would be to compare the DMUs with respect to the values taken by the following kind of quantity: I1

IndicatorCB = % ai ECONmax,i i=1 I2

− % bi ECONmin,i i=1 I3

I4

i=1

i=1

I5

I6

i=1

i=1

+ % gi SOCmax,i − % di SOCmin,i + % oi ENVmax,i − % zi ENVmin,i (1) in which the (positive) coefficients ai, bi, gi, di, oi and zi represent either the prices of marketable economic quantities (i.e. inputs, salaries, etc.) or shadow prices of non-marketable quantities (i.e.


46

pollution, productivity, etc.). It is supposed that each of the quantities appearing in Eq. (1) has been scaled by a quantity reflecting the activity of the DMU. Most relevantly, this would be some measure of the output production. This is necessary to give meaning to the comparisons. The main drawback of a cost – benefit definition such as in Eq. (1) is the quantification of the coefficients, especially those for which no market prices exist. This is indeed a much controversial topic and many would argue that not even shadow prices would exist for such critical factors as biodiversity, or even more simply, existence of a given living species. Therefore in the following we will consider approaches that do not use price coefficients, but instead, weights or intensities, that reflect the importance given to the various factors. Moreover, we consider an indicator defined as a ratio between a weighted sum of quantities that are considered desirable, to a weighted sum of quantities that are viewed as inputs and whose intervention has to be minimized: IndicatorIO = I1

I3

I5

i=1 I2

i=1 I4

i=1 I6

i=1

i=1

i=1

% ai ECONmax,i + % gi SOCmax,i + % oi ENVmax,i % bi ECONmin,i + % di SOCmin,i + % zi ENVmin,i

puts’ (Fa¨re, 1992; Klein and Yaisawarng, 1993; Nestor and Pasurka, 1993; Ball et al., 1994; Fa¨re et al., 1989, 1996; Tyteca, 1996). The basic standpoint of productive efficiency, as applied to environmental performance measurement, is to compare a set of decision making units (DMUs) between themselves, in terms of their environmental or sustainability performance. The comparison is usually restricted to similar units, e.g. plants or firms in a given industrial (sub-) sector, but can be extended to different geographic regions and to different periods of time, which allows, e.g. for the assessment of performance improvements over time. The technique is also known as data envelopment analysis (DEA). Outlined below are three examples of DEA models that consider the sustainability problem under three different perspectives.

3.1. Model 0: generalization of the input–output framework Consider the situation depicted in Fig. 1. This illustrates the simplified case where we consider, in Eq. (2), only one numerator quantity Qmax (i.e. a quantity to be maximized, an output) versus only one denominator quantity Qmin (an input). In Fig. 1, DMU A is efficient because there is no other DMU that uses less of Qmin to produce

(2) in which the subscript IO stands for input – output, and ai, bi, gi, di, oi and zi denote coefficients that represent a priori weights given to the economic, social and environmental factors. These weights should reflect hierarchies and/or priorities in the opinion of the decision makers and may therefore also considerably suffer from a high degree of subjectivity. The latter would also be a serious limitation if we were to generalize the use of so-called ‘scientific’ environmental performance indicators in which the factors are given arbitrary weights (Jaggi and Freedman, 1992; Wehrmeyer, 1993; Tyteca, 1996). As a last proposal for defining indicators of sustainable development, we could generalize the idea of extending the theory of productive efficiency to the consideration of environmental factors, in which the wastes were taken as ‘undesirable out-

Fig. 1. Two dimensional representation of a set of observed decision making units that use quantities Qmin of an input to produce quantities Qmax of an output. The straight lines represent the production or efficiency frontier, under two different assumptions of returns to scale.


more of Qmax, while this is not the case for DMU B, which will be declared inefficient. We translate this situation by giving an efficiency score 1 to DMU A, while the score associated with DMU B will be less than 1. In mathematical terms, generalizing to situations where we have more than one output and one input, we simply consider the framework reflected by Eq. (2) and modify it so that no a priori value is given to the six sets of weighting coefficients. Instead, we consider these as the variables of a mathematical programming scheme and compare one given DMU, indexed ‘0’, to the whole set of DMUs for which data are available. In this way, we offer the possibility of incorporating various different kinds of quantities, with different measurement units (i.e. physical, economic, etc.) and meanings (i.e. stocks, flows, transformations, etc.), that have to be aggregated to reflect their distance to efficiency, without requiring any assumption on the weights used in the aggregation. The weights are the solution of a mathematical program and will be computed in such a way that the distance to the efficiency frontier is minimized. The general formulation of a DEA model that starts from Eq. (2) is the following (see the aforementioned references for details of formulation): maxh0 = I1

I3

I5

i=1 I2

i=1 I4

i=1 I6

i=1

i=1

I1

I3

I5

i=1 I2

i=1 I4

i=1

% ai ECON0max,i + % ci SOC0max,i + % ei ENV0max,i % bi ECON0min,i + % di SOC0min,i + % fi ENV0min,i

i=1

s.t. % ai ECONkmax,i + % ci SOCkmax,i + % ei ENVkmax,i I6

% bi ECONkmin,i + % di SOCkmin,i + % fi ENVkmin,i

i=1

51

i=1

k = 1,..., K

i=1

(3)

ai,bi,ci,di,ei,fi ] 0 Index k designates a DMU in the set of K DMUs for investigation; the constraints define the feasible set, i.e. the set of combinations of the factors taken into account that are feasible in the present state

47

of technology; the frontier of that set is constituted by the DMUs that exhibit the best practice in terms of sustainability. The objective function in Model 0 indicates the problem of DMU ‘0’, i.e. given that the efficiency of all DMUs is set to be smaller than or equal to 1 by the constraints, the efficiency of DMU ‘0’ is either equal to or smaller than 1, in which cases it is considered ‘efficient’ or ‘inefficient’, respectively. The meaning of efficient and inefficient in this context might be taken as ‘sustainable’ and ‘unsustainable’, under the assumption that Model 0 appropriately reflects sustainability. It should be stressed, however, that a DMU that would be declared efficient because its index value is one, is not necessarily sustainable, since it could appear to be efficient because it would be better than all other DMUs on only one factor. Thus it is possible to refer to DMUs that are unsustainable because they are inefficient, but efficient by no means implies sustainable, the former being only a necessary condition for the latter. Among the widely recognized advantages of DEA models are: (1) a clear and obvious standardization, since all units are ranked according to a scale with the convention that the value 1 represents best performance or best practice; (2) an important flexibility, since various versions of DEA models, stressing important aspects in different ways, can be formulated easily; and (3) the robustness of the (non-) linear programming methods used to compute the indicators. Additionally, the fact that no a priori weight has to be given to the factors taken into account is often presented as a main advantage of DEA formulations. Indeed, the weights are self-defined as the variables of the (non-) linear programs used to compute efficiency scores. In that way, some kind of absolute objectivity is associated with DEA methods, because at the outset no judgment of any kind from any person is required. There is at least one definite advantage of such methods, i.e. they provide effective ways to detect DMUs that are inefficient even under the most favourable weight combination, because for every DMU under consideration, DEA methods will actually select the weight combination that minimizes the distance from the DMU to the frontier, which is therefore the most favourable combination within the space of all weight possibilities.


48

This, however, is open to discussion, since in extreme cases, due to the possibility of having zero weights, and therefore all the weight concentrated on one unique output or input, such procedures can, and actually will, result in DMUs that are declared globally efficient even if they are efficient regarding only one of the specified variables. Moreover, DEA detractors argue, why would selfdefined, artificial weights be less arbitrary than any a priori combination? To accommodate for such a drawback, methods exist, that allow one to incorporate judgments of experts or analysts of the situation under concern (Wong and Beasley, 1990). With relation to the methods used in DEA, such judgments can take the form of intervals within which the relative weights given to some production factors should be included. For example, considering the economic inputs, whose value should be minimized, i.e. in Eq. (3), possible weight intervals can take the form of a ki 5

biECONkmin,i I2

5b ki

(4)

% bi ECONkmin,i i=1

which states that, for DMU k, the importance given to input ‘i’ with respect to the whole set of inputs be comprised between a ki and b ki . This implies that zero weights, as well as the possibility of giving the totality of the weight to the sole input i, can be avoided. Other combinations and standpoints can be reflected as well. For example, if for some engineering reason, the relative weight of an undesirable output (e.g. atmospheric pollutant) is associated with the use of a given production input (e.g. fossil fuel), this can also be reflected by a constraint on the ratio between appropriate quantities as in I4

I3

Eq. (4). More generally, engineering standpoints, or managerial perceptions or targets, can be translated using the notion of ‘standards’. Accordingly, the strict endogeneous frontier can be substituted by a standard frontier (Golany and Roll, 1994; Tyteca, 1996). Finally, one should also be aware of one other potential drawback of DEA, i.e. the high sensitivity of the results with respect to the number of factors and units considered; one should therefore be aware that a given result can only be considered with reference to the associated data set. However, this is no longer a drawback if we recall that best practice, or best available technology, is always a relative concept that heavily depends on what actually exists.

3.2. Model 1: social and en6ironmental factors as the main factors for scrutiny Model 0 as stated above only mimics the usual framework of data envelopment analysis, i.e. a situation in which we consider, on one side, the quantities that are considered desirable (those appearing at the numerator), versus the quantities that are viewed as inputs and whose intervention has to be minimized (the denominator). In a further, and more realistic step, we might consider three categories of factors, i.e. the inputs, the outputs, and the undesirable outputs. In this scope, there are various ways in which the problem can be viewed (Tyteca, 1996, 1997), and we first consider one in which we try to minimize negative ecological and social impacts that are opposed to the necessity of producing economic outputs using adequate quantities of inputs (which is a necessary condition for a firm to survive):

I6

I5

i=1 I2

i=1

% di SOC0min,i − % ci SOC0max,i + % fi ENV0min,i − % ei ENV0max,i min h0 =

i=1

i=1 I1

% ai ECON0max,i − % bi ECON0min,i I4

I3

i=1

I6

i=1

I5

% di SOCkmin,i − % ci SOC kmax,i + % fi ENVkmin,i − % ei ENVkmax,i s.t.

i=1

i=1 I1

% ai ECON ai,bi,ci,di,ei,fi ] 0

i=1

i=1 I2

k max,i

− % bi ECON i=1

i=1 k min,i

]1

k= 1,..., K

(5)


Fig. 2. Production set and efficiency frontier for a set of observed decision making units, based on one environmental impact and one social impact. The values of a sustainability indicator are illustrated for observed point B, for the case where a reduction in both impacts is sought (u, reference point B1, Model 1) and the case where only a reduction in the environmental impact is sought (u%, reference point B2, Model 2).

The sign restrictions on the variables indicate that all (economic, social and environmental)

I6

I5

i=1 I2

i=1

I1

i=1

i=1

49

assumption is sometimes released, especially in the case of pollutants: then these are considered only weakly disposable, meaning that beyond a given threshold, a given pollution discharge can be reduced only at the cost of a reduction of the output production or some other variables (Fa¨re et al., 1989; Fa¨re, 1992; Tyteca, 1997). One additional set of constraints could be added to the model, stating that the denominators should be positive, in order to insure that the conditions for DMUs’ survival are met. Fig. 2 illustrates the simplified case where we have one environmental impact and one social impact, the reduction of which would correspond to improved sustainability (e.g. a pollutant release, and an accident record). With Model 1, because both impacts must be reduced, a move towards south-west will yield an improvement, which is feasible until the point reaches the frontier. If we consider, without loss of generality, that the reduction in both impacts is proportional, this will lead us to point B1, where impacts have been reduced by a factor u. The latter can be taken as the indicator value; it corresponds to the inverse of the value of h0 obtained after solving Model 1 as stated in Eq. (5) (Tyteca, 1996, 1997).

3.3. Model 2: en6ironmental factors as the main factors for scrutiny

% fi ENV0min,i − % ei ENV0max,i min h0 =

I3

I4

i=1

i=1

% ai ECON0max,i − % bi ECON0min,i − % ci SOC0max,i + % di SOC0min,i I6

I5

i=1 I2

i=1

I1

i=1

i=1

% fi ENVkmin,i − % ei ENVkmax,i s.t.

I3

I4

i=1

i=1

]1

k= 1,..., K

(6)

% ai ECONkmax,i − % bi ECONkmin,i − % ci SOCkmax,i + % di SOCkmin,i ai,bi,ci,di,ei,fi ]0 parameters are considered strongly disposable (i.e. we can decrease output production down to ‘0’ at no cost, and we can substitute inputs among themselves, or decrease pollution, provided additional levels of appropriate inputs are used). This

In this case, the emphasis is on minimizing the environmental impacts of the DMUs under consideration. The sign conventions at the denominator of the objective function and the constraints are based on the fact that we now consider the social factors as making part of the production

50


factors (inputs) (Tyteca, 1997). In this case, in Fig. 2, a reduction has been sought only in the direction of environmental impacts, and the point reached from point B is point B2, resulting in an indicator value of u%. A similar model could be stated with social factors as the main factors to scrutinize. Thus what is illustrated here is that we do not look for one unique indicator value, but that instead, the methodology outlined in this paper allows us, using the same basic information, to specify as many indicators as required by the specific application considered. The choice among the indicators depends upon political or managerial decisions.

4. Discussion and conclusions Application of the models of Section 3, with only a subset of variables reflecting environmental and economic quantities, was performed using small data sets for a few case studies summarized in Table 3. In these situations we speak of environmental performance indicators. At least the results obtained showed that an analysis such as proposed herein is feasible. Now, to apply more generally the ideas developed in this paper, we have to generalize the environmental performance indicators to much wider situations, i.e. situations in which there are many more variables to account for. The first challenge is to define (as discussed in Section 2) and collect appropriate data that will provide the basis for the new indicators to be built. An important aspect of any kind of indicator is

the necessity to take account of as many relevant characteristics as possible (Gramlich, 1990). For example, a regulation that stresses some peculiar components of the pollutant charge while ignoring others might result in a substitution between pollutants and therefore in an overall increase of the global pollutant charge (Merilaïnen, 1995). On the other hand, taking many factors into account might suffer from both computational intractabilities and data unavailability. Moreover, from a strict technical point of view, accounting for many variables in data envelopment analysis might result in the extreme situation where all firms in a given set are considered efficient. Therefore there must be some kind of tradeoff between the number and the representativity of characteristics accounted for, which can be solved by some preliminary investigation of the data using, e.g. principal component analysis. The latter kind of problem might also find a partial solution if, instead of trying to develop one unique aggregate sustainability indicator (such as, e.g. in Model 1 above), we define two or three partial indicators that stress different aspects of the problem. For example, we would consider two indicators, one centered on environmental preoccupations (such as, e.g. in Model 2 above), the other stressing social preoccupations (as commented under Model 2). In this way we could gain accuracy in the description of the situation, while providing the decision maker with the possibility of meaningful tradeoffs (the priority given to social or environmental preoccupations is left to the public decider).

Table 3 Examples of applications of DEA methods to the measurement of environmental performance and sustainable development Case

Factors accounted for

Reference

Chemical plants Fossil fuel-fired electric utilities Pulp and paper

One output and two pollutants (to air and water) treated as inputs Output: kilowatt-hour production; inputs: labour, fossil fuels, capital; pollutants: SO2, NOx, CO2 Output: pulp production; inputs: labour, energy, fiber, water; pollutants: particulates, TRS, SO2 BOD, TSS, AOX Output: waste diversion rate; input differential cost of recovery; undesirable output: waste residue ratio

Haynes et al. (1994) Fa¨re et al. (1996), Tyteca (1997, 1998) Callens and van den Berghe (1997) Courcelle et al. (1998)

Municipal waste management


A possible disadvantage of DEA methods may be that the way in which the weight factors are determined is not straightforward for the non-specialist, so that decision makers may be guided by parameters they do not fully understand. To cope with this difficulty, we should stress, on the one hand, that graphical representations of the interactions between factors, as in Figs. 1 and 2, can be helpful in understanding and illustrating the way in which tradeoffs are being made and the possible directions towards improvement. On the other hand, as discussed in Section 3, flexibility of the DEA methods easily allows expert judgments and managerial perceptions to be incorporated into the models, thereby improving their acceptability. Additional attention should be drawn to limitations and potential extensions of the methods proposed in this paper. First, whereas the fact that no a priori factors are required for weighting the impacts is often considered an advantage over other methods, this can have significant drawbacks, such as, e.g. the existence of units that will be declared ‘efficient by default’ because they are best on only one aspect, even though they might be much worse in all other respects. As mentioned, this can be solved through use of weight limiting techniques. But even not accounting for that, the main usefulness of the method can be seen in its ability to point to those units that are less efficient, and provide for explanatory factors which may help in the identification of factors of unsustainability. As another example, the frontier obtained from observations on existing decision making units merely reflects best practice, which as such does not imply sustainability. Indeed, as described in Section 3, the data envelopment analysis methods can only yield results that reflect what has been incorporated into the data, i.e. if past observations are being used, the results obtained will be based on how industries made their choices in the recent past. In a sustainable development perspective, there may be important tradeoffs to be made as regards, e.g. the future use of (non-) renewable resources as inputs. Such tradeoffs can be reflected by restrictions on the weight combinations as in Eq. (4), i.e. in that way, the present or past decisions can be gauged with respect to trade offs

51

that appropriately translate the choices we have to formulate for the future. Thus, replacement of the best practice frontier by some kind of ideal frontier reflecting sustainability goals that society may formulate for production units will provide another useful extension of the methods developed herein. Once again, it should be stressed that such a sustainability frontier is by no means static; instead, it should evolve both with the improvement of the knowledge we have about the world we live in and with new goals we set for future generations. A few last comments can be made about the actual use of the indicators. Not only could they be exploited to compare firms of a given industrial (sub-) sector in a national context, but perhaps more importantly in a sustainability perspective, they could serve to compare firms or sectors in different countries (in the First and Third Worlds for example) that can significantly differ in the way they take social and environmental goals into account. This can result in the formulation of adequate corrective actions since the causes of unsustainability can be detected from the indicators. And here we return to the ultimate objective of using indicators, namely, providing the governments with adequate tools to adopt regulations and incentives that will ensure overall sustainability. Acknowledgements The paper benefited from valuable discussions with, among others, Marie-Paule Kestemont (Universite´ Catholique de Louvain, Belgium) and the participants of the Summer University of Southern Stockholm, ‘Management of Sustainable Enterprises’ (Stockholm, Sweden, August 1995). The research benefited from grants from the Belgian National Scientific Research Foundation (FNRS, Brussels) and from the Intercollegiate Center for Management Science (CIM, Brussels). References Ashford, N.A., Meima, R., 1993. Designing the sustainable enterprise. Summary report, Second International Research Conference, The Greening of Industry Network, Cambridge, Massachusetts, November 1993.

52


Ball, V.E., Lovell, C.A.K., Nehring, R.F., Somwaru, A., 1994. Incorporating undesirable outputs into models of production: an application to US agriculture. Cahiers d’Ećon. Sociol. Rurales 31, 60–74. Bergstro¨m, S., 1993. Value standards in sub-sustainable development. On limits of ecological economics. Ecol. Econ. 7, 1– 18. Block, M., Ho¨gstro¨m, P., 1995. Performance measurement towards sustainable development—two applications of sustainable development records. Communication to the Business Strategy and the Environment Conference 1995, University of Leeds, UK, 20–21 September 1995. Callens, I., van den Berghe, S., 1997. Development, analysis and discussion of indicators of sustainability: application for Canadian pulp and paper companies. In: Uso´, J.L., Brebbia, C.A., Power, H. (Eds.), Ecosystems and Sustainable Development, Advances in Ecological Sciences, Vol. 1, Proceedings of the ECOSUD 97 Conference, 14–16 October 1997, Castle of Pen˜iscola, Spain. Computational Mechanics Publications, Ashurst Lodge, Ashurst, Southampton, UK, pp. 83–93. Callens, I., Wolters, L. (coll. Gallez, C.), 1998. Factors of unsustainability: identification, links and hierarchy. Bus. Strategy Environ. 7 (1) 32–42. Courcelle, C., Installe´, M., Kestemont, M.-P., Tyteca, D., 1998. Assessing the economic and environmental performance of municipal solid waste management. Waste Manage. Int., in press. Daly, H.E., Cobb, J.B. Jr., 1989. For the Common Good: Redirecting the Economy Towards Community, the Environment, and a Sustainable Future. Beacon Press, Boston, MA. Devuyst, D., 1993. Environmental impact assessment. In: Nath, B., Hens, L., Devuyst, D. (Eds.), Environmental Management: Instruments for Implementation, vol. 3. VUB Press, Brussels, pp. 145–175. Fa¨re, R., 1992. Productivity Gauging with Undesirable Outputs. Mimeo, Southern Illinois University. Fa¨re, R., Grosskopf, S., Lovell, C.A.K., Pasurka, C., 1989. Multilateral productivity comparisons when some outputs are undesirable: a nonparametric approach. Rev. Econ. Stat. 71, 90 – 98. Fa¨re, R., Grosskopf, S., Tyteca, D., 1996. An activity analysis model of the environmental performance of firms: Application to fossil fuel-fired electric utilities. Ecol. Econ. 18, 161 – 175. Golany, B., Roll, Y., 1994. Incorporating standards via DEA. In: Charnes, A., Cooper, W.W., Lewin, A.Y., Seiford, L.M. (Eds.), Data Envelopment Analysis—Theory, Methodology and Applications. Kluwer, Boston, MA, pp. 313 – 328. Gramlich, E.M., 1990. A Guide to Benefit–Cost Analysis. Prentice Hall, Englewood Cliffs, NJ. Haynes, K.E., Ratick, S., Cummings-Saxton, J., 1994. Toward a pollution abatement monitoring policy: Measurements,

model mechanics and data requirements. Environ. Prof. 16, 292 – 303. Jaggi, B., Freedman, M., 1992. An examination of the impact of pollution performance on economic and market performance: pulp and paper firms. J. Bus. Finance Account. 19, 697 – 713. Klein, J.D. and Yaiswarng, S., 1993. The effects of environmental controls on productive and scale efficiency in the US electric power industry. Discussion paper, Union College, Schenectady, NY. Presented at the Atlantic Economic Conference, Philadelphia, 7 – 10 October 1993. Merilaïnen, S., 1995. Does greening bring changes in managements logics of action? A case study of Finland’s biggest chemical company. Presented at the Greening of Management Workshop, European Institute for Advanced Studies in Management, Brussels, 12 – 13 January 1995. Nestor, D.V., Pasurka, C.A., Jr., 1993. Measuring progress in pollution prevention using frontier production functions: a preliminary assessment. Discussion paper, Environmental Protection Agency, Washington, DC. Norrthon, P., 1995. Measures of sustainability for copper. Communication to the Summer University of Southern Stockholm: ‘Management of Sustainable Enterprises’, Stockholm, Sweden, 9 – 11 August 1995. Pearce, D.W., 1994. Measuring Sustainable Development. Blueprint 3, Earthscan, London. Pearce, D.W., Atkinson, G.D., 1992. Are national economies sustainable? Measuring sustainable development. CSERGE Working Paper GEC 92 – 11, University College London. Pearce, D.W., Atkinson, G.D., 1993. Capital theory and the measurement of sustainable development: an indicator of ‘weak’ sustainability. Ecol. Econ. 8, 103 – 108. Portney, P.R., 1993. Product life cycle analysis: a public policy perspective. Resources for the Future, Discussion paper CRM 93 – 06. Portney, P.R., 1994. The price is right: making use of life cycle analyses. Issues Sci. Technol., Winter 1993 – 1994; 69 – 75. Ragas, A.M.J., Knapen, M.J., van den Heuvel, P.J.M., Eijkenboom, R.G.F.T.M., Buise, C.L., van de Laar, B.J., 1995. Towards a sustainability indicator for production systems. J. Clean. Prod. 3 (1 – 2), 123 – 129. SETAC (Society of Environmental Toxicology and ChemistryEurope), 1992. Life-cycle assessment. SETAC-Europe, Brussels. Tyteca, D., 1996. On the measurement of environmental performance in firms: A literature review and a productive efficiency perspective. J. Environ. Manag. 46, 281 – 308. Tyteca, D., 1997. Linear programming models for the measurement of environmental performance of firms — concepts and empirical results. J. Product. Anal. 8, 183 – 197. Tyteca, D., 1998. Linear programming indicators of sustainable development at the firm level. Application to fossil fuel-fired electric utilities. Presented at the Fourth Biennal Meeting of the International Society for Ecological Economics, ‘Designing Sustainability’, Boston, Ma, 4 – 7 August 1996.

I. Callens, D. Tyteca / Ecological Economics 28 (1999) 41–53 Wehrmeyer, W., 1993. The scientific measurement of environmental performance. Paper presented to ESRC workshop on Environmental Performance Measurement and Reporting, University of Wolverhampton, UK, 25 August 1993. Welford, R., 1995. Environmental Strategy and Sustainable Development. The Corporate Challenge for the 21st Cen-

53

tury. Routledge, London. Wong, Y.-H.B., Beasley, J.E., 1990. Restricting weight flexibility in data envelopment analysis. J. Oper. Res. Soc. 41, 829 – 835. World Commission on Environment and Development, 1987. Our Common Future. Oxford University Press, Oxford.

.

.

Callens Tyteca 1999 - Towards Indicators of Sustainable Development for Firms

Recommend Documents