Вы находитесь на странице: 1из 13

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, Uni6ersité 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: callens@prod.ucl.ac.be-tyteca@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 currently operates in the economy. This is the


issue of futurity’’. This implies longer term, inter-
In the last few decades, environmental prob- generational considerations; therefore, ‘‘longer
lems faced by our modern society have given rise planning horizons need to be adopted and busi-
to a considerable wealth of research. More re- ness policy needs to be proactive rather than
cently, environmental preoccupations appeared to reactive’’ (Welford, 1995).
become part of a more global and fundamental Thus society has to formulate sustainability
context, that of sustainable development. The cul- goals, incorporating all three aforementioned is-
minating point of that new awareness was reached sues, and these goals have to be permanently
at the Earth Summit in Rio in June 1992. Many checked and improved. The extent to which given
nations have subscribed since then to the princi- goals are met can be measured by activity or
ples of sustainable development. Its most wide- performance indicators. The information obtained
spread, general (and vague) definition had the from indicators is especially valuable in the assess-
merit to create quasi-unanimity: it states that ment of tools such as taxes, regulations, or volun-
sustainable development is a ‘‘development that tary agreements, as to their validity and
meets the needs of the present generation without effectiveness towards meeting sustainability objec-
compromising the ability of future generations to tives. For example, the information obtained from
meet their own needs’’ (World Commission on life cycle analysis or life cycle assessment of prod-
Environment and Development, 1987). Such a ucts (SETAC, 1992) can be reflected in the form
broad definition is likely to give rise to various of price signals to reflect their whole environmen-
different interpretations, since people all have dif- tal or societal costs (Portney, 1993, 1994). As
ferent goals and sensitivities and will generally not another example, at the national level, environ-
agree on what to sustain (Norrthon, 1995). Thus mental laws and regulations can be improved by
we must be very careful in defining the goals and using the information obtained from such indices
options of sustainability. Moreover, the so-called as NNP (net national product, i.e. GNP corrected
‘‘goals’’ of sustainability cannot be unique and to account for degradation of natural capital).
defined once and for all; rather, sustainability Indicators can also be used to compare similar
should be viewed as a dynamic process in which existing units (products, technologies, plants,
the targets have to be continuously checked and firms, sectors and countries) in terms of their
improved, or as a philosophy that permanently performance with respect to some specified targets
tends towards improvements. (economic, environmental, etc.). In this case their
According to Welford (1995), ‘‘sustainable de- utility is to identify ‘laggers’ among the units
velopment is made up of three closely connected considered as well as the causes of lagging, and to
issues and each one of these needs to be addressed adopt relevant corrective actions. More generally,
by industry. Firstly, the environment must be indicators can contribute to discover so-called
valued as an integral part of the economic process factors of unsustainability (Callens and Wolters,
and not treated as a free good’’. This implies 1998) and to give recommendations as to means
‘‘minimal use of non-renewable resources and to reduce their influence. Conversely, indicators
minimal emission of pollutants’’, as well as pro- can be used to select the best among a set of
tection of ecosystems in order to avoid the loss of possible alternatives still to be implemented. As
plant and animal species. ‘‘Secondly there is a an example of the latter, one could think of
need to deal with the issue of equity’’. Equity environmental impact assessment methods (De-
‘‘applies not only to relationships between the vuyst, 1993). Other uses of indicators include
First and Third Worlds, but also within countries monitoring the performance and progress of given
between people’’, which, e.g. implies the reduction units (products, firms, etc., see above) over time.
of unemployment. ‘‘Thirdly, sustainable develop- Until now, sustainability indicators have been
ment requires that society, businesses and individ- developed mainly at the most global level, i.e. the
uals operate on a different time scale than state or country level (Daly and Cobb Jr., 1989;
I. Callens, D. Tyteca / Ecological Economics 28 (1999) 41–53 43

Pearce and Atkinson, 1992, 1993; Pearce, 1994). controllable. For example, the geographic and
Due to their central role in human activities and socio-demographic environment in which a plant
development, firms should play an important part is located will play a determinant role. Thus, a
in the attainment of sustainability goals. A recent pulp and paper plant isolated in the Canadian
methodology, based on so-called sustainable de- forest, making extensive use of forest products,
velopment records (SDR), was developed for the and discharging wastes into a large river, may
firm or plant level (Bergström, 1993; Block and turn out to be sustainable because ecosystems are
Högström, 1995). That approach is primarily only locally affected and long-term equilibria are
devoted to the appraisal of a given firm’s contri- not affected. On the other hand, the same plant,
bution to sustainability. Ragas et al. (1995) also using the same resources and with the same pol-
proposed to develop sustainability indicators for luting activities, but located in a heavily popu-
production systems. They start from the concept lated area, is not sustainable because it would
of environmental space, which would allow the deplete natural resources (forest, water) and have
measurement of sustainability, through the defini- negative impacts on the communities living in the
tion of an adequate distribution scheme of the neighbourhood. Therefore we must look for a
various components of environmental space definition in relative terms. As mentioned in the
among production systems. introduction, we might refer to specific targets
The latter approach can be viewed as a ‘top- stated for resource uses and external impacts, e.g.
down’ definition of sustainability indicators, i.e. with reference to the concept of environmental
the measurement is performed with respect to space. Or we might compare similar production
some global, predetermined level of sustainability. units placed in similar contexts, to detect whether
In this paper we have a different, somewhat com- some of them behave in a more appropriate way
plementary ‘bottom-up’ focus, that starts from the than others, with respect to specific economic,
production systems themselves. We set out to social or environmental characteristics. Then the
compare various firms or plants in a given set, in initial question can be more appropriately stated
terms of their performance as regards sustainabil- as ‘in what way would a given enterprise be more
ity, in order to be able to detect which ones of or less sustainable than another’? If we admit that
them are the laggers, what are the reasons for this is a valid question, the next step will be to
lagging (e.g. which are the factors of unsustain- define the kind of information that should appear
ability involved (Callens and Wolters, 1998)) and in the sustainability indicators in order to allow
therefore what are the possibilities towards im- for such comparisons.
provement. In this scheme, efficiency with respect The discussion conducted so far implies that we
to economic, social and environmental resources would hardly define any indicator based on suffi-
is viewed as a necessary (but not sufficient) step cient conditions for sustainable development. In-
towards sustainability. This is discussed in the stead, we start from investigating a set of
following section. necessary conditions that firms must fulfill in or-
der to be sustainable. Necessary conditions are
viewed as being efficient in the use of resources, in
2. Economic, social and environmental efficiency the pollutants released to the environment, in the
as a necessary step towards sustainable social role played by firms as reflected by their
development rate of employment, the working conditions, and
in the care taken with respect to future genera-
Before conceptualizing and implementing sus- tions in the setting of long-term objectives. Since
tainability indicators for firms, we would have to there are no benchmarks, for any of these charac-
raise the following question: what is a sustainable teristics, that would indicate from which level
enterprise? However, the answer to such a ques- firms could be declared sustainable, the focus is
tion in absolute terms appears impossible, because on comparing the firms between themselves. Some
it depends on so many factors that are hardly are more efficient, or inefficient, in all respects;
44 I. Callens, D. Tyteca / Ecological Economics 28 (1999) 41–53

most often, however, they might be efficient in however, setting the boundaries is fundamental:
specific respects but inefficient in others. Therefore then we would have to specify exactly which system
we need a methodology to take appropriate trade- is under investigation (e.g. a plant with its inputs/
offs. This is what is developed in the paper. outputs, a plant or a company with its raw material
Table 1 indicates a set of possible factors for uses and its environmental impacts and immis-
which information would be required in the frame- sions). The list of factors to be accounted for,
work of efficiency. This list is not meant to be among those indicated in Table 1 as an example,
comprehensive and is provided here only as an will depend upon the specification of boundaries.
example. It is likely to vary across the various It can be seen in Table 1 that some of the factors
industrial sectors. The factors considered embrace considered reflect sustainability goals for the firm
various kinds of quantities; for example, some of considered (mainly economic goals, which underlie
them are representatives of stocks, other ones stand its ability to survive), while others (i.e. socio-eco-
for flows, while a third part of them would reflect nomic and ecological factors) mainly reflect sus-
transformations. One other categorization would tainability goals for society. It is also apparent that
distinguish between factors that are viewed as
some of the items cover more than one category of
emissions (e.g. pollutants that are present in a given
aspects. For example, the use of natural (renewable
firm’s effluents), and others that can be character-
or non-renewable) resources clearly has economic
ized as immissions (i.e. wastes that are actually
incidences, ecological impacts, but also social im-
discarded to the environment, whose influence and
pacts, related to resource availability for future
quantification depend on the local characteristics of
the receiving bodies). The concept of immission generations. Some items might have contradictory
does not only apply to environmental, physical influences depending on the aspects considered.
characteristics, but can be adapted and extended as For example, employment would tend to be mini-
well to socio-economic components. For example, mized from a business point of view but should
locating a firm in a given region may induce rather be maximized from a social point of view.
employment not only in the firm itself, but also Also, the intervention of certain factors might
indirectly in services or administrations in charge evolve with the level of scientific knowledge, which
of controlling the ecological or economic activities will influence the way in which they should be
of the firm, or in the transportation sector, etc. accounted for in the elaboration of indicators. As
Thus we are not trying to specify rigid boundaries an example, one may think of the CFCs, once
to the systems whose sustainability is being tested. considered a useful, inert auxiliary in various pro-
Instead, the emphasis herein is on the methodology ductions, but now identified as one of the major
of the approach, whatever the investigated unit pollutants contributing to the ozone layer deple-
may be. When dealing with a specific application, tion.

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

Short term Long term

Economic as- Turnover, value added, output production, resources Profitability, competitiveness, market shares, product
pects used as inputs (including recycled products and en- durability; research and development efforts
ergy)
Social aspects Employment, salaries, labor intensiveness or produc- Welfare, education, availability of (non-) renewable,
tivity, injury risk noise, odour resources (including energy), size (SME vs big), per-
sonnel rotation rate
Ecological as- Natural resources, wastes, pollution, transportation Global impacts: biodiversity, global warming, acid
pects modes and distances deposition, landscape, ultimate waste disposal,
product recycling ability
I. Callens, D. Tyteca / Ecological Economics 28 (1999) 41–53 45

Dealing with sustainability, some authors (Ash- Table 2


ford and Meima, 1993; Welford, 1995) also insist Notation used in the definition of sustainability indicators
on the importance of such factors as the state of Variables to Variables to mini-
technology, information structures, societal values maximize mize
and culture. These are hard factors to measure
directly but they can be captured in an indirect Economic aspects ECONmax,i i= ECONmin,i i= 1,…,
way in some of the items listed in Table 1. For 1,…, I1 I2
Social aspects SOCmax,i i= 1,…, SOCmin,i i= 1,…, I4
example, it might be important to satisfy cultural I3
preferences for small, decentralized centers of pro- Environmental ENVmax,i i= ENVmin,i i= 1,…, I6
duction, ‘‘using local resources and good technol- aspects 1,…, I5
ogy to increase productivity to make products
and provide services that satisfy the fundamental
needs of local people without destroying the envi-
ronment’’ (Ashford and Meima, 1993). Such cul- Table 1 by the symbols defined in Table 2. For
tural or social preoccupations can be reflected by each of the three classes of aspects, i.e. economic,
variables such as employment, labor intensive- social and environmental, we consider factors
ness, and plant size (Table 1) which should there- whose value should be minimized (all other things
fore be given the adequate orientation (i.e. the being equal) and those whose value should be
emphasis should be more on small size, labor maximized, in order to reach efficiency, or more
intensive plants in order to achieve sustainability). generally, in the perspective of sustainability. For
example, natural resources used should clearly be
minimized, from an economic as well as ecological
3. Sustainability indicators based on efficiency standpoint, while durability is a factor we would
standpoints tend to maximize in a sustainability framework.
Note that in the following, we will use the terms
Now we come to the difficult question of how ‘factors’, ‘variables’, ‘quantities’ and sometimes
to account for, and aggregate, such diversified ‘aspects’ as synonyms.
information as listed in Table 1, provided that it is First, if we consider the definition of indicators
available. Data on the factors listed in Table 1 in a strict cost–benefit framework, the problem
can be exploited in two possible ways. First, we would be to compare the DMUs with respect to
could use past observations on the values of these the values taken by the following kind of
factors; the indicator would then be a static tool quantity:
that would help us to identify efficient and less I1
efficient units among a given set. Second, we IndicatorCB = % ai ECONmax,i
could use projected values that would reflect a set i=1
I2
of possible alternatives; in this case the indicators
would be a prospective tool. In both cases, we − % bi ECONmin,i
i=1
have to assume that we have observations on the I3 I4
factors for a set of (existing or possible) decision + % gi SOCmax,i − % di SOCmin,i
making units (DMUs), i.e. plants in a firm or i=1 i=1

firms in an industrial (sub-) sector. Indeed, to be I5 I6

operational, the indicator should compare units + % oi ENVmax,i − % zi ENVmin,i


i=1 i=1
that are really comparable, in the sense that they (1)
produce similar products that are designed to
fulfill analogous usages. We also assume, without in which the (positive) coefficients ai, bi, gi, di, oi
loss of generality, that the available data cover and zi represent either the prices of marketable
one time period (1 year). economic quantities (i.e. inputs, salaries, etc.) or
Let us designate the quantities enumerated in shadow prices of non-marketable quantities (i.e.
46 I. Callens, D. Tyteca / Ecological Economics 28 (1999) 41–53

pollution, productivity, etc.). It is supposed that puts’ (Färe, 1992; Klein and Yaisawarng, 1993;
each of the quantities appearing in Eq. (1) has been Nestor and Pasurka, 1993; Ball et al., 1994; Färe
scaled by a quantity reflecting the activity of the et al., 1989, 1996; Tyteca, 1996). The basic stand-
DMU. Most relevantly, this would be some mea- point of productive efficiency, as applied to envi-
sure of the output production. This is necessary to ronmental performance measurement, is to
give meaning to the comparisons. compare a set of decision making units (DMUs)
The main drawback of a cost – benefit definition between themselves, in terms of their environmen-
such as in Eq. (1) is the quantification of the tal or sustainability performance. The comparison
coefficients, especially those for which no market is usually restricted to similar units, e.g. plants or
prices exist. This is indeed a much controversial firms in a given industrial (sub-) sector, but can be
topic and many would argue that not even shadow extended to different geographic regions and to
prices would exist for such critical factors as different periods of time, which allows, e.g. for the
biodiversity, or even more simply, existence of a assessment of performance improvements over
given living species. Therefore in the following we time. The technique is also known as data envelop-
will consider approaches that do not use price ment analysis (DEA). Outlined below are three
coefficients, but instead, weights or intensities, that examples of DEA models that consider the sustain-
reflect the importance given to the various factors. ability problem under three different perspectives.
Moreover, we consider an indicator defined as a
ratio between a weighted sum of quantities that are 3.1. Model 0: generalization of the input–output
considered desirable, to a weighted sum of quanti- framework
ties that are viewed as inputs and whose interven-
tion has to be minimized: Consider the situation depicted in Fig. 1. This
illustrates the simplified case where we consider,
IndicatorIO = in Eq. (2), only one numerator quantity Qmax (i.e.
a quantity to be maximized, an output) versus only
I1 I3 I5
% ai ECONmax,i + % gi SOCmax,i + % oi ENVmax,i
one denominator quantity Qmin (an input). In
i=1 i=1 i=1 Fig. 1, DMU A is efficient because there is no
I2 I4 I6 other DMU that uses less of Qmin to produce
% bi ECONmin,i + % di SOCmin,i + % zi ENVmin,i
i=1 i=1 i=1
(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 in-
dicators in which the factors are given arbitrary
weights (Jaggi and Freedman, 1992; Wehrmeyer,
1993; Tyteca, 1996).
As a last proposal for defining indicators of
Fig. 1. Two dimensional representation of a set of observed
sustainable development, we could generalize the
decision making units that use quantities Qmin of an input to
idea of extending the theory of productive efficiency produce quantities Qmax of an output. The straight lines
to the consideration of environmental factors, in represent the production or efficiency frontier, under two
which the wastes were taken as ‘undesirable out- different assumptions of returns to scale.
I. Callens, D. Tyteca / Ecological Economics 28 (1999) 41–53 47

more of Qmax, while this is not the case for DMU of technology; the frontier of that set is constituted
B, which will be declared inefficient. We translate by the DMUs that exhibit the best practice in terms
this situation by giving an efficiency score 1 to of sustainability. The objective function in Model
DMU A, while the score associated with DMU B 0 indicates the problem of DMU ‘0’, i.e. given that
will be less than 1. the efficiency of all DMUs is set to be smaller than
In mathematical terms, generalizing to situations or equal to 1 by the constraints, the efficiency of
where we have more than one output and one input, DMU ‘0’ is either equal to or smaller than 1, in
we simply consider the framework reflected by Eq. which cases it is considered ‘efficient’ or ‘ineffi-
(2) and modify it so that no a priori value is given cient’, respectively. The meaning of efficient and
to the six sets of weighting coefficients. Instead, we inefficient in this context might be taken as ‘sustain-
consider these as the variables of a mathematical able’ and ‘unsustainable’, under the assumption
programming scheme and compare one given that Model 0 appropriately reflects sustainability.
DMU, indexed ‘0’, to the whole set of DMUs for It should be stressed, however, that a DMU that
which data are available. In this way, we offer the would be declared efficient because its index value
possibility of incorporating various different kinds is one, is not necessarily sustainable, since it could
of quantities, with different measurement units (i.e. appear to be efficient because it would be better
physical, economic, etc.) and meanings (i.e. stocks, than all other DMUs on only one factor. Thus it
flows, transformations, etc.), that have to be aggre- is possible to refer to DMUs that are unsustainable
gated to reflect their distance to efficiency, without because they are inefficient, but efficient by no
requiring any assumption on the weights used in the means implies sustainable, the former being only a
aggregation. The weights are the solution of a necessary condition for the latter.
mathematical program and will be computed in Among the widely recognized advantages of
such a way that the distance to the efficiency DEA models are: (1) a clear and obvious standard-
frontier is minimized. The general formulation of ization, since all units are ranked according to a
a DEA model that starts from Eq. (2) is the scale with the convention that the value 1 represents
following (see the aforementioned references for best performance or best practice; (2) an important
details of formulation): flexibility, since various versions of DEA models,
stressing important aspects in different ways, can
maxh0 =
be formulated easily; and (3) the robustness of the
I1 I3 I5 (non-) linear programming methods used to com-
% ai ECON0max,i + % ci SOC0max,i + % ei ENV0max,i pute the indicators. Additionally, the fact that no
i=1 i=1 i=1
I2 I4 I6
a priori weight has to be given to the factors taken
% bi ECON0min,i + % di SOC0min,i + % fi ENV0min,i into account is often presented as a main advantage
i=1 i=1 i=1 of DEA formulations. Indeed, the weights are
s.t. self-defined as the variables of the (non-) linear
programs used to compute efficiency scores. In that
I1 I3 I5 way, some kind of absolute objectivity is associated
% ai ECONkmax,i + % ci SOCkmax,i + % ei ENVkmax,i with DEA methods, because at the outset no
i=1 i=1 i=1
I2 I4 I6 judgment of any kind from any person is required.
% bi ECONkmin,i + % di SOCkmin,i + % fi ENVkmin,i There is at least one definite advantage of such
i=1 i=1 i=1
methods, i.e. they provide effective ways to detect
51 k = 1,..., K (3) DMUs that are inefficient even under the most
favourable weight combination, because for every
ai,bi,ci,di,ei,fi ] 0 DMU under consideration, DEA methods will
Index k designates a DMU in the set of K DMUs actually select the weight combination that mini-
for investigation; the constraints define the feasible mizes the distance from the DMU to the frontier,
set, i.e. the set of combinations of the factors taken which is therefore the most favourable combination
into account that are feasible in the present state within the space of all weight possibilities.
48 I. Callens, D. Tyteca / Ecological Economics 28 (1999) 41–53

This, however, is open to discussion, since in Eq. (4). More generally, engineering standpoints,
extreme cases, due to the possibility of having zero or managerial perceptions or targets, can be trans-
weights, and therefore all the weight concentrated lated using the notion of ‘standards’. Accordingly,
on one unique output or input, such procedures the strict endogeneous frontier can be substituted
can, and actually will, result in DMUs that are by a standard frontier (Golany and Roll, 1994;
declared globally efficient even if they are efficient Tyteca, 1996).
regarding only one of the specified variables. More- Finally, one should also be aware of one other
over, DEA detractors argue, why would self- potential drawback of DEA, i.e. the high sensitivity
defined, artificial weights be less arbitrary than any of the results with respect to the number of factors
a priori combination? and units considered; one should therefore be
To accommodate for such a drawback, methods aware that a given result can only be considered
exist, that allow one to incorporate judgments of with reference to the associated data set. However,
experts or analysts of the situation under concern this is no longer a drawback if we recall that best
(Wong and Beasley, 1990). With relation to the practice, or best available technology, is always a
methods used in DEA, such judgments can take the relative concept that heavily depends on what
form of intervals within which the relative weights actually exists.
given to some production factors should be in-
cluded. For example, considering the economic
3.2. Model 1: social and en6ironmental factors as
inputs, whose value should be minimized, i.e. in Eq.
the main factors for scrutiny
(3), possible weight intervals can take the form of
Model 0 as stated above only mimics the usual
biECONkmin,i
a ki 5 I2
5b ki (4) framework of data envelopment analysis, i.e. a
% bi ECONkmin,i situation in which we consider, on one side, the
i=1 quantities that are considered desirable (those ap-
pearing at the numerator), versus the quantities
which states that, for DMU k, the importance given
that are viewed as inputs and whose intervention
to input ‘i’ with respect to the whole set of inputs
has to be minimized (the denominator). In a
be comprised between a ki and b ki . This implies that
further, and more realistic step, we might consider
zero weights, as well as the possibility of giving the
three categories of factors, i.e. the inputs, the
totality of the weight to the sole input i, can be
outputs, and the undesirable outputs. In this scope,
avoided. Other combinations and standpoints can
there are various ways in which the problem can be
be reflected as well. For example, if for some
viewed (Tyteca, 1996, 1997), and we first consider
engineering reason, the relative weight of an unde-
one in which we try to minimize negative ecological
sirable output (e.g. atmospheric pollutant) is asso-
and social impacts that are opposed to the necessity
ciated with the use of a given production input (e.g.
of producing economic outputs using adequate
fossil fuel), this can also be reflected by a constraint
quantities of inputs (which is a necessary condition
on the ratio between appropriate quantities as in
for a firm to survive):
I4 I3 I6 I5
% di SOC0min,i − % ci SOC0max,i + % fi ENV0min,i − % ei ENV0max,i
i=1 i=1 i=1 i=1
min h0 = I1 I2
% ai ECON0max,i − % bi ECON0min,i
i=1 i=1
I4 I3 I6 I5
% di SOCkmin,i − % ci SOC kmax,i + % fi ENVkmin,i − % ei ENVkmax,i
i=1 i=1 i=1 i=1
s.t. I1 I2
]1 k= 1,..., K (5)
% ai ECON k
max,i − % bi ECON k
min,i
i=1 i=1
ai,bi,ci,di,ei,fi ] 0
I. Callens, D. Tyteca / Ecological Economics 28 (1999) 41–53 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 re-
duced only at the cost of a reduction of the
output production or some other variables (Färe
et al., 1989; Färe, 1992; Tyteca, 1997). One addi-
tional 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 sustainabil-
ity (e.g. a pollutant release, and an accident
Fig. 2. Production set and efficiency frontier for a set of record). With Model 1, because both impacts
observed decision making units, based on one environmental must be reduced, a move towards south-west will
impact and one social impact. The values of a sustainability
yield an improvement, which is feasible until the
indicator are illustrated for observed point B, for the case
where a reduction in both impacts is sought (u, reference point point reaches the frontier. If we consider, without
B1, Model 1) and the case where only a reduction in the loss of generality, that the reduction in both im-
environmental impact is sought (u%, reference point B2, Model pacts is proportional, this will lead us to point B1,
2). 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).
The sign restrictions on the variables indicate
that all (economic, social and environmental) 3.3. Model 2: en6ironmental factors as the main
factors for scrutiny

I6 I5
% fi ENV0min,i − % ei ENV0max,i
i=1 i=1
min h0 = I1 I2 I3 I4
% ai ECON0max,i − % bi ECON0min,i − % ci SOC0max,i + % di SOC0min,i
i=1 i=1 i=1 i=1

I6 I5
% fi ENVkmin,i − % ei ENVkmax,i
i=1 i=1
s.t. I1 I2 I3 I4
]1 k= 1,..., K (6)
% ai ECONkmax,i − % bi ECONkmin,i − % ci SOCkmax,i + % di SOCkmin,i
i=1 i=1 i=1 i=1

ai,bi,ci,di,ei,fi ]0
In this case, the emphasis is on minimizing the
parameters are considered strongly disposable (i.e. environmental impacts of the DMUs under con-
we can decrease output production down to ‘0’ at sideration. The sign conventions at the denomina-
no cost, and we can substitute inputs among tor of the objective function and the constraints
themselves, or decrease pollution, provided addi- are based on the fact that we now consider the
tional levels of appropriate inputs are used). This social factors as making part of the production
50 I. Callens, D. Tyteca / Ecological Economics 28 (1999) 41–53

factors (inputs) (Tyteca, 1997). In this case, in the necessity to take account of as many relevant
Fig. 2, a reduction has been sought only in the characteristics as possible (Gramlich, 1990). For
direction of environmental impacts, and the point example, a regulation that stresses some peculiar
reached from point B is point B2, resulting in an components of the pollutant charge while ignor-
indicator value of u%. A similar model could be ing others might result in a substitution between
stated with social factors as the main factors to pollutants and therefore in an overall increase of
scrutinize. Thus what is illustrated here is that we the global pollutant charge (Meriläinen, 1995). On
do not look for one unique indicator value, but the other hand, taking many factors into account
that instead, the methodology outlined in this might suffer from both computational intractabil-
paper allows us, using the same basic information, ities and data unavailability. Moreover, from a
to specify as many indicators as required by the strict technical point of view, accounting for many
specific application considered. The choice among variables in data envelopment analysis might re-
the indicators depends upon political or manage- sult in the extreme situation where all firms in a
rial decisions. given set are considered efficient. Therefore there
must be some kind of tradeoff between the num-
ber and the representativity of characteristics ac-
4. Discussion and conclusions
counted for, which can be solved by some
preliminary investigation of the data using, e.g.
Application of the models of Section 3, with
principal component analysis.
only a subset of variables reflecting environmental
The latter kind of problem might also find a
and economic quantities, was performed using
small data sets for a few case studies summarized partial solution if, instead of trying to develop one
in Table 3. In these situations we speak of envi- unique aggregate sustainability indicator (such as,
ronmental performance indicators. At least the e.g. in Model 1 above), we define two or three
results obtained showed that an analysis such as partial indicators that stress different aspects of
proposed herein is feasible. Now, to apply more the problem. For example, we would consider two
generally the ideas developed in this paper, we indicators, one centered on environmental preoc-
have to generalize the environmental performance cupations (such as, e.g. in Model 2 above), the
indicators to much wider situations, i.e. situations other stressing social preoccupations (as com-
in which there are many more variables to ac- mented under Model 2). In this way we could gain
count for. The first challenge is to define (as accuracy in the description of the situation, while
discussed in Section 2) and collect appropriate providing the decision maker with the possibility
data that will provide the basis for the new indica- of meaningful tradeoffs (the priority given to so-
tors to be built. cial or environmental preoccupations is left to the
An important aspect of any kind of indicator is 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 One output and two pollutants (to air and water) treated as inputs Haynes et al. (1994)
Fossil fuel-fired elec- Output: kilowatt-hour production; inputs: labour, fossil fuels, capi- Färe et al. (1996), Tyteca (1997,
tric utilities tal; pollutants: SO2, NOx, CO2 1998)
Pulp and paper Output: pulp production; inputs: labour, energy, fiber, water; pollu- Callens and van den Berghe
tants: particulates, TRS, SO2 BOD, TSS, AOX (1997)
Municipal waste Output: waste diversion rate; input differential cost of recovery; Courcelle et al. (1998)
management undesirable output: waste residue ratio
I. Callens, D. Tyteca / Ecological Economics 28 (1999) 41–53 51

A possible disadvantage of DEA methods may that appropriately translate the choices we have to
be that the way in which the weight factors are formulate for the future. Thus, replacement of the
determined is not straightforward for the non-spe- best practice frontier by some kind of ideal frontier
cialist, so that decision makers may be guided by reflecting sustainability goals that society may
parameters they do not fully understand. To cope formulate for production units will provide an-
with this difficulty, we should stress, on the one other useful extension of the methods developed
hand, that graphical representations of the interac- herein. Once again, it should be stressed that such
tions between factors, as in Figs. 1 and 2, can be a sustainability frontier is by no means static;
helpful in understanding and illustrating the way instead, it should evolve both with the improve-
in which tradeoffs are being made and the possible ment of the knowledge we have about the world
directions towards improvement. On the other we live in and with new goals we set for future
hand, as discussed in Section 3, flexibility of the generations.
DEA methods easily allows expert judgments and A few last comments can be made about the
managerial perceptions to be incorporated into the actual use of the indicators. Not only could they be
models, thereby improving their acceptability. exploited to compare firms of a given industrial
Additional attention should be drawn to limita- (sub-) sector in a national context, but perhaps
tions and potential extensions of the methods more importantly in a sustainability perspective,
proposed in this paper. First, whereas the fact that they could serve to compare firms or sectors in
no a priori factors are required for weighting the different countries (in the First and Third Worlds
impacts is often considered an advantage over for example) that can significantly differ in the way
other methods, this can have significant draw- they take social and environmental goals into
backs, such as, e.g. the existence of units that will account. This can result in the formulation of
be declared ‘efficient by default’ because they are adequate corrective actions since the causes of
best on only one aspect, even though they might be unsustainability can be detected from the indica-
much worse in all other respects. As mentioned, tors. And here we return to the ultimate objective
this can be solved through use of weight limiting of using indicators, namely, providing the govern-
techniques. But even not accounting for that, the ments with adequate tools to adopt regulations
main usefulness of the method can be seen in its and incentives that will ensure overall
ability to point to those units that are less efficient, sustainability.
and provide for explanatory factors which may
help in the identification of factors of unsustain-
ability. Acknowledgements
As another example, the frontier obtained from
observations on existing decision making units The paper benefited from valuable discussions
merely reflects best practice, which as such does with, among others, Marie-Paule Kestemont (Uni-
not imply sustainability. Indeed, as described in versité Catholique de Louvain, Belgium) and the
Section 3, the data envelopment analysis methods participants of the Summer University of Southern
can only yield results that reflect what has been Stockholm, ‘Management of Sustainable Enter-
incorporated into the data, i.e. if past observations prises’ (Stockholm, Sweden, August 1995). The
are being used, the results obtained will be based research benefited from grants from the Belgian
on how industries made their choices in the recent National Scientific Research Foundation (FNRS,
past. In a sustainable development perspective, Brussels) and from the Intercollegiate Center for
there may be important tradeoffs to be made as Management Science (CIM, Brussels).
regards, e.g. the future use of (non-) renewable References
resources as inputs. Such tradeoffs can be reflected
Ashford, N.A., Meima, R., 1993. Designing the sustainable
by restrictions on the weight combinations as in enterprise. Summary report, Second International Re-
Eq. (4), i.e. in that way, the present or past search Conference, The Greening of Industry Network,
decisions can be gauged with respect to trade offs Cambridge, Massachusetts, November 1993.
52 I. Callens, D. Tyteca / Ecological Economics 28 (1999) 41–53

Ball, V.E., Lovell, C.A.K., Nehring, R.F., Somwaru, A., 1994. model mechanics and data requirements. Environ. Prof.
Incorporating undesirable outputs into models of produc- 16, 292 – 303.
tion: an application to US agriculture. Cahiers d’Écon. Jaggi, B., Freedman, M., 1992. An examination of the impact
Sociol. Rurales 31, 60–74. of pollution performance on economic and market perfor-
Bergström, S., 1993. Value standards in sub-sustainable devel- mance: pulp and paper firms. J. Bus. Finance Account.
opment. On limits of ecological economics. Ecol. Econ. 7, 19, 697 – 713.
1– 18. Klein, J.D. and Yaiswarng, S., 1993. The effects of environ-
Block, M., Högström, P., 1995. Performance measurement mental controls on productive and scale efficiency in the
towards sustainable development—two applications of US electric power industry. Discussion paper, Union Col-
sustainable development records. Communication to the lege, Schenectady, NY. Presented at the Atlantic Eco-
nomic Conference, Philadelphia, 7 – 10 October 1993.
Business Strategy and the Environment Conference 1995,
Meriläinen, S., 1995. Does greening bring changes in manage-
University of Leeds, UK, 20–21 September 1995.
ments logics of action? A case study of Finland’s biggest
Callens, I., van den Berghe, S., 1997. Development, analysis
chemical company. Presented at the Greening of Manage-
and discussion of indicators of sustainability: application
ment Workshop, European Institute for Advanced Stud-
for Canadian pulp and paper companies. In: Usó, J.L.,
ies in Management, Brussels, 12 – 13 January 1995.
Brebbia, C.A., Power, H. (Eds.), Ecosystems and Sustain-
Nestor, D.V., Pasurka, C.A., Jr., 1993. Measuring progress in
able Development, Advances in Ecological Sciences, Vol. pollution prevention using frontier production functions:
1, Proceedings of the ECOSUD 97 Conference, 14–16 a preliminary assessment. Discussion paper, Environmen-
October 1997, Castle of Peñiscola, Spain. Computational tal Protection Agency, Washington, DC.
Mechanics Publications, Ashurst Lodge, Ashurst, Norrthon, P., 1995. Measures of sustainability for copper.
Southampton, UK, pp. 83–93. Communication to the Summer University of Southern
Callens, I., Wolters, L. (coll. Gallez, C.), 1998. Factors of Stockholm: ‘Management of Sustainable Enterprises’,
unsustainability: identification, links and hierarchy. Bus. Stockholm, Sweden, 9 – 11 August 1995.
Strategy Environ. 7 (1) 32–42. Pearce, D.W., 1994. Measuring Sustainable Development.
Courcelle, C., Installé, M., Kestemont, M.-P., Tyteca, D., Blueprint 3, Earthscan, London.
1998. Assessing the economic and environmental perfor- Pearce, D.W., Atkinson, G.D., 1992. Are national economies
mance of municipal solid waste management. Waste sustainable? Measuring sustainable development.
Manage. Int., in press. CSERGE Working Paper GEC 92 – 11, University Col-
Daly, H.E., Cobb, J.B. Jr., 1989. For the Common Good: lege London.
Redirecting the Economy Towards Community, the Envi- Pearce, D.W., Atkinson, G.D., 1993. Capital theory and the
ronment, and a Sustainable Future. Beacon Press, measurement of sustainable development: an indicator of
Boston, MA. ‘weak’ sustainability. Ecol. Econ. 8, 103 – 108.
Devuyst, D., 1993. Environmental impact assessment. In: Portney, P.R., 1993. Product life cycle analysis: a public policy
Nath, B., Hens, L., Devuyst, D. (Eds.), Environmental perspective. Resources for the Future, Discussion paper
Management: Instruments for Implementation, vol. 3. CRM 93 – 06.
VUB Press, Brussels, pp. 145–175. Portney, P.R., 1994. The price is right: making use of life cycle
Färe, R., 1992. Productivity Gauging with Undesirable Out- analyses. Issues Sci. Technol., Winter 1993 – 1994; 69 – 75.
puts. Mimeo, Southern Illinois University. Ragas, A.M.J., Knapen, M.J., van den Heuvel, P.J.M., Eijken-
boom, R.G.F.T.M., Buise, C.L., van de Laar, B.J., 1995.
Färe, R., Grosskopf, S., Lovell, C.A.K., Pasurka, C., 1989.
Towards a sustainability indicator for production sys-
Multilateral productivity comparisons when some outputs
tems. J. Clean. Prod. 3 (1 – 2), 123 – 129.
are undesirable: a nonparametric approach. Rev. Econ.
SETAC (Society of Environmental Toxicology and Chemistry-
Stat. 71, 90 – 98.
Europe), 1992. Life-cycle assessment. SETAC-Europe,
Färe, R., Grosskopf, S., Tyteca, D., 1996. An activity analysis
Brussels.
model of the environmental performance of firms: Appli-
Tyteca, D., 1996. On the measurement of environmental per-
cation to fossil fuel-fired electric utilities. Ecol. Econ. 18, formance in firms: A literature review and a productive
161 – 175. efficiency perspective. J. Environ. Manag. 46, 281 – 308.
Golany, B., Roll, Y., 1994. Incorporating standards via DEA. Tyteca, D., 1997. Linear programming models for the mea-
In: Charnes, A., Cooper, W.W., Lewin, A.Y., Seiford, surement of environmental performance of firms — con-
L.M. (Eds.), Data Envelopment Analysis—Theory, cepts and empirical results. J. Product. Anal. 8, 183 – 197.
Methodology and Applications. Kluwer, Boston, MA, Tyteca, D., 1998. Linear programming indicators of sustain-
pp. 313 – 328. able development at the firm level. Application to fossil
Gramlich, E.M., 1990. A Guide to Benefit–Cost Analysis. fuel-fired electric utilities. Presented at the Fourth Biennal
Prentice Hall, Englewood Cliffs, NJ. Meeting of the International Society for Ecological Eco-
Haynes, K.E., Ratick, S., Cummings-Saxton, J., 1994. Toward nomics, ‘Designing Sustainability’, Boston, Ma, 4 – 7 Au-
a pollution abatement monitoring policy: Measurements, gust 1996.
I. Callens, D. Tyteca / Ecological Economics 28 (1999) 41–53 53

Wehrmeyer, W., 1993. The scientific measurement of environ- tury. Routledge, London.
mental performance. Paper presented to ESRC workshop Wong, Y.-H.B., Beasley, J.E., 1990. Restricting weight flexibil-
on Environmental Performance Measurement and Report- ity in data envelopment analysis. J. Oper. Res. Soc. 41,
ing, University of Wolverhampton, UK, 25 August 1993. 829 – 835.
Welford, R., 1995. Environmental Strategy and Sustainable World Commission on Environment and Development, 1987.
Development. The Corporate Challenge for the 21st Cen- Our Common Future. Oxford University Press, Oxford.

. .

Вам также может понравиться