You are on page 1of 16

y FORUM

A Theoretical Foundation
for Life-Cycle Assessment
Recognizing the Role of Values
in Environmental Decision Making
Edgar G. Hertwich*
Energy & Resources Group
University of California
Berkeley, CA USA

James K. Hammitt
Center for Risk Analysis
Harvard School of Public Health
Boston, MA USA

William S. Pease
School of Public Health
University of California
Berkeley, CA USA

y
Keywords
economic damage index (EDI)
Summary
epistemology of science The presence of value judgments in life-cycle impact as-
global warming potential (GWP)
sessment (LCIA) has been a constant source of contro-
impact assessment
versy. According to a common interpretation, the
ISO 14042
life-cycle assessment (LCA) international standard on LCIA requires that the assess-
ment methods used in published comparisons be “value
free.” Epistemologists argue that even natural science rests
on “constitutive” and “contextual” value judgments. The
example of the equivalency potential for climate change,
the global warming potential (GWP), demonstrates that
Address correspondence to:
Edgar G. Hertwich any impact assessment method inevitably contains not only
Institute for Product Design and constitutive and contextual values, but also preference val-
Industrial Ecology Programme
Norwegian University of Science and
ues. Hence, neither life-cycle assessment (LCA) as a whole
Technology nor any of its steps can be “value free.” As a result, we sug-
Kolbjørn Hejes vei 2b gest a more comprehensive definition of objectivity in LCA
7491 Trondheim, Norway
hertwich@design.ntnu.no that allows arguments about values and their relationship
http://design.ntnu.no/ansatte/hertwich/ to facts. We distinguish three types of truth claims: factual
claims, which are based on natural science; normative
claims, which refer to preference values; and relational
© Copyright 2000 by the Massachusetts
Institute of Technology and Yale claims, which address the proper relation between factual
University

Volume 4, Number 1 * Now at the Norwegian University of Science and Technology, Trondheim, Norway

Journal of Industrial Ecology 13


y FO RUM

knowledge and values. Every assessment


method, even the GWP, requires each
type of claim. Rational arguments can be
made about each type of claim. Factual
truth claims can be assessed using the sci-
entific method. Normative claims can be
based on ethical arguments. The values of
individuals or groups can be elicited using
various social science methods. Relational
claims must follow the rules of logic. Rela-
tional claims are most important for the
development of impact assessment meth-
ods. Because LCAs are conducted to sat- Figure 1 The four steps of life-cycle assessment
isfy the need of decision maker s to (LCA) according to SETAC . (Fava et al. 1993).
consider environmental impacts, relational
claims about impact assessment methods limitations, and provides a basis for developing
should refer to this goal. This article intro- LCA methods and evaluating LCA results. We
duces conditions that affect environmen- focus on impact assessment, the subject of the
tal decision making and discusses how current controversy, but similar arguments apply
LCA—values and all—can be defended to allocation and the selection of system bound-
as a rational response to the challenge of aries in inventory analysis.
moving uncertain scientific information LCA is commonly described as a four-step
into the policy arena. process, as illustrated in figure 1. In goal defini-
tion and scoping, the problem and the intentions
y
of the assessment are defined. The units of com-
parison are chosen, as well as the type of impacts
that should be considered. In inventory analysis,
Introduction information about emissions and resource use
The field of life-cycle assessment (LCA) is during the life cycle of the product is collected.
engaged in a long-standing debate about values In impact assessment, these emissions are com-
and the relationship of values and science in en- pared and aggregated. The last step, interpreta-
vironmental analysis. LCA, a method used to tion and improvement assessment, includes an
comprehensively assess environmental effects of assessment of uncertainties and key assumptions,
product choices1 from the generation of the raw as well as recommendations for actions.
materials to the ultimate disposal of wastes, is Figure 2 outlines the structure of life-cycle
motivated by the desire to authoritatively deter- impact assessment (LCIA). Impact assessment in
mine the least-damaging alternative. Due to the LCA builds on a life-cycle inventory, which lists
presence of dissimilar impacts and the need to the emissions and resource uses that occur at dif-
focus the analytical effort, preference values al- ferent stages in the life cycle of a product. In the
ways enter the analysis, and so LCA results are classification step, these stressors are assigned to
not as value free as one might wish. Instead of different impact categories. In the characteriza-
treating LCA as a scientific tool that is flawed tion step, the stressors listed in each category are
because of the presence of values, we suggest compared and aggregated to a single category in-
that LCA is a decision support tool that not only dicator. In the valuation step, the category indi-
aggregates information but passes judgment re- cators are combined to produce an assessment of
garding the importance of different environ- overall environmental performance.2
mental impacts. In this article, we develop a Inventory analysis, classification, and charac-
theoretical foundation for LCA that outlines terization are commonly treated as “objective”
LCA’s nature, including its inherent power and scientific or technical analysis, and have been

14 Journal of Industrial Ecology


FO RUM y

Figure 2 Life-cycle impact assessment is commonly characterized as a three-step procedure that


determines the overall environmental performance of a product based on a life-cycle inventory. In the
classification step, similar stressors are assigned to the same impact categories. In the characterization step,
the stressors in a single category are compared and aggregated. In the valuation step, category indicators
are compared to each other. The global warming potential (GWP) and the human toxicity potential (HTP)
are characterization methods for the categories of climate change and toxic effects in humans.

the focus of method development. There has Spriensma 1999; Steen 1999). If LCA is seen not
been relatively little research addressing the as a disinterested aggregation of facts, but as part
LCA steps that are seen as value laden. Only re- of the decision-making process, consistent criteria
cently have researchers started to more system- for the evaluation of LCA methods and results can
atically address value questions in inventory be developed. These criteria, as well as the meth-
analysis as well as in the valuation and charac- ods and results developed in accordance with
terization steps of impact assessment (Finnveden them, can be defended as objective because they are
1997; Heijungs 1998; Hofstetter 1998; Newell developed in a process of intelligent, systematic debate.
1998; Tukker 1998; Lundie and Huppes 1999; They are rational because they better meet the needs of
Scheringer 1999). the decision maker, that is, they are superior for
We argue that the dichotomy of value-free and achieving the goals of the assessment.3 We begin
value-laden parts of impact assessment is artificial our analysis by revisiting the values debate and
and hampers the development of characterization showing how values and science are inseparably
methods. Characterization methods invariably in- intertwined. We argue that the presence of value
corporate preference values, as we will show in the judgments does not impede our ability to reason
example of the global warming potential (GWP). about product choice or to compare different en-
More and more, LCA is using impact assessment vironmental stressors. We must distinguish among
methods that combine characterization and valu- arguments about facts, arguments about what we
ation, such as the environmental priority sytem should care about, and arguments about how facts
(EPS) and the Eco-Indicator 99 (Goedkoop and relate to our concerns. Next, we investigate the

Hertwich, Hammitt, and Pease, A Theoretical Foundation for LCA 15


y FO RUM

specific conditions of environmental decision tentials that contain value judgments only for
making. We argue that the public nature of comparative assessments internal to a company
environmental problems and the complexity and or for public statements about a single product
uncertainty of environmental processes funda- that are not comparative. 4 Published compari-
mentally constrain how comparative assessments sons of different products, so-called comparative
are conducted. Any environmental assessment is assertions, must be based on assessment methods
necessarily imperfect (Funtowicz and Ravetz 1990; that are “scientifically and technically valid.”
Scheringer 1999). Ultimately, arguments in life- The aggregation of category results into a single
cycle assessment should refer back to LCA’s pur- indicator through a valuation step is not allowed
pose in decision making. An imperfect assessment for comparative assertions.
is more likely than no assessment at all to lead to One of us (Hertwich) has attended a negoti-
better-informed decisions. ating session for the standard, at which time it
was apparent that most delegates defined scien-
tific validity as the absence of value judgments.
The Values Debate
At one point, the head of an important delega-
LCA has been developed as an analytical tool tion even demanded that LCA results be repro-
to address the environmental impacts of products ducible across different cultures. The demand for
or services. In policy making and corporate deci- scientific validity paradoxically ensures that
sion making, LCA results have been treated as many established impact assessment methods are
facts. In recent years, however, the LCA commu- ruled out as unscientific. According to Udo de
nity has come to realize that life-cycle assessment Haes and colleagues (1999), a lack of scientific
involves value judgments. Value judgments were validity also affects the human toxicity potential,
at first acknowledged as part of the “valuation” a key equivalency potential for impacts of toxic
stage of impact assessment (Fava et al. 1993). In releases on human health. Because of this re-
this last step of impact assessment, the equiva- quirement, the consideration of many important
lency potentials representing different impact cat- environmental concerns is not allowed when de-
egories are combined to form a single indicator of cisions are to be based on the ISO standard.
environmental harm. It is now acknowledged that The negotiations for international LCA stan-
value judgments are also present in other LCA dards received much attention and intense in-
stages. “LCIA is not the strictly technical process dustry participation because of the potential
even in the traditional classification and charac- trade implications of the standards. The World
terization stages (Owens et al. 1997, 6).” Value Trade Organization (WTO) refers to the ISO
judgments are necessary to define different impact standard for the development of methods to dis-
categories and to develop equivalency potentials. tinguish between justified environmental poli-
The presence of value judgments, especially in cies and those that constitute nontariff trade
the characterization step of impact assessment, barriers and, therefore, violate the General
poses a significant problem for LCA researchers. Agreement on Tariffs and Trade (GATT). This
After recognizing the presence of value judg- reliance on the ISO standard by the WTO
ments in the categories of eutrophication, acid means that product policies set by national gov-
deposition, photochemical oxidants, and toxic ernments must use the ISO 14040 series. With
effects, Owens and colleagues (1997, 66) assert the present standard, product policy is not al-
that “aggregation from a number of equivalency lowed to use characterization methods that do
models is not technically valid.” In their view, not pass the test of scientific and technical va-
this lack of technical validity impedes the ability lidity as it is currently defined.
of LCA to establish meaningful differences In a comment to the delegates negotiating
among different products. ISO 14042, Hertwich and Pease (1998) argued
The current draft of the international stan- that it was inappropriate to limit the use of infor-
dard for life-cycle impact assessment developed mation by excluding information that is not
by the International Standards Organization solely justifiable by natural science. The exclu-
(ISO 14042) allows the use of equivalency po- sion of information relating to toxic emissions,

16 Journal of Industrial Ecology


FO RUM y

for example, will invariably lead to undesirable are used during the final “interpretation and
decision outcomes. Hertwich and Pease criti- weighting” step of impact assessment. This as-
cized language that calls for the minimization of sumption is wrong, because value judgments en-
value judgments and argued that LCA should be ter the method development of category-specific
based on a framework that integrates scientific equivalency potentials. This can be demon-
and normative elements. strated by analyzing the GWP, an equivalency
The chair of the ISO committee on LCA potential that the ISO negotiators consider to be
standards along with a number of key delegates scientifically and technically valid.
responded to these criticisms by arguing that the The GWP was developed to compare and ag-
LCA standards presented a good and workable gregate the emissions of greenhouse gases for the
basis for life-cycle assessment (Marsmann et al. purpose of developing a policy response to global
1999). climate change. The GWP is based on the incre-
mental infrared (IR) absorption that results from
Hertwich & Pease express the concern
an increase in the concentration of a GHG. To
that the 14042 document imposes ex-
calculate the GWP, one needs to develop sce-
treme constraints and limitations on
narios for the future emissions of GHGs. The
LCA and LCIA, especially for the case of
IPCC assumes constant concentrations of
comparative assertions. They character-
GHGs. Because the lifetimes of the GHGs are
ize the language used in the committee
not equal, one also needs to decide on how to
draft as being natural science biased and
compare the increase in IR absorption occurring
accuse the committee of ignoring in-
at different times. The official GWP integrates
sights from academic disciplines that ad-
the IR absorption across time up to 20, 100, or
dress value questions.
500 years and ignores increased IR absorption
In fact, much of the effort of this docu-
that occurs after this time horizon (Hammitt et
ment has been to separate natural sci-
al. 1996; Schimel et al. 1996).
ence-based analysis from the use of social
We use the idea of a cause-consequence
science based value judgment. The mem-
chain, or—as the Society for Environmental
bers of the subcommittee are well famil-
Toxicology and Chemistry (SETAC) has called
iar with the advances in social science in
it—an impact web, to investigate the character-
developing valuation methods, and rec-
ization of greenhouse gas emissions (figure 3).
ognize the need for such techniques in
Following Holdren (1980), we have divided the
decision making and communication.
causal chain into five elements (Hertwich et al.
However, we are also very much aware
1997). The first element of an impact chain is a
that the application of these techniques
stressor (emission or resource use). The second
by different groups can yield dissimilar
element is an insult, a change of natural material
results, due to differences in priorities,
flows or a disruption of the natural state, for ex-
opinions and social values. When com-
ample, an increased amount of sulfate particles
paring the environmental performance of
in the air. The third element is an environmen-
two products, applying the value system
tal stress, for example, a sulfate dose presented to
of one group to the decision making of
our lungs or the decrease of a lake pH value in
another group (as for example, in inter-
response to acid rain. The fourth element is a
nationally traded goods) is inappropriate
consequence resulting from the stress, for ex-
and discredits the outcome of an LCA.
ample, respiratory disease in response to sulfate
The argument of Marsmann and colleagues is particle exposure or fish death in response to
based on the assumption that natural science acidification. The last element is the value lost,
and value judgments can be neatly separated. for example, the treatment cost and lost work
Specifically, it assumes that value judgments can days resulting from a respiratory disease or the
be divided into (1) normative statements that decrease in the recreational value of a lake that
reflect the goal of the analysis and are defined by can no longer be used for fishing. The impact
the client at the outset, and (2) judgments that web, as outlined in figure 3, consists of many dif-

Hertwich, Hammitt, and Pease, A Theoretical Foundation for LCA 17


y FO RUM

Figure 3 The impact web describing the connection between stressor and value lost, based on the
cause-consequence chain proposed by Holdren (1980). The left side shows the terms that are used to
describe an impact web. The right side presents the example of acid precipitation and a number of its
impacts. Each level in the web can be described as being caused by the previous level, and the relationship
is described by natural or social processes.

ferent impact chains that result in different types dard discounting instead of accounting for the
of impacts, which occur at different times in dif- contribution of GHGs to global warming in 99
ferent locations, and which affect different indi- years, but not in 101 years.
viduals, populations, or ecosystems. Value judgments enter the development of a
In the cause-consequence chain, the GWP global climate change indicator at two points, at
aggregates emissions at the level of “stress” (fig- least. The first is the selection of a basis for the
ure 4). The selection of this endpoint of analysis comparison. This selection relates to the goals of
is not universally accepted. The economic dam- the analysis as well as to the standards we impose
age index (EDI), a proposal by Hammitt and on scientific evidence to be admissible to the
colleagues (1996), assesses the whole impact analysis. The EDI, as previously mentioned, is
chain down to the level of “value lost” to com- closer to our subject of concern, the damage re-
pare different greenhouse gases (figure 4). This sulting from climate change, and therefore better
proposal evaluates the incremental increase in reflects the goals of the analysis. Scientists have
temperature and then postulates a certain func- less confidence, however, in the relationship be-
tional relationship between the level and speed tween climate forcing and physical damage than
of temperature increase and the resulting dam- they have in the relationship between climate
age, based on integrated assessment models. The forcing and greenhouse gas emissions. The choice
model discounts future effects to compare effects between those two indicators is therefore a trade-
across time. Hammitt and colleagues argue that off between scientific reliability and the ability of
this model is more appropriate than the GWP the indicator to reflect our concerns (Hammitt
because it is closer to the subject of our concern, 1999). The second point at which value judg-
the damages resulting from an increase in GHG ments enter the analysis is in the treatment of
concentration. In addition, this model uses stan- impacts occurring at different times. Howarth and

18 Journal of Industrial Ecology


FO RUM y

tween value judgments and technical assump-


tions. “The choices regarding the GWPs and
ODPs are regarded as technical assumptions,
whereas the adding up of different types of disabil-
ity is regarded as a value choice” (Udo de Haes
1999). 5 The division of value judgments into
value judgments and technical assumptions is ar-
bitrary. The definition of an assumption depends
on the purpose of the analysis. It is hard to argue
that the selection of a basis of comparison for the
GWP is a purely technical assumption whereas
Figure 4 Conceptualization of the climate change
the selection of a basis of comparison for the hu-
impact chain following the proposal for the cause-
man toxicity potential constitutes a value choice.
consequence chain by Holdren (1980). The global
This artificial division should be abandoned,
warming potential (GWP) is based on a modeling
because it is misleading. It does not provide use-
of the chain up to the stress level, whereas the
ful guidance for the development or application
economic damage index (EDI) takes the conse-
of characterization methods. The choice of a dis-
quences of the stress into account. The two
count mechanism and rate, regarded as techni-
assessment methods differ in how they treat effects
cal assumptions by ISO, are value judgments
distributed across time.
embedded in economic analysis (Howarth and
Norgaard 1992; Schelling 1995). The whole
Norgaard (1992) have investigated the question language of the standard should be reconsidered
of the timing of climate change impacts using so- to avoid the bias toward natural science and to
cial welfare analysis, an economic technique de- provide public access to information (Hertwich
veloped to assess the desirability of changes in the and Pease 1999).
distribution of wealth in a society. They built an The values debate raises important epistemo-
overlapping generations model and constructed a logical issues. Some arguments in the debate
social welfare function (indicating the desirability make assertions about the character of science,
of varying distributions of welfare) for different which is an important subject in the philosophy
generations. In their model, the discount rate de- of science. The position of LCA relative to sci-
pends on the allocation of property rights to the ence is also of concern. The debate about the
different generations, that is, the level of welfare ISO standard is about standards for evidence and
achieved by these generations. This shows that for arguments in LCA. In the more detailed in-
the choice of a discount rate (and, by implication, vestigation of these issues in the next section,
the choice of a discounting mechanism) depends we argue that a more differentiated definition of
on how much we value future generations. objectivity is more reasonable than the one used
The GWP presents only a special case of the in the current debate.
comparison of environmental stressors. No two
environmental stressors are equal in their
A Basis for Reasoning in LCA
mechanism of action, local extent, time profile,
or precise endpoint. Any comparison of envi-
“Scientific objectivity requires simply the
ronmental stressors that takes the form of an
possibility of intelligible debate over the
equivalency potential necessarily requires a nor-
merits of rival paradigms.”
mative basis, consisting of value judgments
Israel Scheffler6
about what the critical properties are and what
the desirable (spatial and temporal) distribution The values debate in LCA is significant be-
of impacts across society is. cause it draws into question what LCA is, what
To avoid ruling out well-established methods criteria are employed to evaluate its results and
such as the GWP and the ozone depletion poten- its methods, and how arguments are made.
tial (ODP), ISO delegates now distinguish be- Whereas LCA is seen primarily as a collection

Hertwich, Hammitt, and Pease, A Theoretical Foundation for LCA 19


y FO RUM

and aggregation of environmentally relevant The idea that values play a role in science
facts related to the production, use, and disposal goes back to the underdetermination thesis fa-
of a product, the presence of value choices mously developed by the Scottish philosopher
makes it impossible to develop LCA methods David Hume in 1742 (1978). All forms of this
based only on scientific criteria. The use of val- thesis insist that multiple, mutually incompat-
ues in the characterization step has been re- ible theories can enjoy the same relation to a
jected by many LCA method developers and the given body of evidence. One cannot derive a
ISO committee because it would undermine the genuinely universal statement from any finite
authority and credibility of LCA results. Ac- set of observations. Hume’s work heavily influ-
cording to this line of reasoning, science offers enced the prominent philosopher of science
the only basis for making objective claims, and Karl Popper, who insisted that a positive belief
science is conceived as being value free. If the in a theory could never be rational (Laudan
argument in the previous section of this article is 1998). Values that influence the choice among
correct and value judgments enter any compari- competing theories include consistency with
son of environmental stressors, can we still claim other theories held to be true, simplicity, and ex-
that LCA results are objective or credible? Does planatory power (Lycan 1998; Sober 1998).
the presence of value judgments inescapably un- The influence of values on science receives
dermine our ability to make comparative envi- even more attention in more recent philosophi-
ronmental assertions? cal descriptions of science. According to Kuhn’s
In the next section, we will take a careful (1962) Structure of Scientific Revolutions, scien-
look at the issue of scientific objectivity and ex- tists adhere to a paradigm that consists of the
amine our ability to subject value judgments to theories, experimental techniques, and common
a systematic, rational analysis. We will show beliefs characteristic of their scientific disci-
how a more comprehensive definition of ratio- pline. Scientific revolutions occur when an es-
nality and a more careful distinction of the types tablished paradigm is insufficient to explain
of claims made in LCA can provide a basis for available data and an alternative paradigm that
theory choice and method development. LCA addresses the flaws of the existing paradigm
can be turned into a rational and largely objec- emerges. The adherence to a paradigm is value
tive enterprise as long as its methods and argu- based. Kuhn maintains that “there is no method
ments are consistent with its goal. of inquiry that stands outside any and all para-
digms” (Klee 1997, 141–147).
The cardinal role of values in science is
Scientific Objectivity
nicely illustrated by the following example de-
The proponents of the ISO life-cycle impact scribed by philosopher Kristin Shrader-Frechette
assessment standard view natural science as being (1991, 57):
free of value judgments (Owens et al. 1997;
Marsmann et al. 1999). This view is in conflict There are methodological value judg-
with both the position of philosophers of science ments in science and . . . they substan-
and the observation of historians and sociologists tially affect scientific conclusions. The
of science. The proposition that science is based case of physicist Wolfgang Pauli’s postu-
on or influenced by values is now almost univer- lation of the neutrino illustrates this
sally accepted by philosophers of science. The point. Pauli emphasized the value of in-
hotly contested question is whether, given this ternal consistency with conservation
presence of values, objective knowledge is still principles and therefore postulated the
possible. This is the realism versus antirealism de- existence of the neutrino before it was
bate. With their so-called strong program of anti- observed. Other physicists of his time
realism, social constructivists assert that science is emphasized a different value—external
purely a socially constructed activity. According consistency with principles demanding
to social constructivists, there is no way to distin- empirical proof—and therefore rejected
guish among competing theories (Klee 1997). the neutrino.

20 Journal of Industrial Ecology


FO RUM y

Clearly, the claim that any type of science is ments” (Shrader-Frechette 1991, 40). Contex-
value free cannot sustain scrutiny. It is, hence, tual values often enter the choice of one assump-
also misplaced in the current debate about val- tion, data set, or estimation method over its
ues in LCA. This raises the question whether alternatives. Assumptions are influenced by
there is a better definition of scientific objectiv- what is commonly accepted in a discipline and
ity, one that allows science to be objective. An- region; the selection of data depends on finan-
other question, raised for the purpose of our cial constraints as well as the availability of sec-
investigation, is how such a definition could be ondary or proxy data sets and familiarity with
useful in distinguishing good arguments about them. The latter again depends on the cultural
LCA methods and results from bad ones. This and institutional context.7
definition will have to better account for the
problematic, but important role of values. Preference Values
Instead of lumping all values together, we Preference values are the values that reflect
find it more important to distinguish among dif- what we care about. They could be preferences
ferent types of values and their roles in science for different types of consequences or preferences
and other human endeavors. In her discussion of for procedures or ways of acting. Preference val-
value judgments and scientific objectivity in risk ues are what economists try to assess through
analysis, Shrader-Frechette (1991) distinguished valuation methods such as contingent valuation
among three different types of values: constitu- or hedonic pricing. Preference values reflect not
tive values, contextual values, and bias values. only the utility of various environmental goods,
Bias values are really a special case of preference but also moral values, such as the concern for
values, which are important in decision making equity or for future generations, or esthetic val-
but have no role in scientific investigations, ex- ues, such as the appreciation for certain land-
cept as an object of study. scapes, plants, or animals. These are the values
that LCA refers to in the valuation stage to trade
Constitutive Values off different categories of environmental impact
The acceptance of a scientific theory or para- such as global warming and acidification.
digm as the basis for further scientific investiga-
tion is a constitutive or methodological value Constitutive and contextual value judgments
judgment; the values that underlie theory choice are epistemological in nature; that is, they relate
are simplicity, consistency with other theories, to our ideas of how the world works. Preference
and explanatory power. “Constitutive or method- values concern our ideas of what is good or bad
ological value judgments are an integral part of for ourselves or for society. Preference values
science, since scientists make constitutive value have no place in natural science, and if they do
judgments whenever they follow one method- enter scientific analysis they are called bias val-
ological rule rather than another. Even collecting ues to indicate that they are undesirable.
data requires use of constitutive value judgments Scientific objectivity is clearly not defined by
because one must make evaluative assumptions the absence of value judgments, and scientific re-
about what data to collect and what to ignore, sults do not represent infallible truths. In what
how to interpret the data and how to avoid erro- sense is science objective? It is objective in the
neous interpretations” (Shrader-Frechette 1991, sense that scientists engage in a rational process of
41). Perception and experimentation do not pro- testing and improving their theories. Theories are
vide us with pure facts; knowledge, values, theo- tested in experiments or evaluated according to
ries, and beliefs we hold are important in the rules of logic. Science is characterized by a
determining what we perceive. critical debate, and its theories are updated in the
face of new evidence (Hertwich 1999). Social
Contextual Values constructivists dispute the privileged status of sci-
“Assessors subscribe to particular contextual entific knowledge and the possibility of scientific
values whenever they include personal, social, objectivity, but we take a middle path between the
cultural, or philosophical emphasis in their judg- social constructivists and the nineteenth-century

Hertwich, Hammitt, and Pease, A Theoretical Foundation for LCA 21


y FO RUM

positivists. Like Klee (1997) and Shrader- 3. Relational claims are claims about how
Frechette (1991), we recognize that values and so- facts relate to values. These claims are the
cial conditions influence science, but we believe domain of policy analysis and decision
that scientific progress nonetheless occurs. Even if analysis. They can often be evaluated ac-
specific theoretical choices are not uniquely de- cording to the rules of logic. At the mini-
fined by experiments, “not all theoretical alterna- mum, a reasoned discussion of these
tives are equal,” as Klee asserts. Coherence with claims—appealing to similar, accepted
other theories and further (observational) impli- cases—is possible. Because relational
cations of alternative theories serve as legitimate claims often determine the structure of as-
criteria for theory choice (Klee 1997, 177–178). sessments, they may be the most impor-
tant claims in impact assessment.
The Objectivity of LCA
LCA inevitably uses each type of truth claim.
Science does contain values, but not prefer- The following examples illustrate the usefulness
ence values. LCA, however, requires both sci- of distinguishing among the three different types
ence and preference values because it not only of truth claims.
describes, but also evaluates aspects of reality. Factual claims are made in LCA whenever
The decision of how to treat uncertainty in the scientific results or models are used. The human
characterization step, for example, is based on toxicity potential (HTP), for example, makes
preference values. In the previous section of this factual claims about the fate and partitioning of
article we argued that science is objective be- chemicals. It makes claims about the mecha-
cause it has procedures and criteria that allow us nisms by which pollutants accumulate in expo-
to evaluate the merit of competing theories. In sure pathways. These claims can be tested by
making arguments about LCA, we should distin- experiment. An interesting argument about a
guish among three different types of truth claims. factual claim has been the “flux-pulse” problem
The objectivity of each of these three types of for HTP. Guinée and Heijungs (1993) find it
truth claims needs to be investigated separately. normatively more appropriate to model a pulse
release representing the emissions during the
1. Factual claims are concerned with facts. short time in which a product is manufactured.
They are scientific claims and can be Because dynamic modeling of pollution fate is
evaluated according to the rules and stan- mathematically difficult, however, they prefer
dards established in science. Factual the simpler steady-state model. They argued that
claims can relate to either the natural normalization by a reference chemical takes care
world or the social world. Examples of fac- of the time dimension. This argument was con-
tual claims in LCA are claims regarding tested until Heijungs (1995) and Scheringer
the fate or persistence of chemicals or the (1999) showed that the steady-state model ap-
effect of greenhouse gases on the global propriately accounts for the time-integrated ex-
energy balance. Factual claims relating to posure resulting from an emissions pulse. The
social science may be relevant for eliciting flux-pulse problem was basically resolved with a
a population’s preference values. mathematical proof.
2. Normative claims are claims about what is Normative claims concern preference values.
good or bad, about what should or should The assertion that we should take the number of
not be done. They fall into the domain of affected individuals into account when evaluating
politics, law, religion, and moral philoso- the impacts of toxic chemicals, for example, is a
phy. Examples of normative claims in LCA normative claim.8 Another example of a norma-
are claims regarding the rights of future tive claim is the choice of a discounting mecha-
generations or nonhuman species. Norma- nism and a discount rate for comparing impacts
tive claims can be based on moral reason- across time.9 The types of arguments we can make
ing. Values can be elicited using various about normative claims are different from the
valuation methods (Hertwich 1999). types of arguments we can make about factual or

22 Journal of Industrial Ecology


FO RUM y

relational claims. We can argue that normative symmetry, it may be desirable to also introduce a
claims are reasonable and that they are consistent requirement of “normative validity.”
with other, well-established normative judgments. An LCA method is scientifically valid if it
We can test the consistency of a set of normative uses scientific models and data in a way that is
claims.10 Normative claims, however, cannot be not in conflict with scientific understanding. It
tested in the same manner as a scientific theory. requires both that factual claims be identified as
Relational claims concern the relationship of such and that they be valid as scientific state-
facts and values, especially the relevancy of facts ments. The claim that the emission of 1 kg of
to our values and the consistency of the math- methane (CH4) is as bad as the emission of 24.5
ematical relationship between the two. LCA kg of carbon dioxide (CO2) is not a factual claim,
makes a relational claim whenever it says that if as it cannot be evaluated by the scientific
we are concerned about a specific endpoint such method. Rather, it is a relational claim that rests
as human health, we should care about one stres- on factual claims regarding the lifetime and IR
sor more than another. Such a claim is based on absorption of the gases and normative claims re-
normative judgments about what constitutes rel- garding the relevance of increased IR radiation
evant criteria for representing a specific end- and the rights of future generations. The defini-
point, and it contains factual information about tion of “scientifically valid” suggested here to re-
the different stressors. A relational claim can be solve the problems with the ISO 14042 standard
expressed as an argument that one method is would, for example, allow the GWP to be de-
better than another method based on various cri- fined as scientifically valid because the models
teria for method choice. We would argue that and the data used for its determination are scien-
the EDI is a better indicator for climate change tifically valid. The EDI is also scientifically valid,
than the GWP because it measures an endpoint but it reflects different normative claims. The re-
closer to our concern. This claim is based on quirement of scientific validity does not uniquely
constitutive value judgments about theory define the proper impact assessment method.
choice (Sober 1998) regarding the consistency of Technical validity refers to the validity of re-
intertemporal trade-offs and the relevancy of the lational claims. A method lacks technical valid-
endpoint. ity if it can be shown to be logically flawed or if it
Each of these types of truth claims can be does not appropriately account for the normative
evaluated objectively. This allows us to make ar- elements it purports to include. Most of the issues
guments about the merit of different assessment that arise in LCA method development are tech-
methods. Not all theoretical alternatives are nical issues, not scientific issues. The questions of
equivalent. The difference between LCA and what elements should be included in a method
natural science, however, is that in LCA there and how they relate to each other are technical.
are multiple legitimate sets of preference values “Normative validity” could ensure that LCA
and alternative, logically consistent ways of method development only takes into account
making judgments about facts (Renn 1998). normative claims that reflect reasonable prefer-
ence values. There are simply too many alterna-
tive analysis methods that would need to be
The Validity of LCA Methods
developed if every imaginable normative claim
The ISO standard uses the terms “scientific required attention. The approaches to norma-
validity” and “technical validity” to describe the tive arguments in LCA presented by Scheringer
acceptability of LCA methods. These terms are (1999) and Heijungs (1998) are important start-
useful only if they are defined in a way so that ing points to further develop this requirement.
“valid” LCA methods are objectively better In defining requirements for LCA methods,
than “invalid” methods for informing environ- it is important to take into account the charac-
mental decisions. We would like to suggest that ter of environmental problems and the limita-
scientific validity should refer to the validity of tions of social decision making. These factors
factual claims and technical validity to the va- define, and limit, the relational and normative
lidity of relational claims in LCA. To ensure claims possible in LCA.

Hertwich, Hammitt, and Pease, A Theoretical Foundation for LCA 23


y FO RUM

The Conditions of on preferences for strategies to address un-


Environmental Decision Making certainty (e.g., utility functions, or the
Maximin principle) and attitudes toward
Life-cycle assessment is a decision support
risk (e.g., risk aversion).
tool that addresses environm ental problems.
5. The trade-off between different objectives
The development of LCA needs to consider
and the judgment under uncertainty are
both the character of environmental problems
affected by systematic flaws in human rea-
and the principal and cognitive constraints to
soning. These flaws have become known
human decision making. Together, these ele-
as “bounded rationality” (Simon 1957)
ments define the conditions of environmental
and “framing effects” (Kahneman et al.
decision making that any assessment method in-
1982). Empirical investigations show that
variably encounters.
they are pervasive and affect individual
1. Environmental effects are, in the terminol- and group decision making (Kleindorfer et
ogy of economics, external effects, and en- al. 1993). These factors need to be consid-
vironmental goods, such as clean air and a ered in developing environmental deci-
stable climate, are public goods, that is, sion support.
nonrival, nonexclusive goods (Latin 1982;
Tietenberg 1992). Environmental deci- Latin (1982, 234) reviews uncertainty and
sions are, therefore, decisions that concern data gaps as well as the cognitive limitations of
society at large, not just individuals. the human mind that prevent us from under-
2. Social decision making is problematic be- standing environmental repercussions of con-
cause competing individual preferences sumption choices in all their complexity and
cannot, in general, be reconciled to pro- concludes:
duce universally agreed-upon social pref-
The ideal allocation of environmental
erences. In addressing the problem of
resources, the one that perfectly in-
ranking different objectives, Arrow
formed and perfectly rational consumers
(1951) has shown that it is impossible to
would select, is not only unattainable but
construct generally acceptable group
indeterminable.
rankings without being “dictatorial”; that
is, one member of a group will always A perfect assessment is both practically and
dominate (rule over) other members of theoretically impossible. The consequences of
the group. Similar problems affect the environmental stressors cannot be described un-
more difficult task of constructing a social ambiguously in terms of their extent, location,
welfare function (Tresch 1981, 25–26). and timing. Even if such a description were pos-
3. The overwhelming complexity of environ- sible, the large number of different impacts
mental processes implies that (1) we have would pose a challenge to the assessment of
only a limited understanding of the pro- trade-offs. An ideal valuation process is impos-
cesses that lead to deleterious consequences; sible because of the nature of social decision
(2) a mapping and utilization of even that making and the limitations of human judgment.
limited understanding is infeasible for indi- These limitations have important implications
vidual assessments; (3) any assessment will for LCA. The fact that the environment is a
be data intensive and require the combina- public good justifies the development of LCA
tion of knowledge from different fields. based on public attitudes and concerns, as well
4. Where scientific models are available to as the publication of both definitive and specu-
describe causal links between stressors and lative knowledge regarding environmental im-
impacts, data used in the models are often pacts. The impossibility of identifying a single
uncertain. Given this uncertainty, deter- socially preferred alternative suggests that we
ministic statements about environmental need to tolerate and accommodate conflicting
impacts are not possible. How impact as- goals and interests. Discussions of both Arrow’s
sessment addresses uncertainty depends “impossibility” theorem (Sen 1995) and the

24 Journal of Industrial Ecology


FO RUM y

limitations of our cognitive abilities (Slovic et nonoptimal. It is the decision, not the analysis,
al. 1985) point to the necessity of an open, that is inconsistent with our values or “the
democratic process. facts,” and the analysis is just a step leading to
The specific characteristics of LCA are a re- the decision. The question then is whether or
flection of the general conditions of environ- not analysis reduces the likelihood of the deci-
mental decision making. Both indicate that sion being inconsistent with our values or the
technical analysis cannot produce decision out- magnitude of this inconsistency. Even if the
comes: It can inform, but not replace the deci- analysis is not able to identify one single superior
sion maker. No uniquely correct answers exist alternative, or if, by accident, the analysis rec-
and a mediation among competing normative ommends a choice that is only nearly optimal,
claims is necessary. As two reviewers of this ar- not totally optimal, the analysis may still be jus-
ticle pointed out, these conclusions have already tified. It is a question of degree of information
been reached in similar debates about other en- that can be offered and of whether the decision
vironmental assessment tools, such as risk analy- outcome informed by analysis is better than a
sis (Jasanoff 1993; PCCRARM 1997; Renn decision outcome not informed by analysis, not
1998) and the valuation of ecosystems (O’Hara whether the decision outcome is perfect.
1996; Daily 1997). The realities of decision making allow us to
put LCA in perspective. We are looking neither
for universal agreement nor for perfect analysis
A Rationale for LCA
that reflects the minute details of environmental
The constraints outlined in the previous sec- processes, but for an improvement of the current
tion raise the question of why we bother to con- decision-making process that often does not in-
duct LCAs at all. LCAs are conducted to satisfy corporate environment considerations. This im-
the aspiration of decision makers to consider the en- provement makes LCA both worthwhile and a
vironment in their decision making. Most of the manageable task.
decisions that LCA are meant to inform are de-
cisions that are made anyway, regardless of
Conclusions
whether an LCA has been conducted or not.
The information presented in an LCA may in- As a decision support tool, life-cycle assess-
fluence a decision’s outcome, but not the fact ment invariably combines preference values and
that a decision needs to be made. The inevita- science. It is the very purpose of a decision sup-
bility of a decision and the supplementary nature port tool to allow the decision maker to choose in
of LCA speak directly to some of the issues accordance with her preferences. LCIA presents
raised in the previous section and have impor- a comparison and aggregation of different envi-
tant repercussions for the design of impact as- ronmental stressors that utilize scientific data and
sessment methods. models. The use of such scientific knowledge
Decisions made without the benefits of infor- does not imply that the comparison itself is a re-
mation provided by LCA are not informed about sult of natural science. Even if, due to the pres-
potential environmental “side effects.” Unin- ence of normative elements, LCA is not a strictly
formed decisions are more likely to be scientific tool, LCA can still be objective.
nonoptimal to many or all of the individuals af- Like science, LCA can be objective in the
fected. Informed decisions, even if they are im- sense that arguments for method choice fulfill
perfect as Arrow’s theorem suggests, are likely to clear criteria and follow rules deemed as reason-
be better (in the view of most, if not all indi- able by the community involved in the develop-
viduals) than uninformed decisions. ment of this tool. It can be objective not in the
The problem here is that because of the com- sense that all arguments or methods have the
plexity of environmental (and economic) pro- same standing, but that we can distinguish be-
cesses, the lack of data, and the pervasive tween valid and invalid arguments.11
uncertainty, an analysis may recommend an al- To investigate the merit of impact assessment
ternative that further analysis may show to be methods and results, it is necessary to distinguish

Hertwich, Hammitt, and Pease, A Theoretical Foundation for LCA 25


y FO RUM

among different types of truth claims. Norma- Acknowledgments


tive claims should be expressed explicitly and
We would like to thank Cathy Koshland,
have some amount of credible support. Factual
Tom McKone, Martin Scheringer, Sarah
claims relating to data and scientific models can
Trainor, the editor, and four reviewers for help-
be evaluated according to the criteria used in
ful comments. Funding was provided by the En-
science. Relational claims about the proper
vironmental Defense Fund and the Norwegian
combination of factual information and values
Research Council.
in the analysis must follow the rules of logic.
Whereas the approach suggested here can
turn LCA method development and application Notes
into objective exercises, its reliance on prefer-
1. LCA can and has been applied to other things in-
ences and the conditions of environmental deci- cluding services, processes, and facilities. We re-
sion making preclude the existence of uniquely fer to its use with product choices in this article
correct methods and results. Competing claims for purposes of brevity.
of the environmental superiority of alternative 2. We use the older SETAC term valuation over the
products are therefore permissible. One reviewer term weighting used by ISO, because we find it more
suggested that this creates a “tower of Babel” appropriate. Weighting does not capture multi-
problem for LCA, and that the public would lose criteria decision making methods (Guitouni and
confidence when confronted with competing as- Martel 1998) or (nonlinear) utility functions
sessments. We contend that distinguishing (Raiffa 1968), which decision analysts might see as
the most appropriate tools for this step.
among the three types of claims and evaluating
3. In this article, the term “rational” comes from
the validity of each allows us to better discrimi-
economics and decision analysis and means best
nate between good and bad LCA methods and suited to meet the goals. The term “objectiv e”
results. In this way, fewer results will be deemed comes from the philosophy of science and means
valid than if the focus is limited to the scientific subject to reasoning . Martin Scheringer offered us
aspects, and divergences will be easier to explain. the following more specific definition: A state-
We have shown that the distinction between ment is objective if it is based on transparent and re-
value-free and value-based elements of LCA is a producible methods and rules, so that anyone who
false dichotomy. Value choices are present in all starts at the same point and uses the same methods
parts of an LCA. The presence of preference val- and rules will end up with the same statement. This
ues does not disqualify any method; rather, it is a assumes that there is agreement on the appropriate
methods and rules. If there is not, these have to be
necessary feature of any decision support tool. In
subjected to the same process. Because LCA is a
the place of this false dichotomy, we propose a
tool designed to meet the goal of protecting the
distinction among factual claims, relational environment, what is objectively better in LCA
claims, and normative claims. Different criteria method development is necessarily rational.
must be used to evaluate the validity of each type 4. The International Standards Organization (ISO)
of claim. We have also introduced fundamental has developed four standards on LCA: ISO 14040
conditions of environmental decision making: defines LCA generally, ISO 14041 addresses in-
the societal nature of environmental problems, ventory analysis, ISO 14042 addresses impact as-
the impossibility of ideal societal decisions, the sessment, and ISO 14043 addresses normalization
complexity and uncertainty of environmental weighting and the interpretation of LCA results.
processes, and the imperfections of individual 5. Without this distinction, the standard as it is now
written would prohibit the use of any currently
reasoning. These conditions are unavoidable and
available impact assessment method for compara-
need to be taken into account when developing
tive assertions and thus make it impossible to
LCA methods. Because it is the overall goal of publish the results of ISO-compliant LCAs. The
LCA to improve environmental decision mak- use of LCA in setting product policy could there-
ing, we suggest that the ultimate criterion for fore be challenged and forbidden by the WTO.
method choice is whether a given method is bet- 6. Cited in Shrader-Frechette (1991, 52).
ter than its alternatives in improving the deci- 7. Some scientists see such a practice as a potential
sion.

26 Journal of Industrial Ecology


FO RUM y

source of bias and accept it only for pragmatic mate policy: Atmospheric stabilization, benefit-
reasons. Others use contextual information in- cost analysis, and near-ter m greenh ouse-gas
tentionally by utilizing Baysian methods, which emissions. Climatic Change 41(3/4): 447–468.
are designed to include “prior” information, in- Hammitt, J. K., A. K. Jain, J. L. Adams, and D. J.
cluding guesses (Malakoff 1999). Wuebbles. 1996. A welfare-based index for as-
8. This is by no means a universally accepted claim. sessing environmental effects of greenhouse-gas
A number of U.S. environmental regulations are emissions. Nature 381(23 May): 301–303.
based on the premise that most exposed individu- Heijungs, R. 1995. Harmonization of methods for im-
als should be protected, regardless of how many pact assessment. Environmental Science and Pollu-
individuals are exposed. tion Research 2(4): 217–224.
9. Normative claims are, therefore, included even Heijungs, R. 1998. Towards eco-efficiency with LCA’s
in the characterization step of impact assessment. prevention principle: An epistemological foun-
10. A common approach to reasoning about norma- dation of LCA using axioms. In Product innova-
tive issues is through definition s, axioms, and tion and eco-efficiency, edited by J. E. M.
theorems. Heijungs (1998) proposes to use this Klostermann and A. Tukker. Dordrecht, Nether-
approach to “derive” LCA methods and sketches lands: Kluwer Academic Publishers.
how this could be done. Hertwich, E. G. 1999. Toxic equivalency: Accounting
11. For a more detailed elaboration of objectivity see for human health in life-cycle impact assessment .
Shrader-Frechette ’s (1991) discussion of scien- Energy and Resources Group. Berkeley: Univer-
tific proceduralism and epistemological discus- sity of California.
sions of the scientific method and theory choice Hertwich, E. G. and W. S. Pease. 1998. ISO 14042
(Klee 1997; Lycan 1998; Sober 1998). restricts use and development of impact assess-
ment. International Journal of Life Cycle Assess-
ment 3(4): 180–181.
References
Hertwich, E. G. and W. S. Pease. 1999. Rebuttal to
Arrow, K. J. 1951. Social choice and individual values. Marsmann et al. on ISO 14042. The LCA Global
New York: John Wiley & Sons. Village. http://www.ecomed.de.
Daily, G. C., ed. 1997. Nature’s services: Societal depen- Hertwich, E. G., W. S. Pease, and C. P. Koshland.
dence on natural ecosystems. Washington DC: Is- 1997. Evaluating the environmental impact of
land Press. products and production processes: A compari-
Fava, J., F. Consoli, R. Denison, K. Dickson, T. son of six methods. The Science of the Total Envi-
Mohin, and B. Vigon, eds. 1993. A conceptual ronment 196: 13–29.
fram ework for life-cy cle impact assessm ent. Hofstetter, P. 1998. Perspectives in life cycle impact as-
Pensacola, FL: Society of Environmental Toxi- sessment: A structured approach to combine models
cology and Chemistry. of the technosphe re, ecosphere and valuespher e.
Finnveden, G. 1997. Valuation methods within Boston: Kluwer.
LCA—Where are the values? International Jour- Holdren, J. P. 1980. Integrated assessment for energy-re-
nal of Life Cycle Assessment 2(3): 163–169. lated environmental standards: A summary of issues
Funtowicz, S. O. and J. R. Ravetz. 1990. Uncertainty and findings . LBL Report 12799. Berkele y:
and quality in science for policy. Dordrecht, Neth- Lawrence Berkeley Laboratory.
erlands: Kluwer Academic Publishers. Howarth, R. B. and R. B. Norgaard. 1992. Environ-
Goedkoop, M. and R. Spriensma. 1999. The eco-indi- mental valuation under sustain able develop-
cator 99. Amersfoort: PRe Consultant s. ment. American Economic Review 82: 473–477.
http:\\www.pre.nl. Hume, D. 1978. A treatise of human nature. Oxford:
Guinée, J. and R. Heijungs. 1993. A proposal for the Clarendon Press.
classifi cation of toxic substan ces within the Jasanoff, S. 1993. Bridging the 2 cultures of risk analy-
framework of life cycle assessment of products. sis. Risk Analysis 13(2): 123–129.
Chemosphere 26(10): 1925–1944. Kahneman, D., P. Slovic, and A. Tversky. 1982. Judg-
Guitouni, A. and J. M. Martel. 1998. Tentative guide- ment under uncertain ty: Heuristics and biases.
lines to help choosing an appropriate MCDA Cambridge: Cambridge University Press.
method. European Journal of Operational Research Klee, R. 1997. Introduction to the philosophy of science.
109(2): 501–521. New York, Oxford: Oxford University Press.
Hammitt, J. K. 1999. Evaluation endpoints and cli- Kleindorfer, P. R., H. C. Kunreuther, and P. G. H.

Hertwich, Hammitt, and Pease, A Theoretical Foundation for LCA 27


y FO RUM

Schoemaker. 1993. Decision sciences: An integra- Raiffa, H. 1968. Decision analysis . Readin g, MA:
tive perspective . Cambridge: Cambridge Univer- Addison-Wesley.
sity Press. Renn, O. 1998. The role of risk perception for risk
Kuhn, T. S. 1962. The structure of scientific revolutions . management. Reliability Engineering & System
Chicago: University of Chicago Press. Safety 59(1): 49–62.
Latin, H. A. 1982. Environmental deregulation and Schelling, T. C. 1995. Intergenerational discounting.
consumer decision making under uncertainty. Energy Policy 23: 395–401.
Harvard Environmental Law Review 6: 187–239. Scheringer, M. 1999. Persistenze und Reichweite von
Laudan, L. 1998. Underdetermination . In Routledge Umweltchem ikalien. Weinheim, Germany:
encyclopedi a of philosophy , edited by E. Craig. Wiley-vch.
London: Routledge. 9: 527–529. Schimel, D., D. Alves, and I. Enting. 1996. Radiative
Lundie, S. and G. Huppes. 1999. Environmental as- forcing of climate change. In Climate Change
sessment of products— The ranges of societal 1995—The science of climate change, edited by the
preferences method. International Journal of Life Intergovernmental Panel on Climate Change.
Cycle Assessment 4(1): 7–15. Cambridge: Cambridge University Press.
Lycan, W. G. 1998. Theoretical (epistemic) virtues. Shrader-Freche tte, K. S. 1991. Risk and rationality:
In Routledge encyclopedia of philosophy , edited by Philosophical foundations for populist reforms. Ber-
E. Craig. London, New York: Routledge. keley: University of California Press.
Malakoff, D. 1999. Bayes offers a ‘new’ way to make Simon, H. A. 1957. Models of man. New York: John
sense of numbers. Science 286(5444) : 1460– Wiley & Sons.
1464. Sober, E. 1998. Simplicity (in scientific theories). In
Marsmann, M., S. O. Ryding, H. Udo de Haes, J. Fava, Routledge encyclopedia of philosophy , edited by E.
W. Owens, K. Brady, K. Saur, and R. Schenck. Craig. London: Routledge.
1999. In reply to Hertwich & Pease Int. J. LCA Steen, B. 1999. A systematic approach to environmental
3(4), S. 180–181 “ISO 14042 restricts use and priority strategies in product development (EPS).
development of impact assessment.” International Version 2000— General system characte ristics.
Journal of Life Cycle Assessment 4(2): 65. CPM 1999:4. Gothenburg: Chalmers Univer-
Newell, S. A. 1998. Strategic evaluation of environmen- sity. http:\\www.cpm.chalmers.se.
tal metrics: Making use of life cycle inventories. Tietenberg, T. H. 1992. Environmental and natural re-
Technology, Management and Policy. Cambridge: source economics. New York: HarperCollins.
Massachusetts Institute of Technology. Tresch, R. W. 1981. Public finance: A normative theory.
O’Hara, S. U. 1996. Discursive ethics in ecosystem Plano, TX: Business Applications.
valuation and environmental policy. Ecological Tukker, A. 1998. Uncertainty in life cycle impact as-
Economics 16: 95–107. sessment of toxic releases. International Journal of
Owens, J. W., L. Barnthouse, J. Fava, K. Humphreys, Life Cycle Assessment 3(5): 246–258.
B. Hunt, S. Noesen, J. A. Todd, B. Vigon, K. Udo de Haes, H. A. 1999. ISO’s compromise on com-
Weitz, et al. 1997. Life-cycle impact assessment: parative assertions in Life Cycle Impact Assess-
The state-of-the art. Pensacola, FL: Society of En- ment. Journal of Industrial Ecology 2(3): 4–7.
vironmental Toxicology and Chemistry. Udo de Haes, H. A., O. Jolliet, G. Finnveden, M.
PCCRARM. 1997. Risk assessment and risk manage- Hauschild, W. Krewitt, and R. Müller-Wenk.
ment in regulatory decision-making . Washington, 1999. Best available practice regarding impact
DC: The presidential/congressional commission categories and category indicators in Life Cycle
on risk assessment and risk management. Impact Assessment. Brussels, Belgium: SETAC.

28 Journal of Industrial Ecology

You might also like